article
stringlengths 0
745k
| abstract
stringlengths 0
92.7k
|
|---|---|
the computation of the double differential scattering cross section ( ddcs ) at low incident neutron energies is required to solve neutron transport problems .
wigner and wilkins @xcite used a two - body kinematic approach to study potential scattering in a free gas . under the same classic assumptions , blackshaw and murray @xcite studied the case of an energy - dependent cross section and
further generalizations were investigated by ouisloumen and sanchez @xcite and rothenstein and dagan @xcite . a general quantum formalism due primarily to van hove @xcite expresses the ddcs of potential scattering by a bound nucleus as a fourier transform of a 2-point correlation function .
kazarnovski et al .
@xcite and word and trammell @xcite extended the van hove theory to study resonant processes .
the resonant ddcs becomes a fourier double - laplace transform of a 4-point correlation function . in the case of a harmonic crystal ,
the numerical computation of the multifold integral is difficult because of the highly oscillating behavior of the 4-point correlation function .
a simplified multiphonon model , known as the `` uncoupled phonon approximation '' ( upa ) was proposed by naberejnev @xcite .
however , the validity of this approximation is questioned especially within the short time range ( case of high temperature for instance ) .
the present work suggests a new approximation that gives a correct limit at short time and provides a simple formula to compute the scattering kernel .
when the scattering amplitude is independent of incident neutron energy , it is known since the work of van hove @xcite that the scattering ddcs is related to the fourier transform of a two point correlation function @xmath2 : @xmath3 @xmath4 is the energy transfer with @xmath5 and @xmath6 the energy of the neutron before and after the scattering event .
the fourier transform is defined as @xmath7 @xmath8 is the usual scattering function @xmath9 which depends on the momentum transfer @xmath10 where @xmath11 and @xmath12 are the initial and final neutron wave vector respectively .
the theory was later generalized to treat the case of resonant scattering @xcite . in the incoherent approximation ,
the scattering cross section includes the resonant , the potential - resonant interference and potential terms .
it can be demonstrated that the ddcs of the resonant term for instance is related to the fourier - double - laplace transform @xmath13 of a 4-point correlation function @xmath14 : @xmath15 @xmath16 , @xmath17 and @xmath18 are the energy , the neutron and total width of the resonance . @xmath19 and @xmath20 is the usual statistical spin factor .
the fourier - double - laplace transform is defined as @xmath21 similarly , the resonant - potential interference term is expressed as a fourier - single - laplace transform of a 3 point correlation function @xmath22 .
a quantum calculation shows that @xmath14 is a function of time - dependant displacement operators : @xmath23 note that @xmath24 and @xmath25 .
@xmath26 denotes the thermal average at temperature @xmath27 . in this paper , time @xmath28 , @xmath29 and @xmath30
are expressed in unit of energy and @xmath31 . the right hand side of equation 1 and 3 should be multiplied by @xmath32 which is implicitly assumed in the paper . @xmath33 and @xmath34 are the mass of the target nucleus and incident neutron and @xmath35 respectively .
the scattering is assumed isotropic in the center - of - mass frame and multiple scattering effects are neglected ( single collision approximation ) .
equation 3 assumes that the hamiltonian of the target and compound nucleus is the same .
this condition neglects the mass change of the target nucleus when the neutron is absorbed and re - emitted : @xmath36 . in the case where the target nucleus behaves like a free gas ,
the displacement operator is proportional to the momentum operator @xmath37 . after averaging over a maxwellian distribution of momentum at temperature @xmath27 , @xmath38 within the free gas model ( fgm ) becomes @xmath39 - \frac{kt}{2 m }
[ -\vec k_i ( t - t ' ) + \tau ( \vec k_f - \vec k_i ) ] ^2 \}.\ ] ] when the target nucleus is bound to a crystal lattice , the 4-point correlation function is more complex . under the harmonic approximation ,
the bloch theorem @xcite transforms @xmath38 into @xmath40 for a cubic lattice with a phonon density of states @xmath41 , @xmath42}.\ ] ] in the short time approximation for @xmath28 , @xmath29 , and @xmath30 the ddcs can be simplified using @xmath43 with @xmath44 plugging [ eqrho ] into [ eqclm ] , we get the free gas formula [ eqfgm ] with @xmath45 instead of @xmath27 . consequently , the crystal - lattice model lead to the free gas model , with an effective temperature , when @xmath28 , @xmath29 , and @xmath30 are small ( short time approximation ) .
this is the same effective temperature as lamb @xcite derived for the doppler broadening of a capture resonance in a solid in the weak binding limit .
the upa approximation proposed in @xcite was an attempt to compute the resonant scattering kernel for a harmonic crystal .
it neglected @xmath29 and @xmath30 in the coupling terms @xmath46 , @xmath47 and @xmath48 in equation [ eqclm ] and applied a short time approximation to @xmath49 and @xmath50 . and
@xmath50 with a discrete phonon spectrum ] the correlation function becomes @xmath51 \right).\end{aligned}\ ] ] the integral over @xmath28 and @xmath52 can be performed separately . since the first exponential is the van - hove function @xmath2 ,
the differential cross section can be factored out as @xmath53 the detailed mathematical expression for the so - called upa cross section @xmath54 can be found in @xcite .
the upa model separates the phonon dynamic and the nuclear interaction and provides a rather simple way to compute ddcs .
however , within the upa approximation , the 4-point correlation for small @xmath28 , @xmath29 , and @xmath30 becomes @xmath55 - \frac{kt_{eff}}{2 m } [ \vec k_i \vec k_f ( t^2 + t'^2 ) + \tau^2 ( \vec k_f - \vec k_i ) ^2 ] \},\ ] ] which differs markedly from the required equation [ eqfgm ] discussed in the previous section .
consequently , the ddcs within the upa approximation does not give the correct limit for small @xmath29 and @xmath30 .
our model seeks to get a factorization of the ddcs similar to the upa model : @xmath56 however , in order to get the correct limit at short time , we computed the @xmath57 term using the short time approximation of the correlation function ( equation [ eqfgm ] ) , without the previous upa approximation . for values of @xmath58 ,
the following notation is used , @xmath59\ /\
\gamma.\ ] ] it is found that the differential cross section , including resonant , potential and resonant - potential interference can be calculated almost analytically and can be split into nuclear and transfer terms .
the details of the demonstration are presented in the appendix .
the final result is : @xmath60.\ ] ] for @xmath61 and @xmath58 , @xmath62 for @xmath63 and @xmath58 @xmath64 where @xmath65 and @xmath66 , are the real and imaginary part of complex complementary error function @xmath67\ erfc\ [ -i(x+iy)]$ ] , related to the classic voigt functions .
@xmath68 = \frac{y}{\pi}\int_{-\infty}^{\infty}\frac{\exp ( -t^2)}{(x - t)^2 + y^2 } dt\ ] ] @xmath69 = \frac{x}{\pi}\int_{-\infty}^{\infty}\frac{\exp ( -t^2)}{(x - t)^2 + y^2 } dt.\ ] ] when @xmath70 , the same calculations lead to equation [ eqfinal ] with : @xmath71\ /\ \gamma.\ ] ] we recognize the usual doppler - broadened cross section @xmath72 , so that the differential cross section becomes @xmath73 where @xmath74 and @xmath75 depend on the cosine of the scattering angle @xmath76 : @xmath77 and for @xmath70 @xmath78 equation [ eqfinalresult ] is valid only when @xmath79 is positive .
the previous equation can be further simplified noting that @xmath80[1+\delta e/(2 e_i ) ] \approx t ( 1+\mu ) /2 $ ] and when @xmath81 we have @xmath82 and this approximation produces satisfactory results when used over the whole range of values of @xmath83 . making these approximations
, we get a simple formula @xmath84/2 -e_i\mu / a \right).\ ] ] equation [ eqfinalresult2 ] provides a simple way to compute the differential cross section within the free gas model .
it also gives an approximate way to account for solid state effects by using the known scattering function @xmath85 for the harmonic crystal .
note that equation [ eqfinalresult2 ] was demonstrated for a single resonance and it is not known if this formula can be generalized to any form of free scattering amplitude .
low - enriched uranium oxide uo@xmath86 is widely used as nuclear - reactor fuel . with the present model
, the scattering kernel of @xmath87u in uo@xmath86 has been calculated near the first resonance at 6.67 ev by numerical integration of the differential cross section over scattering angles .
the weighted phonon spectrum for @xmath87u in uo@xmath86 has been taken from the measurement of dolling et al . @xcite .
the spectrum features two acoustic modes around 14 and 21 mev .
the @xmath88 van - hove scattering function has been computed using the usual phonon expansion methods .
the @xmath87u resonance parameters evaluated by moxon and sowerby @xcite were used .
the present model has been compared with the classic free gas kernel published by blackshaw and murray @xcite and studied by ouisloumen and sanchez @xcite .
fig . 1 and 2
show the resulting scattering kernel at 300 k at two incident neutron energies , 6.674 ev ( the peak of the first resonance ) and 6.520 ev which is the minimum of the cross section due the interference between the resonant and potential amplitudes .
compared to the upa model ( see figures in reference @xcite ) , the present crystal model gives results closer to the free gas model .
the only distinctive features are the small peaks around the elastic peak , associated with the one - phonon excitation of the acoustic modes .
note that when the @xmath89 of the fgm is used in equation [ eqfinal ] , the numerical computations give the same results as ouisloumen and sanchez .
comparison of scattering kernels calculated with the present formalism ( black line ) and the free gas model of ouisloumen and sanchez @xcite ( grey line ) at incident neutron energy of 6.674 ev .
the dotted line is the kernel with a static model . ]
comparison of scattering kernels calculated with the present formalism ( black line ) and the free gas model of ouisloumen and sanchez @xcite ( grey line ) at incident neutron energy of 6.520 ev .
the dotted line is the kernel with a static model and the dashed line is the scattering cross section ( @xmath90 in barns ) . ]
an approximate formula is proposed to treat solid state effects in neutron - crystal interactions .
the ddcs is the product of a doppler - broadened scattering cross section and the usual van - hove scattering function .
@xmath91/2 -e_i\mu / a \right).\ ] ] this formula gives the correct free - gas limit for short - time range ( high - temperature cases ) .
a rigorous calculation of the ddcs within the crystal lattice model is still an open issue .
the present work develops the case of a single isolated resonance but further generalizations to multilevel forms of collision matrix may be possible . in reactor applications ,
the model predicts small solid - state effects in the scattering of neutrons in uo@xmath86 at 300k , contrary to previous studies using the upa approximation .
measurements of the secondary spectrum of neutrons scattered elastically in resonances would be valuable to check existing scattering models .
it is a pleasure to acknowledge fruitful discussions with r. dagan , c. r. lubitz , and a. santamarina .
for @xmath58 , @xmath38 in equation [ eqfgm ] can be transformed into @xmath92 \nonumber \\ & - & \frac{kt}{2 m } [ \left(\tau - \frac{\vec k_i \vec \delta k ( t - t')}{\vec{\delta k}^2 } \right ) ^2
\vec \delta k^2 - \frac{k_i^2 k_f^2 - ( \vec k_i \vec k_f)^2}{\vec{\delta k}^2 } ( t - t')^2 ] \}.\end{aligned}\ ] ] using the variable @xmath93 , we recognize @xmath94 the pair correlation function for a free gas : @xmath95 therefore , the fourier transform of @xmath38 ( integration over @xmath28 ) will lead to the @xmath96 terms with a phase factor to account for the change of variable @xmath97 .
the ddcs takes the form of equation [ eqfact ] with @xmath98 with @xmath99 and @xmath100 to compute the double laplace transform , the following identity is used : @xmath101 equation [ eqint2 ] becomes @xmath102 c.c denotes the conjugate complex .
this equation is then reduced into a single integral which can be further simplified using @xmath103\ ] ] we recognize in [ eqint3 ] and [ eqc ] the real of the complex complementary error function .
equations [ eqfinal ] , [ eqfinal2 ] and [ eqfinal3 ] are obtained in a similar way when the potential and potential - resonant interference terms are accounted for .
|
in the theory of resonant scattering , the double differential cross section involves the computation of a multifold integral of a 4-point correlation function , which generalizes the traditional 2-point correlation function of van - hove for potential scattering . in the case of a neutron - crystal interaction , the numerical computation of these multifold integrals
is cumbersome . in this paper ,
a new approximation is suggested .
it is based on a factorization of the differential cross section into one function describing the exchange of kinetic energy between the neutron and the bound nucleus ( phonons dynamic ) and a function related to the nuclear scattering amplitude .
this formalism is then applied to the modeling of resonant scattering of a neutron by @xmath0 in a @xmath1 crystal lattice .
|
the detection of redshifted 21 cm signals from the early universe is attracting many attentions in the study of cosmology , because it provides a window to probe the baryonic gas and the first generation of light sources in the cosmic dark ages ( e.g. furlanetto et al .
observational projects of highly sensitive meter radio telescopes , such as the low frequency array ( lofar ) ( rottgering 2003 ) , the square kilometer array ( ska ) ( van de weygaert & van albada 1996 ) , the mileura widefield array ( mwa ) and the 21 cm array ( 21cma ) ( pen et al .
2004 ) , are ongoing or being planned .
a major challenge to detect the primordial 21 cm signals is from the contamination of radio signals of the foreground .
the brightness temperature fluctuations of the 21 cm signals would be seriously contaminated by high - redshift radio point sources , free - free emissions of the igm , and the noises from the artificial radio interference in the vhf band ( di matteo et al .
2002 , 2004 ; oh & mack 2003 ) .
the power spectrum of the brightness temperature fluctuations of the redshifted 21 cm map may even be dominated by the fluctuations of the foreground ( di matteo et al .
thus , we should search for useful features of the primordial 21 signals to distinguish it from the noisy foreground .
for instance , the patchy structures of the 21 cm brightness temperature field are found to be highly non - gaussian , which would be helpful to identify the 21 cm signals from the noisy foreground ( he et al .
2004 , mellema et al .
. moreover , the correlation between the 21 cm signals and other emissions of the sources would also be useful for identifying high redshift sources .
in this context , some studies focused on the 21 cm signals from individual uv ionizing sources in the reionization epoch .
their features may provide a direct identification of the ionized patches of the reionization ( e.g. , tozzi et al . 2000 ) .
very recently , the 21 cm emission and absorption around isolated ionizing sources has been calculated ( chuzhoy et al .
2006 ; cen 2006 ; chen & miralda - escude 2006 ) . however , their results , especially the time - dependence of the profile of the 21 cm emission and absorption , still do not quite agree with each other .
this problem comes partially from the fact that they dropped the time derivative term of the radiative transfer equation , i.e. , the speed of light is taken to be infinite
. this approximation would be reasonable if the retardation effect is negligible .
however , it is not the case for the 21 cm problem , for which we note that the size of the hii regions and the age of the sources used in these calculations have already been close to , or even violated , the retardation constraint @xmath9 . the retardation effect , or the effect of finite velocity of light , is not trivial . the time and space dependencies of the ionized and heated regions are substantially affected by the retardation effect ( white et al . 2003 ;
shapiro et al . 2006a ; qiu et al 2007 ) .
therefore , it is necessary to re - investigate this problem by fully considering the kinetics of photons in the phase space .
we will use the weno algorithm for the radiative transfer calculation , which has been developed by qiu et al .
( 2006 , 2007 ) .
the weno algorithm has been shown to have high order of accuracy and good convergence in capturing discontinuities and complicated structures ( shu 2003 ) .
the weno algorithm is able to incorporate the temperature and species equations and to provide stable and robust solutions .
it is also able to calculate the fractions of neutral hydrogen and helium within the ionized spheres , which generally are extremely small , but would be important to understand the possible correlation of the 21 cm and ly@xmath10 signals . in 2 we present the ionization and temperature profiles of the igm around a uv ionizing source .
3 shows the 21 cm emission and absorption regions , and their dependency on intensities , frequency - spectrum and life time of the uv photon sources .
discussions and conclusions are presented in 4 .
the algorithm is summarized in the appendix .
the whole process of reionization is very complex due to the inhomogeneous and clustering nature of the structure formation ( e.g. ricotti et al .
2002 ; razoumov & norman 2002 ; ciardi et al .
therefore , the configuration of the ionized regions generally would be non - spherical . however , in order to be comparable with previous studies ( chuzhoy et al . 2006 ; cen 2006 ; chen & miralda - escude 2006 ) , we will consider the ionization and heating of a uniform igm around a point source .
this simplified model is reasonable to reveal the features of the formation and evolution of the 21 cm emission and absorption regions .
the radiative transfer equation of the specific intensity @xmath11 is ( e.g. bernstein 1988 ) @xmath12 where @xmath13 is the photon frequency and @xmath14 is the hubble parameter at time @xmath15 .
supposing the source is at the center @xmath16 , we have @xmath17 , where @xmath18 is the energy of photons emitted from the central source per unit time within the frequency range from @xmath13 to @xmath19 . the source term can equivalently be described by a boundary condition as @xmath20 we will consider two kinds of frequency sepctrum : a power law spectrum @xmath21 with the total intensity ( energy per unit time ) @xmath22 and a blackbody spectrum .
the effect of the expansion of the universe on the number density is negligible if the time scale concerned is a few myr at the epoch @xmath23 .
however , the effect of the redshift of the photon energy is essential to calculate the flux of ly@xmath10 photons ( cen 2006 )
. therefore , we will keep the @xmath14 term on the right hand side of eq.(1 ) . the absorption coefficient in eq.(1 )
is given by @xmath24 and @xmath25 where @xmath26 runs over the components @xmath27 , @xmath28 and @xmath29 .
@xmath30 and @xmath31 are , respectively , the cross section of ionization and the number density of component @xmath26 .
the number density @xmath32 is calculated by solving the species equations involving ionization and recombination processes .
we use the on - the - spot approximation to take into account the effect of the diffusing photons and treat the coupling of h and he by the radiation of hei recombination continuum and bound - bound transitions following the method in osterbrock ( 1989 ) .
the corresponding coefficients for the ionization and recombination processes are taken from fukugita & kawasaki ( 1994 ) and theuns et al .
( 1998 ) .
the igm is heated by the uv photons with a rate in unit erg @xmath33 s@xmath34 as @xmath35 where @xmath36 is the threshold of frequency to ionize component @xmath26 .
the fraction of photon - electron energy converted to heat @xmath37 depends on the free electron fraction @xmath38 and the photon frequency @xmath13 .
we use the fitting formula of @xmath37 given by shull & van steenberg ( 1985 ) .
the corresponding coefficients for radiative cooling processes are also taken from theuns et al .
( 1998 ) .
the algorithm of the numerical solution of the radiative transfer equation coupled with the temperature and species equations of hi , hei and heii are described in qiu et al .
( 2006 , 2007 ) .
a fifth - order finite difference weno scheme and a third - order tvd runge - kutta time discretization are applied to the spatial and time derivative , respectively .
a semi - implicit method is used for the temperature and species equations .
the details and tests of the numerical algorithm are described in qiu et al .
( 2006 , 2007 ) and we give a brief description of the algorithm in the appendix .
we first calculate the profile of an ionized sphere around a point source at @xmath1 . as an example
, we consider a source of a power law spectrum with @xmath39 erg s@xmath34 and a spctral index @xmath4 , which are the typical values of qsos at the euv band ( e.g. zheng et al .
the initial ionization fraction of hydrogen and helium is set to be @xmath40 , which is the remaining ionization fraction after recombination , and the initial temperature of the baryonic gas is set to be 10 k. a @xmath41cdm cosmology model with @xmath42 , @xmath43 , and @xmath44 is adopted .
the profiles of the ionized sphere at time 0.575 , 1.15 , 1.725 , 2.3 and 3 myr are shown in figure 1 , in which @xmath45 , @xmath46 , @xmath47 , @xmath48 and @xmath49 , where @xmath50 and @xmath51 .
figure 1 shows that although the main ionization zones have steep eges ( i - front ) , there are still a small fraction of hydrogen and helium been ionized beyond the i - fronts .
this ionization is produced by high energy photons which can effectively penetrate into the igm further than photons near the ionization thresholds .
the fraction @xmath52 has a long tail before it approaches the initial value @xmath40 ; the fraction of heii shows a similar behavior as that of hii .
the i - fronts of heii and heiii generally are different from that of hii , and the decrease of the fraction of heiii with @xmath53 is much faster than that of hii and heii .
we define the i - front @xmath54 as the location where @xmath55 .
the speed of the i - front @xmath56 is plotted as the solid line in figure 2 .
it shows that the evolution of the ionized sphere has two phases separated by a characteristic time @xmath57 . in the first phase @xmath58 , @xmath59 , and in the second phase @xmath60 , the speed decreases with @xmath15 in a power law , @xmath61
this feature of two phases was first pointed out by white et al .
( 2003 ) with an approximate solution by neglecting recombinations .
this feature is actually not sensitive to the choice of the number @xmath40 ; if we take @xmath62 , the two phases feature still holds ( qiu et al .
2007 ) . within the i - front
, @xmath63 is on the order of @xmath64 and corresponds to a column density of the order of 10@xmath65 @xmath66 on a physical length scale of the order of 100 kpc .
this value of column density is just of the same order as the weak ly@xmath10 absorptions ( bi & davidsen 1997 ) .
therefore , the ionized sphere is not completely gunn - peterson transparent , but may produce ly@xmath10 absorption features on the spectra of background sources . on the other hand , within the ionization front , @xmath67 , and
therefore , the gunn - peterson optical depth of heii ly@xmath10 is larger than that of hi .
it leads to significant fluctuations of the ratio between the two optical depths ( e.g. shull et al . 2004
; liu et al . 2006 ) .
we also calculate the ionized sphere of sources with intensities @xmath68 and @xmath69 erg s@xmath34 and a power law index @xmath4 .
their i - front speeds are also plotted in figure 2 .
it shows that all the three curves of @xmath70 have the two phases . especially , in the second phase the power law @xmath71 is approximately scaled with respect to @xmath72 , in other words , the characteristic time @xmath73 .
this scaling can not be found with static solutions of the radiative transfer equation or by assuming the speed of light is infinite ( e.g. cen , apjs , 2002 ; whalen & norman 2006 ) .
figure 3 shows the corresponding kinetic temperature distribution for the ionizing source with @xmath39 erg s@xmath34 and a power law index @xmath4 at time 0.575 , 1.15 , 1.725 , 2.3 and 3 myr .
the kinetic temperature basically keeps constant around @xmath74 k within the ionized sphere , which is determined by the balance between the photon - ionization heating and radiative cooling .
an interesting feature is that the temperature at the center ( @xmath75 ) is lower than the constant temperature .
this is because the ionizing photon flux at the center is very strong , the number of @xmath63 at @xmath76 is extremely low ( see figure 1 ) , and therefore , the heating rate is low .
figure 3 shows that the gas is also substantially heated beyond the i - front . actually , the energy spectrum of the radiation is significantly hardened around the i - front , because soft photons are effective to ionize the igm , and hard photons are more effective to heat the igm ( qiu et al 2007 ) . after the time @xmath57
, soft photons are exhausted within the ionized sphere @xmath77 ; in contrast , hard photons are still abundant in the region @xmath78 , as hard photons can penetrate further beyond the ionized sphere .
the region between the i - front and the front of light @xmath79 is a pre - heating layer .
the temperature profile is different at different time and more elongated at a later time .
this is because the speed of the i - front is much less than the speed of light after @xmath57 , while the spped of the front of heating is about the same as the speed of light .
we introduce a radius @xmath80 by the solution of @xmath81 k , which is about the cmb temperature at @xmath1 .
we call the high temperature region to be @xmath82 and the low temperature region @xmath83 . the latter will play the role of the 21 cm absorption .
since the speed of the radius @xmath80 is also less than @xmath84 , the low temperature region increases quickly with time .
therefore , the kinetic temperature distribution shows a long tail in the range @xmath85 k. we also calculate the kinetic temperature distribution of sources with intensities @xmath86 and @xmath69 erg s@xmath34 and a power law index @xmath4 .
the profiles of the temperature distribution is substantially @xmath87-dependent .
the 21 cm ( @xmath88 1420 mhz ) line emission or absorption associated with the spin - flip transition of the electron in neutral hydrogen is significant if the number density @xmath89 of neutral hydrogen is high and the spin temperature @xmath90 , which describes the number ratio between the spin - up and the spin - down states , deviates from the temperature of cosmic microwave background ( cmb ) @xmath91 .
the preheating region discussed in 2.3 , i.e. , the region between the i - front and the front of light @xmath79 , would satisfy these two conditions .
we may call this region as a 21 cm shell .
we do not consider the bulk motion of the baryonic fluid .
thus , when @xmath90 is much larger than the hydrogen hyperfine energy @xmath92 k , the optical depth of the 21 cm absorption @xmath93 is given by ( wild 1952 ; field 1959 ) @xmath94 ( 1+z)^{1/2},\end{aligned}\ ] ] since @xmath95 , the observed brightness temperature excess ( or deficit ) for the mean gas density of the universe at the redshifted frequency @xmath96 is @xmath97 ( 1+z)^{1/2 } \hspace{3 mm } { \rm k}.\ ] ] the decoupling of the spin temperature @xmath90 from the cmb temperature @xmath91 could be achieved by the collision of neutral hydrogen and scattering of ly@xmath10 photons , which is the wouthuysen - field mechanism ( wouthuysen 1952 ; field 1959 ) .
the spin temperature @xmath90 is then given by a weighted average of the cmb temperature @xmath91 , the kinetic temperature @xmath98 , and the color temperature @xmath99 of background photons at ly@xmath10 frequency ( field 1958 , 1959 ) @xmath100 where @xmath101 is the ly@xmath10 photon coupling coefficient and @xmath102 is the collision coupling coefficient . at redshift @xmath103 ,
the collision coupling is only efficient in overdense regions ( e.g. furlanetto & loeb 2002 ; shapiro et al .
2006b ) , and is negligible for cosmological mean densities . in the low density igm , the wouthuysen - field mechanism would play the major role to decouple @xmath90 from @xmath91 .
the resonant scattering of ly@xmath10 photons by neutral hydrogen leads to that the color temperature of the radiation spectrum near ly@xmath10 frequency approaches the kinetic temperature of the baryonic gas . therefore ,
if there are enough ly@xmath10 photons , we have @xmath104 . there are two important ways to contribute ly@xmath10 photons : one is from the continuum spectrum photons redshifted to ly@xmath10 frequency , which is governed by the @xmath14 term in eq.(1 ) . the other one is from the excitation of neutral hydrogen by high energy electrons , and we take the fitting formula of the fraction of excitation for high energy electrons from shull & van steenberg ( 1985 ) .
figure 4 shows the profiles of the spin temperature @xmath90 and the brightness temperature @xmath105 for a point source with @xmath39 erg s@xmath34 and a power law index @xmath4 at time 0.575 , 1.15 , 1.725 , 2.3 and 3 myr .
the spin temperature @xmath90 for @xmath106 k is set to be @xmath98 ; this will not affect our results since hi is highly ionized for @xmath106 k. in the 21 cm shell , i.e. , the region between the i - front and the front of light @xmath79 , the brightness temperature @xmath107 .
in the early stage , when the speed of the i - front @xmath108 is close to the speed of light , there is no 21 cm shell .
the 21 emission and absorption regions become broad when the difference between @xmath54 and @xmath79 is significant .
figure 4 shows that the 21 cm layer always contains an emission region @xmath109 and an absorption region @xmath110 .
the maximum brightness temperature in the emission region is @xmath111 mk , while the minimum temperature of the absorption region is @xmath112 mk .
the absorption region grows faster than the emission region . before 1 myr
, the emission region is comparable with the absorption region , but at 3 myr , the emission region is about 2 comoving mpc while the absorption region exceeds 10 comoving mpc .
a prominent feature of the brightness temperature in figure 4 is that the absorption region has a sharp cut - off at the front of light @xmath79 .
this comes from the fact that in our model the wouthuysen - field coupling is efficient within the entire range of the photon propagation ; consequently , in the region @xmath113 , the ly@xmath10 photons are abundant enough to lock @xmath90 with @xmath98 .
these photons come from the redshifted photons originally in the energy range between ionization energy and ly@xmath10 line . in this case
, the edge of the 21 cm absorption region is an indicator of the lifetime of the ionzing source .
we also calculate the profiles of the spin temperature @xmath90 and the brightness temperature @xmath105 of sources with intensities @xmath68 and @xmath69 erg s@xmath34 and a power law index @xmath4 .
the results are presented in figure 5 .
comparing figures 4 and 5 , one can see that the maximum ( emission ) and minimum ( absorption ) of @xmath105 basically are independent of the source intensity , while the profile of the 21 cm brightness temperature and its evolution depend strongly on the intensity .
this is because the heating processes are sensitive to the evolution of photons in the phase - space , which also can not be described by a static approach .
figure 6 shows the ratio between the sizes of emission and absorption regions , which decreases with time , because the growth of the absorption region ( brightness temperature @xmath110 ) is faster than that of the emission region . for a source with @xmath114 erg s@xmath34 ,
the size of the emission region is larger than that of the absorption region at an early stage @xmath115 myr , while for sources with @xmath116 and @xmath117 erg s@xmath34 , the size of the emission region is always smaller than that of the absorption region .
figure 7 shows the profiles of the spin temperature @xmath90 and the brightness temperature @xmath105 for an ionizing source at @xmath1 with the spectral index @xmath118 and 4 .
the intensities for both cases are normalized to be @xmath39 erg s@xmath34 , i.e. , the same as the source of 2.2 .
comparing figures 3 and 7 , we see that the size of the 21 cm emission region of both @xmath118 and 4 is narrower than that of @xmath4 .
this is due to the lack of high energy photons for heating .
on the other hand , the absorption regions are more notable than that in the case of @xmath4 , as the sources of @xmath118 and 4 have stronger ly@xmath10 photon fluxes than @xmath4 .
the lights from first galaxies may have a thermal spectrum .
let us consider first galaxies consisting of stars with a blackbody spectrum and an effective temperature @xmath7 k , which corresponds to population iii stars with mass 200 - 500 m@xmath119 ( schaerer et al . 2002 ) .
thermal radiation at temperature @xmath7 k does not contain many photons with energy larger than 100 ev .
therefore , the pre - heating beyond the i - front is very weak , and the 21 cm region is small .
figure 8 shows the spin temperature and the brightness temperature for a first galaxy with the same luminosity as the source for figure 4 .
we see that the size of the 21 cm emission region of figure 8 is much narrower than that of figure 4 . however , the 21 cm absorption region is still significant , because the ly@xmath10 photons are still efficient to support the wouthuysen - field mechanism .
we now consider a source which is the same as that used for figures 1 and 3 , but the source stops to emit photons after 3 myr .
figure 9 shows the kinetic temperature @xmath98 , the spin temperature @xmath90 and the brightness temperature @xmath105 at time 3 , 3.45 , 4.03 and 4.60 myr .
we see that the i - front and @xmath80 still move ahead in the period of 3 to 3.45 myr , but they slow down after 3.45 myr . it is simply due to the fact that there are no more photons to support ionization and heating .
consequently , the 21 cm emission shell shrinks and almost disappears at 4.03 myr .
an interesting result is that the 21 absorption region is alive even when @xmath120 myr , i.e. , 1.6 myr after the photon source has died out .
this is once again due to the abundance of ly@xmath10 photons .
when @xmath121 myr , no high energy photons can reach @xmath122 mpc , while the wouthuysen - field mechanism is working to the size @xmath79 .
figure 9 shows that the absorption region at @xmath123 myr is limited within two sharp cut - offs : @xmath124 .
it is easy to check that @xmath125 and @xmath126 , which is just the range given by the retardation of the first and last light signals from the source with a luminous lifetime of 3 myr . while including the thermal broadening , eq.(5 )
should be replaced by @xmath127 where the factor @xmath128 is the normalized doppler broadening line profile as @xmath129 with @xmath130 . for gas with temperature
@xmath131 k , we have @xmath132 , and the comoving scale is thus @xmath133 . in the @xmath41cdm model at @xmath134 , we have @xmath135 h@xmath34 mpc .
if the thickness of the 21 cm emission shell is less than @xmath136 , the doppler broadening will smooth out the 21 cm emission . as discussed in previous sections ,
the 21 cm emission regions are narrow for weak sources and high spectral index .
therefore , the 21 cm emission would not be observable for sources with @xmath87 much less than @xmath137 ergs s@xmath34 or @xmath138 ergs s@xmath34 , but @xmath139 .
we have studied the formation and evolution of the ionized and heated region around a uv ionizing source in the reionization epoch . in the first stage ,
the ionizing and heating fronts of the igm are coincident and propagate with a speed close to the speed of light .
the evolution enters the second stage when the frequency spectrum of uv photons is hardened due to the loss of photons by reionization . in this stage ,
the propagating speed of the i - front is less than the speed of light , but high energy photons and ly@xmath10 photons can still reach the front of light @xmath79 .
the spherical shell between the i - front @xmath54 and @xmath79 is the region in which 21 cm signals are produced .
the inner shell from @xmath54 to @xmath80 gives 21 cm emissions ; the outer shell from @xmath80 to @xmath79 produces 21 cm absorptions .
both the 21 cm emission and absorption regions are sensitively dependent on the intensity , frequency - spectrum and life - time of the ionizing source .
if the intensity is much less than @xmath140 ergs s@xmath34 and the spectral index @xmath141 is larger than 3 , the emission region is very narrow , and the signal will probably be smoothed out by thermal broadening . for sources with intensity
@xmath142 ergs s@xmath34 and a spectral index @xmath143 , the comoving size @xmath144 is about 1 - 5 mpc .
the brightness temperature excess upon the cmb is about 30 mk .
therefore , it yields a radio signal near the band of 70 mhz and is resolvable with observations of angular resolution better than 0.5 arcmin and spectral resolution better than 40 khz .
the 21 cm absorption region has very different behaviors from the emission regions .
absorption region mainly depends on the coupling of the spin temperature and the kinetic temperature , which is efficient if there are enough ly@xmath10 photons .
the ionizing source with a power law spectrum or thermal spectrum at @xmath145 can always provide enough ly@xmath10 photons until the largest distance @xmath79 .
that is , the size of the absorption region directly measures the age of the source . for a source with lifetime @xmath146
, the absorption region can survive even when @xmath15 is larger than @xmath146 . in this region , the brightness temperature deficit with respect to the cmb is of the order of a few ten mk .
the comoving size of the absorption region can be as large as 10 - 20 mpc .
moreover , the small fraction of hi and heii remained in the ionized sphere can lead to absorption features on the spectra of background sources .
therefore , the correlated 21 cm emission , absorption and ly@xmath10 signals would be helpful to identify ionized patches in the early universe .
although all the calculations of this paper are on spherical regions , many features revealed with the spherical profiles would also hold for non - spherical profiles .
this is because many features of the formation and evolution of 21 cm signals depend only on the growth of the ionized and heated regions , but not on the details of their configuration .
for instance , it can be shown that the growth of a non - spherical ionized region around clustered sources also consists of two phases as that shown in figure 2 . in the first phase , the speed of the i - front is also close to the speed of light , even though in these cases we can not use one ionized radius @xmath147 to describe the ionized region . moreover ,
similar to the spherical ionized regions , the characteristic time @xmath57 is longer for clustered sources with higher intensity @xmath87 .
therefore , the stronger the sources , the later the formation of the 21 cm regions .
this property is regardless of the geometry of the clustered sources ( qiu et al .
2007 , in preparation ) . in our calculations , the dynamical behavior of cosmic baryonic gas
is ignored .
such an approximation would be reasonable when the speed of the sound wave are very different from the speed of the formation and evolution of the 21 cm emission regions .
most features of the 21 cm emissions and absorptions should still hold when the hydrodynamics of the baryonic gas is taken into account .
we thank the anonymous referee for useful comments and suggestions .
this work is supported in part by the us nsf under the grants ast-0506734 and ast-0507340 .
j.l acknowledges the financial support from the international center for relativistic astrophysics .
l.f acknowledges support from the national science foundation of china under the grant 10573036 .
to approximate the spatial derivative in equation ( 1 ) , the fifth - order finite difference weno scheme is used .
specifically , to calculate @xmath148 , the variable @xmath13 is fixed and the approximation is performed along the spatial coordinate @xmath53 @xmath149 where the numerical flux @xmath150 on a uniform mesh @xmath151 at time step @xmath152 is obtained with the procedure given below .
we can use the upwind fluxes without flux splitting in the fifth - order weno approximation because the wind direction is fixed ( positive ) .
first , we denote @xmath153 the numerical flux from the regular weno procedure is obtained by @xmath154 where @xmath155 are the three third order fluxes on three different stencils given by @xmath156 and the nonlinear weights @xmath157 are given by @xmath158 with the linear weights @xmath159 given by @xmath160 and the smoothness indicators @xmath161 given by @xmath162 @xmath163 is a parameter to avoid the denominator to become 0 and is taken as @xmath164 .
the computational domain in @xmath13 is on a non - uniform mesh @xmath165 the integration of heating rate eq .
( 4 ) and photon - ionization rate is approximated by a fourth order quadrature formula @xmath166 where @xmath167 is the threshold frequency of ionization , and the weights @xmath168 are given by @xmath169 the time derivative @xmath170 is calculated using the third - order tvd runge - kutta discretization : @xmath171 we adapt the multi - timescale strategy and use a semi - implicit scheme in the evolution of temperature and species .
it greatly releases the time step restriction , therefore saves computational cost .
we refer to qiu et al .
( 2007 ) for more details of the numerical implementation . , @xmath172 , @xmath173 , @xmath174 , and @xmath175 , the fractions of hi , hii , hei , heii , heiii around a point source at redshift @xmath1 with an intensity @xmath176 erg s@xmath34 and a power low index @xmath4 . from left to right ,
the time is 0.575 , 1.15 , 1.725 , 2.3 and 3 myr , respectively.,width=480,height=672 ] and the brightness temperature @xmath105 for point sources with a spectral index @xmath4 , but with intensities @xmath177 ( left ) and @xmath69 erg s@xmath34 ( right ) . in each figure , from left to right , the time is 0.575 , 1.15 , 1.725 , 2.3 and 3 myr , respectively . , title="fig:",width=307,height=480 ] and the brightness temperature @xmath105 for point sources with a spectral index @xmath4 , but with intensities @xmath177 ( left ) and @xmath69 erg s@xmath34 ( right ) . in each figure , from left to right , the time is 0.575 , 1.15 , 1.725 , 2.3 and 3 myr , respectively . , title="fig:",width=307,height=480 ] and the brightness temperature @xmath105 for a point source with spectral indexes @xmath118 ( left ) and 4 ( right ) .
the intensities are normalized to have the same number of ionizing photons as the model in 2.2 . in each panel , from left to right , the time is 0.575 , 1.15 , 1.725 , 2.3 and 3 myr , respectively . , title="fig:",width=307,height=480 ] and the brightness temperature @xmath105 for a point source with spectral indexes @xmath118 ( left ) and 4 ( right ) .
the intensities are normalized to have the same number of ionizing photons as the model in 2.2 . in each panel , from left to right , the time is 0.575 , 1.15 , 1.725 , 2.3 and 3 myr , respectively . , title="fig:",width=307,height=480 ] and the brightness temperature @xmath105 for a source with a thermal spectrum of @xmath178 with the same luminosity as in figure 4 . in each panel , from left to right , the time is 0.575 , 1.15 , 1.725 , 2.3 and 3 myr , respectively , width=307,height=480 ] , the spin temperature @xmath90 , and the brightness temperature @xmath105 for the same model as in figure 4 , but it stops to emit photons after @xmath179 myr . in each panel , from left to right , the time is 3 , 3.45 . 4.03 and 4.60 myr , respectively .
@xmath180 and @xmath181 are the boundaries of the profile @xmath105 at time 4.60 myr.,width=307,height=480 ]
|
we investigate the 21 cm signals from the uv ionizing sources in the reionization epoch .
the formation and evolution of 21 cm emission and absorption regions depend essentially on the kinetics of photons in the physical and frequency spaces . to solve the radiative transfer equation
, we use the weno algorithm , which is effective to capture the sharp ionization profile and the cut - off at the front of light @xmath0 and to handle the small fraction of neutral hydrogen and helium in the ionized sphere .
we show that a spherical shell of 21 cm emission and absorption will develop around a point source once the speed of the ionization front ( i - front ) is significantly lower than the speed of light .
the 21 cm shell extends from the i - front to the front of light ; its inner part is the emission region and its outer part is the absorption region .
the 21 cm emission region depends strongly on the intensity , frequency - spectrum and life - time of the uv ionizing source . at redshift @xmath1 , for a uv ionizing source with an intensity @xmath2 and a power law spectrum @xmath3 with @xmath4 ,
the emission region has a comoving size of 1 - 3 mpc at time @xmath5 myr .
nevertheless , the emission regions are very small , and would be erased by thermal broadening if the intensity is less than @xmath6 , the frequency spectrum is thermal at temperature @xmath7 k , or the frequency spectrum is a power law with @xmath8 . on the other hand ,
the 21 cm absorption regions are developed in all these cases . for a source of short life - time
, no 21 cm emission region can be formed if the source dies out before the i - front speed is significantly lower than the speed of light . yet , a 21 cm absorption region can form and develop even after the emission of the source ceases .
|
[ definition:1.1 ] let @xmath6 and @xmath7 be positive integers .
a _ block of length @xmath7 in base @xmath6 _ is an ordered @xmath7-tuple of integers in @xmath8 .
a _ block of length @xmath7 _ is a block of length @xmath7 in some base @xmath6 .
block _ is a block of length @xmath7 in base @xmath6 for some integers @xmath7 and @xmath6 .
[ definition:1.2 ] given an integer @xmath9 , the _ @xmath6-ary expansion _ of a real @xmath10 in @xmath4 is the ( unique ) expansion of the form @xmath11 such that @xmath12 is in @xmath8 for all @xmath3 with @xmath13 infinitely often .
denote by @xmath14 the number of times a block @xmath15 occurs with its starting position no greater than @xmath3 in the @xmath6-ary expansion of @xmath10 .
[ definition:1.3 ] a real number @xmath10 in @xmath4 is _ normal in base @xmath6 _ if for all @xmath7 and blocks @xmath15 in base @xmath6 of length @xmath7 , one has @xmath16 a number @xmath10 is _ simply normal in base @xmath6 _ if ( [ eqn : bnormal1 ] ) holds for @xmath17 .
borel introduced normal numbers in 1909 and proved that almost all ( in the sense of lebesgue measure ) real numbers in @xmath4 are normal in all bases .
the best known example of a number that is normal in base @xmath18 is due to champernowne @xcite .
the number @xmath19 formed by concatenating the digits of every natural number written in increasing order in base @xmath18 , is normal in base @xmath18 .
any @xmath20 , formed similarly to @xmath21 but in base @xmath6 , is known to be normal in base @xmath6 . since then
, many examples have been given of numbers that are normal in at least one base .
one can find a more thorough literature review in @xcite and @xcite .
the @xmath5-cantor series expansion , first studied by georg cantor in @xcite , is a natural generalization of the @xmath6-ary expansion .
[ definition:1.4 ] @xmath22 is a _ basic sequence _ if each @xmath1 is an integer greater than or equal to @xmath2 .
[ definition:1.5 ] given a basic sequence @xmath5 , the _
@xmath5-cantor series expansion _ of a real @xmath10 in @xmath4 is the ( unique ) expansion of the form @xmath23 such that @xmath12 is in @xmath24 for all @xmath3 with @xmath25 infinitely often .
we abbreviate ( [ eqn : cseries ] ) with the notation @xmath26 with respect to @xmath5 .
clearly , the @xmath6-ary expansion is a special case of ( [ eqn : cseries ] ) where @xmath27 for all @xmath3 .
if one thinks of a @xmath6-ary expansion as representing an outcome of repeatedly rolling a fair @xmath6-sided die , then a @xmath5-cantor series expansion may be thought of as representing an outcome of rolling a fair @xmath28 sided die , followed by a fair @xmath29 sided die and so on .
for example , if @xmath30 for all @xmath3 , then the @xmath5-cantor series expansion of @xmath31 is @xmath32 if @xmath33 for all @xmath3 , then the @xmath5-cantor series expansion for @xmath34 is @xmath35 for a given basic sequence @xmath5 , let @xmath36 denote the number of times a block @xmath15 occurs starting at a position no greater than @xmath3 in the @xmath5-cantor series expansion of @xmath10 . additionally , define @xmath37 a. rnyi @xcite defined a real number @xmath10 to be normal with respect to @xmath5 if for all blocks @xmath15 of length @xmath38 , @xmath39 if @xmath27 for all @xmath3 , then ( [ eqn : rnormal ] ) is equivalent to _ simple normality in base @xmath6 _ , but not equivalent to _ normality in base @xmath6_. thus , we want to generalize normality in a way that is equivalent to normality in base @xmath6 when all @xmath27 .
[ definition:1.7 ] a real number @xmath10 is _
@xmath5-normal of order @xmath7 _ if for all blocks @xmath15 of length @xmath7 , @xmath40 we say that @xmath10 is _ @xmath5-normal _ if it is @xmath5-normal of order @xmath7 for all @xmath7 .
a real number @xmath10 is _
@xmath5-ratio normal of order @xmath7 _ if for all blocks @xmath15 and @xmath41 of length @xmath7 , we have @xmath42 @xmath10 is _ @xmath5-ratio normal _ if it is @xmath5-ratio normal of order @xmath7 for all positive integers @xmath7 .
we make the following definitions : [ definition:1.8 ] a basic sequence @xmath5 is _ @xmath7-divergent _ if @xmath43 .
@xmath5 is _ fully divergent _ if @xmath5 is @xmath7-divergent for all @xmath7 .
@xmath5 is _
@xmath7-convergent _ if it is not @xmath7-divergent . [ definition:1.6 ] a basic sequence @xmath5 is _ infinite in limit _ if @xmath44 .
for @xmath5 that are infinite in limit , it has been shown that the set of all @xmath10 in @xmath4 that are @xmath5-normal of order @xmath7 has full lebesgue measure if and only if @xmath5 is @xmath7-divergent @xcite . therefore
, if @xmath5 is infinite in limit , then the set of all @xmath10 in @xmath4 that are @xmath5-normal has full lebesgue measure if and only if @xmath5 is fully divergent .
suppose that @xmath5 is @xmath38-divergent .
given an arbitrary non - negative integer @xmath45 , f. schweiger @xcite proved that for almost every @xmath10 with @xmath46 , one has @xmath47 j. galambos proved an even stronger result in @xcite .
he showed that for almost every @xmath10 in @xmath4 and for all non - negative integers @xmath45 , @xmath48 we provide the following main results : 1 . a notion of strong @xmath5-normality
is provided and we construct an explicit example of a basic sequence @xmath5 and a real number that is @xmath5-normal , but not strongly @xmath5-normal ( theorem [ thm : construction ] ) .
2 . ( theorem [ thm : typicalasymptotics ] )
if @xmath5 is a basic sequence that is infinite in limit and @xmath15 is a block of length @xmath7 , then for almost every real number @xmath10 in @xmath4 , we have @xmath49 3 . if @xmath5 is infinite in limit , then almost every real number is @xmath5-normal of order @xmath7 if and only if @xmath5 is @xmath7-divergent ( theorem [ thm : aeqnok ] ) .
if @xmath5 is @xmath7-convergent for some @xmath7 , then the set of numbers that are @xmath5-normal is empty ( proposition [ proposition : fewnormal2 ] ) .
if @xmath5 is infinite in limit , then the set of @xmath5-ratio normal numbers is dense in @xmath4 ( corollary [ coro : rationormalisdense ] ) .
in this section , we will introduce a notion of normality that is stronger than @xmath5-normality .
this notion of normality will arise naturally later in this paper and will be useful for studying the typicality of @xmath5-normal numbers .
we will first need to make definitions similar to those of @xmath36 and @xmath50 . given a real number @xmath51 , a basic sequence @xmath5 , a block @xmath15 of length @xmath7 , a positive integer @xmath52 $ ] , and a positive integer @xmath3
, we will denote by @xmath53 the number of times that the block @xmath15 occurs in the @xmath5-cantor series expansion of @xmath10 with starting position of the form @xmath54 for @xmath55 .
if @xmath3 and @xmath7 are positive integers , define @xmath56 suppose that @xmath5 is a basic sequence and that @xmath57 , and @xmath7 are positive integers with @xmath52 $ ] .
we will write @xmath58 let @xmath7 be a positive integer
. then a basic sequence @xmath5 is _
strongly @xmath7-divergent_-divergent basic sequences must be strongly @xmath7-divergent . the following example of a @xmath2-divergent basic sequence that is not strongly @xmath2-divergent was suggested by c. altomare ( verbal communication ) : let the basic sequence @xmath59 be given by @xmath60 ] if for all positive integers @xmath61 with @xmath52 $ ] , we have @xmath62 . a basic sequence @xmath5 is _ strongly fully divergent _ if it is strongly @xmath7-divergent for all @xmath7 .
given a real number @xmath51 , a basic sequence @xmath5 , a block @xmath15 of length @xmath7 , a positive integer @xmath52 $ ] , and a positive integer @xmath3 , we will denote by @xmath53 the number of times the block @xmath15 occurs in the @xmath5-cantor series expansion of @xmath10 with positions of the form @xmath54 for @xmath55 .
suppose that @xmath5 is a basic sequence .
a real number @xmath10 in @xmath4 is _
strongly @xmath5-normal of order @xmath7 _ if for all blocks @xmath15 of length @xmath63 and all @xmath64 $ ] , we have @xmath65 a real number @xmath10 is _
strongly @xmath5-normal _ if it is strongly @xmath5 normal of order @xmath7 for all @xmath7 .
we will use the following lemmas frequently and without mention : given a real number @xmath51 , a basic sequence @xmath5 , a block @xmath15 of length @xmath7 , a positive integer @xmath52 $ ] , and a positive integer @xmath3 , we have @xmath66 @xmath67 this follows directly from the definitions of @xmath36 and @xmath68 .
[ lemma : sumlittleo ] if @xmath69 are non - negative functions on the natural numbers , then @xmath70 [ thm : snisn ] if @xmath5 is a basic sequence and @xmath10 is strongly @xmath5-normal of order @xmath7 , then @xmath10 is @xmath5-normal of order @xmath7
. let @xmath63 be a positive integer and let @xmath15 be a block of length @xmath7 . since @xmath10 is strongly @xmath5-normal of @xmath7 , we know that for all @xmath64 $ ] , @xmath71 . thus , we see that @xmath72 @xmath73 so @xmath74 . therefore , @xmath10 is @xmath5-normal of order @xmath7 .
[ coro : snisn ] suppose that @xmath5 is a basic sequence .
if @xmath10 is strongly @xmath5-normal , then @xmath10 is @xmath5-normal . in this subsection
, we will work towards giving an example of a basic sequence @xmath5 and a real number @xmath10 that is @xmath5-normal , but not strongly @xmath5-normal of order @xmath2 .
we will use the conventions found in @xcite . given a block @xmath15
, @xmath75 will represent the length of @xmath15 .
given non - negative integers @xmath76 , at least one of which is positive , and blocks @xmath77 , the block @xmath78 will be the block of length @xmath79 formed by concatenating @xmath80 copies of @xmath81 , @xmath82 copies of @xmath83 , through @xmath84 copies of @xmath85 .
for example , if @xmath86 and @xmath87 , then @xmath88 .
we will need the following definitions : [ definition:1.9 ] a _ weighting _
@xmath89 is a collection of functions @xmath90 with @xmath91 such that for all @xmath7 , @xmath92 $ ] and @xmath93 .
[ definition:1.10 ] the _ uniform weighting in base @xmath6 _ is the collection @xmath94 of functions @xmath95 such that for all @xmath7 and blocks @xmath15 of length @xmath7 in base @xmath6 @xmath96 [ definition:1.11 ] let @xmath61 and @xmath6 be positive integers such that @xmath97 .
a weighting @xmath89 is _ @xmath98-uniform _ if for all @xmath7 and blocks @xmath15 of length @xmath7 in base @xmath61 , we have @xmath99 given blocks @xmath15 and @xmath100 , let @xmath101 be the number of occurrences of the block @xmath15 in the block @xmath100 .
[ definition:1.12 ] let @xmath102 be a real number such that @xmath103 and let @xmath7 be a positive integer .
assume that @xmath89 is a weighting .
a block of digits @xmath100 is _ @xmath104-normal _ is a generalization of the concept of @xmath105-normality , originally due to besicovitch @xcite . ]
if for all blocks @xmath15 of length @xmath63 , we have @xmath106 for the rest of this subsection , we use the following conventions .
given sequences of non - negative integers @xmath107 and @xmath108 with each @xmath109 and a sequence of blocks @xmath110 , we set @xmath111 @xmath112 and @xmath113 moreover , if @xmath114 , we set @xmath115 given @xmath116 and @xmath117 it is assumed that @xmath10 and @xmath5 are given by the formulas above .
[ definition:1.13 ] a _ block friendly family _ is a @xmath118-tuple @xmath119 with non - decreasing sequences of non - negative integers @xmath120 , @xmath121 , @xmath122 and @xmath123 , for which @xmath124 , @xmath125 and @xmath126 , such that @xmath127 is a sequence of @xmath128-uniform weightings and @xmath129 strictly decreases to @xmath130 .
[ definition:1.14 ] let @xmath119 be a block friendly family .
if @xmath131 , then let @xmath132 .
otherwise , let @xmath133 .
a sequence @xmath110 of @xmath134-normal blocks of non - decreasing length is said to be _
@xmath135-good _ if for all @xmath7 in @xmath136 , the following three conditions hold : @xmath137 @xmath138 @xmath139 we now state a key theorem of @xcite .
[ thm : oldmain ] let @xmath135 be a block friendly family and @xmath140 a @xmath135-good sequence . if @xmath141 , then @xmath10 is @xmath5-normal of order @xmath7 .
if @xmath142 , then @xmath10 is @xmath5-normal .
if @xmath6 and @xmath143 are positive integers where @xmath6 is greater than or equal to @xmath2 and @xmath144 is odd , then we let @xmath145 be one of the blocks formed by concatenating all the blocks of length @xmath143 in base @xmath6 in such a way that there are at least twice as many copies of the block @xmath146 at odd positions as the block @xmath147 .
for example , we could pick @xmath148 @xmath149 which has @xmath150 copies of @xmath146 at the odd positions and @xmath151 copies of @xmath147 at the odd positions .
note that @xmath152 .
the next lemma is proven identically to lemma 4.2 in @xcite : [ lemma : l4.2 ] if @xmath153 and @xmath154 , then @xmath145 is @xmath155-normal .
[ thm : construction ] may be used to construct other explicit examples of @xmath5-normal numbers that satisfy some unusual conditions .
given a basic sequence @xmath5 , we say that @xmath10 is _ @xmath5-distribution normal _ if the sequence @xmath156 is uniformly distributed mod @xmath38 .
@xcite uses theorem [ thm : oldmain ] to give an example of a basic sequence @xmath5 and a real number @xmath10 such that @xmath10 is @xmath5-normal , but @xmath157 , so @xmath10 is not @xmath5-distribution normal .
] there exists a basic sequence @xmath5 and a real number @xmath10 such that @xmath10 is @xmath5-normal , but not strongly @xmath5-normal of order @xmath2 .
let @xmath158 , @xmath159 , and @xmath160 .
for @xmath161 , let @xmath162 , @xmath163 , and @xmath164 . set @xmath165 , @xmath166 , @xmath167 and @xmath168 . for @xmath161 , put @xmath169 , @xmath170 , @xmath171 , @xmath172 , and @xmath119 .
thus , since @xmath173 where @xmath174 and @xmath175 , @xmath176 is @xmath177-normal by lemma [ lemma : l4.2 ] . in order to show that @xmath178 is a @xmath135-good sequence we need to verify ( [ eqn : good1 ] ) , ( [ eqn : good2 ] ) , and ( [ eqn : good3 ] ) . since @xmath142 , we let @xmath7 be an arbitrary positive integer .
we will make repeated use of the fact that @xmath179 .
we first verify @xmath180 : @xmath181 we next verify @xmath182 . since @xmath183 , @xmath184 and @xmath185 we have @xmath186 @xmath187
lastly , we will verify @xmath188 .
since @xmath189 , @xmath190 , and @xmath191 we have @xmath192 @xmath193 @xmath194 since @xmath195 is @xmath128-uniform , @xmath178 is a @xmath135-good sequence and by theorem [ thm : oldmain ] , @xmath10 is @xmath5-normal . since the length of each block @xmath176 is even ,
so there will always be at least twice as many copies of the block @xmath146 as the block @xmath147 in any initial segment of digits of @xmath10 , so @xmath10 is not strongly @xmath5-normal of order @xmath2 .
for this section , we must recall a few basic notions from probability theory . given a random variable @xmath196
, we will denote the expected value of @xmath196 as @xmath197}$ ] .
we will denote the variance of @xmath196 as @xmath198}$ ] .
lastly , @xmath199 will represent the probability that @xmath200 .
we consider @xmath10 as a random variable which has uniform distribution on the interval @xmath4 .
if @xmath201 with respect to @xmath5 , then we consider @xmath202 to be random variables .
so for all @xmath3 , we have @xmath203 [ lemma : eareindependent ] if @xmath5 is a basic sequence , then the random variables @xmath202 are independent .
suppose that @xmath204 and @xmath205 are distinct positive integers and that @xmath206 for all @xmath207 .
then @xmath208 @xmath209 suppose that @xmath5 is a basic sequence , @xmath6 is a natural number , @xmath15 is a block of length @xmath7 , and @xmath210 is an integer with @xmath211 $ ] .
we set @xmath212 @xmath213 @xmath214 } , v_m^{(k)}={\hbox{var}\left [ \zeta^q_{b , i , p}(x ) \right ] } \hbox { , and } t_{n , p}^{(k)}=\sum_{i=0}^{\rho(n , k ) } v_{ik+p}^{(k)}.\ ] ] [ lemma : rnindependent ] for all non - negative integers @xmath6 , the random variables @xmath215 are independent .
this follows directly from lemma [ lemma : eareindependent ] as the random variables @xmath216 are independent .
[ lemma : productrip ] if @xmath217 is a block of length @xmath7 , then @xmath218 by definition , @xmath219 or in other words , @xmath220 if @xmath221 and @xmath222 otherwise .
[ coro : rareindep ] for all blocks @xmath217 of length @xmath7 and non - negative integers @xmath223 $ ] , @xmath224 , and @xmath225 with @xmath226 , the random variables @xmath227 and @xmath228 are independent . using lemma [ lemma : rnindependent ] and lemma [ lemma : productrip ]
, we see that @xmath229 } = { \hbox{e}\left [ \left(\pi_{j=0}^{k-1 } \zeta^q_{b_j , i_1 k+p_1+j}(x ) \right ) \cdot \left(\pi_{j=0}^{k-1 } \zeta^q_{b_j , i_2 k+p_2+j}(x ) \right ) \right]}\ ] ] @xmath230 } \right ) \cdot \left(\pi_{j=0}^{k-1 } { \hbox{e}\left [ \zeta^q_{b_j , i_2 k+p_2+j}(x ) \right]}\right)\ ] ] @xmath231 } \cdot { \hbox{e}\left [ \pi_{j=0}^{k-1 } \zeta^q_{b_j , i_2 k+p_2+j}(x ) \right]}={\hbox{e}\left [ \zeta^q_{b , i_1,p_1}(x ) \right ] } \cdot { \hbox{e}\left [ \zeta^q_{b , i_2,p_2}(x ) \right]}.\ ] ] [ lemma : compofexpected ] if @xmath217 is a block of length @xmath7 , then @xmath232 @xmath233 we first compute the expected value of @xmath234 . by lemma [ lemma : rnindependent ] and lemma [ lemma : productrip ]
, we see that @xmath235}={\hbox{e}\left [ \zeta^q_{b_1,ik+p}(x ) \cdot \zeta^q_{b_2,ik+p+1}(x ) \cdots \zeta^q_{b_k , ik+p+k-1}(x ) \right]}\ ] ] @xmath236 } \cdot { \hbox{e}\left [ \zeta^q_{b_2,ik+p+1}(x ) \right ] } \cdots { \hbox{e}\left [ \zeta^q_{b_k , ik+p+k-1}(x ) \right]}\ ] ] @xmath237 next , we recall that @xmath238}={\hbox{e}\left [ \zeta^q_{b , i , p}(x)^2 \right]}-{\hbox{e}\left [ \zeta^q_{b , i , p}(x ) \right]}^2 $ ] .
since @xmath234 may only be @xmath130 or @xmath38 , we see that @xmath239 , so @xmath240}=\frac { 1 } { q_{ik+p } q_{ik+p+1 } \cdots q_{ik+p+k-1}}-\left ( \frac { 1 } { q_{ik+p } q_{ik+p+1 } \cdots q_{ik+p+k-1 } } \right)^2.\ ] ] lastly , we remark that @xmath241 by lemma [ lemma : compofexpected ] and will use this fact frequently and without mention .
we will need the following : [ thm : iteratedlog1 ] let @xmath242 be independent random variables .
assume that there exists a constant @xmath243 such that @xmath244 for all @xmath207 .
let @xmath245 } , u_j={\hbox{var}\left [ x_j \right]}$ ] , and @xmath246 . if @xmath247 , then , with probability one , @xmath248 [ coro : iteratedlog2 ] under the same assumptions of theorem [ thm : iteratedlog1 ] , with probability one , @xmath249 we will also need the borel - cantelli lemma : [ thm : borelcantelli](the borel cantelli lemma ) if @xmath250 , then @xmath251 given a basic sequence @xmath5 , we will define @xmath252 .
[ lemma : tnp ] if @xmath5 is a basic sequence and @xmath253 , and @xmath61 are positive integers with @xmath52 $ ] , then @xmath254 @xmath255 @xmath256 to show the other direction of the inequality , we recall that since @xmath5 is a basic sequence , @xmath257 for all @xmath258 , so for all @xmath259 @xmath260 @xmath261 @xmath262 [ lemma : lemmastronglynormal ] if @xmath5 is infinite in limit and @xmath15 is a block of length @xmath7 , then for almost every real number @xmath10 in @xmath4 , we have @xmath263 we consider two cases .
the first case is when @xmath264 .
we see that @xmath265 so by theorem [ thm : borelcantelli ] , we have @xmath266 .
thus , for almost every @xmath51 , @xmath267 and ( [ eqn : lemmastronglynormal1 ] ) holds .
second , we consider the case where @xmath268 . by lemma
[ lemma : tnp ] , we have @xmath269 .
note that @xmath270 by corollary [ coro : iteratedlog2 ] , @xmath271 for almost every @xmath51 . by lemma [ lemma : tnp ] ,
@xmath272 , so the lemma follows .
lemma [ lemma : lemmastronglynormal ] allows us to prove the following results on strongly normal numbers : [ thm : aeqsnok ] suppose that @xmath5 is strongly @xmath7-divergent and infinite in limit .
then almost every @xmath51 is strongly @xmath5-normal of order @xmath7 .
let @xmath15 be a block of length @xmath63 and @xmath64 $ ] . then by lemma [ lemma : lemmastronglynormal ] , for almost every @xmath51 , we have that @xmath273 @xmath274 however , @xmath5 is strongly @xmath7-divergent , so @xmath275 and @xmath276 since there are finitely many choices of @xmath258 and @xmath61 and only countably many choices of @xmath15 , the result follows .
[ coro : aeqsn ] if @xmath5 is strongly fully divergent and infinite in limit , then almost every real @xmath51 is strongly @xmath5-normal .
we now work towards proving a result much stronger than corollary [ coro : aeqsn ] on the typicality of @xmath5-normal numbers .
we will need the following lemma in addition to lemma [ lemma : lemmastronglynormal ] : [ lemma : bigoineq ] if @xmath5 is a basic sequence and @xmath7 and @xmath61 are positive integers with @xmath52 $ ] , then @xmath277 we first note that @xmath278 . since @xmath279
, we see that @xmath280 next , note that @xmath281 by ( [ eqn : iso1 ] ) and ( [ eqn : nastyineq ] ) , @xmath282 thus , the lemma follows by combining ( [ eqn : iso1 ] ) and ( [ eqn : firstbigo ] ) .
[ thm : typicalasymptotics ] if @xmath5 is a basic sequence that is infinite in limit and @xmath15 is a block of length @xmath7 , then for almost every real number @xmath10 in @xmath4 , we have @xmath49 we first note that @xmath283 thus , by ( [ eqn : bong ] ) and lemma [ lemma : lemmastronglynormal ] , for almost every @xmath51 , we have @xmath284 thus , the theorem follows by applying lemma [ lemma : bigoineq ] to ( [ eqn : bong2 ] ) .
we recall the following standard result on infinite products : [ lemma : infiniteproduct ] if @xmath285 is a sequence of real numbers such that @xmath286 for all @xmath3 , then the infinite product @xmath287 converges if and only if the sum @xmath288 is convergent .
[ thm : aeqnok ] suppose that @xmath5 is a basic sequence that is infinite in limit .
then almost every real number in @xmath4 is @xmath5-normal of order @xmath7 if and only if @xmath5 is @xmath7-divergent .
first , we suppose that @xmath5 is @xmath7-divergent .
then by theorem [ thm : typicalasymptotics ] , for almost every @xmath51 , we have @xmath289 we now suppose that @xmath5 is @xmath7-convergent.we will now use similar reasoning to that found in @xcite .
set @xmath290 ( @xmath7 zeros ) .
we will show that the set of real numbers in @xmath4 whose @xmath5-cantor series expansion does not contain the block @xmath15 has positive measure .
call this set @xmath291 .
we see that @xmath292 set @xmath293 .
since @xmath5 is @xmath7-convergent , we have @xmath294 .
thus , @xmath295 by lemma [ lemma : infiniteproduct ] .
[ coro : aeqn ] suppose that @xmath5 is a basic sequence that is infinite in limit .
then almost every real number in @xmath4 is @xmath5-normal if and only if @xmath5 is fully divergent .
we are now in a position to compare the prevelance of @xmath5-normal numbers to @xmath5-ratio normal numbers , depending on properties of the basic sequence @xmath5 . in particular
, we will show that if @xmath5 is infinite in limit , then the set of @xmath5-ratio normal numbers is dense in @xmath4 even though the set of @xmath5-normal numbers may be empty .
suppose that @xmath5 is a @xmath7-convergent basic sequence and define @xmath296 we make the observation that since @xmath297 for all @xmath3 , @xmath298 for all @xmath7 .
thus , there exists a @xmath299 such that for all @xmath300 , we have @xmath301 .
thus , no blocks of length @xmath300 can occur in any @xmath5-normal number and the set of @xmath5-normal numbers is empty .
let @xmath306 be any fully divergent basic sequence that is infinite in limit .
then we know that there exists a @xmath307-normal number by corollary [ coro : aeqn ] .
let @xmath308 with respect to @xmath307 be @xmath307-normal and let @xmath309 .
set @xmath310 , @xmath311 , and @xmath312 .
suppose that @xmath15 and @xmath41 are two blocks of length @xmath7 and let @xmath313 .
thus , if @xmath314 , then @xmath315 is equivalent to @xmath316 and @xmath317 is equivalent to @xmath318 .
since @xmath10 is @xmath307-normal , there are infinitely many occurences of every block .
additionally , @xmath319 for all @xmath3 , so @xmath320 is @xmath5-ratio normal .
|
fix a sequence of integers @xmath0 such that @xmath1 is greater than or equal to @xmath2 for all @xmath3 . in this paper , we improve upon results by j. galambos and f. schweiger showing that almost every ( in the sense of lebesgue measure ) real number in @xmath4 is @xmath5-normal with respect to the @xmath5-cantor series expansion for sequences @xmath5 that satisfy a certain condition .
we also provide asymptotics describing the number of occurrences of blocks of digits in the @xmath5-cantor series expansion of a typical number . the notion of strong @xmath5-normality , that satisfies a similar typicality result ,
is introduced .
both of these notions are equivalent for the @xmath6-ary expansion , but strong normality is stronger than normality for the cantor series expansion . in order to show this , we provide an explicit construction of a sequence @xmath5 and a real number that is @xmath5-normal , but not strongly @xmath5-normal .
we use the results in this paper to show that under a mild condition on the sequence @xmath5 , a set satisfying a weaker notion of normality , studied by a. rnyi in @xcite , will be dense in @xmath4 .
|
there is a window in the gamma - ray spectrum which has yet to be systematically explored by any telescope .
current state - of - the - art atmospheric cherenkov telescopes have energy thresholds of 250 gev or greater .
conversely , the egret experiment on the compton gamma ray observatory detects few astrophysical photons above 20 gev .
exploring the gamma - ray window between 20 and 250 gev is a primary goal of stacee .
a more complete discussion of the scientific motivation for exploring this window can be found elsewhere ( ong 1997 ) .
to first order , the energy threshold of an atmospheric cherenkov telescope is limited by night sky fluctuations and scales inversely with the square root of the telescope mirror collection area . the most straightforward way to achieve a low gamma - ray energy threshold
is to use a very large mirror area .
two approaches to significantly increasing mirror area have been considered .
the first is to build large ( e.g. 10 m diameter ) mirrors from scratch . in this way
the mirrors can be custom - built for cherenkov astronomy but the costs will be relatively high .
a second approach is to make use of large mirror facilities that already exist .
more than a decade ago , it was pointed out that large solar mirrors ( heliostats ) could be used as the primary collector in an atmospheric cherenkov telescope ( danaher _ et al .
later , a design incorporating a secondary optic was suggested by members of our group ( t " umer _ et al .
_ 1990 ) . in the past three years ,
the celeste ( par ' e 1997 and qu ' ebert 1997 ) and stacee ( chantell _ et al .
_ 1997 , covault _ et al .
_ 1997 ) groups have carried out tests at solar mirror facilities to demonstrate that low energy cherenkov telescopes can in fact be constructed at these locations .
both experiments have been funded and are now under construction .
a third experiment called graal ( plaga 1997 ) is under development .
the designs of celeste and stacee are similar in many respects , but here we concentrate on the issues relevant for stacee .
the stacee collaboration explored the possibility of using the heliostat mirrors of the solar two power plant near barstow ca , usa .
initial prototype work at the solar two site led to the first detection of atmospheric cherenkov radiation at such a facility ( ong _ et al .
solar two is currently an operating power plant , and a more favorable u.s .
location is the research heliostat field of the national solar thermal test facility ( nsttf ) at sandia national laboratories in albuquerque nm .
the overall design of stacee is shown in figure 1 .
cherenkov light created in gamma - ray air showers reaches the ground with a broad lateral extent @xmath0 m across .
the nsttf consists of 220 heliostat mirrors ( each 37.2@xmath1m@xmath2 in area ) .
stacee uses 48 heliostats for a total reflector surface of @xmath3m@xmath2 .
the heliostats reflect the cherenkov light onto secondary mirrors on a central tower .
three @xmath4 m diameter secondary mirrors viewing different parts of the heliostat field are used .
light reflected from each secondary mirror is imaged onto a photomultiplier tube ( pmt ) camera .
each heliostat is viewed by a single pmt . in this paper
, we describe the essential features of the stacee design .
we first discuss the nsttf site itself .
we then outline the essential hardware components of stacee , including the : 1 .
heliostats , 2 .
secondary mirror structures , 3 . cameras , and 4 . electronics .
the nsttf is located approximately 15 km southeast of albuquerque at an altitude of 1700 m above sea level .
the site is dry and has excellent sky clarity .
weather patterns accumulated over the last thirty years predict an average precipitation of 13.4 cm per year and a yearly average of 170 fully clear days and 111 mostly clear days ( noaa 1997 ) . we have measured the intensity of night sky background light at the site and have found an average value of @xmath5 photons m@xmath6 sec@xmath7 sr@xmath7 over the wavelength range spanned by a typical bi - alkali pmt ( 300 - 600 nm ) .
the nsttf is located at sandia national laboratories on kirtland air force base .
the facility is fully supported and operated by the u.s .
department of energy ( doe ) . as such
, the doe maintains the heliostat field , heliostat control tower , and solar tower .
heliostat repair is done as a matter of course .
since the nsttf is a research facility and not an operating power plant , there is little potential for interference between stacee work and solar energy research .
we are able to schedule our observations well in advance with minimal interruption from other activities at the site .
heliostat mirrors at the nsttf consist of 25 square facets made of rear - surfaced glass .
facets for selected heliostats are aligned and focused using a television camera mounted at the top of the central tower .
each heliostat mirror can be positioned in azimuth and elevation with 13-bit precision ( 0.05@xmath8 ) .
the optical and pointing performance of the heliostats is found to be satisfactory for atmospheric cherenkov astronomy and has been described elsewhere ( covault _ et al .
_ 1997 , chantell _ et al .
_ 1997 ) . of the 220 heliostats at the nsttf site
, we have selected 48 heliostats to be used for stacee ( 16 heliostats for each of three secondary mirrors ) .
ideally , heliostats are selected so as to provide uniform coverage over the entire area of the field .
however , selected heliostats must also not crowd each other in the focal plane of the secondary mirror .
we also select heliostats to minimize off - axis aberrations .
finally , field - of - view requirements prevent consideration of heliostats that are too close to the central tower .
figure 2 shows the view from the tower into the northeastern corner of the heliostat array corresponding to one of the three secondary mirrors .
this view is imaged onto the focal plane of the mirror .
the rectangular border indicates the boundary of the camera .
thick circles indicate the position of each pmt can entrance aperture for the 16 selected heliostats in this part of the array .
note that in actual operation , non - selected heliostats will be turned edge - on and thus will not reflect light up to the secondary mirror .
all heliostats at the nsttf are operated using a single master control system ( mcs ) based on an hp-1000 mini - computer .
software to control heliostats and to track the sun onto various targets has been previously developed by the nsttf staff .
the stacee group has modified this software to track astrophysical sources , accounting for the fact that air showers originate near the top of the atmosphere and not at infinity .
the mcs is operated directly by stacee scientists during observations the mcs communicates with individual heliostats over a network of serial lines .
heliostat pointing coordinates are updated at least once every 1.2 sec .
several video displays provide real - time information on every aspect of the heliostat field and warn the operator if there is a malfunction .
status and pointing information for every heliostat is logged to a file every 5 seconds .
a schematic view of one of the three secondary mirror structures is shown in figure 3 .
the structure consists of a commercial pallet stacker which is secured to a steel base plate , or skid .
the stacker raises and lowers the mirror assembly consisting of a jib - crane and spider , both made of aluminum .
the jib - crane allows for azimuthal control by means of a linear actuator .
the spider holds the mirror and is attached to the jib - crane by a ball joint and two actuators for elevation control .
the actuators are used only for alignment and the mirror is not moved during observations .
the 2 m diameter mirror is spherical in shape and consists of seven approximately hexagonal segments , as shown in figure 4 .
each segment is a single piece of lime float glass that has been slumped , ground , polished from a 0.5 " thick blank to a 2 m focal length .
the glass is coated by evaporation of aluminum onto its front surface and by a protective layer of alsio .
each mirror segment is attached to the spider by means of a triangular aluminum base - plate which is glued to the mirror and is attached by bolts to the spider frame .
mirror adjustments are made by a three - point mounting system on each facet .
the overall mirror size is set by the focal properties of the heliostats themselves .
solar images have been measured for fifteen individual heliostats and their sizes are reported in covault _
the maximum image size at the tower is 1.7 m ( fwhm ) and the image sizes scale as expected with the distance from the tower .
the camera consists of two principal components : an adjustable stand , and pmt can assemblies mounted on a slotted plate , as shown in figure 5 .
the stand is a rigid structure made of aluminum bolted to a movable carriage .
the carriage allows the entire camera to slide away from the tower edge to the proper focal point of the secondary mirror .
the slotted plate is tilted in elevation to allow for correct focusing of the heliostat images in the center of the field - of - view of the mirror onto the entrance apertures of their respective can elements .
each pmt can element holds an assembly which comprises an acrylic light concentrator followed by a photomultiplier tube .
the pmt and concentrator are held together in a hollow metal cylinder which has an edge ring at one end and a threaded disk at the other .
the disk pushes the concentrator and pmt together ; they are separated by a silicone rubber cookie which provides good optical and mechanical coupling .
this assembly is approximately 13 cm in diameter and 40 cm long .
the concentrators are dielectric total internal reflecting concentrators ( dtircs ) ( ning _ et al .
_ 1987 ) .
the can elements are mounted in a sleeve arrangement .
the sleeves are secured to the slotted plate with a system that allows for continuous x - y adjustment as well as canting so that pmts away from the optical axis can be pointed towards the center of the mirror . for cans
furthest from the optic axis of the mirror the canting angle is 17@xmath8 . operationally , one positions and cants the sleeves before sliding in the heavy pmt can elements .
the dtirc is approximated by a converging cylinder of length between 10.6 cm and 13.2 cm .
the cylinder is made of a single piece of uv transparent acrylic that has been machined on a numerically controlled mill to the proper dimensions .
the front surface of the dtirc is a sphere and the rear surface is a circular aperture .
the dtirc is designed so the field - of - view of each pmt is restricted .
three different types of dtircs are used corresponding to pmts that view different parts of the heliostat field .
pmts viewing heliostats closest to the tower are matched with dtircs that have a field - of - view of 16@xmath8 ( half - angle ) , whereas pmts viewing heliostats farthest from the tower are matched with dtircs having a field - of - view of 26@xmath8 .
the photomultiplier tubes are chosen for high speed and moderate gain .
speed is important to minimize the level of pile - up from pulses created by night sky photons .
a moderate gain is needed so that a sufficient amount of charge is generated , but at the same time the total current draw of the tube must be kept to an acceptable level ( @xmath9a ) .
the pmts ( philips xp2282 ) have eight dynode stages and a typical gain of @xmath10 at a high voltage value of -1500 v. each tube has a transistorized base ( philips vd 182k / c ) that has been optimized for good linearity and speed . the pulse width of a single photoelectron signal at
the output of the base is @xmath11 nsec ( fwhm ) and the time transit spread of the pulse is @xmath12 nsec .
the tube has a 2 " diameter bi - alkalai photocathode and a front window made of borosilicate glass .
the electronics serve two major purposes .
first , the pmt signals must be combined together in a coincidence circuit to form the experiment trigger .
the trigger detects the occurrence of a cherenkov air shower amidst the fluctuations of the night sky background light .
second , electronics are needed to measure the arrival time and pulse - height of the cherenkov signal reaching each of the pmts .
the shower direction is reconstructed from the arrival time information and the pulse - height information is used to estimate the shower energy . a block diagram of the electronics is shown in figure 6 .
commercial components are used wherever possible .
the pmts are powered by a high voltage mainframe ( lecroy 4032 ) .
the pmt signals are capacitively coupled ( time constant = 75 nsec ) , amplified ( phillips 776 ) , and discriminated ( lecroy 4413 ) .
copies of the pmt signals are generated by a linear fan - out ( phillips 748 ) and these copies are sampled by waveform digitizers .
the waveform information is read out via ethernet by the data acquisition ( daq ) computer .
the trigger is formed by a coincidence of the discriminated pmt signals that are delayed appropriately to bring the signals into time .
the delays account for the fact that the cherenkov light reaches different pmts at different times because of variations in the times of flight from individual heliostats to the secondary mirror .
in addition , delays are needed to account for time of flight differences due to the orientation of the shower wavefront .
the trigger is composed of two distinct parts : level 1 ( l1 ) and level 2 ( l2 ) the discriminated signals are delayed by programmable delay units ( lecroy 4518 ) and fixed length time of flight cables ( rg 58u ) and combined to form the l1 trigger .
the l1 trigger consists of multiplicity logic acting on parallel groups of four pmts ( i.e. n of 4 logic ) .
the pmts are grouped to form l1 triggers based on the proximity of their respective heliostats .
the l1 trigger signals are passed through additional programmable delay ( caen 469 ) to the l2 multiplicity logic , which makes a majority decision based on the number of in - time l1 signals ( i.e. n of 12 logic ) .
the l2 trigger signifies an event trigger which stops the waveform digitizers and initiates an event read - out .
the waveform digitizers are 1 ghz sampling units with 8-bit resolution .
two commercially available products are being considered .
one is a digitizer made in the vxi standard ( tektronix tvs641 ) and the other is a flash adc in the vme standard ( etep 301 ) .
the daq computer will be a silicon graphics workstation ( indy r4400 or equivalent ) .
the workstation will control the camac components via an ethernet crate controller ( hytec 1365 ) .
universal time will be recorded via a camac - based global positioning system ( gps ) clock ( hytec gps92 ) .
stacee is an experiment designed to carry out gamma - ray observations in the unopened window between 20 and 250 gev using the atmospheric cherenkov technique .
the experiment uses large heliostat mirrors of the national solar thermal test facility to achieve a very large cherenkov photon collection area .
additional components of stacee include secondary mirrors and structures , cameras , and electronics .
the baseline design for stacee has been finalized and the full experiment is now under construction .
stacee is a collaboration of scientists from the university of chicago , mcgill university , the university of california ( santa cruz ) , the university of california ( riverside ) , barnard college , yale university , and california state university ( los angeles ) .
we acknowledge the help of e. pod , m. houde , the engineering staff of the yerkes observatory , and the personnel of the mcgill physics department machine shop .
we wish to thank p. fleury , e. par ' e , j. qu ' ebert , and d.a .
smith for many useful discussions .
this work is supported in part by the national science foundation , the institute of particle physics of canada , the natural sciences and engineering research council , and the california space institute .
rao acknowledges the support of the grainger foundation and the university of chicago .
cec is a cottrell scholar of the research corporation .
|
the solar tower atmospheric cherenkov effect experiment ( stacee ) is designed to explore the gamma - ray sky between 20 and 250 gev using the atmospheric cherenkov technique .
stacee will use large solar heliostat mirrors to reflect cherenkov light created in gamma - ray air showers to secondary mirrors on a central tower .
the secondary mirrors image this light onto photomultiplier tube cameras that are read out by fast electronics . here
we outline the important features of the stacee design .
we present an overview of the experimental site , describe the method of heliostat selection and control , and discuss the current designs for the detector components , including the secondary mirror structures , cameras , and electronics . *
the stacee project + * ren ' e a. ong and corbin e. covault + _ enrico fermi institute , university of chicago , chicago , il 60637 , usa + _ ( for the stacee collaboration ) to appear in proc .
towards a major atmospheric cherenkov detector - v ( kruger park )
|
we consider the photosynthetic apparatus of the purple bacterium _ rhodobacter sphaeroides _ using the model described previously @xcite . as shown in fig .
[ fig : structure ] , it includes antenna complexes lh1 and lh2 that increase the amount of light absorbed per rc .
two rcs are surrounded by the s - shaped lh1 complex consisting of @xmath0 tightly packed bchls , which is itself surrounded by lh2 complexes , the structures being taken from crystal structures @xcite .
although each lh2 contains two rings of bchls , b800 and b850 , we only consider the 18-member b850 because eet between b800 and b850 is fast and efficient .
overall , the main inter - complex eet pathway is lh2 @xmath1 lh1 @xmath1 rc .
lh2 has a beautiful 9-fold symmetry , with the transition dipoles of the b850 bchls almost in the plane of the ring ( about off ) , pointing alternately left and right as they go around .
although lh1 is less symmetric than lh2 , its bchls are approximately arranged on a flat ring within each of its halves , with their transition dipole moments again roughly parallel to the plane of the ring ( at most off ) and also alternating left - right . strong coupling between nearest - neighbor bchls results in exciton delocalization within each complex , i.e. , lh2 , lh1 , or rc
the excitonic states are eigenstates of a frenkel - type hamiltonian of the particular complex @xcite , which in weak light where there is at most one exciton present takes the form @xmath2 where @xmath3 is the number of pigments within the complex , @xmath4 is the `` site '' energy of state @xmath5 corresponding to an exciton on bchl @xmath6 , and @xmath7 is the coupling between sites @xmath6 and @xmath8 .
different computational methods can predict substantially different energies and couplings @xcite . here , we compute the intra- and inter - complex couplings @xmath7 using transition charges from electrostatic potentials ( tresp ) @xcite , a method that is not only fast , but also as accurate as is realistically possible across the full range of ( bacterio)chlorophyll separations and orientations @xcite . the only exception is the rc special pair , whose coupling we take to be @xmath9 @xcite . furthermore , for each complex , we choose the site energies so that the energy of the brightest state matches the observed absorption maximum of that complex . because couplings between different complexes are weak
, we neglect inter - complex excitonic delocalization . accordingly , optical pumping and dynamics will be entirely through the eigenstates of the different complexes , as opposed to individual sites @xcite .
in particular , eet between two weakly coupled aggregates is described by multichromophoric frster resonant energy transfer ( mc - fret ) @xcite .
mc - fret simplifies to the more tractable generalized frster resonant energy transfer ( gfret ) @xcite in several cases , including if the emission and absorption spectra of the complexes are diagonal in the excitonic basis or if the system - environment coupling is weak compared to the coupling between bchls in the same complex @xcite .
assuming the latter , the gfret transfer rate between eigenstates of two complexes is @xmath10 where @xmath11 , @xmath12 and @xmath13 are the components of the excitonic states @xmath14 and @xmath15 in the site basis , and @xmath16 is the coupling between sites @xmath6 and @xmath8 .
@xmath17 is the spectral overlap between the normalized emission spectrum @xmath18 of the donor and the normalized absorption spectrum @xmath19 of the acceptor . @xmath18 and
@xmath19 can be calculated using multichromophoric fret theory @xcite , but here we follow @xcite in taking both to be normalized gaussians , giving @xmath20 , where @xmath21 is the energy difference between the states and @xmath22 . finally , to ensure detailed balance , we use eq . [ eq : fretrate ] only for downhill transitions ( @xmath23 ) , otherwise taking @xmath24 with @xmath25 . because the site - to - site couplings in the generalized fret expression ( eq . [ eq : fretrate ] ) are combined with amplitudes @xmath12 and
@xmath13 that can be positive or negative ( or complex ) , the overall rate @xmath26 can be larger or smaller than an analogous incoherent sum of site - to - site fret rates .
when the excitonic states are delocalized so that the amplitudes cause a cooperative enhancement of the transfer rate , the effect is called supertransfer @xcite . in natural light , optical pumping and relaxation
take place continuously , meaning that the ensemble of complexes will reach a steady state , finding which is sufficient to determine all observables of interest @xcite . because sunlight is incoherent
, it creates excitons in energy eigenstates @xcite , i.e. , it does not induce coherences in the energy basis .
strictly speaking , the incoherent pumping is into eigenstates of the combined system and bath , which , when reduced to the system alone , may not coincide with eigenstates of the system hamiltonian @xcite .
here we assume that the system - bath coupling is not large enough for this discrepancy to be significant .
consequently , the dynamics of the apparatus can be described using a pauli master equation , @xmath27 , where @xmath28 is the vector of all eigenstate populations , including the ground state .
the rate matrix @xmath29 contains the absorption , relaxation , and inter - complex transfer rates , @xmath30 here , nonradiative recombination to the ground state @xmath31 is assumed to occur at rate @xmath32 @xcite and internal conversion to lower - lying excitonic levels with rate @xmath33 @xcite .
radiative recombination is taken to occur with rate @xmath34 where @xmath35 @xcite and @xmath36 are , respectively , the radiative decay rate and site energy of bchl in solution , while @xmath37 is the oscillator strength ( or brightness ) of state @xmath15 relative to a single bchl . the optical pumping rate is @xmath38 , where @xmath39 is the mean photon number at energy @xmath40 at the effective black - body temperature of solar radiation , @xmath41 . finally , @xmath42 ( if @xmath15 is a state of the rc ) is the rate of charge separation in the rc @xcite . at steady state @xmath43 and the problem
simplifies to finding the zero - eigenvalue eigenstate of @xmath29 , whose existence and uniqueness are guaranteed because @xmath29 describes an irreducible continuous - time markov chain . the overall eet efficiency , @xmath44 is the quantum yield of photoexcited excitons that drive charge separation in the rc at steady state .
we study how the orientations of bchls within lh2 and lh1 aggregates affect the efficiency of exciton transfer in the purple - bacterial light - harvesting apparatus . in doing so
, we fix the site energies whose role we examined previously @xcite and the positions of the central mg atom in each bchl . rotating the bchls has a complex influence on the performance of the complexes because changing the orientations of the dipole moments affects inter - pigment couplings and thus the nature and energy of the eigenstates .
these changes , in turn , affect the inter - complex eet rates and the overall efficiency .
we consider two types of changes : rotating bchls within their planes as well as randomizing their orientations completely .
the bacteriochlorin ring within each bchl is approximately planar , meaning that it would occupy roughly the same space if it were rotated about an axis passing through the mg atom and perpendicular to the ring ( see inset to fig .
[ fig : in - plane ] ) .
thus , in - plane bchl rotations might be considered more plausible evolutionary alternatives than some other rotations , since the bacteriochlorin ring would not require large adjustments in the surrounding protein .
although this argument neglects the bchl s phytyl tail , the tail is of secondary importance because it would be flexible enough to bend out of the way in many cases of steric hindrance .
most importantly , the simple rotation provides substantial intuition about the role of bchl orientations that can be used to understand more complicated rearrangements .
figure [ fig : in - plane ] shows how in - plane bchl rotations affect the overall efficiency .
two cases are shown , with all the bchls in either lh2 or lh1 rotated by the same angle @xmath45 .
the x - ray geometry corresponds to @xmath46 , whose efficiency ( 73% ) is nearly optimal .
indeed , rotating the bchls can reduce the efficiency significantly , as low as 15% in the case where lh1 bchls are set to be perpendicular to the lh1 plane .
the reduction in efficiency upon bchl rotation can be understood by considering the brightnesses of the aggregate energy levels .
if the aggregates were far apart compared to their size , each could be considered as a supermolecule , with fret rates proportional to the oscillator strengths of the excitonic states . here , the small inter - complex distances mean that the supermolecule approximation does not capture all the details of eq .
[ eq : fretrate ] , but it is nevertheless a useful conceptual tool .
we stress that the brightness of the states relates to the efficiency because it is a proxy for supertransfer , not because it implies that more light is absorbed in the first place . on the contrary
, the oscillator sum rule implies that the total absorption cross - section of all the states will be constant regardless of their individual brightnesses , assuming the solar intensity is approximately constant over the absorption spectrum . as depicted in fig .
[ fig : in - plane](a ) and ( c ) , the natural geometries of both lh2 and lh1 give rise to very bright states near the bottom of the energy spectrum ( within @xmath47 of the lowest state ) .
this is important because the high internal - conversion rate @xmath48 ensures rapid thermalization , meaning that only the low - lying states contribute to eet and that their brightness regulates supertransfer .
as the dipole moments rotate away from the plane of the rings , the low - lying bright states are gradually lost until , at the minima shown in fig .
[ fig : in - plane](b ) and ( d ) , the thermally accessible states carry very little oscillator strength .
thus , eet is slowed down and efficiency decreases .
figure [ fig : in - plane ] shows that the efficiency is more sensitive to the reorientation of bchls in lh1 than in lh2 .
this is because lh1 @xmath1 rc transfer is the kinetic bottleneck of the entire process , largely because it is energetically uphill ( for the complete energy diagrams of lh2 , lh1 , and rc , see ref .
@xcite ) . therefore , decreases in the lh1 @xmath1 rc eet rate caused by reorientation immediately translate to a reduced efficiency . by contrast , lh2 @xmath1 lh1 transfer is energetically downhill and relatively fast , meaning that it can proceed with high efficiency even if the rate decreases somewhat .
thus , the broad plateau in the efficiency as a function of lh2 rotation angle reflects the need for a significant rotation of the bchls before the rate is decreased enough for it to affect the overall efficiency . even at the minimum ,
the efficiency only decreases from 73% to 49% , reflecting the decisive effect of downhill energetic funnelling @xcite . to further investigate the influence of geometry on eet efficiency in purple bacteria , we considered aggregates in which the orientations of the bchls in lh2 and/or lh1 were completely randomized , the orientations being chosen using a standard spherical point - picking algorithm . because random rotations could cause nearest - neighbor bchls to collide with each other ,
we only accepted geometries in which the distance between any two atoms in different bchls is greater than @xmath49 , which is the shortest distance between bchls in lh1 aggregate and is approximately twice the van der waals radius of a hydrogen atom .
the distributions of efficiencies for the random orientations are shown in fig .
[ fig : efficiencydistributions](a c ) . in particular , the efficiency is not sensitive to the orientation of bchls in the lh2 complexes , always attaining a value close to the original @xmath50 ( fig .
[ fig : efficiencydistributions](a ) ) .
this indicates that no geometric fine tuning is necessary to achieve a high efficiency and that lh2s are tolerant to orientational disorder .
by contrast , bchl orientations in lh1 have a large effect on the efficiency ( fig .
[ fig : efficiencydistributions](b ) )
. importantly , the mean efficiency is @xmath51 , with none of the @xmath52 samples coming close to the original @xmath50 , making the natural geometry an outlier by 5.5 standard deviations .
it follows that the whole light - harvesting apparatus has an unusually high efficiency , as is seen when the bchls in both lh2 and lh1 are randomized ( fig .
[ fig : efficiencydistributions](c ) ) .
the natural lh1 geometry is an outlier because it occupies a corner of an enormous , 168-dimensional space ( three angles per bchl ) .
if only small perturbations to the original bchl angles were considered ( up to ) , the efficiency would only change by up to a few percent ( see fig .
[ fig : scatterplots ] ) .
indeed , it is unlikely that bchl orientations are fine - tuned to less than several degrees , considering the constant fluctuations at physiological temperatures .
the stark difference between the effect of randomization on lh2 and lh1 can be understood , as in the case of in - plane rotations , in terms of the brightnesses of the states and of the rate - limiting nature of the lh1 @xmath1 rc step . to establish this conclusion , fig .
[ fig : scatterplots ] shows the relationships between the overall efficiency and the two properties that have the greatest influence on eet , energy funnelling and coherent excitonic delocalization @xcite .
energetic alignment enters @xmath53 in eq .
[ eq : fretrate ] , strongly favoring downhill eet , while the oscillator strength is a useful , if approximate , proxy for the delocalization and supertransfer contained in @xmath54 . fig .
[ fig : scatterplots](a ) and ( b ) show there is no appreciable correlation between efficiency and the energy of the lowest excited states in either lh2 or lh1 . in lh2 ,
random reorientations increase the energy of the lowest excitonic state a few hundred wavenumbers with no effect on the efficiency because the eet to lh1 remains downhill . in lh1
, it might be expected that an increase in energy upon randomization would increase the efficiency by reducing the uphill energy barrier for transfer to the rc . however , the range of energetic variation is comparable to @xmath47 , resulting in minor changes to the efficiency compared to other effects .
[ fig : scatterplots](c ) and ( d ) examine the correlation between efficiency and the brightness of the low - lying states in lh2 and lh1 , defined as the sum of oscillator strengths of the states lying within @xmath55 of the lowest state . in both lh2 and lh1 , randomization destroys the symmetry and decreases the oscillator strength by a factor of 34 .
nevertheless , there is a large difference between the effect of brightness on efficiency for the two complexes : for lh2 there is no effect , while for lh1 it leads to a large decrease in efficiency .
the same difference is seen when considering the rates of forward eet ( from lh2 to lh1 and from lh1 to rc ) : although the decrease in brightness reduces forward eet rates in both lh2 and lh1 , only the decrease in lh1 affects the efficiency ( fig .
[ fig : scatterplots](e ) and ( f ) ) .
indeed , in lh1 the decrease in brightness is sufficient to decrease the eet rate despite the improved energetic landscape . as with in - plane rotations , lh1 is more sensitive to changes in the brightness of its states because the rate of lh1 @xmath1 rc transfer is low to begin with ; at , it is the lowest eet rate in the entire light - harvesting apparatus @xcite and is comparable to the recombination rate @xmath56 .
decreasing it further by reducing brightness tightens the bottleneck , directly resulting in a decrease in efficiency ( fig .
[ fig : scatterplots](f ) ) .
by contrast , the lh2 @xmath1 lh1 rate is high enough even with the decrease in brightness that there is little risk of the exciton being lost while on lh2 .
bright states are a manifestation of excitonic delocalization , and their crucial contribution to the nearly optimal efficiency in the natural geometry can be corroborated by turning delocalization off .
[ fig : efficiencydistributions](d f ) shows the distribution of efficiencies for complexes that are trimmed by removing every second bchl .
doing so doubles the nearest - neighbor distances , weakening intra - complex couplings and suppressing excitonic delocalization , meaning that eet takes place by ordinary , site - to - site fret @xcite .
in particular , trimming lh1 not only reduces the efficiency from the natural delocalized case , but it also makes it so that the original orientation of bchls is no longer an efficiency outlier .
this confirms our claim that the feature which makes the original orientation of the bchls within lh1 an outlier is the coherent excitonic delocalization and the resulting supertransfer .
in summary , our study of geometric effects in purple - bacterial energy transfer reveals that , if the site energies are fixed , altering pigment orientations can significantly reduce the efficiency .
the effect is due to the fragility of low - lying excitonic states , whose high brightness in the natural geometry yields the high efficiency through supertransfer . the magnitude of the improvement a natural geometry that is 5.5 standard deviations better than the mean random geometry
is one of the largest photosynthetic efficiency enhancement we are aware of that has been attributed to a coherent effect .
the natural geometry s exceptional efficiency among plausible evolutionary alternatives suggests that it may have conferred an evolutionary advantage .
if so , delocalization would likely be a spandrel , a feature that was originally a byproduct of evolution , but was later exploited to improve fitness @xcite .
we argued previously that delocalization was not required for high efficiencies in purple - bacterial light harvesting , suggesting that it arose as a byproduct of the tight bacteriochlorophyll packing that enhances the absorption cross - section per rc @xcite .
but if ring - like structures with delocalization were already present , it is plausible that subsequent evolution adjusted the directions of the dipole moments to take advantage of supertransfer . of course
, these speculations would need to be tested by future work , especially through a comparison of corresponding structures in different taxa of purple bacteria .
more generally , our results confirm the predicted importance of supertransfer as one of the few coherent mechanisms possible in incoherent light @xcite and promise to increase its deployment in artificial light - harvesting complexes .
we thank saleh rahimi - keshari for useful discussions . we were supported by the australian research council through a discovery early career researcher award ( de140100433 ) and the centre of excellence for engineered quantum systems ( ce110001013 ) .
65ifxundefined [ 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 _ _ , ed .
( , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( )
|
the remarkable rotational symmetry of the photosynthetic antenna complexes of purple bacteria has long been thought to enhance their light harvesting and excitation energy transport .
we study the role of symmetry by modeling hypothetical antennas whose symmetry is broken by altering the orientations of the bacteriochlorophyll pigments .
we find that in both lh2 and lh1 complexes , symmetry increases energy transfer rates by enabling the cooperative , coherent process of supertransfer .
the enhancement is particularly pronounced in the lh1 complex , whose natural geometry outperforms the average randomized geometry by 5.5 standard deviations , the most significant coherence - related enhancement found in a photosynthetic complex .
photosynthetic organisms use light - harvesting antenna complexes to absorb light and funnel the resulting excitation energy into a reaction center ( rc ) , where the energy is used to drive charge separation @xcite . despite the diversity of antenna complexes , the efficiency of excitation energy transfer ( eet @xcite ) through them is generally high , prompting hopes that understanding eet mechanisms in these complexes will generate new ideas for improving artificial light harvesting @xcite . in searching for design principles in photosynthetic architectures ,
it is important to not assume that a particular photosynthetic system is optimized simply because it is a product of billions of years of natural selection . if nothing else , the dramatically different antenna architectures in different plant and bacterial taxa @xcite can not all be optimal . in other words , the optimality of photosynthetic light harvesting is a hypothesis to be tested , with there being a distinct possibility that a particular arrangement is not optimal but is merely good enough to ensure the particular organism s survival .
a way to determine whether an eet architecture is optimal is to examine its performance if its structure is changed in significant ways @xcite .
this kind of analysis has been carried out for the photosynthetic apparatus of several species .
for example , in a model of the cyanobacterial photosystem i ( psi ) , randomizing the orientations of the chlorophyll ( chl ) molecules about their mg atoms altered the overall quantum yield by less than 1% , and , indeed , the already high yield could be further increased by adding small variations in site energies @xcite .
similarly , in a kinetic model of photosystem ii ( psii ) , randomizing the chl orientations changed the yield by up to a few percent , with the x - ray geometry being near the middle of the distribution @xcite . to us
, these findings suggest that chl orientations in neither psi nor psii are fine - tuned to an optimal geometry , especially considering that the uncertainties in the approximations employed exceeded the maximum claimed few - percent enhancement .
eet through the fenna - matthews - olson ( fmo ) complex of green sulfur bacteria has also been widely studied , and it has been claimed to be close to optimal with respect to not only geometric changes but also changes in environmental parameters such as the temperature or the reorganization energy @xcite .
however , the optimality of eet in fmo is not settled , since treating photoexcitation realistically is likely to substantially affect the efficiency @xcite . here , we consider the light - harvesting apparatus of purple bacteria , well known for their highly symmetric antenna complexes lh2 and lh1 @xcite , shown in fig .
[ fig : structure ] .
eet through these complexes has been studied extensively @xcite , often using kinetic models .
most models recognize the importance of the strong couplings between bacteriochlorophylls ( bchls ) in lh2 and lh1 , which gives rise to considerable excitonic delocalization @xcite .
delocalization is particularly relevant in lh2 and lh1 , where it can lead to supertransfer @xcite , an enhancement of eet rates over site - to - site hopping because the dipole moments of individual pigments are oscillating with a definite phase .
limited work has been done on whether the efficiency of purple - bacterial light harvesting is optimal with respect to certain parameters . in particular
, it has been suggested that the internal symmetry of lh2 is particularly well - suited to maximizing the packing density and minimizing frustration within a larger lattice @xcite .
in addition , we have previously considered the efficiency of purple - bacterial light harvesting after changes to site energies and after suppressing delocalization by trimming away every second bchl @xcite .
trimming frequently reduced the efficiency by a large margin , confirming the influence of delocalization on light harvesting .
however , we also showed that delocalization is not necessary for high efficiency , since a decrease in performance could always be compensated by altering the site energies to create an energy funnel into the rc . here
, we investigate the efficiency of purple - bacterial light - harvesting in natural light conditions as the bchl orientations are changed , finding that the performance of the natural geometry is one of the highest among thousands of reorientations .
there is also a marked difference between the robustness of lh2 and lh1 to pigment reorientation . whereas changes to lh2 hardly affect the efficiency ,
the natural orientations in lh1 are a significant outlier , lying 5.5 standard deviations from the mean .
we attribute this sensitivity to supertransfer , which reaches its maximum near the natural orientations . indeed , because supertransfer is a consequence of excitonic delocalization , we show that the efficiency is less sensitive to geometric effects if delocalization is turned off by removing every second pigment .
|
in june 2013 , brazil faced the largest and most significant mass protests in a generation , comparable in size to the protests that triggered the collapse of the military dictatorship in 1984 @xcite .
the 2013 protests had been exacerbated by the broader disenchantment of the population towards the party system in brazil @xcite .
banners with sentences such as `` no party represents me '' or `` we do nt have a party , we are brazil ! ''
were commonly seen among the protesters . in response to these protests ,
the government proposed a program of political reform @xcite .
however , more than one year has passed and very little has been done .
why are people so unhappy with the brazilian party system ? to illustrate its incapability , consider the following examples .
19 months before rio de janeiro stages south america s first olympic games , an evangelical pastor without any link to sports was nominated as brazil s new sports minister .
he replaced communist aldo rebelo , who oversaw preparations for the world cup and was highly criticized for subsidizing the construction of white - elephant football stadiums @xcite .
aldo rebelo , who in 1994 proposed a bill that prohibits the adoption of any technological innovation in local , state and federal agencies , is , ironically , brazil s new minister of science and technology .
moreover , in 2010 elections , tiririca , a well known entertainer whose career began as a circus clown in brazil , was first elected to represent so paulo in congress , winning the most votes of any candidate in the country with the slogans `` it ca nt get any worse '' and `` what does a congressman do ?
in fact , i do not know , but vote for me and i will tell you '' .
one of the main causes of brazil s political inefficiency is its highly fragmented party system @xcite .
this is a system with many political parties and with no one party being able to obtain an absolute majority in the representative assembly .
the more fragmented the party system is , the less likely it is that the president s party will control a majority of seats in the legislature .
therefore , presidents are usually forming informal coalition governments , needing to build cross - party coalitions to implement most major policies @xcite . under these circumstances , many ( if not most ) deputies spend the bulk of their time arranging jobs and pork - barrel projects for their constituents in exchange for legislative support @xcite .
also , parties rarely organize around national - level questions , which means that congress rarely deals with serious social and economic issues @xcite . as a consequence , individualism , clientelism , and personalism , rather than programmatic appeals ,
dominate electoral campaigns @xcite .
more generally , party system fragmentation impacts the electoral dynamic , the process of coalition formation , governing , and ultimately , the survival of political systems in presidential democracies @xcite . in brazil
, party system fragmentation has reached one of the highest levels ever found in the world @xcite .
after 2014 elections , the number of parties represented in congress grew from 23 to 28 .
besides brazil , many countries have party systems with high levels of fragmentation , such as bolivia , bulgaria , denmark , ecuador , finland , france , guatemala , india , israel , italy , netherlands and thailand @xcite . in theory ,
the number of parties of a country can be explained by its electoral and social structures @xcite . with regard to the electoral structure , while plurality elections favor two - party competition , proportional representation ( pr ) electoral systems create fragmented party systems @xcite . concerning the social structures , the more socially heterogeneous a country is , the more electoral parties it will have @xcite .
social heterogeneity is measured either by the number of linearly independent ideological dimensions ( e.g. religious and socio - economic ) being discussed in the society or the number of social cleavages ( e.g. centre - periphery and state - church ) a country has @xcite .
thus , is it possible to measure if a fragmented party system is a reflex of a socially heterogeneous society ? in this direction , how can we determine whether an electoral system is optimally fragmented ? and
if it is not optimally fragmented , how can we optimize it ? to answer these questions ,
i propose _ arrange _ , a _
d**a**ta d**r**iven method fo**r * * * * a**ssessing and reduci**ng * * party fragm**e**ntation_in a country . , _
arrange_uses as input the roll call data , i.e. , the votes given by congressmen on bills and amendments , as a proxy for political preferences and ideology .
the idea , regularly employed by political scientists @xcite , is that congressmen who give the same vote regularly share the same political ideology and , therefore , should belong to the same party . using this insight , _ arrange_reorganizes the party system of a given country by trying to find the minimum amount of social cleavages that divides its congressmen into coherent voting blocks , i.e. , sets of congressmen whose members votes similarly .
these coherent voting blocks would be the new political parties of the analyzed country and would allow us to assess its actual level of fragmentation .
if _ arrange_divides the congressmen into a much lower number of political parties than the actual number , then it is possible to conclude that party fragmentation in this particular country is much higher than it should be . if this is the case , _
arrange_immediately provides a new congressmen party configuration
that both reduces the country s level of party fragmentation and increases intra - party similarity , which could potentially increase the efficiency of the party system @xcite . in summary ,
the main contributions of this paper are :
from roll votes primarily focuses on the estimation of cleavages and ideologies across congressmen and parties @xcite to characterize and predict legislative behavior @xcite . recently , with the advancement of political weblogs and online social networks , researchers are also extracting political knowledge from user generated data on the web .
there are studies that focus on mining political opinions @xcite and profiles @xcite from the texts users post on social media applications .
others , such as @xcite , extract political opinions from general texts , such as statement records of u.s .
senators and online news .
more recently , leman akoglu @xcite classified the political polarity of individuals using roll call votes of u.s .
congressmen and texts posted on political forums .
the idea behind all these studies is that political preferences tend to be stable over time and can be predicted accurately . .
within this sub - field , since duverger s seminal paper @xcite , many studies focused on predicting and understanding the factors that determine the number of parties that compete in a given polity @xcite . in summary , there are two lines of thought : one that emphasizes the role of electoral laws in structuring coalition incentives , and another that emphasizes the importance of preexisting social cleavages .
another fundamental problem in this sub - field is to count the number of parties by taking into account their relative size @xcite .
if , for instance , a party has a very small percentage of seats in congress ( e.g. one seat of one thousand ) , then it should be counted accordingly .
the metric that considers this is called _ the effective number of parties_. conceptually , the effective number of parties is simply the number of `` viable '' or `` important '' political parties in a party system that includes parties of unequal sizes .
since laakso and taagepera s seminal work @xcite , several ways of computing the effective number of parties were proposed @xcite .
the number of effective parties is a frequent metric for assessing party system fragmentation in a country @xcite .
the high interest in these problems comes from the fact that the actual number of parties usually determines the number of _ effective _ parties , or how _ fragmented _ a party system is @xcite .
highly fragmented party systems can affect governance drastically @xcite .
the more fragmented the party system is , the less likely it is that the president s party will control a majority of seats in the legislature .
simone bohn @xcite reviewed the literature and concluded that party system fragmentation impacts the electoral dynamic , the process of coalition formation , governing , and ultimately , the survival of political systems in presidential democracies .
thus , in this paper , we measure party fragmentation by counting both the actual and the effective number of parties .
another crucial factor for governance is party discipline , i.e. , the ability of a political party to get its members to support the policies of their party leadership .
mainwaring and shugart @xcite assessed the effects of this on the costs of governing . if parties are not disciplined , presidents will be forced to rely on ad - hoc coalitions based on the distribution of patronage to individual legislators , which raises the costs of governing and reduces policy coherence .
limongi and figueiredo @xcite argued that `` institutional engineering '' should focus on electoral formulas that reduce party fragmentation and increase party discipline .
brazil is a special case in politics for its high level of party fragmentation , being consistently analyzed in the literature @xcite .
thus , using brazil as the use case makes this work specially challenging but rewarding .
all the data used in this work was collected from the open data ( _ dados abertos _ ) project of the house of representatives ( or chamber of deputies ) of brazil . in total
, i collected roll votes on bills of @xmath0 thousand congressman that worked in the house of representatives of brazil from november , 4th , 1998 to december , 3rd , 2014 .
the reason for this particular time interval is related to the purpose of this work . since party discipline is a fundamental metric of evaluation , i only collected the bills in which the party leaders declared the desired vote for their fellow partisans .
more than @xmath1 of the votes given during this period had a declared party leader vote .
moreover , note that congressmen vote for bills and their amendments .
an amendment is a proposition presented as ancillary to the bill , to amend its form or content .
thus , bills and amendments compose a total of thousand propositions to be voted by the congressmen .
each congressman may or may not agree with the vote of his / her party leader .
there are , in total , thousand declared votes of the leaders of the 36 parties that had congressmen elected for the house of representatives during the analyzed period .
as discussed previously , reducing party fragmentation only makes sense if party discipline does not decrease significantly .
one way to measure the discipline of a congressman is to compute the fraction of votes given by him / her through his political life that agreed with his / her party leader .
however , in brazil it is common for congressmen to switch parties .
also , it is well known that some parties demand ( or inspire ) higher levels of discipline than others @xcite .
thus , instead of analyzing the discipline of the whole political life of a congressman , i will analyze his / her discipline as a member of a single party , i.e. , a congressman may have different levels of discipline if he / she was member of different parties during his / her political life . for
simplicity , from now on i will assume that the vote given by the leader of the party was given by the party itself , e.g. , i will call _ party vote _ the vote given by the leader of the party .
thus , given the set of @xmath2 congressmen @xmath3 and the @xmath4 political parties that compose the set @xmath5 , i define a * partisan * @xmath6 as the tuple formed by a congressman @xmath7 and a political party @xmath8 .
the set containing all @xmath9 partisans is defined as @xmath10 .
these partisans job is to vote for the set @xmath11 of @xmath12 bills and amendments that were put to vote in the house of representatives during the analyzed period . from now on
, i will use the term * propositions * to refer to both bills and amendments . since
a partisan @xmath13 had not necessarily voted for all propositions , i define the set @xmath14 as the set of propositions that were voted by partisan @xmath13 .
i also define @xmath15 as the set of partisans which are members of party @xmath8 , i.e. , @xmath16 .
before the partisans give their votes for a given proposition @xmath17 , their parties have to announce their votes for @xmath17 . again , since a party @xmath8 had not necessarily voted for all propositions , i define the set @xmath18 as the set of propositions that were voted by party @xmath8 . for each proposition @xmath19 of a given party @xmath8 , there is a vote @xmath20 associated with it . in the same way , for each proposition @xmath21 of a given partisan @xmath13 , there is a vote @xmath22 associated with it .
thus , a party @xmath8 and a partisan @xmath13 have , respectively , a set of votes @xmath23 and @xmath24 , where @xmath25 and @xmath26 .
the set of all votes given by partisans and parties is simply @xmath27 .
a vote @xmath20 given by party @xmath8 on a proposition @xmath17 may be of four types : _ y _ ( yes ) , _ n _ ( no ) , _ o _ ( obstruction ) and _ f _ ( free ) , i.e. , @xmath28 .
if the vote is @xmath29 ( @xmath4 ) , the party approves ( disapproves ) the proposition . if the vote is @xmath30 , the party is trying to avoid the vote on the proposition , i.e. , its partisans are called to withdraw from the plenary
. finally , if the vote is @xmath31 , its partisans are free to vote at will .
similarly , a partisan @xmath13 vote @xmath32 on a proposition @xmath17 may be of three types : _ y _ , _ n _ and _ o _ , i.e. , @xmath33 . if the vote is @xmath29 ( @xmath4 ) , the partisan approves ( disapproves ) the proposition .
if the vote is @xmath30 , the partisan withdrew from plenary . in this work ,
the vote is the fundamental feature that determines a preference or ideology .
since for all propositions in our dataset we have both the vote of the partisan and the party , here i define a general function @xmath34 that receives two votes as input and outputs @xmath35 if the votes are in accordance or @xmath36 otherwise .
this function is defined as : @xmath37 note that both @xmath38 and @xmath39 can be a vote of a partisan or a party .
also note that a @xmath31 vote implies accordance , since the party that gave that vote does not particularly care about its members votes .
once we know how to compare votes , we can propose a way to compare the similarity @xmath40 between two vote sets @xmath41 and @xmath42 .
it is given as : @xmath43 in summary , @xmath40 sums all the votes in agreement between the vote sets @xmath41 and @xmath42 considering only the propositions that are in both sets . from equation [ eq : sim ] , we can define the three levels of discipline that we will use throughout this paper .
first , i define * partisan discipline * as the discipline @xmath44 of a partisan @xmath6 towards his / her party @xmath8 , calculated as : @xmath45 second , i define * party discipline * as the discipline @xmath46 throughout all the votes that were given by partisan members of the party @xmath8 , calculated as : @xmath47 finally , i define * overall discipline * as the discipline @xmath48 throughout all the votes given in the house of representatives during the analyzed period , calculated as : @xmath49 \in \mathcal{a}}{|\mathcal{b}_a|}\times d_{a\rightarrow p}}{\sum_{[a:=(u , p ) ] \in \mathcal{a}}|\mathcal{b}_a| } \label{eq : overaldiscipline}\ ] ]
in this section i will show a summarized view of the fundamental characteristics of brazilian party system that are relevant to the purpose of this work .
s. fig .
[ fig : partysize ] , i show the historical participation of all parties in the house of representatives during the analyzed period . _ party participation _
, which i will interchangeably call _ party size _ , is represented by the total number of propositions that were voted by the members of the party ( horizontal axis ) and the total number of partisans that are and were members of the party ( vertical axis ) .
observe the heterogeneity of this universe .
from the three biggest parties ( pmdb , psdb and pt ) , with hundreds of partisans who voted for thousands of propositions , to the two smallest ones ( pstu and pmr ) , which together have only three partisans and propositions , there are another @xmath50 parties with very distinct levels of representation . in order to verify
if there is any correlation between participation and party discipline , i show in fig .
[ fig : partydiscipline ] , for each party @xmath8 , the total number of votes given by partisans members of @xmath8 ( horizontal axis ) and the party discipline @xmath46 ( vertical axis ) .
note that there is no apparent relationship between party discipline and participation .
in fact , the pearson s correlation coefficient between party discipline and the total number of votes is , but since the @xmath51 for testing the hypothesis of no correlation is , we can not affirm the correlation is significant . nevertheless , as already observed by @xcite using a smaller dataset ,
party discipline in brazil is consistently high : no party has a historical party discipline below @xmath52 and only parties have a figure below @xmath53 . in fig .
[ fig : partisandiscipline ] , we show the behavior of partisans discipline in brazil during the analyzed period .
first , observe in fig .
[ fig : partisandiscipline]a the cumulative distribution function ( cdf ) of all partisans disciplines .
note that the curve representing partisan discipline in brazil is not very far away from the ideal curve , where all partisans have discipline of @xmath54 .
thus , together with party discipline , partisan discipline is also usually high in brazil : only @xmath55 of partisans have discipline lower or equal to @xmath56 . in fig .
[ fig : partisandiscipline]b , we plot the heatmap of partisans using their disciplines and total number of votes .
the color bar at the right indicates the number of partisans in a given area of the map .
observe that the vast majority of partisans are located in the upper - right of the heatmap , i.e. , they have given many votes and have high partisan discipline .
although party fragmentation in brazil has reached one of the highest levels ever found in the world @xcite , we have seen that party and partisan discipline is consistently high .
does this mean that the level of party fragmentation in brazil is necessary ? according to the seminal _ doctrine of responsible party government _
@xcite , parties must differ sufficiently between themselves , providing the electorate with a proper range of choice between alternative actions . given that ,
i reformulate the previous question : is the actual level of party fragmentation in brazil a consequence of a high number of sufficiently different parties ? instead of performing a deep clustering analysis to answer this question
, i will apply the principal component analysis ( pca ) @xcite technique to the matrix @xmath57 composed by the voting vectors @xmath58 of each partisan @xmath59 .
pca is a widely used statistical technique for unsupervised dimension reduction .
it transforms the data into a new coordinate system such that the greatest variance is achieved by projecting the data into the first coordinate , namely principal component , the second greatest variance is achieved by projecting into the second coordinate , the second component , and so on .
in order to draw more interesting conclusions from this analysis , i will also apply pca to the matrix @xmath60 composed by the voting vectors of the u.s .
congressmen . for this
, i will use the same dataset used in @xcite , which consists of votes on 1655 bills in the house of representatives in years 2009 - 2010 by 451 representatives . in the matrices
the @xmath61 votes are represented as @xmath35 , the @xmath62 votes as @xmath63 , and the @xmath31 , @xmath30 and non - attendance as @xmath36 . to make the comparison more precise
, i will only use the votes in years 1999 - 2000 for constructing @xmath57 .
this period comprises votes on the bills by 767 contemporary partisans of 18 parties in the first two years of president cardoso in power .
also , it is the two year period that gives the higher explained variance by the first two components of the pca . in fig .
[ fig : pca ] , i show the first two components of the pca for both @xmath57 and @xmath60 , where each point represents a congressman and each symbol represents a party .
observe that , for both usa and brazil the first two components explain a significant part of the variance : @xmath64 and @xmath65 , respectively .
however , only the usa pca can visually divide the members of different parties . for brazil
, many members of different ( same ) parties are located together ( apart ) .
this suggests that parties in brazil are not sufficiently different to justify one of the highest levels of party fragmentation ever found in the world @xcite .
in this section i describe the method _ arrange _ , which has , basically , two steps .
first , based on the votes given by party leaders , it tries to find pairs of parties that can be merged into one .
the idea is that parties that always give the same vote could be merged into a single party .
then , _ arrange_attempts to assign partisans to new parties with the objective of minimizing the total number of parties receiving partisans and preserving party discipline .
finally , i describe the quality outputs of _ arrange_. formally , the problem tackled by _
arrange_is : given a set of parties @xmath66 , a set of partisans @xmath67 , a set of bills and amendments @xmath68 and the set of all votes given by parties and partisans @xmath27 , find the minimum set of parties @xmath69 to which the partisans in @xmath67 can be assigned in a way that overall , party and partisan disciplines are maximized .
[ problem1 ] in order to assign partisans to other parties , it is necessary to formally define the ways it can be done .
thus , here i define an * option * as the descriptor of the party that can receive an external partisan as member . more formally , given a partisan @xmath70 , his / her current party @xmath71 , and his / her set of propositions @xmath72
, an option @xmath73 is a tuple composed by the partisan @xmath13 , a party @xmath74 , and the similarity value @xmath75 between @xmath13 and @xmath76 .
more importantly , the option @xmath77 exists _ if and only if _ @xmath78 , i.e. , if the party @xmath76 has voted for all propositions in @xmath72 . the set @xmath79 is composed by all the options of partisan @xmath13 or , more formally rcl _ a & = & \{[o_a : = ( a:=(u , p_i ) , p_j , sim(a , p_j ) ) ] : + & & p_i , p_j p_j p_i _ a _
p_j}. moreover , an option @xmath73 of partisan @xmath70 is characterized as a _ good option _ if @xmath80 .
the set of good options @xmath81 for partisan @xmath70 is defined as rcl _ a^ * & = & \{[o_a = ( a:=(u , p_i ) , p_j , sim(a , p_j ) ) ] _ a : + & & sim(a , p_j ) sim(a , p_i ) } [ eq : goodoptions ] in fig .
[ fig : numberofoptions]a , i show the histogram of the number of good options @xmath82 for all partisans of our dataset .
consistent with the high levels of party and partisan discipline , observe that the majority of partisans do not have a single good option , i.e. , for @xmath83 of the partisans there is no other party in brazil that offers more similar voting vectors .
in fact , only @xmath84 of the partisans have more than three good options .
this result suggests that it is very difficult to reduce the number of parties in brazil by moving partisans from one party to another .
the main reason for the low number of good options is related to the short lifetime of many parties in brazil , as it can be observed in fig .
[ fig : partysize ] . since an option exists if and only if the propositions voted on by the partisan is a subset of the propositions voted on by the parties , many partisans with a long history of votes can not find options for them among the small parties . thus
, here i propose a method for creating new parties by merging existing ones .
the method is based on a simple idea : if two parties are not contemporary or are contemporary , but all votes given by them are equal , then these parties can be merged into a new one . more formally , two parties @xmath71 and @xmath76 can be merged into a new party @xmath85 if one of the two conditions below is met : * * c1 .
* @xmath86 , i.e. , parties @xmath71 and @xmath76 have not voted for a common proposition . *
* c2 . * @xmath87 ,
i.e. , parties @xmath71 and @xmath76 have given the same vote for all common propositions . given these two conditions ,
the first thing we have to do is find all pairs of parties to which at least one of the two conditions is valid .
once this is done , for every pair of parties @xmath88 that can be merged , we create a merged party @xmath89 , for which the sets of propositions and votes are , respectively , @xmath90 and @xmath91 .
all parties created in this step are put in set @xmath92 .
after this , we repeat this process by verifying , for each party @xmath93 , all parties @xmath94 that can be merged to @xmath8 .
we merge @xmath8 to @xmath95 into @xmath89 as previously , but this time adding the merged parties to @xmath96 .
once this process is done , we copy @xmath96 to @xmath92 , empty @xmath96 , and restart the process of finding parties @xmath94 eligible to be merged to parties @xmath93 .
the process ends when set @xmath96 does not receive a new merged party .
all parties @xmath93 that were not merged are put in the final set of merged parties @xmath97 .
this whole process is described in algorithm [ alg : partymerger ] .
* new * ` party ` @xmath89 @xmath100 @xmath101 @xmath102 @xmath103 ` false ` @xmath104 @xmath103 ` true ` @xmath105 @xmath106 @xmath107 @xmath108 basically , what algorithm [ alg : partymerger ] does is to find all possible @xmath109combinations @xmath110 of set @xmath66 for all @xmath111 .
it is well known that @xmath112 , making the worst - case complexity for this problem to be @xmath113 .
nevertheless , in practice , finding parties in @xmath66 that can be merged with parties in @xmath92 gets significantly harder as @xmath114 increases , which , in practice , makes algorithm [ alg : partymerger ] computationally feasible for the problem in question . for the case of the 36 brazilian parties ,
@xmath114 went up to 6 and the algorithm stopped , generating a set @xmath108 containing 95 parties , in which only 5 were not a product of a merge , namely pdt , pmdb , pps , psdb and pt . besides these , algorithm [ alg : partymerger ] generated 12 parties of size 2 , 17 parties of size 3 , 20 parties of size 4 , 31 parties of size 5 and 10 parties of size 6 . in fig .
[ fig : numberofoptions]b , i show the histogram of the number of good options @xmath82 for all partisans considering the new set of merged parties @xmath108 .
observe that the number of partisans that do not have a single good option dropped from @xmath115 to @xmath116 , all of them being members of the five parties that were not merged .
moreover , the number of partisans that have more than three good options grew from @xmath117 to @xmath118 .
thus , i conclude that merging parties significantly raises the chances of reducing the number of parties in brazil by moving partisans from one party to another without decreasing party discipline .
now that most of the partisans have multiple good options , we can find ways of redistributing them among the parties in @xmath108 . the idea in this redistribution is to find a set of parties @xmath119 that has a lower cardinality than @xmath66 , i.e.
, @xmath119 is a set of parties able to receive all partisans @xmath120 as members and the number of parties @xmath121 in @xmath69 has to be lower than the actual number of parties @xmath122 .
however , this has to be done cautiously , since there are two partially conflicting goals : 1 .
minimize the number of parties ; 2 .
maximize party and partisan discipline .
these goals are conflicting because minimizing the number of eligible parties to receive partisans as members implies reducing the options for moving partisans and , as a consequence , the number of good options .
if a partisan does not have a good option , he / she is obliged to stay in his / her party in order to not decrease his / her partisan discipline . on the other hand , maximizing the discipline implies in maximizing the size of the set of good options and , therefore , the number of parties to be considered has to be as high as possible . in order to solve this conflict ,
i model this redistribution problem as a * set cover problem * ( * scp * ) @xcite .
scp is a well studied problem for the field of approximation algorithms @xcite , being also one of karp s 21 np - complete problems shown to be np - complete . in summary , given a set of elements @xmath123 ( called the universe ) and a set @xmath124 of @xmath12 sets whose union equals the universe , the scp is to identify the smallest subset of @xmath124 whose union equals the universe .
more formally , given a universe @xmath125 and a family @xmath126 of subsets of @xmath125 , a _
cover _ is a subfamily @xmath127 of sets whose union is @xmath125 . in our case ,
the universe @xmath125 is the set of partisans @xmath67 and the family @xmath126 of subsets of @xmath125 is a family of subsets of partisans @xmath128 where each subset @xmath129 is composed by the partisans that are eligible for moving to party @xmath97 . in order to build @xmath129 , it is necessary to recalculate the set of options @xmath130 and good options @xmath81 of each partisan @xmath13 with respect to the merged parties in @xmath108 .
once this is done , i define that each set @xmath129 is composed by all the partisans @xmath13 that have a good option @xmath131 , i.e. , @xmath132 \in \mathcal{o}_a^ * \}$ ] . since this problem is np - complete ,
i recur to a greedy algorithm to solve it .
literature shows that the greedy algorithm is essentially the best possible polynomial time approximation algorithm for set cover under plausible complexity assumptions @xcite .
the greedy algorithm for set covering chooses sets according to one rule : at each stage , choose the set that contains the largest number of uncovered elements .
it can be shown @xcite that this algorithm achieves an approximation ratio of @xmath133 , where @xmath134 is the size of the set to be covered and @xmath135 is the @xmath12-th harmonic number : @xmath136 . for our specific problem ,
this algorithm works as follows , being described in algorithm [ alg : greedy ] .
first , we create two empty sets : @xmath69 , which will contain the final set of parties , and @xmath137 , which receives the covered partisans during the process . then , while @xmath137 does not contain all partisans , we find the party @xmath97 to which @xmath129 contains the largest number of uncovered partisans .
then , we add all partisans @xmath138 to @xmath137 , make @xmath129 an empty set ( so it is not selected again ) , and add @xmath8 to the final set of parties @xmath69 .
note that so far this process _ guarantees _ that all partisans will , at least , have the same partisan discipline as their actual ones , since only good options are used to build the sets @xmath139 . nevertheless , it is possible that relaxing this constraint a little might diminish the total number of parties that compose @xmath69 considerably .
for instance , we may allow a partisan to be member of a party if their similarity is at most @xmath140 smaller than his / her similarity with his / her actual party .
this relaxation increases the number of partisans that are eligible to be member of other parties and , therefore , may reduce the size of @xmath69 .
thus , here i introduce the parameter @xmath141 , which is the maximum allowed difference between the actual partisan discipline and the future one .
we accommodate this in _ arrange_by simply changing the way the set of good options is constructed . with the introduction of @xmath141 , the set of good options @xmath81
is defined as : rcl _ a^ * & = & \{[o_a : = ( a:=(u , p_i ) , p_j , sim(a , p_j ) ) ] _ a : p_i + & & p_j ^m sim(a , p_j ) sim(a , p_i ) - } .
[ eq : goodoptionsdelta ] in summary , after generating the set of merged parties @xmath108
using algorithm [ alg : partymerger ] , it is necessary to create the sets @xmath129 for all @xmath97 considering @xmath141 . these sets @xmath129 will contain all partisans that are eligible for being members of party @xmath8 given @xmath141 .
this is done by selecting a value for @xmath141 and then running the procedure of algorithm [ alg : greedy ] .
if , for instance , @xmath142 , then no partisans are allowed to decrease their discipline when moving to a different party . on the other side , if @xmath143 , partisans are allowed to be members of any party that voted for all propositions that they voted on , i.e. , partisan discipline is not a constraint . after running algorithm [ alg : greedy ]
, we will have a minimum set of parties @xmath69 able to accommodate all partisans in @xmath67 .
then , all we have to do is to assign a party @xmath144 to each partisan @xmath120 .
this is done by selecting the option @xmath145 that gives the maximum similarity value @xmath75 for all parties @xmath146 and making @xmath76 the new party of partisan @xmath13 .
@xmath147 ~:~ p \in \mathcal{p } ~\wedge~
q \in \mathcal{p}^m ~\wedge~ \mathcal{b}_a \subseteq \mathcal{b}_{q}\}$ ] @xmath148 \in \mathcal{o}_a ~:~ sim(a , q ) \geq sim(a , p)-\delta\}$ ] @xmath149 @xmath150 @xmath151 @xmath152 @xmath153 @xmath154 @xmath155 @xmath69 the main goal of _ arrange_is to reduce party fragmentation by reducing the number of parties that are able to accommodate all elected partisans . thus , the main output of _
@xmath156 ( or @xmath157 ) . nevertheless , it is necessary to assess the quality of the _ cover _ @xmath69 , since it is essential to reduce party fragmentation by achieving desirable levels of party and partisan discipline .
but what are the desirable levels of party and partisan discipline ? in fig .
[ fig : partisandiscipline]a , i showed the cdf of the actual partisan discipline distribution during the analyzed period .
the most desirable , or _
ideal _ , partisan discipline distribution is shown as a dashed red line , which represents the situation where all partisans have discipline of @xmath54 .
thus , a new partisan discipline distribution , generated by algorithm [ alg : greedy ] with parameter @xmath141 and defined by the random variable @xmath158 , is considered _ desirable _ if its cdf @xmath159 is closer to the ideal than the actual one , defined by the random variable @xmath160 and cdf @xmath161 . more formally , considering that the area under the ideal cdf curve is @xmath36 for both partisan and party discipline distributions , i propose the following definition : a new discipline distribution defined by random variable @xmath158 and its cdf @xmath159
is considered a _ * desirable discipline distribution * _ if @xmath162 , where @xmath161 is the cdf of the actual discipline distribution defined by random variable @xmath160 .
[ def : desirable ] in other words , if the area under @xmath159 is smaller than the area under @xmath161 , then @xmath158 represents a _ desirable discipline distribution_. moreover , given that discipline ranges from @xmath36 to @xmath35 , i propose the following lemma : given a random variable @xmath160 representing the actual discipline distribution and a random variable @xmath158 representing a new discipline distribution , if the expected value @xmath163 $ ] of @xmath158 is higher than the expected value @xmath164 $ ] of @xmath160 , then @xmath158 represents a _ desirable discipline distribution_. from probability theory , we know that @xmath165 = \int_{0}^{\infty}1-f_x(x)~dx$ ] for a given random variable @xmath166 and cdf @xmath167 @xcite . since discipline has values from @xmath36 to @xmath35 , we can write @xmath165 = 1-\int_{0}^{1}f_x(x)~dx$ ] , or @xmath168 $ ] for discipline distributions . then , we can replace definition [ def : desirable ] for @xmath169 < 1-e_{x_0}[x]$ ] or @xmath163 > e_{x_0}[x]$ ] .
thus , now i can formally define three binary quality signals for the new partisan configuration over the new set of parties @xmath69 generated by algorithm [ alg : greedy ] with parameter @xmath141 .
these signals indicate , respectively , if the new partisan configuration has desirable levels of partisan , party and overall discipline , and is defined as : * * @xmath170 * : @xmath35 if the overall discipline @xmath171 of the new configuration is greater than the overall discipline @xmath172 of the actual configuration ; @xmath36 otherwise .
recall that the overall discipline @xmath48 is defined in equation [ eq : overaldiscipline ] . *
* @xmath173 * : @xmath35 if the expected ( average ) partisan discipline of the new configuration @xmath174 $ ] is greater than the expected party discipline of the actual configuration @xmath175 $ ] or , in other words , if the new partisan discipline distribution is a _
desirable partisan discipline distribution _ ; @xmath36 otherwise . * * @xmath176 * : @xmath35 if the expected ( average ) party discipline of the new configuration @xmath177 $ ] is greater than the expected party discipline of the actual configuration @xmath178 $ ] or , in other words , if the new party discipline distribution is a _ desirable party discipline distribution _ ; @xmath36 otherwise .
in this section i show the results for _ arrange_for a hundred equally distributed values of @xmath141 between @xmath36 and @xmath35 . from now on i will call a configuration @xmath179 the distribution of partisans among parties generated by _ arrange_for a particular @xmath141 value .
moreover , in all plots i will indicate whether the quality signals described in previous section were @xmath35 or @xmath36 .
namely , i will use yellow stars for results where all quality signals were @xmath35 ( @xmath180 ) , blue diamonds when only signals @xmath173 and @xmath176 were @xmath35 ( @xmath181 ) , red circles when only signal @xmath176 was @xmath35 and , finally , green squares when all quality signal were @xmath36 . in summary , _
arrange_was able to generate @xmath50 distinct configurations that , compared with the _ status quo _ , have ( i ) a significantly smaller number of parties , ( ii ) higher discipline of partisans towards their parties and ( iii ) more even distributions of partisans into parties . besides comparing with the _ status quo _ , _
arrange_will be compared with two random models : _ random - sq_and _ _ random-__@xmath141 .
the competitors can be summarized as : 1 .
_ status quo_. it is the existing state of affairs , i.e. , the actual and historical situation in brazil during the analyzed period .
random - sq_. it randomly redistributes the partisans among the existing parties . for each partisan @xmath120 ,
the model randomly pick an option @xmath182 \in \mathcal{o}_a$ ] and assigns @xmath13 to party @xmath8 . by using the options set
@xmath130 i guarantee that the party @xmath8 allocated to @xmath13 has voted for every proposition voted by @xmath13 , i.e. , @xmath183 .
_ random-__@xmath141 .
works in the same way as _ random - sq _ , but instead of allocating partisans to the actual set of parties @xmath66 , it randomly redistributes the partisans among the minimum set of parties @xmath69 generated by algorithm [ alg : greedy ] . the random models are used to quantify the payoffs obtained by using _ arrange_when it generates the minimum party set @xmath69 ( _ random - sq _ ) and it efficiently allocates partisans to the parties of this set ( _ _ random-__@xmath141 ) .
although not always visible , for all results of both models i also show the @xmath184 confidence interval . in fig .
[ fig : numberofparties ] , i show the number of parties @xmath156 generated by _ arrange_for different values of @xmath141 .
i also show the _ status quo _ , i.e. , the actual number of parties @xmath4 of which partisans were members during the analyzed period , and the number of parties generated by _
random - sq_. first , note that the number of parties generated by _
arrange_is significantly lower than the _ status quo _ , decreasing as @xmath141 increases .
even when no partisans are allowed to decrease its discipline ( @xmath142 ) , @xmath185 , a number @xmath186 lower than @xmath187 , the _ status quo_. moreover , _
arrange_was able to generate a configuration ( @xmath188 ) with only @xmath189 parties ( @xmath190 reduction ) and with all quality signals equal to @xmath35 ( @xmath180 ) , i.e. , with overall , party and partisan disciplines greater than the _
status quo_. when only @xmath173 and @xmath176 are @xmath35 ( @xmath181 ) , _
arrange_could generate a configuration ( @xmath191 ) with @xmath192 parties , a @xmath193 reduction .
finally , when only @xmath176 is @xmath35 ( @xmath176 ) , _ arrange_could provide a @xmath194 reduction in the number of parties by generating a configuration ( e.g. @xmath195 ) with only @xmath196 parties .
it is worth mentioning that the number of parties in brazil is so excessive that even _
random - sq_was able to generate a configuration with fewer parties than the _ status quo_. concerning discipline
, i show in fig .
[ fig : discipline ] the overall discipline ( fig .
[ fig : discipline]a ) and the average party ( fig .
[ fig : discipline]b ) and partisan ( fig .
[ fig : discipline]c ) disciplines of the configurations produced by _ arrange_and its competitors .
observe that for all discipline metrics the results produced by _
arrange_are very similar with the _ status quo _ , even when all quality signals are @xmath36 .
the discipline values decrease significantly only for @xmath141 values close to @xmath35 , when the number of parties generated by _
arrange_is @xmath197 or @xmath35 .
it is also interesting to note that the random models are able to produce configurations with considerably high levels of discipline , which shows that brazilian political parties are , on average , very similar to each other . in order to analyze how well distributed
are the partisans among parties , we compute the gini coefficient @xcite for each configuration generated by _ arrange_and its competitors .
the gini coefficient was initially proposed to describe the income inequality in a population @xcite .
it assumes values from @xmath36 , which expresses perfect equality , where all parties have the same number of partisans , to @xmath35 , which expresses maximal inequality among values , where all partisans are allocated to a single party .
observe in fig .
[ fig : ginicoefficient ] that _
arrange_is able to produce configurations in which the partisans are more evenly distributed than the _ status quo_for all values of @xmath141 .
while the gini coefficient is @xmath198 for the _ status quo _ , it decreased to ( @xmath180 ) , to , and to @xmath199 for ( all quality signals equal to @xmath36 , but with discipline values similar with the _ status quo _ ) .
random - sq_has a slightly high gini coefficient than the _ status quo_because a few big parties are more likely to randomly receive new partisans , since they appear as an option in the option sets @xmath130 for all @xmath120 .
another relevant characteristic to be measured in political party systems is switching , i.e. , party changes among partisans . as stated by @xcite
, switching effectively destroys the meaning of party labels , raises voters information costs , and eliminates party accountability , being a threat to the very core of democratic representation .
thus , in fig .
[ fig : partychanges ] , i plot the number of party changes among partisans that would occur if the configurations generated by _
arrange_and the random models were the reality .
observe that in all scenarios generated by _
arrange_the number of changes is lower than the _ status quo _
, _ random - sq_and _ _ random-__@xmath141 .
this is another evidence of the importance of reducing party fragmentation to have ideologically well defined parties .
now i will take a closer look at particular configurations generated by _ arrange _ , namely .
these are the configurations that have the lowest number of parties and achieved , respectively , quality signals @xmath180 , @xmath181 and only @xmath176 . in fig .
[ fig : npartiesperyear]a , i show the number of active parties per year for the _ status quo_and these three configurations .
first , observe that the number of parties changes constantly over the years in brazil , this being a harmful consequence of its highly fragmented party system . on the other hand , observe that the configurations generated by _
( i ) much more stable over the years and ( ii ) have a significantly smaller number of parties than the _ status quo_.
besides counting the actual number of parties per year , it is also important to measure the _ effective number of parties_. as described previously , it is a concept which provides for an adjusted number of political parties in a country s party system , weighting the partisan count per party by their relative strength @xcite . in our case , the relative strength refers to their seat share in the parliament .
this measure is especially useful to detect trends toward fewer or more numerous parties over time @xcite .
the number of parties equals the effective number of parties only when all parties have equal strength . in any other case ,
the effective number of parties is lower than the actual number of parties .
it is also a frequent metric for the fragmentation of a party system @xcite .
moreover , although several indexes for computing the effective number of parties exist @xcite , in this paper i use the _ golosov _ index @xmath200 , where @xmath4 is the actual number of parties , @xmath201 is the proportional share of each party @xmath71 , and @xmath202 is the highest share of a party @xcite . for the best of my knowledge
, @xmath203 is the most recent one and its results confirm it works better than earlier proposed alternatives in measuring the effective number of components in highly fragmented and highly concentrated party systems , which is the case of brazil .
thus , in fig . [
fig : npartiesperyear]b , we show the _ effective number of parties _
@xmath203 per year , calculated for configurations and for the _ status quo_. first , observe that @xmath203 is significantly lower and more stable for than for the _ status quo_. while @xmath203 grows constantly after the year @xmath204 for the _ status quo _ , it remains practically constant for the three configurations generated by _
arrange_. moreover , if we consider only , brazil would go from having one of the most fragmented party systems in the world @xcite to having one of the least fragmented @xcite , averaging effective parties per year . in fig .
[ fig : pca]b i showed the first two components of the pca for the matrix @xmath57 composed by the votes of the partisans of brazil during the years of 1999 and 2000 . in fig .
[ fig : pcanew ] , i show this same result , but i replace the _ status quo_party labels by the ones generated by _ arrange_in .
observe that all three configurations have a visually better clustering of partisans of the same party than the _ status quo_. this suggests that _ arrange_is also able to provide configurations in which parties are more different among themselves than in the _ status quo_.
i leave a deeper quantitative analysis for future work .
in practical terms , the main contribution of _ arrange_to a government and its population is the ability to provide a quantitative assessment of fragmentation in its party system . given that the effective number of parties is expected to reflect the number of issue ( or ideological ) dimensions in a country @xcite , highly fragmented party systems overestimate this number of issue dimensions , providing a distorted view to the population that harms democracy @xcite .
thus , by mapping plenary votes into ideological preferences , _ arrange_quantitatively provides an `` ideal '' number of parties for a country given the ideological preferences of its congressmen , revealing the _ presumable _ true number of issue dimensions that exists in this particular country .
i used the expression `` _ _ presumable _ _ true '' because , as verified by @xcite , it is not always true that a disciplined and cohesive party represents an ideological cleavage ( or group ) in society , i.e. , its members may simply be a group of congressmen obeying its leader in order to obtain a particular benefit .
it is also worth mentioning that _ arrange_also provides the list of parties that should ideally exist and which partisans should be their members .
although i know a democratic government can not implement this solution easily , it can be used to support significant reforms in its political system . with this in mind , i could not fail to mention in this paper that one of the possible reasons for brazil s high level of party fragmentation is the so called _ fundo partidrio _
, which are funds distributed by the federal government to brazilian political parties for them to spend indiscriminately .
a share of the amount paid by the federal government through the _ fundo partidrio _ is the same for every party , but another part is proportional to the number of elected congressmen , senators and governors by each organization . in 2014 , _ pt _ , the party with the highest share ( @xmath205 ) , received r@xmath206 million _ reais _ from the fund . on the other hand , _ pros _
, the party with the lowest share ( @xmath207 ) , received r$493,873.68 thousand _ reais _ in 2014 @xcite .
consider that , in 2014 , the exchange rate of the _ real _ varied from @xmath208 to @xmath209 u.s .
it is out of the scope of this paper to point _
fundo partidrio _ as the main culprit for brazil s high level of party fragmentation , but it poses as a clear incentive for the creation of many parties in brazil . in spite of the fact that this work considers brazil as its use case
, the methods and results shown here can be easily replicated to other countries that have highly fragmented party systems .
carsten anckar studied party system fragmentation in 77 countries @xcite and reported high levels of fragmentation for many countries besides brazil , such as bolivia , bulgaria , denmark , ecuador , finland , france , guatemala , india , israel , italy , netherlands and thailand . nevertheless , it is important to point out that this work was motivated and eased by the open data initiative of the brazilian government , that provides public data related to politics and also to many other areas , such as demographics , government spending , budget and road accidents .
in this work , i proposed the method _
arrange_to assess and reduce fragmentation in multi - party political systems . from roll votes data of partisans and their respective party leaders ,
_ arrange_redistributes the partisans into new parties considering two conflicting objectives : to minimize the number of parties and to maximize party discipline . when applied to brazilian historical roll call data , _
arrange_was able to generate distinct configurations that , compared with the _ status quo _ , have ( i ) a significantly smaller number of parties , ( ii ) higher discipline of partisans towards their parties and ( iii ) more even distributions of partisans into parties .
these results show that brazil has and had many redundant parties , i.e. , parties that are very similar ideologically .
thus , if today brazil has one of the highest levels of party system fragmentation in the world @xcite , this work proved it could be much lower .
finally , it is important to point out that _ arrange_is a general method and could be directly applied to analyze fragmentation in any of the many highly fragmented party systems that exists in the world @xcite .
mische a. `` come to the streets , but without parties '' : the challenges of the new brazilian protests ; 2013 .
available from : https://mobilizingideas.wordpress.com/2013/09/04/come - to - the - streets - but - without - parties - the - challenges - of - the - new - brazilian - protests/. pennacchiotti m , popescu am . .
in : proceedings of the 17th acm sigkdd international conference on knowledge discovery and data mining - kdd 11 ; 2011 .
available from : http://dl.acm.org/citation.cfm?doid=2020408.2020477 .
akoglu l. quantifying political polarity based on bipartite opinion networks ; 2014 .
available from : https://aaai.org/ocs/index.php/icwsm/icwsm14/paper/view/8073/8100 .
poole kt , rosenthal h. .
american journal of political science .
1991;35(1):228278 .
desposato sw . .
american journal of political science .
2006 jan;50(1):6280 . available from : http://doi.wiley.com/10.1111/j.1540-5907.2006.00170.x .
ranney a. .
jefferson memorial lectures .
university of california press ; 1975 .
available from : http://books.google.com.br/books?id=6mjbavjavqcc .
|
in june 2013 , brazil faced the largest and most significant mass protests in a generation .
these were exacerbated by the population s disenchantment towards its highly fragmented party system , which is composed by a very large number of political parties . under these circumstances ,
presidents are constrained by informal coalition governments , bringing very harmful consequences to the country . in this work
i propose _ arrange _ , a _
d**a**ta d**r**iven method fo**r * * * * a**ssessing and reduci**ng * * party fragm**e**ntation_in a country .
_ arrange_uses as input the roll call data for congress votes on bills and amendments as a proxy for political preferences and ideology . with that , _ arrange_finds the minimum number of parties required to house all congressmen without decreasing party discipline . when applied to brazil s historical roll call data , _
arrange_was able to generate distinct configurations that , compared with the _ status quo _ , have ( i ) a significant smaller number of parties , ( ii ) a higher discipline of partisans towards their parties and ( iii ) a more even distribution of partisans into parties . _ arrange_is fast and parsimonious , relying on a single , intuitive parameter .
|
in many applications it is desirable to shape a beam or pulse of light so that it has specific properties along the optical axis . in particular , for applications in microscopic imaging @xcite , optical tweezers @xcite , laser micro - machining @xcite , dressing of optical filaments @xcite , filament formation @xcite and long - range laser ablation @xcite , to name a few , it is important to have a well controlled beam with a nearly uniform intensity along the optical axis .
however , due to their wave nature localized packets of light will broaden spatially through diffraction .
for example , gaussian beams of width @xmath1 and wavenumber @xmath2 double in spatial extent over the rayleigh range @xmath3 @xcite . in a linear , isotropic medium the electric field of a beam of light polarized in the @xmath4 direction can be modeled by a wave packet of the form @xmath5 , where @xmath6 denotes the spatial coordinate along the optical axis , @xmath7 are the cartesian coordinates transverse to the optical axis , @xmath8 is time and @xmath9 the carrier angular frequency of the wave . within the slowly varying _ ansatz _ on @xmath10 , the propagation of a beam with initial data @xmath11 at the @xmath12 plane
is described by the paraxial wave equation : @xmath13 where @xmath14 is the transverse laplacian defined by @xmath15 @xcite .
exact solutions to equation ( [ intro : governingequation ] ) can be expressed in terms of an integral transform of the initial profile @xmath16 ; see [ appendix : solution ] .
this integral transform is in fact equal to the fresnel diffraction integral @xcite .
constructing solutions to ( [ intro : governingequation ] ) in free space that resist diffraction over length scales much larger than the rayleigh range has attracted considerable interest within the optics community .
diffraction free beams are solutions to equation ( [ intro : governingequation ] ) satisfying the condition that for all @xmath6 , @xmath17 . if we make the _ ansatz _
@xmath18 then diffraction free beams can be found by solving the two dimensional helmholtz equation .
indeed , plane waves , bessel beams @xcite , mathieu beams @xcite and parabolic beams @xcite are examples of such solutions in various orthogonal coordinate systems .
note , however , that the diffraction free beams described above are not realizable in practice since they carry infinite energy .
one technique for obtaining finite energy approximations of such solutions is apodization at the input plane @xmath12 .
that is , if @xmath19 solves ( [ intro : governingequation ] ) then apodized approximations to @xmath19 are found by solving ( [ intro : governingequation ] ) with the initial data @xmath20 where @xmath16 is a function of finite extent . in @xcite so called `` bessel - gauss '' beams were created by apodizing bessel beams with a gaussian function and a general framework for constructing gaussian apodized solutions to ( [ intro : governingequation ] ) was developed in @xcite and @xcite .
bessel - gauss beams have received considerable interest because they can be produced experimentally , for example by a diffraction grating with concentric grooves @xcite .
alternatively , a gaussian beam passing through an axicon lens produces a ring beam with a truncated gaussian cross - section that results from the division of the beam around the axicon tip @xcite .
the effect of these optical elements is to create a linear superposition of gaussian beamlets whose individual wave vectors @xmath21 lie on a cone of aperture angle @xmath22 .
the mutual interference of these beamlets generates the bessel pattern and due to influx of energy supplied by the oscillatory wings of the bessel function to the central core , these waves are robust to nonlinear losses and physical barriers @xcite .
however , for a truncation of width @xmath23 the bessel - gauss beam only accurately approximates the ideal bessel beam in a `` bessel - zone '' of width @xmath24 near the tip of the axicon @xcite and then transitions to a ring - beam in the far - field @xcite .
it was also shown by bagini et .
@xcite that bessel - gauss beams can be created remotely by focusing ring beams .
specifically , in cylindrical coordinates @xmath25 bessel - gauss beams can be generated remotely by solving equation ( [ intro : governingequation ] ) with the initial data @xmath26 where @xmath27 is the peak value of the electric field , @xmath28 is the radius of the ring beam , @xmath1 is the characteristic width of the ring , @xmath29 is the focal distance , @xmath30 is the zero order bessel function of the first kind and @xmath31 is a parameter related to the focusing of the beam @xcite .
note that for this initial data the argument of the bessel function is complex valued and in particular if @xmath32 no focusing it reduces to a modified bessel function of the first kind . in figure
[ fig : genbesselbeams ] we provide a contour plot of the @xmath33 intensity profile for such a `` generalized '' bessel - gauss beams and select four radial cross sections to highlight the transition from a ring - beam into a bessel - gauss beam and back to a ring beam .
as can be seen from figure [ fig : genbesselbeams ] , the intensity profile along the optical axis ( @xmath34 ) is not uniform in @xmath35 within the bessel - zone and is in fact given by a lorentzian - gaussian type function @xcite .
the non - uniformity of the on - axis intensity limits the utility of bessel - beams in some applications . in this paper
we explore the problem of applying additional phase shaping to the ring beam in order to create beams with specific properties along the optical axis . specifically , we consider the problem of constructing a phase function @xmath36 at the input plane @xmath12 that focuses a ring beam intensity profile @xmath37 of radius @xmath38 and width @xmath1 into a target intensity profile @xmath39 of width @xmath40 along the optical axis centered at a target distance @xmath29 . in figure
[ fig : besselsubfig2 ] we illustrate the geometry and notation for the problem we are considering .
the constants @xmath41 denote peak values of the electric field and @xmath42 , @xmath43 are smooth , non negative compactly supported functions with a maximum value of @xmath44 supported on the intervals @xmath45 and @xmath46 defined by @xmath47 respectively .
simultaneous amplitude and phase shaping of the beam will achieve the task and indeed this problem was studied in @xcite .
however , in practice this is achieved through holographic techniques which are difficult to implement and have low power efficiency @xcite . in figure
[ fig : besselsubfig1 ] we illustrate one possible method for how phase shaping of ring beams can be achieved .
first , a ring beam of radius @xmath38 and width @xmath1 is created by passing a collimated beam through an axicon or a concentric diffraction grating . the phase function
is then applied , perhaps through a lens with a radially dependent thickness , which focuses the beam into the desired profile .
the effect of adding a shaper phase @xmath36 can be realized by solving equation ( [ intro : governingequation ] ) with the initial data @xmath48 .
the exact solutions to equation ( [ intro : governingequation ] ) can be expressed in terms of a hankel transform and the field along the optical axis is given by the following relationship : @xmath49 ( see [ appendix : solution ] ) . in this paper
we consider the problem of optimizing the on - axis profile of the beam in the sense that we are interested in minimizing the @xmath0 norm : @xmath50 over the space of measurable phase functions @xmath51 .
note that while the optimal axis is defined by the half - line @xmath52 we have defined the @xmath0 norm over the entire real axis .
we make this definition as it allows us to use the @xmath0 isometry between a function and its fourier transform .
moreover , this definition of the functional penalizes the beam wasting light outside the support of the target intensity distribution .
since we are assuming that @xmath53 is compactly supported away from zero it follows from the change of variables @xmath54 and @xmath55 that @xmath56(\omega),\ ] ] where @xmath57 denotes the fourier transform from @xmath58-space into @xmath59-space .
the optimal design problem can then abstractly be posed as finding a measurable phase function @xmath60 that minimizes the functional @xmath61 $ ] defined by @xmath62=\left\| g(\omega)- |\mathcal{f } \left[g(s ) \exp(i \varphi(s))\right](\omega)\right|\|_{l^2},\ ] ] where @xmath63 and @xmath64 are positive compactly supported functions .
this abstract formulation can be identified with the specific application we are interested in by setting @xmath65 in the two - dimensional setting similar functionals have been studied within the context of shaping beams into desired intensity patterns in the focal plane of a lens @xcite .
the variational problem we are considering is closely related to the problem of phase retrieval from two intensity measurements , i.e. the problem of determining the complex argument of a function given both knowledge of the modulus of a function and the modulus of its fourier transform .
phase retrieval arises in applications in various fields ranging from astronomy @xcite , ultrashort pulses @xcite , radial beam shaping @xcite , pulse shaping @xcite , microscopy @xcite , to image reconstruction @xcite .
a detailed review of the history of this problem and a number of numerical techniques that have been developed to solve this problem can be found in @xcite . within the context of phase retrieval , the most common technique for optimizing @xmath66
is the alternating projection algorithm pioneered by gerchberg and saxton @xcite and its variants such as the hybrid input - output algorithm discovered by fienup @xcite . in the original gerchberg - saxton algorithm
the phase is recovered as follows : 1 .
choose an initial guess @xmath67 .
2 . define @xmath68\right)$ ] .
3 . define @xmath69\right)$ ] .
4 . loop through items 2 and 3 until @xmath70 $ ] is sufficiently close to zero . through plancherel s identity
it can be shown that the gerchberg - saxton algorithm is an error reducing algorithm in the sense that @xmath71\leq i[\phi_{n}]$ ] @xcite .
however , this property alone does not guarantee convergence of the algorithm . while projection algorithms converge on convex sets @xcite , the projections employed by the gerchberg - saxton algorithm are analogous to projections onto the boundary of the unit ball in @xmath72 which is clearly not convex
this lack of convexity commonly leads to stagnation of the algorithm away from the global minimum which must be overcome by additional ad hoc means @xcite .
indeed , additional techniques such as the hybrid input - output algorithm are commonly implemented along with the gerchberg - saxton to force the algorithm out of local minimum @xcite .
other techniques for phase retrieval have also been developed including in the two diemnsional setting a variational principle for the transport of intensity distributions between two planes orthogonal to the optical axis @xcite . in phase
shaping , however , the infimum of @xmath66 may be significantly bounded away from zero and therefore when applied to our problem convergence of the gerchberg - saxton algorithm can not be assessed . indeed , without any further _ a priori _ information about the minimum value of @xmath66 stagnation of the algorithm is the only indication of possible convergence . moreover ,
if the gerchberg - saxton stagnates at a large value of @xmath66 it may indeed be at the global minimum for the problem . the paper is organized as follows . in section 2 we use elementary fourier analysis to deduce some basic results concerning the functional @xmath66 and translate these results in terms of the practical parameters for the phase shaping problem we are interested in . using plancherel s identity
we obtain as a necessary condition for accurate beam shaping a normalization for the peak target intensity @xmath73 in terms of the design parameters : @xmath74 we also prove the existence of a minimum for this problem and in doing so obtain the regularity result that the optimal phase is the argument of an analytic function .
finally , in using the uncertainty principle we prove _ ansatz free _ lower bounds on the minimum value of the functional .
indeed we prove the following necessary condition for accurate beam shaping : @xmath75 although we discovered this result independently , romero and dickey are to the best of our knowledge the first to make the connection between the uncertainty principle and limitations in beam shaping arising from diffraction @xcite .
our work goes beyond their treatment in that we prove in a mathematically precise manner the exact range of parameters for which the minimum value is bounded away from zero .
in section 3 we use the stationary phase method to a construct a phase @xmath36 such that solutions to equation ( [ intro : governingequation ] ) with the initial data @xmath76 satisfy @xmath77 the phase is constructed as follows : 1 . the following initial value problem is solved : @xmath78 2 .
the phase is found by integrating : @xmath79 a similar approach was taken by friberg in the analysis of a logarithmic axicon lens @xcite . while friberg showed that the logarithmic axicon lens maps the support of @xmath53 into @xmath43 in the short wavelength limit ( @xmath80 ) , it fails to match the target intensity .
our approach is more general in that for our construction the intensity distribution is exactly matched in the short wavelength limit .
we would like to point out , however , that we are not the first to discover the use of the method of stationary phase for beam shaping . indeed , after we successfully extended friberg s analysis in @xcite to our setting we discovered that romero and dickey developed essentially the same algorithm in the two - dimensional setting in @xcite and further developed the theory in @xcite . in section 4
we explore how the method of stationary phase coupled with the gerchberg - saxton algorithm can be applied to practical problems .
we consider three specific applications .
we first study the problem of creating a near - uniform intensity along the optical axis down range from a source .
specifically , we take the input beam to be a gaussian ring and the target profile to be a @xmath81-th order super - gaussian .
we show that the stationary phase approximation yields a good initial guess that is further improved upon by the gerchberg - saxton algorithm .
we also highlight the role of the uncertainty principle and show that accurate phase shaping is impossible for sufficiently small values of @xmath1 or @xmath40 .
we also consider in this section the problem of creating beams with oscillatory intensity profiles .
we show that that as a consequence of the uncertainty principle there is a cutoff in the period of these oscillations below which phase shaping can not adequately match the target intensity .
we illustrate this by applying our algorithm to periodic intensity profiles with small scale oscillations .
finally , we show in this section how the method of stationary phase coupled with linear temporal chirping can be used to form pulses with specified temporal widths at the target distance @xmath29 . specifically , by taking a separable spatio - temporal _ ansatz _ on the initial data we show that spatial phase shaping can be employed to correct for the additional spatial spreading of the beam arising from linear temporal chirping .
we conclude with a summary and discussion of our results .
in this section we use the abstract variational formulation of our problem to obtain quantitative information about our beam shaping problem .
let @xmath82 denote the space of lebesgue measurable functions on @xmath83 and @xmath84 the space of smooth compactly supported functions on @xmath85 .
recall from the introduction that we are interested in minimizing the functional @xmath86 defined by @xmath87=\left \|g-\left |\mathcal{f}\left[g\exp(i \varphi)\right ] \right| \right\|_{l^2},\ ] ] with @xmath88 having support bounded away from the origin . in terms of the particular beam shaping problem
we are interested in the _ shaper phase _ is given by @xmath89 , the _ input profile _ of the electric field is given by @xmath90 and the _ target profile _ is given by @xmath91 .
in particular we will assume that the supports of @xmath64 and @xmath63 are given by the intervals @xmath92 and @xmath93 defined by @xmath94 and @xmath95 respectively where @xmath96 .
recall plancherel s theorem which states that @xmath97\right \|_{l^2}$ ] .
the following lemma immediately follows from plancherel s identity and applications of the triangle inequality and reverse triangle inequality .
[ form : thm : lowerbound1 ] let @xmath86 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . then , @xmath100 \leq
\|g\|_{l^2 } + \sqrt{2\pi } \|g\|_{l^2}.\ ] ] if the lower bound in lemma [ form : thm : lowerbound1 ] is zero , i.e. @xmath101 , then in terms of the design parameters we are interested in we have that @xmath102 this provides a natural normalization for @xmath103 in terms of @xmath27 and the @xmath0 norms for the specific profiles of the input and target intensity patterns .
it is important to note that in our initial formulation of the problem we did not specify any constraints on the target electric field .
throughout the rest of this paper we will assume this normalization and hence that @xmath101 which will be a critical assumption in the stationary phase analysis as well .
we can also use plancherel s identity to obtain functionals that are defined over a larger admissible set but have the same infimum as @xmath66 .
these functionals are easier to analyze and we use them to prove the existence of a minimizer for our original problem . for @xmath88 , define the functional @xmath104 by @xmath105=\left \| g\exp(i \psi)- \mathcal{f}\left [ g\exp(i \varphi ) \right ] \right\|_{l^2}.\ ] ] letting @xmath106(\omega)\right)$ ] it follows for a fixed @xmath107 that @xmath108 $ ] is minimized by setting @xmath109 .
therefore , @xmath110=i[\varphi]$ ] and consequently @xmath111=\inf_{\varphi \in \mathcal{m } } i[\varphi].\ ] ] let @xmath112 denote the unit ball in the @xmath113 norm .
that is , @xmath114 define another functional @xmath115 by @xmath116=2 \re\left(\left < gw,\mathcal{f}\left[g\zeta\right ] \right>\right).\ ] ] since we are assuming @xmath101 it follows that @xmath117=2\left\|g\right\|_{l^2}+i^{\prime}[\exp(i\varphi),\exp(i\psi)].\ ] ] therefore , to study minimizers of @xmath118 it is equivalent to study minimizers of @xmath119 .
furthermore , we can use the bilinearity of @xmath119 to prove the existence of a minimizer for our functional of interest .
[ thm : existence ] let @xmath115 be the functional defined by equation ( [ form : deffunc3 ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively .
there exists @xmath120 such that for all @xmath121 @xmath122\leq i^{\prime}[\zeta , w].\ ] ] to prove the result we proceed in two steps .
first we use the direct method to show that minimizers @xmath123 of @xmath119 over @xmath124 exist .
second we show that these minimizers satisfy @xmath125 , @xmath126 @xmath127 and hence @xmath128 defined by @xmath129 , @xmath130 are the desired minimum .
let @xmath131 be a minimizing sequence for @xmath119 over @xmath124 .
that is , @xmath132=\inf_{\zeta,\omega\in b_1^{\infty}}i^{\prime}[\zeta,\omega].\ ] ] by the banach - alaoglu theorem @xmath112 is weak-@xmath133 compact and hence @xmath124 is also compact in the product topology .
therefore , reindexing a subsequence if necessary , we can assume without loss of generality that there exists @xmath134 such that @xmath135 and @xmath136 @xcite .
let @xmath137 $ ] and @xmath138 $ ] . from weak-@xmath133 convergence
it follows that for fixed @xmath59 that @xmath139 furthermore , for all @xmath59 it follows from holder s inequality that @xmath140 therefore , by lebesgue s dominated convergence theorem it follows that @xmath141 finally , from weak-@xmath133 convergence of @xmath142 and the strong convergence of @xmath143 it follows that @xmath144=\lim_{n\rightarrow \infty}i^{\prime}[\zeta_n,\omega_n]=\inf_{\zeta,\omega\in b_1^{\infty}}i^{\prime}[\zeta,\omega].\ ] ] that is , @xmath145 minimizes @xmath119 over @xmath124 . since @xmath146
we deduce from the paley - wiener theorem that @xmath147 is analytic and hence has only isolated zeros @xcite .
let @xmath148 and @xmath149 .
therefore , @xmath144=-2\int_{\mathbb{r } } \left| h(\omega ) \right|
\left| w(\omega ) \right| g(\omega)\cos\left ( \theta(\omega)-\psi(\omega)\right)\,d\omega\ ] ] is minimized if and only if @xmath150 a.e .
and there exists an integer valued function @xmath151 such that @xmath152 . from parseval
s equality a similar argument proves that @xmath153 a.e .. let @xmath154 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively .
there exists @xmath155 such that for all @xmath107 @xmath156 < i[\varphi].\ ] ] let @xmath157 be the functional defined by equation ( [ form : deffunc2 ] ) and let @xmath128 be the global minimizer for @xmath118 . if we let @xmath158\right)(\omega)\ ] ] then without loss of generality we can assume @xmath159 and hence @xmath156=\overline{i}[\overline{\varphi},\overline{\psi}]\leq \inf_{\varphi\in \mathcal{m}}i[\varphi].\ ] ] if @xmath154 is the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively then [ cor : lowerbound1 ] @xmath160>0.\ ] ] let @xmath155 be the global minimizer of @xmath66 over @xmath82 . since @xmath146 the paley - weiner theorem implies that @xmath161 $ ] is analytic .
therefore , @xmath161 $ ] can not be compactly supported on @xmath83 and since @xmath162 is compactly supported on @xmath99 it follows that @xmath163\geq \int_{\mathbb{r}\backslash s_g(z_d ,
w_t)}\left| \mathcal{f}[g\exp\left(i \overline{\varphi}\right)]\right|^2(\omega)\,d\omega>0.\ ] ] [ cor : regularity ] let @xmath154 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 respectively . if @xmath155 minimizes @xmath66 then @xmath164 is analytic .
since @xmath165 minimizes @xmath66 over @xmath82 it follows from theorem [ thm : existence ] that there exists @xmath166 such that @xmath167 minimizes @xmath119 .
consequently , from the same argument used in the proof of theorem [ thm : existence ] and parseval s theorem it follows that @xmath168(s)\right).\ ] ] since @xmath162 it follows from the paley - wiener theorem that @xmath169 is analytic .
it follows from the previous corollary that the minimizers of @xmath66 are well behaved in the sense that they are smooth except possibly for jump discontinuities of @xmath170 .
the uncertainty principle quantifies the impossibility of localizing both a function and its fourier transform @xcite .
therefore , if the length of the support of @xmath63 and @xmath64 are both sufficiently small then the minimum value of @xmath66 will necessarily be significantly bounded away from zero .
in particular , while corollary [ cor : lowerbound1 ] quantifies that the minimum value of @xmath66 is not zero the uncertainty principle can give us quantitative information about how large the minimum value can be .
this is made precise by the following theorem .
[ thm : unc1 ] let @xmath171 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 respectively . if @xmath172 , we have the lower bound @xmath160\geq \|g\|_{l^2}\left(1-\frac{1}{\sqrt{\pi}}\sqrt{\frac{2kw_tr_0w_0}{4z_d^2-w_t^2}}\right).\ ] ] by the reverse triangle inequality it follows that @xmath173&= \left\|g-\mathcal{f}\left[g\exp\left ( i \varphi\right)\right]\right\|_{l^2}\\ & \geq \left(\int_{s_g(z_d , w_t ) } \left(g(\omega)-\left|\mathcal{f}\left[g\exp\left ( i \varphi \right)\right](\omega)\right| \right)^2\,d\omega \right)^{\frac{1}{2}}\\ & \geq \left| \left(\int_{s_g(z_d , w_t ) } g^2(\omega)\,d\omega \right)^{\frac{1}{2 } } - \left(\int_{s_g(z_d , w_t ) } \left|\mathcal{f}\left[g\exp\left ( i \varphi \right)\right](\omega)\right|^2 \,d\omega \right)^{\frac{1}{2 } } \right|\\ & = \left| \|g\|_{l^2 } - \left(\int_{s_g(z_d , w_t ) } \left|\mathcal{f}\left[g \exp\left ( i \varphi \right)\right](\omega)\right|^2 \,d\omega \right)^{\frac{1}{2 } } \right|.\\\end{aligned}\ ] ] furthermore , by direct calculation , an application of the cauchy - schwarz inequality and lemma [ form : eqn : normalization ] we have that @xmath174(\omega)\right|^2 \,d\omega \right)^{\frac{1}{2 } } & \leq \sqrt { \frac{2k w_t}{4z_d^2-w_t^2}}\left \| \mathcal{f}\left[g \exp\left ( i \varphi \right)\right ] \right\|_{l^{\infty}}\\ & \leq \sqrt { \frac{2k w_t}{4z_d^2-w_t^2}}\|g(s)\|_{l^1}\\ & = \sqrt { \frac{2k w_t}{4z_d^2-w_t^2}}\int_{s_g(r_0,w_0)}\left|g(s)\right|\,ds\\ & \leq \sqrt { \frac{4k w_tr_0w_0}{4z_d^2-w_t^2}}\|g \|_{l^2}\\ & = \sqrt { \frac{4k w_tr_0w_0}{4z_d^2-w_t^2}}\frac{1}{\sqrt{2\pi}}\|g \|_{l^2}.\end{aligned}\ ] ] using the same notation as in @xcite we define the dimensionless parameter @xmath31 by @xmath175 it then follows from theorem [ thm : unc1 ] , that @xmath176\geq \|g\|_{l^2}\left(1-\sqrt{\frac{\beta}{\pi}}\right)}$ ] , for any choice of the phase function @xmath60 . in particular , if @xmath177 , there is no way of using phase shaping to match the target profile .
we emphasize that the result is only a lower bound for the functional @xmath61 $ ] , and in particular it does not follow that if the parameters @xmath178 and @xmath29 are such that @xmath179 , then the infimum of @xmath66 is zero . also , this result is somewhat crude in that it is independent of the exact forms of @xmath63 and @xmath64 . below is a refinement of theorem [ thm : unc1 ] that takes into account the @xmath180 norms of @xmath64 and @xmath63 .
[ thm : unc2 ] let @xmath171 be the functional defined by equation ( [ form : deffunc ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively . 1 .
if @xmath172 then @xmath160\geq \sqrt{2\pi}\|g\|_{l^2 } \left(1- \sqrt{\frac{2kw_t}{4z_d^2-w_t^2}}\frac{\|g\|_{l^1}}{\left\|g\right\|_{l^2}}\right).\ ] ] 2 .
if if @xmath172 then @xmath160\geq \left\|g\right\|_{l^2}\left ( 1 - \sqrt{\frac{r_0w_0}{\pi}}\frac{\|g\|_{l^1 } } { \left\| g \right\|_{l^2}}\right).\ ] ] item 1 follows from the proof of theorem [ thm : unc1 ] and plancherel s identity . to prove item 2 we note that from plancherel s identity and the triangle inequality that @xmath181\geq \sqrt{2\pi } \left|\left(\int_{s_g(r_0,w_0 ) } \left| \mathcal{f}^{-1}[g\exp\left(i \psi \right)](s)\right|^2\,ds\right)^{\frac{1}{2}}-\|g\|_{l^2}\right|.\ ] ] following the same arguments as in the proof of theorem [ thm : unc1 ] it follows that @xmath182(s)\right|^2\,ds\right)^{\frac{1}{2}}\leq \frac{1}{2\pi } \sqrt{2r_0w_0}\|g \|_{l^1}.\ ] ] the result follows from applying plancherel s identity to @xmath183 .
theorems [ thm : unc1 ] and [ thm : unc2 ] are estimates for the global error in the beam shaping problem in terms of the design parameters .
the same techniques can be used to obtain local information as well .
specifically , the uncertainty principle implies that if @xmath63 has sufficiently small scale features within its support then the infimum of @xmath66 will be bounded away from zero . to make this statement precise we look at the local error for an interval of width @xmath184 centered at @xmath185 on the optical axis
specifically , for all @xmath186 satisfying @xmath187 define @xmath188 by @xmath189&=\left \|g-\left |\mathcal{f}\left[g\exp(i \varphi)\right ] \right| \right\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}\\ & = \left(\int_{s_g(z_d^{\prime},w_t^{\prime})}\left(g(\omega)-\left |\mathcal{f}\left[g\exp(i \varphi)\right](\omega ) \right|\right)^2\,d\omega\right)^{\frac{1}{2}},\end{aligned}\ ] ] which is a local measure of the @xmath0 error on the interval @xmath190 along the optical axis .
the following corollary follows from the same arguments used to prove theorem [ thm : unc1 ] .
[ cor : unc1 ] let @xmath191 be the functional defined by equation ( [ form : func4 ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively .
suppose @xmath192 satisfy @xmath193 and @xmath194 .
if @xmath172 and @xmath195 then @xmath196\geq \|g\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}-\frac{1}{\sqrt{\pi}}\sqrt{\frac{2kw_t^{\prime}r_0z_d^{\prime}}{4\left(z_d^{\prime}\right)^2-\left(w_t^{\prime}\right)^2}}\|g\|_{l^2}.\ ] ] [ cor : unc2 ] let @xmath191 be the functional defined by equation ( [ form : func4 ] ) with @xmath98 having supports @xmath92 and @xmath99 defined by equations ( [ form : sg ] ) and ( [ form : sg ] ) respectively .
suppose @xmath192 satisfy @xmath193 and @xmath194 .
if @xmath172 and @xmath197 then @xmath196\geq \|g\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}-\sqrt{\frac{2kw_t^{\prime}}{4\left(z_d^{\prime}\right)^2-\left(w_t^{\prime}\right)^2}}\|g\|_{l^1}.\ ] ] [ form : remark ] the local error @xmath198 also provides a lower bound for the global error @xmath66 . indeed ,
if we let @xmath199 then a sharper estimate for the lower bound on the global error can be found by considering : @xmath200&\geq \sup _ { ( z_d^{\prime } , w_t^{\prime})\in \mathcal{a } } \inf_{\varphi \in \mathcal{m}}i^{w_t^{\prime}}_{z_d^{\prime}}[\varphi]\\ & \geq \sup_{(z_d^{\prime},w_t^{\prime})\in \mathcal{a } } \left(\|g\|_{l^2(s_g(z_d^{\prime},w_t^{\prime}))}-\frac{1}{\sqrt{\pi}}\sqrt{\frac{2kw_tr_0z_d^{\prime}}{4\left(z_d^{\prime}\right)^2-\left(w_t^{\prime}\right)^2}}\|g\|_{l^2}\right ) .
\nonumber\end{aligned}\ ] ]
recall that the method of stationary phase can be used to obtain asymptotic expansions for integrals of the form @xmath201 where @xmath202 , @xmath203 is real , @xmath204 is a smooth real valued function and @xmath147 is a smooth but not necessarily analytic complex valued function . if @xmath204 has one critical point at @xmath205 , i.e. @xmath206 , and @xmath207 then @xmath208 as @xmath209 @xcite . in this section
we show how the method of stationary phase can be used to approximate the on - axis profile of the electric field in the limit of short wavelengths and we show how this approximation can be used to construct initial guesses for the optimal phase .
first , we must define precisely what we mean by short wavelengths in terms of the input parameters and the length scales in the parameter . using the geometry of the problem as in figure [ fig : besselsubfig1 ]
, we define the dimensionless quantities @xmath210 equation ( [ intro : opticalaxis ] ) gives the on - axis electric field : @xmath211 where we used the fact that the input profile @xmath212 is only supported on @xmath213 $ ] .
assuming @xmath214 , we expand the phase in powers of @xmath215 and @xmath216 to obtain @xmath217 note that the first three terms in the expansion of @xmath218 do not depend on both @xmath219 and @xmath220 jointly .
consequently , they can be absorbed as phase factors in @xmath221 or @xmath222 as below , and they _ do not affect the intensity _ of the input @xmath221 or the on - axis profile @xmath222 . dropping the higher order terms in @xmath215 and @xmath216 , we get @xmath223 \\ & \times e^{i \left(\phi - \frac{k r_0 w_0 w_t \zeta \varrho } { z_d^2 } \right ) } ( r_0 + w_0 \varrho)\,d\varrho.\end{aligned}\ ] ] with @xmath219 . in order to use phase variations for optical beam shaping ,
it is clear that the optical path differences over the relevant region @xmath224 should be many times the wavelength , so that the beam can interfere with itself .
this allows for the possibility of controlling the intensity of the beam as a function of position .
since the variations in @xmath220 and @xmath219 are @xmath225 by construction , the preceding calculation determines the correct notion of the short wavelength limit in our problem , namely @xmath226 we include an additional factor of 2 in defining @xmath227 so that for @xmath228 , the dimensionless parameter @xmath31 , whose definition was motivated by theorem [ thm : unc1 ] _ is identical _ to the non - dimensional wave number @xmath227 arising from the analysis of beam shaping using phase variations .
this argument therefore gives an alternative physical interpretation of the parameter @xmath31 it is the phase variation within the beam arising from the different optical path lengths for different parts of the beam in our geometry ( fig .
[ fig : besselsubfig1 ] ) .
we now assume @xmath229 , @xmath230 , @xmath231 and formally apply the method of stationary phase to approximate the on - axis electric field .
define dimensionless quantities by @xmath232 then by equation ( [ intro : opticalaxis ] ) it follows that the on - axis profile of solutions to the paraxial wave equation is given by @xmath233 if @xmath234 is a given smooth function , then for a fixed value of @xmath235 stationary points of the phase can be found by solving the equation @xmath236 for @xmath237 . if @xmath234 is a monotone decreasing function in @xmath237 and @xmath238 is either strictly convex or concave then there exists one solution to this equation for each value of @xmath239 and we can implicitly define a function @xmath240 by @xmath241 therefore , if we let @xmath242 then by the method of stationary phase it follows that if @xmath243 then @xmath244 as @xmath245 . equation ( [ stat : approximation2 ] ) can be used to construct an accurate approximation to an optimal @xmath238 .
formally , if we set @xmath251 then the following initial value problem must be satisfied : @xmath252 the phase corresponding to @xmath253 can then be found by integration : @xmath254 equations ( [ stat : zdifferentialequation ] ) and ( [ stat : phasedifferentialequation ] ) are a simple system of differential equations whose solution we will use as an approximation to the near optimal phase .
solutions to ( [ stat : zdifferentialequation ] ) must also satisfy the additional physical constraint that @xmath255 , that is , the stationary phase approximation maps the support of @xmath53 to the support of @xmath43 .
the following lemma shows that this condition is met if the normalization ( [ form : eqn : normalization ] ) is satisfied and again highlights the important role this normalization takes .
if @xmath253 is a solution to the initial value problem ( [ stat : zdifferentialequation ] ) and @xmath103 and @xmath27 satisfy equation ( [ form : eqn : normalization ] ) then @xmath255 .
suppose @xmath253 is a solution to the initial value problem ( [ stat : zdifferentialequation ] ) and @xmath256 . from the normalization ( [ form : eqn : normalization ] ) and the fact that @xmath257 solves ( [ stat : zdifferentialequation ] ) it follows that @xmath258 now , if @xmath259 then @xmath260 if @xmath261 then since @xmath249 is a continuous monotone increasing function of @xmath262 there exists @xmath263 such that @xmath264 . consequently , @xmath265 therefore it follows that @xmath266 .
the full electric field near the bessel - zone @xmath267 can also be approximated in the focal region @xmath268 using the method of stationary phase . in the dimensionless
coordinates we obtain through the method of stationary phase that : latexmath:[\[\label{stat : offaxis } that is , the beam essentially forms a bessel - beam in the focal region .
however , the function @xmath30 is also a rapidly oscillating function and the above approximation is only valid for radii satisfying @xmath270 .
this expression is not applicable outside the focal region , and of course does not can not capture the transition of the ring beam into a bessel - gauss beam and back into a ring beam .
in this subsection we use the method of stationary phase coupled with the gerchberg - saxton algorithm to obtain approximations to the optimal phase when the input profile @xmath42 is gaussian and the target profile @xmath271 is super - gaussian . specifically , for @xmath81 a positive integer and @xmath272 we assume that @xmath273 since we are using input and target functions that decay exponentially fast we can use laplace s method @xcite to obtain accurate estimates for the normalization .
directly applying laplace s method it follows that @xmath274 and explicitly calculating we have that @xmath275 therefore , by equation ( [ form : eqn : normalization ] ) it follows that @xmath276 where @xmath277 is bounded between @xmath278 and @xmath279 .
equation ( [ bb : peakintensity ] ) gives the ratio of the peak intensity of the on - axis profile of the beam to the peak intensity of the input beam .
the input and target functions defined above are not smooth bump functions and strictly speaking the theory and numerical algorithm we developed in the previous two sections does not apply .
however , the functions given by equation ( [ bb : functions ] ) are schwartz class and they can be very accurately approximated by non - smooth functions with compact support which in turn can be approximated by smooth functions with compact support @xcite .
in particular , we truncate these functions by forcing them to be zero for values in which the function is less than @xmath280 .
this gives us the following estimates for the width of the supports @xmath281 with the above normalization , equations ( [ stat : zdifferentialequation ] ) and ( [ stat : phasedifferentialequation ] ) can be numerically solved to construct a guess for an optimal phase function @xmath51 .
using @xmath51 to initialize the gerchberg - saxton algorithm , further corrections to the stationary phase construction can be obtained .
to illustrate the potential applicability of our techniques to remote laser ablation , we applied the algorithm outlined in the previous section with 100 iterations of the gerchberg - saxton algorithm for various input and target widths with the following parameters fixed @xmath282 the exact profile along the optical axis is then obtained by applying the fast fourier transform ( fft ) to equation ( [ form : fouriertrans ] ) . in figure
[ fig : besseltargetwidth ] we plot the normalized value of @xmath66 for both the stationary phase analysis and the gerchberg - saxton algorithm for @xmath283 and @xmath184 ranging from @xmath284 to @xmath285 . for this range of parameters
, @xmath31 varies from @xmath286 to @xmath287 .
the inset plot in figure [ fig : besseltargetwidth ] illustrates a subset of the profiles obtained for @xmath288 . in figure
[ fig : besselinputwidth ] we plot the normalized value of @xmath66 for both the stationary phase analysis and the stationary phase analysis coupled with the gerchberg - saxton algorithm for @xmath289 ranging from @xmath290 to @xmath291 and @xmath292 . for this range of parameters
, @xmath31 varies from @xmath293 to @xmath294 .
the inset plot in figure [ fig : besselinputwidth ] illustrates the intensity profile obtained for @xmath295 .
figures [ fig : besseltargetwidth ] and [ fig : besselinputwidth ] also illustrate the important role that @xmath31 not only plays in terms of the lower bound presented in theorem [ thm : unc1 ] but also the accuracy of the stationary phase approximation . for @xmath296 @xmath297 , the lower bound in theorem [ thm : unc1 ]
is relevant and as is guaranteed by theorem [ thm : unc1 ] neither the stationary phase approximation nor the gerchberg - saxton algorithm , or indeed any other algorithm can yield accurate shaping of the beam . for @xmath184 ranging from @xmath298 to @xmath299 ( @xmath300 to @xmath301 )
theorem [ thm : unc1 ] no longer applies and the stationary phase analysis yields a guess that is further improved upon by the gerchberg - saxton algorithm . for @xmath302 ( @xmath301 ) ,
the stationary phase analysis yields a very accurate guess that is only slightly improved upon by the gerchberg - saxton algorithm . in figures [ fig : phaseplot1 ] and [ fig :
phaseplot2 ] we plot surface and contour plots of the phase function found by coupling the stationary phase approximation with the gerchberg - saxton algorithm for the parameters given by equation ( [ bb : parameters ] ) with @xmath283 and @xmath184 ranging from @xmath303 to @xmath285 . specifically , we plot @xmath304 the term @xmath305 corresponds to a linear shift in the fourier variable @xmath59 and also has an interpretation in optics as the effect of a lens that focuses a beam at the distance @xmath29 @xcite .
interestingly , when @xmath306 ( @xmath286 ) the phase @xmath307 is essentially constant .
that is , except for the focusing term @xmath305 additional phase shaping can not improve upon the effect of a focusing lens .
however , for @xmath308 ( @xmath287 ) there is a significant difference between the numerically obtained phase and the focusing term . from figures [ fig : statphaseplot1 ] and [ fig : statphaseplot2 ]
it is again evident that the phase obtained from that gerchberg - saxton algorithm converges in the limit of @xmath309 to the one obtained using the stationary phase alone . in laser ablation while it is important to have a beam with a near uniform on - axis intensity profile , the amount of power delivered by the beam is also of critical importance . in particular , numerical experiments
indicate that for critical collapse of a bessel - gauss beam to occur the power contained in the central lobe of the beam must be near the critical power of the townes profile @xcite . for a fixed value of @xmath35
the power delivered by a beam in a circular cross section of radius @xmath310 about the optical axis is proportional to the following integral : @xmath311 using the asymtotic formula ( [ stat : offaxis ] ) for the off - axis profile electric field , the power in the central core of the bessel beam at the target distance @xmath312 can be approximated in the short wavelength limit .
if we assume that @xmath313 , then for @xmath312 the electric field is approximately zero at the value @xmath314 . consequently , the power delivered in the central core of the bessel beam at the distance @xmath29 is approximated by @xmath315 from the normalization ( [ form : eqn : normalization ] ) it follows that in the short wavelength limit we have that : @xmath316 where @xmath317 is the initial power of the beam .
therefore , from equation ( [ bb : wtnorm ] ) it follows that @xmath318 this equation illustrates the trade - off between the amount of power delivered within the central lobe and the accuracy of the shaping .
in particular , for @xmath319 and @xmath320 the shaping will be very accurate but the amount of power delivered by the beam in the central core will be much smaller than the initial power of the beam . for the range of parameters used to generate figure [ fig : besseltargetwidth ] , for example , this approximation for @xmath321 ranges from @xmath322 to @xmath323 . in this subsection
we explore the possibility of using beam shaping to create intensity distributions with an oscillatory pattern for length scales that are applicable to micromachining .
again we use a gaussian input beam : @xmath324 with @xmath325 and @xmath326 . for integer values of of @xmath327
we take the target profile to be : @xmath328 with @xmath329 , @xmath330 and we take @xmath331 . as before , we can then apply the method of stationary phase coupled with the gerchberg - saxton algorithm to numerically optimize @xmath66 .
now , it follows that @xmath332 for these parameters and hence theorem [ thm : unc1 ] is not relevant .
however , as the value of @xmath327 increases the target intensity develops fine scale oscillations of approximate width @xmath333 and therefore by corollary [ cor : unc1 ] we expect the accuracy of the beam shaping to decrease with increasing @xmath327 . as in section 2
define @xmath334 and for @xmath335 define the interval @xmath336 by @xmath337 to capture the influence of the small scale oscillations on the error we define the following local quantity : @xmath338 which is a local measure of the error in matching the target intensity within one period . from corollary [ cor : unc1 ]
it follows that if @xmath339 then the small scale oscillations can not be matched by any phase function . in figure
[ fig : bumps ] we plot the on - axis intensity profiles for the beams obtained by using the stationary phase as an initial guess for the gerchberg - saxton algorithm with @xmath340 iterations .
we selected the values @xmath341 to illustrate the influence of @xmath327 on the accuracy of the shaping . as we can see , with increasing @xmath327 the method of stationary phase yields a guess that matches the support of the function but fails to capture the small scale oscillations .
for these values of @xmath327 we have that @xmath342 , @xmath343 , @xmath344 and @xmath345 and indeed for @xmath346 , as is to be expected , the gerchberg - saxton algorithm can not match the target intensity .
it is interesting to note , however , for large @xmath327 that although the gerchberg - saxton algorithm fails to capture the amplitude of the oscillations it does a decent job recovering the period . .
the blue dashed - dotted curves are the on - axis intensities obtained through the stationary phase analysis alone , the solid red curves are the on - axis intensities obtained by coupling the stationary phase algorithm with the gerchberg - saxton algorithm and the green dashed curves are the target intensity profile .
the function @xmath347 quantifies the accuracy of the beam shaping and in particular for @xmath339 the amplitude of the oscillations is not matched . ] in this subsection we show that for a linear medium and under the paraxial approximation the techniques we developed in this paper can be combined with linear temporal chirping to remotely deliver pulses with a desired temporal width at a target distance .
the governing equation for such pulses is the temporal paraxial wave equation : @xmath348 where @xmath349 is the group velocity dispersion for a pulse with wavenumber @xmath2 @xcite .
if we assume separable initial data of the form @xmath350 then the solution to equation ( [ eqn : temppw ] ) can be expressed in terms of the hankel transform of a convolution : @xmath351 again , we assume that the input spatial profile is given by @xmath352 , where @xmath353 , @xmath354 is supported on the interval @xmath355 and @xmath51 is a measurable function corresponding to a shaper phase .
we assume that the input temporal profile is a linearly chirped gaussian pulse : @xmath356 where @xmath357 is a measure of the temporal width of the initial pulse and @xmath358 is a linear chirping parameter . with this initial data ,
the on - axis intensity is given by @xmath359 with @xmath360 a quadratic function defined by @xmath361 the function @xmath360 quantifies how the temporal width broadens or contracts as it propagates ; indeed the characteristic temporal width is @xmath362 .
we assume that the target profile along the optical axis is also given by a gaussian pulse of the form @xmath363 where @xmath364 , @xmath365 with support @xmath366 and @xmath367 . to ensure that @xmath368 matches the desired temporal width at @xmath29 and
the change in @xmath368 over the support of @xmath43 is as small as possible we must have that @xmath369 solving these equations yields the conditions : @xmath370 define @xmath371 by @xmath372 and as we did before define @xmath373 , @xmath374 .
the problem of designing the optimal shaper phase with this temporal chirping then corresponds to minimizing the functional @xmath171 defined by @xmath61=\|g-|\mathcal{f}(g \exp(i \varphi))| \|_{l^2}$ ] which problem corresponds to the variational problem we considered before except for the additional term @xmath375 accounting for the additional modification to the spatial intensity arising from temporal broadening . in figure
[ fig : temporalprofile ] we plot the on - axis spatial - temporal profile for a gaussian pulse propagating in air obtained by using the method of stationary phase . for this figure we again assumed that @xmath376 with @xmath377 @xmath378 following the experimental results in @xcite we took the group velocity dispersion of air for this wavelength of light to be @xmath379 and we assumed a target temporal width of @xmath380fs . for these values
the initial temporal width is @xmath381fs and the chirping parameter is @xmath382fs .
in this paper we studied the problem of phase shaping an annular beam into a desired intensity profile along the optical axis by analyzing minimizers of the functional @xmath383 defined by equation [ form : deffunc ] .
formally , this functional is equivalent to the ones arising in phase retrieval from two intensity measurements .
typically , due to its speed and ease of implementation , the gerchberg - saxton algorithm is used to construct approximate minimizers . while the gerchberg - saxton algorithm is an error reducing algorithm
there are no known general convergence results for this algorithm .
in particular the algorithm typically stagnates away from the global minimum which in phase retrieval can be overcome by additional modifications to the algorithm such as the hybrid input - output method @xcite or those in @xcite .
moreover , in phase retrieval the infimum of @xmath66 is zero and convergence of the gerchberg - saxton algorithm and its modifications can be assessed by the proximity of @xmath66 to zero .
however , as a consequence of the uncertainty principle the minimum value of @xmath66 may be significantly bounded away from zero .
consequently it is not clear if modifications to the gerchberg - saxton algorithm are applicable to beam shaping problems nor is it clear what a sufficient convergence criterion of such an algorithm would be .
the central result of our work is the identification of the dimensionless quantity @xmath384 and the associated lower bounds on the functional @xmath66 .
we identified three scaling regimes for @xmath31 for which different quantitative and qualitative behaviors of the optimal profile occur .
these regimes are as follows . 1 .
for @xmath385 the uncertainty principle guarantees that the optimal profile will differ significantly from the target profile .
this result has the consequence that neither the gerchberg - saxton algorithm nor any other numerical algorithm will yield accurate shaping of the beam . 2 .
for @xmath231 the method of stationary phase yields an approximation to the problem that is very close to the optimal profile .
this asymptotic regime can be considered within the geometrical optics setting in the sense that the light rays originating from the input plane can be accurately mapped to the target intensity profile through the method of stationary phase .
indeed , for the problem of creating a beam with a nearly uniform on - axis intensity , the gerchberg - saxton algorithm coupled with the method of stationary phase yields nearly optimal profiles .
in particular , the method of stationary phase can be used as a robust method for overcoming the stagnation issues of the gerchberg - saxton algorithm commented on in @xcite .
3 . for the intermediate regime where @xmath386 the phase produced by the method of stationary phase
is significantly improved upon by the gerchberg - saxton algorithm . however , in this asymptotic regime the uncertainty principle no longer applies .
moreover , the value of @xmath66 for @xmath51 obtained through the method of stationary phase is not close to zero .
consequently , within this regime there is no clear convergence of the algorithm . following the reasoning in remark [ form : remark ]
we could obtain more accurate lower bounds on the functional that account for the specific forms of the input and target intensity profile instead of simply the widths of their supports .
however , in the regime @xmath387 a universal scaling law for the error in terms of the wavelength is not possible .
the functional we considered was derived under the paraxial assumption and hence is applicable for linear beams with the same range of wavelengths and geometrical length scales as in the fresnel approximation . indeed , for the range parameters we considered in the applications , the fresnel number ranges from @xmath388 to @xmath389 .
consequently , for these parameters the integral transform given by equation ( [ intro : opticalaxis ] ) very accurately approximates the propagation of the on - axis electric field for linear beams .
therefore , the results of this paper and in particular the uncertainty principle are fundamental and place restrictions on the types of intensity profiles that can be achieved through linear phase shaping alone
. for specific applications and in particular those requiring high power beams it may be necessary to consider the off - axis electric field and nonlinear interactions . equation ( [ stat : offaxis ] ) gives an asymptotic result indicating that the intensity pattern near the target distance will form a bessel - like function .
numerical evidence indicates that for critical collapse of the beam to occur the power in the central core of the bessel function must be near that of the townes profile @xcite .
for the applications we considered , it follows from equation ( [ form : eqn : normalization ] ) that the initial peak intensity of the ring beam @xmath390 is much smaller than the target peak intensity @xmath73 of the generated bessel beam and hence kerr nonlinearities will be negligible until the ring beam is focused into the bessel - like pattern .
predicting the filamentation patterns of a beam after critical collapse is a challenging problem and perhaps the nonlinear geometrical optics method presented in @xcite could be used in conjunction with our methods near the target distance .
the full effect of the kerr nonlinearities will be addressed in subsequent publications . additionally , for propagation of the beam over ranges of @xmath391 it will be necessary to consider the effects of turbulence in the atmosphere which is beyond the scope of this paper .
consider the initial value problem with radially symmetric initial data : @xmath392 where @xmath53 is a smooth function with compact support .
if we let @xmath393(k_{\perp})$ ] and @xmath394(r)$ ] denote the hankel transform and its inverse defined by @xmath395(k_{\perp})=\int_0^{\infty } g(r)j_0(rk_{\perp})\,dr \text { and } \mathcal{h}^{-1}[g](r)=\int_0^{\infty } g(k_{\perp})j_0(rk_{\perp})k_{\perp}dk_{\perp}\ ] ] then the solution to the initial value problem is given by @xmath396(k_{\perp } ) \exp\left(-\frac{i k_{\perp}^2 z}{2k } \right)j_0\left(rk_{\perp}\right)k_{\perp } dk_{\perp}.\ ] ] noting that @xmath397(r)=-\frac{ik}{z } \exp\left ( \frac{i k r^2}{2z}\right)\ ] ] it follows from the convolution theorem for hankel transforms @xcite that @xmath398 * acknowledgments : * the authors acknowledge funding support under the muri afosr grant fa9550 - 10 - 0561 . c.d .
acknowledges funding support from afosr under the grant fa9550 - 10 - 1 - 0394 an s.v . acknowledges support from the nsf grant 0807501 .
the authors wish to thank ewan wright for many useful discussions and bringing to our attention reference @xcite and we would like to thank colm dineen who constructed figure [ fig : besselsubfig1 ] .
the authors would also like to thank the two anonymous referees who reviewed this paper .
their comments greatly improved the organization and the presentation of our results .
+ * references * t. planchon , l. gao , d. e. milkie , m. w. davidson , j. a. galbraith , c. g. galbraith , e. betzig , rapid three - dimensional isotropic imaging of living cells using bessel beam plane illumination , nature methods 8 ( 2011 ) 417423 .
k. stelmaszczyk , p. rohwetter , g. mjean , j. yu , e. salmon , j. kasparian , r. ackermann , j .-
wolf , l. woste , long - distance remote laser - induced breakdown spectroscopy using filamentation in air , applied physics letters 85 ( 18 ) ( 2004 ) 39773979 .
p. rohwetter , k. stelmaszczyk , l. wste , r. ackermann , g. mjean , e. salmon , j. kasparian , j. yu , j .-
wolf , filament - induced remote surface ablation for long range laser - induced breakdown spectroscopy operation , spectrochimica acta part b : atomic spectroscopy 60 ( 7 ) ( 2005 ) 10251033 .
t. graf , d. christodoulides , m. s. mills , j. moloney , s. c. venkataramani , e. m. wright , propagation of gaussian - apodized paraxial beams through first - order optical systems via complexcoordinate transforms and ray transfer matrices , j. opt .
am . a 29 ( 9 ) ( 2012 ) 18601869 .
j. wrzesinski , d. pestov , v. v. lozovoy , j. r. gord , m. dantus , s. roy , et al . , group - velocity - dispersion measurements of atmospheric and combustion - related gases using an ultrabroadband - laser source , opt . express 19 ( 6 ) ( 2011 ) 51635171 .
n. gavish , g. fibich , l. t. vuong , a. l. gaeta , predicting the filamentation of high - power beams and pulses without numerical integration : a nonlinear geometrical optics method , physical review a 78 ( 4 ) ( 2008 ) 043807 .
|
we investigate the problem of shaping radially symmetric annular beams into desired intensity patterns along the optical axis . within the fresnel approximation ,
we show that this problem can be expressed in a variational form equivalent to the one arising in phase retrieval . using the uncertainty principle
we prove various rigorous lower bounds on the functional ; these lower bounds estimate the @xmath0 error for the beam shaping problem in terms of the design parameters .
we also use the method of stationary phase to construct a natural _ ansatz _ for a minimizer in the short wavelength limit .
we illustrate the implications of our results by applying the method of stationary phase coupled with the gerchberg - saxton algorithm to beam shaping problems arising in the remote delivery of beams and pulses .
beam shaping ; localized waves ; phase retrieval ; paraxial wave equation ; fresnel approximation
|
techniques involving laser cooling and trapping of neutral atoms are of immense interest for many scientific applications , including those that are capable of probing new physics @xcite and searching for exotic quantum phase transitions using ultracold atoms @xcite . in particular , trapping atoms using optical lattices have many advantages since they offer long storage time @xcite and their energy levels can be easily accessed using lasers @xcite .
it is conducive to carry out measurements in a transition of an optically trapped atom without realizing the stark shifts due to the applied laser field .
one can achieve this by trapping the atom at the wavelengths for which the differential stark shift of the transition gets nullified .
these wavelengths are popularly known as magic wavelengths ( @xmath3s ) @xcite .
they play crucial role in state - insensitive quantum engineering to set - up many high precision experiments .
they are useful in the atomic clock experiments to acquire relative uncertainty as small as @xmath4 @xcite , in the quantum information and communication studies @xcite , investigating fundamental physics @xcite , so on .
alkali atoms are mostly preferred for performing experiments using laser cooling and trapping techniques .
the reason being that the low - lying transitions in these atoms are conveniently accessible by the available lasers .
since cs atom has wider hyperfine splittings in its ground state , it has been considered for making microwave clock and for quantum computation . for laser cooling of these atoms ,
its @xmath5-@xmath2 transitions are mainly being used .
owing to a large number of applications of trapped cs atoms , it would be imperative to find out all plausible magic wavelengths of these transitions such that lasers can be appropriately chosen at these wavelengths to reduce systematics significantly in the above possible measurements due to the stark shifts .
had experimentally demonstrated @xmath3 at 935.6 nm for the @xmath5-@xmath2 transition in cs @xcite using linearly polarized light . following this , magic wavelengths for many alkali atoms including cs atom
were determined using linearly polarized light by arora _
_ @xcite by evaluating the dynamic polarizabilities of these atoms using the relativistic coupled cluster ( rcc ) method .
the estimation of ac stark shifts using circularly polarized light may be advantageous to look for magic wavelengths owing to the predominant role played by the vector polarizabilities .
since the vector polarizability contribution is absent in the use of linearly polarized light , this can open - up new windows to manipulate the magic wavelengths for a wide range of applications .
recently , we had investigated magic wavelengths in the lighter alkali atoms for circularly polarized light and realized many possible magic wavelengths , especially for the @xmath6-@xmath7 transitions with the ground state principal quantum number @xmath8 @xcite .
however , magic wavelengths for circularly polarized light in the cs atom have not been explored sufficiently .
[ cols="^,^,^,^,^,^,^,^,^,^ " , ] to reduce the uncertainties in the evaluation of these polarizability values , we use e1 matrix elements for @xmath9 transitions extracted from the very precisely measured lifetimes of the @xmath2 and @xmath1 states of cs by rafac _
et al_. @xcite .
we also use e1 matrix elements for the @xmath10 transitions compiled in ref .
@xcite , which are derived from the measured lifetime of the @xmath11 state .
similarly , the e1 matrix element of @xmath12 transition has been derived by combining the measured differential stark shift of the d1 line with the experimental value of the ground state dipole polarizability of amini _ et al_. @xcite .
we adopt a procedure similar to that is given in ref .
@xcite to determine the e1 matrix elements from the measured stark shifts .
the values of these matrix elements along with their experimental uncertainties are listed in table [ pol1 ] .
otherwise , the required e1 matrix elements for considered transitions up to @xmath13 , @xmath14 and @xmath15 states are obtained by employing sd method as described in the previous section .
the uncertainties in these matrix elements are calculated by comparing matrix elements for the @xmath9 and @xmath10 transitions calculated using our method and available experimental values .
the maximum difference between the experiment and our results for these matrix elements is 6% .
therefore , we assign maximum a 10% uncertainty to all the matrix elements given in table [ pol1 ]
. we have used 70 b - spline functions confined within a cavity of radius @xmath16 a.u . to construct the single - particle orbitals .
we use experimental values of the excitation energies of these transitions from the national institute of science and technology ( nist ) database @xcite to reduce further the uncertainties in the evaluation of the polarizabilities .
our calculated value of @xmath17 for the ground state is 399.5 a.u . , which matches very well with other theoretical values 399 a.u . and 399.9 a.u
estimated by borschevsky _
_ @xcite and derevianko _ et al . _
@xcite , using the other variants of rcc method , respectively .
these results are also in very good agreement with the experimental result 401.0(6 ) a.u . measured by amini _
. _ @xcite using the time - of - flight technique .
similarly our calculation gives @xmath17 of the @xmath18 state to be 1335 a.u .
, which is slightly larger than the other calculated value 1290 a.u . of wijngaarden _
@xcite but agrees quite well with another calculated value 1338 a.u . reported by arora _
_ @xcite and the measured value 1328.4(6 ) a.u . reported in ref . @xcite . in the work of wijngaarden _
et al . _ , polarizabilities were evaluated using the oscillator strengths from the method of bates and damgaard @xcite .
the scalar and tensor polarizabilities of the @xmath19 state using our method are obtained to be 1644 a.u . and @xmath20 a.u .
, respectively .
they are also in very good agreement with the experimental values reported in ref .
@xcite and are in reasonable agreement with the theoretical values reported by arora _
_ @xcite and wijngaarden _ et al .
the above analysis shows that we have obtained very accurate values of the dipole polarizabilities using our method of evaluation .
this justifies that determining dynamic polarizabilities using the same procedure can also provide competent results .
hence , @xmath21 values of the @xmath22 transitions in cs can be determined without any ambiguity using these accurate values of the dipole polarizabilities .
we now proceed to determine @xmath21s for the @xmath22 transitions in the cs atom . for this purpose ,
we plot the dynamic dipole polarizabilities of the @xmath23 and @xmath24 states in figs .
[ magiccirc4 ] , [ magiccirc3 ] , [ magiccirc1 ] and [ magiccirc2 ] .
they are shown in the upper and lower halves for linearly and circularly polarized light respectively .
we use left circularly polarized light ( @xmath25 ) while determining the magic wavelengths .
note that it is not required to produce results for right circularly polarized light ( @xmath26 ) separately , because the results for @xmath21s will be same as that for left circularly polarized light with the counter sign of @xmath27 sublevels .
for this purpose we consider only left handed circularly polarized light with all positive and negative @xmath27 values .
we consider all possible @xmath27 values of the @xmath23 and @xmath24 states .
it is evident from the above plots that for the considered wavelength range , the dynamic polarizability for the @xmath23 state is generally small except in the close vicinity of the resonant @xmath28 and @xmath29 transitions .
on the other hand , polarizabilities of the @xmath30 states have significant contributions from several resonant transitions .
thus , the polarizability curves of the @xmath30 states cross with the polarizability curve of the @xmath23 state in between these resonant transitions .
the wavelengths at which intersections of these polarizability curves take place are recognized as @xmath31s in the above figures for both linearly and circularly polarized light .
we have also tabulated these values for the @xmath22 transitions in tables [ magic3 ] , [ magic1 ] , [ magic2 ] and [ magic22 ] along with their respective uncertainties in the parentheses .
these uncertainties are estimated by considering maximum possible errors in the estimated differential polarizabilities between the involved states in a transition .
the corresponding values of the dynamic polarizabilities are also mentioned in the above tables to provide an estimate of the kind of trapping potential required at those magic wavelengths .
we also list the resonant wavelengths ( @xmath32 ) in the tables to highlight the respective placements of these @xmath33s .
as seen from table [ magic3 ] , we find @xmath34s for the @xmath35 transition around 630 nm , 660 nm and 760 nm for both linearly and circularly polarized light .
other @xmath33s are located at 1522 nm for the linearly polarized light and around 1966 nm with @xmath18(@xmath27=1/2 ) for the circularly polarized light .
@xmath33s at 632 , 756.1 and 1966 nms does not support state - insensitive trapping for @xmath36 sublevel of the @xmath18 state and hence switching trapping scheme as discussed in ref .
@xcite can be used here . in this approach
, the change of sign of @xmath37 will lead to the same result for the positive values of @xmath27 sublevels of the @xmath30 state .
@xmath33s at 1522 nm and 1966.1 nm support the red detuned trap , while the other above mentioned @xmath33s support the blue detuned traps .
values of @xmath34s for the linearly polarized light are also compared with @xmath33s of arora _ et al . _
reported in ref .
both findings agree with each other , as the method of calculation in both is almost similar .
similarly , we tabulate @xmath33s for the @xmath38 transition in same table .
it can be evidently seen from the table that @xmath33s are red shifted from the @xmath33s for @xmath35 .
we have also determined an extra @xmath33 at 880.79 nm , which supports a red detuned trap .
we list @xmath33s for the @xmath39 transition for linearly and circularly polarized light separately in tables [ magic1 ] and [ magic2 ] respectively . in case of the linearly polarized light
, at least ten @xmath33s are systematically located between the resonant transitions with @xmath40 , while only seven @xmath33s are located for the @xmath41 sublevel .
it , thus , implies that use of linearly polarized light does not completely support state insensitive trapping of this transition and results are dependent on the @xmath27 sublevels of the @xmath19 state .
this is also in agreement with the results presented in ref .
@xcite . the experimental magic wavelength at 935.6 nm , as demonstrated by mckeever _ et .
@xcite , matches well with the average of the last two magic wavelengths obtained at 933 nm ( for @xmath40 ) and 941.7 nm ( for @xmath41 ) . as shown in table
[ magic2 ] , we get a set of ten magic wavelengths for circularly polarized light in between the eleven @xmath19 resonances lying in the wavelength range 600 - 1600 nm .
those magic wavelengths for which the values corresponding to @xmath42 sublevels are absent , do not support state - insensitive trapping , and we recommend the use of a switching trapping scheme for this transition as proposed in @xcite . for the @xmath43 transition , we enlist the @xmath33s in table [ magic22 ] .
these @xmath33s are slightly red shifted to those demonstrated for @xmath43 transition .
we have investigated possible magic wavelengths within the wavelength range 600 - 2000 nm for the @xmath22 transitions in the cs atom considering both linearly and circularly polarized light .
our values for linearly polarized light were compared with the previously estimated values and they are found to be in good agreement . with circularly polarized light , we find a large number of magic wavelengths that are in the optical region and would be of immense interest for carrying out many precision measurements at these wavelengths where the above transitions are used for the laser cooling purposes .
we have used very precise electric dipole matrix elements , extracting from the observed lifetimes and evaluating using the relativistic coupled - cluster method , to evaluate the dynamic polarizabilities of the cs atom very precisely .
these quantities are used to determine the above magic wavelengths . by comparing static values of the polarizabilities with their respective experimental results , accuracies of
the polarizabilities and magic wavelengths were adjudged . in few situations
, we found it would be advantageous to use the magic wavelengths of circularly polarized light over linearly polarized light .
as an example , magic wavelengths for circularly light are missing for some @xmath44 sublevel but they are present for the corresponding @xmath45 sublevel or vice - versa . in this case
, one can switch the polarization of the light and can successfully locate the positions of the magic wavelengths .
the work of s.s . and b.a .
is supported by csir grant no .
03(1268)/13/emr - ii , india . k.k . acknowledges the financial support from dst ( letter no .
dst / inspire fellowship/2013/758 ) .
the employed sd method was developed in the group of professor m. s. safronova of the university of delaware , usa .
|
we demonstrate magic wavelengths , at which external electric field produces null differential stark shifts , for the @xmath0 transitions in the cs atom due to circularly polarized light .
in addition , we also obtain magic wavelengths using linearly polarized light , in order to verify the previously reported values , and make a comparative study with the values obtained for circularly polarized light .
a number of these wavelengths are found to be in the optical region and could be of immense interest to experimentalists for carrying out high precision measurements . to obtain these wavelengths ,
we have calculated dynamic dipole polarizabilities of the ground , @xmath1 and @xmath2 states of cs .
we use the available precise values of the electric dipole ( e1 ) matrix elements of the transitions that give the dominant contributions from the lifetime measurements of the excited states .
other significantly contributing e1 matrix elements are obtained by employing a relativistic coupled - cluster singles and doubles method .
the accuracies of the dynamic polarizabilities are substantiated by comparing the static polarizability values with the corresponding experimental results .
|
the behavior of ( finite lumps of ) quark matter is of great interests in cosmology , neutron stars , cosmic ray physics and heavy ion collisions @xcite .
absolutely stable nonstrange quark matter contradicts ordinary nuclei consisting of neutrons and protons .
however , the existence of stable strange quark matter is still an open question and it may be realized in the form of strangelets , small lumps of strange quark matter .
the ( meta)stability of nonstrange and strange quark matter has been investigated within the mit bag model @xcite , quark mass density dependent model @xcite , nambu jona - lasinio ( njl ) model @xcite , and so on @xcite . in the mit bag model , which assumes that asymptotically free quarks are confined in a bag , the bag constant and the current quark masses are phenomenological input parameters .
farhi and jaffe @xcite found a reasonable range of these parameters in which strange quark matter is stable , while nonstrange quark matter is unstable as compared to a gas of @xmath0 .
the quark mass density dependent model takes account of the confinement mechanism by introducing density dependent quark masses .
_ @xcite concluded that the critical density for a phase transition between nuclear and quark matter is sensitively dependent on the confinement mechanism . in the njl model ,
the bag constant and constituent quark masses are generated dynamically in marked contrast to the above mentioned models . in ref .
@xcite , it has been argued that stable nonstrange quark matter lies in chirally restored phase and , then , the model shows similar behavior to the mit bag model ( see , also , ref .
the stability of strange quark matter has been also studied and concluded that strange quark matter is not absolutely stable when realistic values of current quark masses and coupling constants are used @xcite . on the other hand , unlike bulk systems , finite size effects can not be ignored in the study of finite lumps of quark matter .
so far , the mit bag model and fermi gas model including finite size effects have been widely used to study their properties @xcite . for finite quark matter
, it is known that the lumps are not energetically favored by finite size effects .
the disfavor also arises from large strange quark masses .
therefore , finite size effects on dynamical quark masses ( i.e. the behavior of chiral symmetry ) might have significant effects on their properties . in this paper
, we adopt the njl model that is a simple tractable model respecting chiral symmetry
. then we consider isospin symmetric ( @xmath1 with @xmath2 ) quark droplets .
finite size effects are taken account of by making use of the multiple reflection expansion ( mre ) @xcite which is an approximation for the density of states in finite systems .
the mre itself has been derived for free massless particles or nonrelativistic particles .
however , the mre has been used to calculate the thermodynamic quantities of _ massive _ quarks and reproduced well the results of the mit bag model @xcite .
for instance , it can be checked that the vector ( @xmath3 ) and scalar ( @xmath4 ) densities in the mit bag model are very similar to those computed in the mre with little @xmath5 dependence for bags of radius @xmath6 a few fm .
therefore , we expect that the mre also works in the njl model in the mean field approximation with a constant @xmath7 inside the bag and apply the mre to the effective potential of the model
. this paper is organized as follows . in sec .
ii , we formulate the effective potential of spherical quark droplets that consist of an equal number of up and down quarks . the schwinger dyson equation and related thermodynamic quantities are also derived . in sec .
iii , we present numerical results .
section iv is devoted to conclusions .
in this section , the effective potential and related thermodynamic quantities of spherical quark droplets are derived in the two flavor njl model .
we consider up and down quarks to be massless .
then , the @xmath8 chirally symmetric lagrangian is given by @xmath9,\end{aligned}\ ] ] which is our starting point .
here , @xmath10 denotes a quark field with two flavors ( @xmath11 ) and three colors ( @xmath12 ) , and @xmath13 is a dimensionful coupling constant ( @xmath14 ) .
the pauli matrices @xmath15 act in the flavor space .
let us first discuss the effective potential for bulk systems . in the mean - field ( hartree ) approximation ,
the effective potential @xmath16 at finite temperature @xmath17 and quark chemical potential @xmath18 is written as @xcite @xmath19 \left[1+e^{-\beta(e_k-\mu)}\right],\end{aligned}\ ] ] where @xmath20 , @xmath21 , @xmath22 , with @xmath7 being the dynamically generated ( constituent ) quark mass .
since the model is not renormalizable , we have to specify a regularization scheme . throughout this paper
, we use a sharp cutoff @xmath23 mev in the three dimensional momentum space with the coupling constant @xmath24 @xcite . with these parameters , the vacuum values of the pion decay constant @xmath25 , the constituent quark mass @xmath7 and the quark condensate are obtained as @xmath26 mev , @xmath27 mev and @xmath28 .
further discussions about the stability of nonstrange quark matter can be found in refs .
@xcite .
we now turn to discussions on finite size droplets . within the framework of the mre @xcite
, we write the density of states for a spherical system as @xmath29 , where @xmath30 with @xmath31 being the radius of the sphere . the functions @xmath32 and @xmath33 represent the surface and the curvature contributions to the fermionic density of states , respectively .
the functional form of the surface contribution is given by @xmath34 on the other hand , the curvature contribution for an arbitrary quark mass has not been derived in the mre . in this paper , we adopt the following _ ansatz _ by madsen @xcite : @xmath35\label{eqn : fc}.\end{aligned}\ ] ] note that eqs .
( [ eqn : fs ] ) and ( [ eqn : fc ] ) have the following @xmath36 limits : @xmath37 thus , the effective potential for the spherical system which is regarded as a function of @xmath7 , @xmath18 and @xmath31 is given by @xmath38 \left[1+e^{-\beta(e_k-\mu)}\right],\nonumber\\ & \underset{t \to 0}{\longrightarrow}&\frac{m^2}{4 g } -\nu\int_{k_f}^{\lambda}\frac{k^2dk}{2\pi^2}\rho_{{\mbox{\scriptsize mre}}}e_k -\nu\mu\int_0^{k_f}\frac{k^2dk}{2\pi^2}\rho_{{\mbox{\scriptsize mre}}},\end{aligned}\ ] ] where the fermi momentum , @xmath39 , is related to the chemical potential by @xmath40 .
henceforth we restrict ourselves to zero temperature . for computation of a finite system ,
we choose a fixed baryon number @xmath41 and a radius of the sphere @xmath31 .
the quark mass @xmath7 is then determined by the schwinger
dyson equation ( sde ) that is the extremum condition of @xmath16 with respect to @xmath7 : @xmath42 .
the resulting sde is given by @xmath43 in addition to the sde , since the baryon number is fixed to @xmath41 , we have another equation , i.e. @xmath44 where @xmath45 is the volume of the droplet and @xmath46 denotes the baryon number density in the droplet that is one third of the quark number density @xmath47 : @xmath48 equations ( [ eqn : sd1 ] ) and ( [ eqn : sd2 ] ) have to be solved self - consistently . therefore , by virtue of eq .
( [ eqn : sd2 ] ) , one can reduce these equations to the following set of coupled equations : @xmath49 notice that we consider @xmath39 in eq .
( [ eqn : rsd1 ] ) to be a variable independent of @xmath7 .
however , under the constraint ( [ eqn : rsd2 ] ) @xmath39 depends on @xmath7 .
therefore , these equations are also obtained by minimizing the energy density @xmath50 with the constraint of fixed baryon number .
once we know the density dependent ( i.e. the radius dependent ) quark mass @xmath7 , we can compute various physical quantities .
for instance , the pressure @xmath51 inside the droplet are given by @xmath52 where @xmath53 has been introduced to ensure @xmath54 when @xmath55 ( i.e. @xmath56 ) .
we note that , unlike the bulk case , the sde ( [ eqn : rsd1 ] ) does not have a trivial solution @xmath57 .
the reason is that @xmath58 , the density of states of the mre , depends on the quark mass @xmath7 , and hence the r.h.s . of the sde ( [ eqn : rsd1 ] )
is no longer proportional to @xmath7 , in contrast to the bulk case .
in this section , we present the numerical results . unless stated otherwise , the baryon number @xmath41 is fixed to @xmath59 . for a given baryon number , a nontrivial solution to the coupled equations ( [ eqn : rsd1 ] ) and ( [ eqn : rsd2 ] )
is obtained as a function of the radius @xmath31 .
then , as shown in fig .
1 , the pressure of chirally symmetric and broken phases are calculated from eq . ( [ eqn : press ] ) . as a result , the critical radius at which the chiral phase transition occurs is identified as @xmath60 fm .
the corresponding critical baryon number density @xmath61 is rather small @xmath62 with @xmath63 being normal nuclear matter density @xmath64 .. ] the resultant quark mass as a function of the radius is shown in fig .
2 . since the baryon number is fixed , by changing the radius @xmath31 , the baryon number density @xmath46 changes . therefore ,
2 implies that the phase transition as @xmath46 is varied is of first - order when the effect of the mre is included . for comparison
, we also present the result without the mre showing that the transition at @xmath65 fm is of second - order .
the difference between these two curves obviously stems from the finite size effects .
this result is roughly interpreted as follows . in the mre ,
the finite size effects reduce the density of states and , therefore , the _ actual _ fermi momentum increases for a given baryon number .
the increasing tendency of the fermi momentum is more pronounced when the baryon number is small ( see fig .
3 ) . as a consequence ,
the chirally broken phase is energetically disfavored for small baryon numbers .
figure 4 shows the quark mass as a function of the radius for the cases of @xmath59 and @xmath66 .
it is clear that , for relatively large baryon numbers , the first - order transition is weakened .
in addition , at @xmath66 we find @xmath67 fm and @xmath68 .
these results , off course , indicate that the finite size effects become less important for large baryon numbers .
let us consider the radius of the stable droplet from the usual pressure balance relation between the inside of the droplet and the vacuum ( the outside of the droplet ) .
the pressure inside the droplet with and without the mre are shown in fig . 5 for @xmath59 , where the vacuum pressure is taken to be zero .
when @xmath31 is small , the pressures are positive and chiral symmetry is restored inside the droplet . as @xmath31 is increased the pressures decrease . now without the finite size effects ( mre ) the pressure keeps decreasing in the negative pressure region toward the cusp point , where the chiral phase transition takes place . beyond the cusp point , the pressure curve follows the one in the chirally broken phase and turns into the positive region ( but with small values ) , asymptotically approaching the zero value .
there are two points where the pressure curve passes the zero , where the system is in equilibrium .
the left point ( @xmath69 fm ) corresponds to the absolute minimum of the energy per baryon , while the right point ( @xmath70 fm ) corresponds to the local maximum .
consequently , the stable droplet of radius @xmath71 fm lies in the chirally symmetric phase . on the other hand , with the mre
the pressure crosses zero only at @xmath72 fm .
we have calculated the pressure up to @xmath73 fm and confirmed that the pressure of the chirally broken phase is always negative .
therefore , we find @xmath74 fm , which is somewhat small as compared to that without the mre , and the stable droplet lies in the chirally symmetric phase .
the radius of a stable droplet is presented in fig .
6 as a function of @xmath41 . as is evident from the figure , as long as @xmath41 is sufficiently large
, @xmath75 varies as @xmath76 .
this result implies that , as mentioned before , the finite size effects are less important at large baryon numbers .
incidentally , we have observed the saturation property ( @xmath77 ) when @xmath41 is large .
this is a consequence of the balance between the kinetic energy of quarks and the ( bag like ) volume type potential energy .
we have studied the behavior of chiral symmetry in finite systems within the @xmath8 symmetric nambu jona - lasinio model .
the multiple reflection expansion was used to take account of finite size effects .
we considered spherical quark droplets with baryon numbers @xmath78 .
the schwinger dyson equation for the quark mass was solved and , then , it turned out that the finite size effects enhance chiral symmetry restoration . the critical radii with and without the mre differed significantly from each other .
if the finite size effects are important also for strange quarks , it may have significant effects on the ( meta)stability of strangelets . on one hand ,
the enhancement of restoring chiral symmetry lowers the constituent mass of strange quarks .
this affects the argument made by buballa and oertel @xcite who pointed out that the large strange quark mass at the relevant densities is the main reason that makes strange quark matter unstable . on the other hand ,
the finite size effects increase the energy of strangelets @xcite .
these two competing aspects should be carefully treated in realistic calculations in order to find the radius of the stable droplet , we calculated the pressure inside the droplet . without the finite size effects ( mre )
the pressure shows the characteristic behavior as shown in fig .
such behavior has been also obtained @xcite with interactions / parameters which differ from those in this paper .
in contrast , with the mre its behavior changes considerably .
we found that the pressure of the chirally broken phase is always negative and , therefore , the chirally broken phase is unstable against collapse .
however , the radius of the stable droplet does not change much as compared to that without the mre .
after all , it was found that chiral symmetry is restored in the stable droplet .
this result reminiscent of the mit bag model supports the former results without the mre @xcite .
we have already referred to the validity of the mre in the first section .
it should be also noted that the mre contains several problems concerning its reliability . in the first place
, @xmath58 should have terms proportional to @xmath79 , @xmath80 , and so on , though their functional forms are not known .
these terms would be dominant at small radii .
furthermore , it is well known that the mre causes unphysical negative density of states at small radii .
neergaard and madsen @xcite have investigated the validity of the mre and proposed a solution to these problems . however , for systems of @xmath81 , these problems have rather minor effects and the results presented here should hold when an improved density of states is employed . in the present analysis
we have adopted the simplest njl model with scalar and pseudoscalar channels . at the quantitative level
the inclusion of vector and axial - vector channels could be important which brings repulsive correlations @xcite .
it might modify the present result such that , for instance , the transition point would be shifted to higher density side .
however , we expect that our main conclusion drawn here on the relative effect of finite cavity as compared to the bulk case should be rather stable .
this is , however , an interesting question to be investigated in the future work . finally , we comment on the outlook for future studies .
physics of strangelets and the quark droplets at finite temperature should have important implications for the cosmological quark - hadron phase transition and heavy ion collision experiments .
it is worthwhile doing more detailed analysis including charge neutrality , chemical equilibrium by weak interactions and the color singlet constraint .
furthermore , it is also interesting to include other dynamical effects as well as the chiral phase transition , e.g. , color superconductivity and color flavor locking @xcite .
they may alter the present phase structure at small radii , because , at sufficiently low temperatures and high densities , the chirally broken phase would undergo a phase transition into the color superconducting phase .
see , for example , _ proceedings of the international symposium on strangeness in quark matter _ , santorini , 1997 , j. phys .
g * 23 * , no . 12
( 1997 ) ; j. madsen , in _ proceedings of the eleventh chris engelbrecht summer school _ , capetown , south africa , 1998 , edited by j. cleymans _ et al . _
( springer - verlag , berlin , 1999 ) , and the references cited therein .
see , for example , k. rajagopal and f. wilczek , in _ at the frontier of particle physics
handbook of qcd : boris ioffe festschrift _ , edited by m. shifman ( world scientific , singapore , 2001 ) ; m. alford , ann .
sci . * 51 * , 131 ( 2001 ) . for @xmath59 .
the solid line shows the pressure of chirally symmetric phase and the dotted line shows that of chirally broken phase .
two lines cross at @xmath82 fm .
notice that the vacuum ( the outside of the droplet ) pressure is taken to be zero . ]
( solid line ) and that without the mre ( dotted line ) in unit of @xmath83 as a function of @xmath84 ( left panel ) .
the increase of the fermi momentum due to the finite size effects @xmath85 as a function of @xmath41 ( right panel ) .
the quark mass @xmath7 and the radius @xmath31 are taken to be their typical values ; @xmath86 mev and @xmath87 fm . ] ) as a function of @xmath31 .
the curves correspond to the pressure with ( solid line ) and without ( dotted line ) the mre . without the mre
the pressure becomes positive for @xmath88 fm , though it is difficult to see in this figure . ]
|
chiral phase properties of finite size hadronic systems are investigated within the nambu jona - lasinio model .
finite size effects are taken into account by making use of the multiple reflection expansion .
we find that , for droplets with relatively small baryon numbers , chiral symmetry restoration is enhanced by the finite size effects .
however the radius of the stable droplet does not change much , as compared to that without the multiple reflection expansion .
|
the search for supersymmetry and its breaking , in addition to the direct searches at lep , b - factories , tevatron and the large hadron collider ( lhc ) , is actively pursued using the wmap limits on the relic density constraints . however the sensitivity of the lightest supersymmetric particle relic density calculation to the variation of the cosmological expansion rate before big - bang nucleosynthesis ( bbn ) , even if modest and with no consequences on the cosmological observations , can modify considerably the relic density , and therefore change the constraints on the supersymmetric parameter space @xcite . in the standard cosmology
the dominant component before bbn is radiation , however energy density and entropy content can be modified . in the following
we consider the impact of different scenarios of alternative cosmologies .
the precision of the wmap data should therefore not make us forget the hypothesis which are implied by the use of standard cosmology .
we discuss in the following the implications of precision b - physics , direct searches and cold dark matter relic abundance for the case of anomaly mediated models , from a minimal anomaly mediation supersymmetry breaking @xcite , to mixed moduli - anomaly mediated @xcite and to hypercharge anomaly mediation @xcite .
we also discuss similar supersymmetry breaking scenarios in the case of the next - to - minimal supersymmetric standard model . in section [ sec : models ] we discuss the different anomaly mediated supersymmetry breakings in the minimal supersymmetric standard model ( mssm ) in terms of the parameter spaces for these models and discuss the limits which can be applied due to the present data from particle physics experiments and from relic dark matter density assuming the standard model of cosmology . in section [ sec : models2 ] similar scenarios are considered in the next - to - minimal supersymmetric standard model ( nmssm ) and the corresponding particle and cosmological bounds are discussed . in section [ sec : cosmologies ] we discuss how alternative cosmological models can affect dramatically the bounds on the parameter space of the models we have considered while letting unchanged the observable cosmology .
four different alternatives to the standard cosmology are discussed which share this behaviour .
section [ sec : constr ] and section [ sec : lhc ] discuss respectively the constraints implied by these different cosmological scenarios and the perspectives at the lhc for a list of benchmark points which are representative of the available parameter space for these amsb models .
our conclusions are given in section [ sec : conclusion ] .
the superconformal anomaly mediated supersymmetry breaking ( amsb ) mechanism @xcite is one of the most well - known and attractive set - ups for supersymmetry breaking .
supersymmetry breaking effects in the observable sector have in this framework a gravitational origin .
superconformal symmetry is classically preserved in theories without dimensional parameters and it is in general broken by the quantum effects . as anomalies only depend on the low - energy effective theory , the same will be true for the soft terms .
usually the amsb scenario can not be applied to the mssm , as it leads to tachyonic sleptons .
however the presence of an intermediate threshold can displace the soft terms and avoid this problem . in superconformal gravity one
introduces a chiral superfield playing the role of the compensating multiplet for super - weyl transformations , called the weyl or conformal compensator .
the f - term vacuum expectation value of the conformal compensator is turned on by the supersymmetry breaking in the hidden sector and the soft breaking of supersymmetry in the visible sector appears through the chiral anomaly supermultiplet .
as the soft susy breaking terms arise from the anomaly , the supersymmetry breaking terms do not dominate at tree - level .
several soft susy breaking scenarios can be realised starting from this setup .
we discuss in the following some of these realisations .
the minimal amsb ( mamsb ) scenario @xcite has very attractive properties , since the soft susy breaking terms are calculated in terms of one single parameter , namely the gravitino mass @xmath3 , and the soft terms are renormalisation group invariants which can be calculated for any scale choice .
however , the amsb scenarios suffer from the problem that slepton squared masses are found to be negative , leading to tachyonic states .
a solution to this problem is to consider that the scalar particles acquire a universal mass @xmath4 at the gut scale , which when added to the amsb soft susy breaking terms , makes them positive .
therefore , the mamsb model relies on only four parameters : @xmath5 this scenario has been thoroughly studied in the literature , but is known to have cosmological consequences incompatible with the wmap observations of the dark matter density @xcite .
we perform here an updated analysis of the mamsb parameter space constraints from flavour physics and cosmological relic density . for this study
, we generate mass spectra and couplings using isajet 7.80 @xcite .
the calculation of flavour observables and the computation of the relic density are performed with superiso relic v3.0 @xcite .
we use the constraints given in table 9 of the latest version of the superiso manual .
the first flavour observable that we consider here is the branching ratio of @xmath6 , which has been been thoroughly studied in the literature and is still under scrutiny .
this observable is very interesting , as its sm contributions only appear at loop level , and its theoretical uncertainties as well as the experimental errors are now under control .
it provides strong constraints on the supersymmetric parameter space , especially for large @xmath7 , where it receives large enhancements from its supersymmetric contributions .
we use the following interval at 95% c.l .
@xcite : @xmath8 another interesting observable is the branching fraction of @xmath9 , which is also a loop level observable , and which can receive extremely large contributions from susy at large @xmath10 , and can receive an enhancement of several orders of magnitude compared to the sm branching ratio . this decay mode has not yet been observed , and we have at 95% c.l .
@xcite : @xmath11 we also consider a set of tree - level observables which are very sensitive to the charged higgs mass as well as @xmath10 , and we use the following 95% level intervals , which include the theoretical and experimental errors @xcite : @xmath12 the observable @xmath13 is related to the decay of @xmath14 and is detailed in @xcite .
+ for the relic density constraint , we use the wmap constraints @xcite increased by 10% of theoretical error to account for the uncertainties in the calculation of the relic density : @xmath15 in the following , we disregard the case of negative @xmath16 since it is disfavoured by the muon anomalous magnetic moment constraint , and we scan over the intervals @xmath17 $ ] gev , @xmath18 $ ] tev and @xmath19 $ ] .
figure [ fig_amsb ] presents projection plots of the parameter space into the possible different planes .
the green region in the plots corresponds to the parameter zone which is not excluded by flavour constraints or mass limits .
the red stars corresponds to points leading to a favoured relic density but excluded by other constraints , whereas black stars are favoured by all the presented constraints , including the relic density constraint . as can be seen
, no black star is visible in these plots , and the whole parameter space presented here is disfavoured either by flavour or direct constraints , or by the relic density constraint which tends to favour the low @xmath3 region . disregarding the relic density constraint , a large zone at low @xmath3
is excluded .
constraints on the minimal amsb parameter space .
the exclusion regions are plotted in the order given in the legend .
the red zones are excluded by the inclusive branching ratio of @xmath20 , the yellow ones correspond to charged lsp , the olive - green areas are excluded by direct collider constraints , the light blue zones are excluded by br(@xmath21 ) , the dark blue zones by br(@xmath22 ) , the magenta zones by @xmath23 , the orange zones by br(@xmath24 ) and the grey zones by br(@xmath25 ) .
the green area are in agreement with all the previously mentioned constraints .
the stars are points favoured by the relic density observable , in red if disfavoured by any other constraints and in black if in agreement with all the constraints simultaneously.,title="fig:",width=604 ] + constraints on the minimal amsb parameter space .
the exclusion regions are plotted in the order given in the legend .
the red zones are excluded by the inclusive branching ratio of @xmath20 , the yellow ones correspond to charged lsp , the olive - green areas are excluded by direct collider constraints , the light blue zones are excluded by br(@xmath21 ) , the dark blue zones by br(@xmath22 ) , the magenta zones by @xmath23 , the orange zones by br(@xmath24 ) and the grey zones by br(@xmath25 ) .
the green area are in agreement with all the previously mentioned constraints .
the stars are points favoured by the relic density observable , in red if disfavoured by any other constraints and in black if in agreement with all the constraints simultaneously.,title="fig:",width=604 ] + constraints on the minimal amsb parameter space .
the exclusion regions are plotted in the order given in the legend .
the red zones are excluded by the inclusive branching ratio of @xmath20 , the yellow ones correspond to charged lsp , the olive - green areas are excluded by direct collider constraints , the light blue zones are excluded by br(@xmath21 ) , the dark blue zones by br(@xmath22 ) , the magenta zones by @xmath23 , the orange zones by br(@xmath24 ) and the grey zones by br(@xmath25 ) .
the green area are in agreement with all the previously mentioned constraints .
the stars are points favoured by the relic density observable , in red if disfavoured by any other constraints and in black if in agreement with all the constraints simultaneously.,title="fig:",width=604 ] in fig .
[ fig_amsb_omega ] , we show the relic density values as a function of the amsb parameters .
the green zones correspond to regions favoured by the flavour and direct constraints , whereas the other points are either excluded by these constraints or by cosmological considerations ( charged relic or sneutrino relic , which interact therefore strongly ) .
two green zones clearly appear on the @xmath4 and @xmath10 plots , for @xmath26 around @xmath27 and @xmath28 .
these areas are far from the wmap dark matter allowed interval , making the mamsb scenario disfavoured by the standard cosmology . relic density in function of the amsb parameters .
the green points are favoured by all the constraints , the yellow points corresponds to a charged lsp , the blue points correspond to left sneutrino lsp , and the red points are excluded by the other constraints ( flavour and direct limits).,title="fig:",width=604 ] + relic density in function of the amsb parameters .
the green points are favoured by all the constraints , the yellow points corresponds to a charged lsp , the blue points correspond to left sneutrino lsp , and the red points are excluded by the other constraints ( flavour and direct limits).,title="fig:",width=604 ] + relic density in function of the amsb parameters .
the green points are favoured by all the constraints , the yellow points corresponds to a charged lsp , the blue points correspond to left sneutrino lsp , and the red points are excluded by the other constraints ( flavour and direct limits).,title="fig:",width=604 ] another possibility to solve the negative slepton squared masses of the original amsb scenario has been proposed : the hypercharge anomaly mediated supersymmetry breaking ( hcamsb ) scenario @xcite , in which the mssm resides on a d - brane and the hypercharge gaugino mass is generated in a geometrically separated hidden sector @xcite . in this way , additional contribution to the gaugino mass @xmath29 is generated , and the large value of @xmath29 then increases the weak scale slepton masses beyond tachyonic values , solving the generic amsb problem @xcite .
the hcamsb scenario has four parameters : @xmath30 where @xmath31 is the hcamsb contribution to @xmath29 . in order to study the parameter space ,
we generate mass spectra and couplings using isajet 7.80 @xcite and compute the flavour observables and relic density with superiso relic v3.0 @xcite .
we use the constraints described in the previous subsection . in fig .
[ fig_hcamsb ] , we scan over the whole parameter space , and project the results in two - dimensional planes .
the results are somehow similar to those of the mamsb scenario : the constraints exclude low @xmath3 values .
a large part of the parameter space is favoured by flavour and direct constraints , but unfortunately no part of the parameter space respects at the same time the relic density and the other constraints , so that hcamsb is also disfavoured by the standard cosmology . in fig .
[ fig_hcamsb_omega ] , we show the relic density values as a function of the hcamsb parameters . again , the results are similar to those of the mamsb scenario , with two distinct zones which are not excluded by flavour and direct constraints , corresponding to a relic density @xmath26 around @xmath27 and @xmath32 .
constraints on the hcamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the hcamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the hcamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] relic density in function of the hcamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the hcamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the hcamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] contrary to the two previous amsb scenarios , the mixed modulus anomaly mediated susy breaking ( mmamsb ) scenario @xcite provides viable dark matter candidates , in addition to solving the negative slepton mass problem naturally .
this scenario is based on type - iib superstrings with stabilised moduli @xcite . in this scenario ,
an interesting result is that the soft susy breaking terms receive comparable contributions from both anomaly and modulus , resulting in positive slepton masses .
we examine here the minimal mmamsb scenario and @xmath33 , which represent respectively the modular weights of visible sector of the matter fields and the gauge kinetic function , and which can modify the mass spectra .
] , which relies on four parameters : @xmath34 @xmath35 here parametrises the relative contributions of modulus mediation and anomaly mediation to the soft breaking terms : the largest @xmath35 is , the more mediation comes from modulus @xcite . in fig .
[ fig_mmamsb ] , the parameter space is scanned over , and it is projected in two - dimensional planes .
the resulting plots are different from those of mamsb and hcamsb .
indeed , a large part of parameter space escapes the flavour and direct constraints , and zones around @xmath36 or at low @xmath3 also fulfil the relic density constraint .
constraints on the mmamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the mmamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the mmamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] in fig .
[ fig_mmamsb_omega ] , we present the relic density values in function of the mmamsb parameters .
the relic density of points favoured by flavour and direct constraints takes values between @xmath28 and @xmath37 .
we can notice however that most of the points are in the interval @xmath38 $ ] , and some points fit in the wmap interval .
for this reason , mmamsb is a scenario which can appear as attractive , as it fulfils simultaneously the standard cosmology and particle physics constraints .
relic density in function of the mmamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the mmamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the mmamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ]
an interesting extension of the mssm is the nmssm , which brings a solution to the @xmath39-problem @xcite .
it has an extended higgs sector involving additional higgs bosons , modifying the relic density and flavour physics constraints .
moreover , the couplings being modified , the nmssm can escape direct constraints , and new parameter zones can be allowed by the constraints used in the previous section .
we consider here the simplest version of the nmssm , where the term @xmath40 of the mssm superpotential is replaced by @xmath41 in order for the superpotential to be scale invariant .
the soft breaking terms @xmath42 are _ a priori _ independent . using the minimisation conditions for the potential ,
the scalar mass parameters @xmath43 can be replaced by the vacuum expectation values of the doublet @xmath44 and @xmath45 , with @xmath46 the singlet field mass parameter can also be replaced by the singlet expectation value @xmath47 .
expanding the singlet field @xmath48 around @xmath47 gives rise to an effective parameter @xmath49 .
one can also define an effective doublet mass such as @xmath50 once the mssm - like parameters ( and in particular @xmath51 ) have been fixed by the specification of the amsb scenario , we are left with four additional independent nmssm - specific parameters : @xmath52 we scan over the intervals @xmath53 $ ] gev , @xmath54 $ ] , @xmath55 $ ] gev and @xmath56 $ ] .
we review in the following the differences between the different amsb scenarios when applied to the nmssm .
minimal nmssm - amsb ( mnamsb ) parameter points are generated here using nmssmtools 2.3.4 @xcite , and flavour constraints and relic density are computed with superiso relic v3.0 @xcite . in fig .
[ fig_namsb ] , the parameter space of mnamsb is presented . as in fig .
[ fig_amsb ] , no point satisfy simultaneous flavour , direct and relic density constraints .
the direct limits are less constraining than in the mssm , but the flavour constraints are stronger and exclude a larger part of the namsb parameter space in comparison with amsb
. constraints on the mnamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the mnamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the mnamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] relic density in function of the mnamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the mnamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the mnamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] fig .
[ fig_namsb_omega ] reveals more differences between the mamsb and mnamsb models .
first , the zone not excluded by flavour and direct constraints having a relic density around @xmath28 does not exist in the namsb , a new green zone appears for @xmath57 , and its relic density around @xmath58 is much closer to the wmap constraint .
however , as in the mamsb scenario , the mnamsb model is globally disfavoured by the standard cosmology .
we generate the nmssm - hcamsb ( nhcamsb ) parameter points using nmssmtools 2.3.4 @xcite .
the obtained constraints are shown in fig .
[ fig_nhcamsb ] .
again , we see that no parameter point satisfies at the same time the relic density constraint and the direct and flavour constraints . similarly to the mnamsb scenario ,
nhcamsb is less constrained by the direct mass limits , but is more excluded by the flavour constraints .
constraints on the nhcamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the nhcamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the nhcamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] relic density in function of the nhcamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the nhcamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density in function of the nhcamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] fig .
[ fig_nhcamsb_omega ] shows the relic density in function of the different parameters .
the green region that exists in the hcamsb scenario for a relic density around @xmath32 disappears in the nhcamsb , and a new region opens up around @xmath58 .
however , as the mnamsb scenario , the nhcamsb scenario remains also disfavoured by the standard cosmology .
the nmssm - mmamsb ( nmmamsb ) scenario leads to similar results as the mmamsb scenario . as can be seen in fig .
[ fig_nmmamsb ] , there exists many points satisfying all the constraints , including relic density , especially for values of @xmath35 near to 0 .
constraints on the nmmamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the nmmamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] + constraints on the nmmamsb parameter space .
the colour codes are the same as in fig .
[ fig_amsb].,title="fig:",width=604 ] fig .
[ fig_nmmamsb_omega ] reveals a difference , as the calculated relic density takes values between @xmath27 and @xmath59 , which is more restrictive in comparison to the mssm case .
relic density as a function of the nmmamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density as a function of the nmmamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ] + relic density as a function of the nmmamsb parameters .
the colour codes are the same as in fig .
[ fig_amsb_omega].,title="fig:",width=604 ]
we have seen in the previous section that the relic density imposes severe constraints to the parameter spaces , excluding a major part of the amsb scenarios . however , the relic density calculation is generally based on the simplistic standard model of cosmology . in this section ,
we reinterpret the previous results by considering four different alternatives to the cosmological standard scenario . for this study
, we focus our interest on six different benchmark points , which are described in table [ points ] and are representative of the allowed parameter space in the different models . the mass spectra associated to these points
are shown in fig .
[ fig_spectra ] .
.benchmark points for testing alternative cosmology scenarios .
all these points have @xmath60 and are in agreement with all the flavour and direct search constraints but would be excluded by wmap constraints based on the standard cosmology . for the point f
extra parameters are needed to specify a point in the parameter space , which are chosen to be : @xmath61 , @xmath62 and @xmath63 gev .
[ points ] [ cols="^,^,^,^,^,^,^,^ " , ] point a point b + mass spectra of the six benchmark points . note that the scales are not identical for all spectra.,title="fig:",width=302]mass spectra of the six benchmark points .
note that the scales are not identical for all spectra.,title="fig:",width=302 ] + point c point d + mass spectra of the six benchmark points .
note that the scales are not identical for all spectra.,title="fig:",width=302 ] mass spectra of the six benchmark points .
note that the scales are not identical for all spectra.,title="fig:",width=302 ] + point e point f + mass spectra of the six benchmark points .
note that the scales are not identical for all spectra.,title="fig:",width=302 ] mass spectra of the six benchmark points .
note that the scales are not identical for all spectra.,title="fig:",width=302 ] + in the following , we consider that dark matter is composed of exclusively one particle produced thermally .
+ the density number of supersymmetric particles is determined by the boltzmann equation @xcite : @xmath64 where @xmath65 is the number density of supersymmetric particles , @xmath66 is the thermally averaged annihilation cross - section , @xmath67 is the hubble parameter , @xmath68 is the relic particle equilibrium number density .
the expansion rate @xmath67 is determined by the friedmann equation : @xmath69 where @xmath70 is the total energy density of the universe .
the entropy evolution reads : @xmath71 where @xmath72 is the total entropy density . solving and evolving simultaneously
( [ boltzmann ] ) , ( [ friedmann ] ) and ( [ entropy_evolution ] ) enable to compute the relic density in our present universe . in the standard cosmology , the dominant component before big - bang nucleosynthesis is considered to be radiation , which is constituted of all relativistic particles .
this assumption is however relaxed in many cosmological scenarios .
the last two equations can indeed be written more generally as @xcite @xmath73 @xmath74 parametrises a modified evolution of the total density of the universe , beyond radiation density @xmath75 .
@xmath76 parametrises here effective entropy fluctuations due to unknown properties of the early universe .
the radiation energy and entropy density evolutions are known and can be written as usual : @xmath77 in the following , we consider two scenarios in which the energy density is modified , and two scenarios with a modified entropy content .
the quintessence model is one of the most well - known models for dark energy .
it is based on a cosmological scalar field which has presently a negative constant pressure @xmath78 and a positive constant energy density @xmath70 such as @xmath79 @xcite .
this behaviour is achieved when the kinetic term of the scalar field equilibrates the potential .
however , in the early universe , the scalar field has a dominating kinetic term , leading to a positive pressure such as @xmath80 . during this period ,
the energy density was varying very quickly , such as @xmath81 .
we study here a quintessence scenario in which the quintessence field before big - bang nucleosynthesis was dominating the expansion of the universe . in this case
@xcite : @xmath82 where @xmath83 is the proportion of quintessence to radiation at the bbn temperature ( @xmath841 mev ) .
we consider that @xmath83 is a free parameter , which can be constrained using the bbn abundance constraints . to compute the abundance of the elements produced during the primordial nucleosynthesis
, we use the code alterbbn @xcite integrated into superiso relic . comparing the abundances to the observational constraints
, we obtain limits on @xmath83 .
the second scenario we consider here is a late decay of an inflaton field .
the inflaton field is a scalar field which is considered to be responsible for the rapid inflation of the early universe .
generally , the inflaton is considered to decay into standard particles much before the relic decoupling from the primordial soup .
however , several models evoke the possibility of a late decay of the inflaton , around the time of bbn . from @xcite ,
there exist cosmological models in which the late decay of a scalar field reheats the universe to a low reheating temperature , which can be smaller than the freeze - out temperature , without spoiling primordial nucleosynthesis .
the decay of this scalar field into radiation increases the entropy and modifies the expansion rate of the universe .
we consider here such a scenario in which we neglect the eventual entropy production , and we takes @xcite @xmath85 the exponent is here increased from 6 to 8 in comparison to the quintessence field , as mentioned also in @xcite .
such a modification of the expansion rate can be also achieved in a universe with extra - dimensions modifying the friedmann equations @xcite . in this third scenario
, we assume that a primordial entropy production due to an unknown component occurs .
in general , such an entropy production should be accompanied with energy production , but to better estimate the deviation of the relic density in such a cosmological scenario , we neglect here the energy production , and we consider that the universe has , in addition to the standard radiation entropy density , a dark entropy density evolving like @xcite @xmath86 where @xmath87 is the ratio of effective dark entropy density over radiation entropy density at the time of bbn . the corresponding entropy production is related to @xmath88 by the relation @xmath89 \ ; , \label{sigmad}\ ] ] with @xmath90 and @xmath91 in this last scenario , the reheating temperature @xmath92 is smaller than the neutralino freeze out temperature ( @xmath93 gev ) @xcite , @xmath92 should be considered as a cosmological parameter that can take any value around a few mev and then the neutralinos decoupled from the plasma before the end of the reheating process so their relic density will differ from its standard value .
we consider that the inflaton decays around the time of big - bang nucleosynthesis , generating entropy . from the standard
late reheating scenarios , we assume that the entropy production evolves like @xcite @xmath94 for @xmath95 mev and that this entropy production stops at the time of bbn .
we again neglect the energy production or non - thermal production of particles in order to better understand the effects of a reheating entropy production on the relic density . in term of entropy density
, we have @xmath96 the different scenarios do not have impact on the cosmological observations , but they can modify the abundance of the elements . we compute with alterbbn @xcite the abundance of the elements in the different scenarios for each benchmark points , varying the single parameter for a given scenario , and we apply the following conservative constraints @xcite : @xmath97 for the helium abundance @xmath98 and the primordial @xmath99 , @xmath100 , @xmath101 and @xmath102 ratios . in fig . [ fig_bbn ]
, we consider the relic density calculated for each of the benchmark points ( a - f , from top to bottom ) and for different cosmological scenarios ( from left to right ) .
relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162 ] + relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162 ] + relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162 ] + relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162 ] + relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162 ] + relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162]relic density in function of the cosmological model parameters .
the red lines correspond to points excluded by the bbn constraints in eq .
( [ bbnconstraints ] ) , whereas the blue lines correspond to a region with a correct abundance of the elements .
the green squares correspond to points allowed by the bbn constraints and giving a relic density in agreement with the wmap dark matter interval.,title="fig:",width=162 ] these plots reveal that the quintessence and inflaton scenarios globally increase the relic density , while the entropy and reheating scenarios decrease it .
the comparison with the bbn constraints is also represented , and the red part of the curves is excluded at 95% c.l . , while the blue part gives a correct abundance of the elements . as general features ,
the quintessence and inflaton scenarios can increase the relic density by three orders of magnitudes without interfering with the bbn constraints , and the entropy and reheating scenarios can decreased to a factor of @xmath103 .
therefore , apart from point b which has an extremely low relic density value in the standard cosmological scenario , all the other benchmark points can have a relic density value compatible with the cosmological observations if a non minimal cosmological scenario is considered .
we have shown with different well - known cosmological scenarios , that the relic density constraints can be very strongly relaxed .
therefore , we propose to compare the relic density calculated in the standard model of cosmology to the following interval @xmath104 to take into consideration the fact that it is possible to increase any relic density calculated in the standard cosmology by three orders of magnitude , and to decrease it by six orders , with non - standard cosmological scenarios in agreement with the current cosmological data . we can re - apply the relic density constraints , and the results are shown in figs .
[ fig_amsb_reloaded]-[fig_nmmamsb_reloaded ] . as in the figures of section [ sec : models ] ,
the green zones correspond to regions in agreement with all flavour and direct constraints , but not necessarily with the relic density constraint .
we added in the figures black points , which correspond to regions also in agreement with the new dark matter interval .
it is clear that the allowed regions are therefore much larger than with the initial relic density interval , but a surprising result is that even with the very large interval we use here for the relic density , the relic density constraint still excludes large part of the parameter spaces . in particular , in the mamsb and hcamsb scenarios and their nmssm counterparts ,
the relic density constraints clearly exclude the region @xmath105 tev .
the mmamsb scenario however is not constrained anymore when using the new dark matter interval .
constraints on the minimal amsb parameter space .
the exclusion regions are plotted in the order given in the legend .
the red zones are excluded by the inclusive branching ratio of @xmath20 , the yellow ones correspond to charged lsp , the olive green area is excluded by direct collider constraints , the light blue zones are excluded by br(@xmath21 ) , the dark blue zones by br(@xmath22 ) , the magenta zones by @xmath23 , the orange zones by br(@xmath24 ) and the grey zones by br(@xmath25 ) .
the green areas are in agreement with all the previously mentioned constraints .
the black area corresponds to parameters in agreement with all constraints , including the revised relic density interval given in eq .
( [ newwmap]).,title="fig:",width=604 ] + constraints on the minimal amsb parameter space .
the exclusion regions are plotted in the order given in the legend .
the red zones are excluded by the inclusive branching ratio of @xmath20 , the yellow ones correspond to charged lsp , the olive green area is excluded by direct collider constraints , the light blue zones are excluded by br(@xmath21 ) , the dark blue zones by br(@xmath22 ) , the magenta zones by @xmath23 , the orange zones by br(@xmath24 ) and the grey zones by br(@xmath25 ) .
the green areas are in agreement with all the previously mentioned constraints .
the black area corresponds to parameters in agreement with all constraints , including the revised relic density interval given in eq .
( [ newwmap]).,title="fig:",width=604 ] + constraints on the minimal amsb parameter space .
the exclusion regions are plotted in the order given in the legend .
the red zones are excluded by the inclusive branching ratio of @xmath20 , the yellow ones correspond to charged lsp , the olive green area is excluded by direct collider constraints , the light blue zones are excluded by br(@xmath21 ) , the dark blue zones by br(@xmath22 ) , the magenta zones by @xmath23 , the orange zones by br(@xmath24 ) and the grey zones by br(@xmath25 ) .
the green areas are in agreement with all the previously mentioned constraints .
the black area corresponds to parameters in agreement with all constraints , including the revised relic density interval given in eq .
( [ newwmap]).,title="fig:",width=604 ] constraints on the hcamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the hcamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the hcamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] constraints on the mmamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the mmamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the mmamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] constraints on the mnamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the mnamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the mnamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] constraints on the nhcamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the nhcamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the nhcamsb parameter space with the revised relic density interval given in eq .
( [ newwmap ] ) .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] constraints on the nmmamsb parameter space including the revised relic density interval .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the nmmamsb parameter space including the revised relic density interval .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ] + constraints on the nmmamsb parameter space including the revised relic density interval .
the colour codes are the same as in fig .
[ fig_amsb_reloaded].,title="fig:",width=604 ]
the benchmark points selected for testing alternative cosmology scenarios are in agreement with precision flavour and direct search constraints .
it turns out that the phenomenology expected at the lhc is quite peculiar .
the mass spectra for the benchmark points we considered show that the lightest neutralino is the lsp and the lightest chargino is in most cases very close in mass to the neutralino ( points a , b , c , f )
. this will give rise to peculiar signatures due to the very limited number of open channels for the decay modes .
we analyse in more detail in the following each of the six benchmark points previously selected .
production of charginos and neutralinos takes place at the lhc via cascade decays of squark and gluinos and via the direct production channels @xmath106 where the s - channel exchange of an off - shell w or z or photon , and the contribution of susy - qcd diagrams are important .
indeed these cross - sections receive important susy - qcd corrections , typical values are given in @xcite or can be obtained using a monte carlo program including the relevant k - factors or the detailed next - to - leading matrix elements .
the minimal amsb scenario is disfavoured by the standard cosmology as very far from the wmap dark matter allowed interval .
the region allowed in our more general analysis typically favours points in which the lightest chargino and neutralino are very close in mass and not so heavy . for point
a , @xmath107 gev and @xmath108 gev so that the mass splitting is only 170 mev . due to this small mass
splitting , the open decay modes for the @xmath109 are ( neglecting very small modes below the @xmath110 level ) @xmath111 , where @xmath112 so that @xmath113 and @xmath114 have the same branching through the three body decays while most of the signal with @xmath115 is through the two body production of a charged pion .
the next lightest particle is the @xmath116 , decaying mainly to @xmath109 and w or @xmath117 and sm - like higgs as for the light higgs for this point ( 118.15 gev ) the decay is kinematically allowed : @xmath118 where @xmath113 , @xmath39 . since the branching of the mode @xmath119 is 100% and the mode @xmath120 is non - negligible one can study the clean trilepton signal usually suggested at hadron colliders @xcite .
the benchmark point b is in the minimal amsb scenario too .
the situation is similar to the previous point with in this case the lightest chargino and neutralino lighter than the sm - like higgs boson . for point b ,
@xmath121 gev and @xmath122 gev so that the mass splitting is only 63 mev .
the decay modes for the lightest chargino are similar to the previous case @xmath123 and for the second neutralino @xmath124 the sm - higgs boson ( with a mass of 115.7 gev ) decays with a sizeable branching to pairs of lightest charginos and pairs of lightest neutralinos ( branching @xmath125 and @xmath126 respectively ) while the largest mode is @xmath127 with a branching of @xmath128 .
the hcamsb scenario is disfavoured by standard cosmology and excluded by the wmap dark matter allowed interval as for the minimal amsb scenario discussed above .
also in this case the lightest chargino and neutralino are very close in mass .
the masses are @xmath129 gev and @xmath130 gev with a splitting of only 17 mev .
in this situation the lightest chargino decays are very close to the numbers for point a. the decay modes for the lightest chargino are : @xmath131 and for the second neutralino @xmath132 in this case the trilepton mode @xmath133 is especially interesting as the branching fraction @xmath111 is 100% and the one for @xmath134 is 26% . the mixed modulus amsb supersymmetry breaking scenario allows for viable dark matter candidates .
benchmark point d has the neutralino lsp with a mass @xmath135 gev .
the next - to - lightest supersymmetric particle is the stau with a mass @xmath136 gev while the lightest chargino and the second lightest neutralino are heavier and almost degenerate with a mass of 1095 and 1098 gev respectively .
the stau decays to tau and the lightest neutralino @xmath137 while the lightest chargino decays mainly to @xmath138 the second lightest neutralino decays to stau and tau while as the sm - like higgs is light enough to be produced ( 124.3 gev ) , the decay @xmath139 is also kinematically open @xmath140 for the same model as in the previous paragraph we have also an example of a higher mass spectrum in which all supersymmetric partners are quite heavy with the neutralino lsp above 7 tev and all other particles above 10 tev .
nmssm with minimal amsb scenario is one of the models considered in the previous sections .
it is disfavoured by the relic density constraints if standard cosmology is assumed .
point f is in the allowed zone for flavour , low energy and direct search constraints and is allowed if standard cosmology constraints are relaxed as discussed in the previous sections . from the point of view of lhc searches ,
this point is similar to point a. for point f , @xmath141 gev and @xmath142 gev so that the mass splitting is 99 mev .
apart from these two particles , the next supersymmetric particle ( in terms of mass ) is the second neutralino ( see fig .
[ fig_spectra ] ) with a mass of 721 gev .
we have considered in details the constraints on different possible realisations of superconformal anomaly mediation breaking mechanisms in supersymmetry .
these constraints include the usual lep , b - factories , tevatron and lhc searches , but also precision constraints of cosmological origin , namely the wmap limits on the relic density of cold dark matter .
we have discussed the standard cosmological approach and also alternative cosmological scenarios which do not change the cosmological observations but which can affect strongly the constraints on the parameter space of these supersymmetric models based on the relic abundance of dark matter .
we therefore show how the dark matter constraints can be weakened in order to avoid strong model dependent assumptions in the choice of the cosmological model .
based on different benchmark points for amsb models , we performed a detailed analysis of the constraints imposed by particle data and cosmology ( both standard and alternative ) and finally we gave the typical mass spectra and decay modes relevant for the lhc searches .
the main lesson that can be learnt in such an exercice is that usual bounds on the parameter space of these models are too restrictive and bear a strong hidden cosmological model dependence in the assumption of the standard cosmological scenario
. concerning the lhc searches , points which are excluded in the standard analysis but permitted in this more general approach , may be quite relevant for testing not only the particle theory models themselves but also alternative cosmological scenarios at the lhc , as in many cases relatively low mass supersymmetric particles are allowed with a peculiar spectrum where the lightest neutralino and chargino are very close in mass .
99 a. arbey and f. mahmoudi , `` susy constraints from relic density : high sensitivity to pre - bbn expansion rate , '' phys .
b * 669 * ( 2008 ) 46 [ arxiv:0803.0741 ] .
a. arbey and f. mahmoudi , `` susy constraints , relic density , and very early universe , '' jhep * 1005 * ( 2010 ) 051 [ arxiv:0906.0368 ] ; a. arbey and f. mahmoudi , `` lhc and ilc data and the early universe properties , '' nuovo cim .
c * 33 * ( 2010 ) 151 [ arxiv:1002.4096 ] .
l. randall and r. sundrum , `` out of this world supersymmetry breaking , '' nucl .
b * 557 * ( 1999 ) 79 [ hep - th/9810155 ] ; g.f .
giudice , m.a .
luty , h. murayama and r. rattazzi , `` gaugino mass without singlets , '' jhep * 9812 * ( 1998 ) 027 [ hep - ph/9810442 ] ; a. pomarol and r. rattazzi , `` sparticle masses from the superconformal anomaly , '' jhep * 9905 * ( 1999 ) 013 [ hep - ph/9903448 ] ; d.w .
jung and j.y .
lee , `` anomaly - mediated supersymmetry breaking demystified , '' jhep * 0903 * ( 2009 ) 123 [ arxiv:0902.0464 ] . k. choi , k.s .
jeong and k.i .
okumura , `` phenomenology of mixed modulus - anomaly mediation in fluxed string compactifications and brane models , '' jhep * 0509 * ( 2005 ) 039 [ hep - ph/0504037 ] .
r. dermisek , h. verlinde and l.t .
wang , `` hypercharged anomaly mediation , '' phys .
* 100 * ( 2008 ) 131804 [ arxiv:0711.3211 ] .
h. baer , r. dermisek , s. rajagopalan and h. summy , `` neutralino , axion and axino cold dark matter in minimal , hypercharged and gaugino amsb , '' jcap * 1007 * ( 2010 ) 014 [ arxiv:1004.3297 ] . f. e. paige , s. d. protopopescu , h. baer and x. tata , `` isajet 7.69 : a monte carlo event generator for p p , anti - p p , and e+ e- reactions , '' hep - ph/0312045 . f. mahmoudi , `` superiso : a program for calculating the isospin asymmetry of @xmath143 in the mssm , '' comput .
commun .
* 178 * ( 2008 ) 745 [ arxiv:0710.2067 ] ; f. mahmoudi , `` superiso v2.3 : a program for calculating flavor physics observables in supersymmetry , '' comput .
commun .
* 180 * ( 2009 ) 1579 [ arxiv:0808.3144 ] ; ` http://superiso.in2p3.fr ` f. mahmoudi , `` superiso v3.0 , flavor physics observables calculations : extension to nmssm , '' comput .
commun .
* 180 * ( 2009 ) 1718 , ` http://superiso.in2p3.fr ` a. arbey and f. mahmoudi , `` superiso relic : a program for calculating relic density and flavor physics observables in supersymmetry , '' comput
. phys .
commun .
* 181 * ( 2010 ) 1277 [ arxiv:0906.0369 ] , ` http://superiso.in2p3.fr/relic ` f. mahmoudi , `` new constraints on supersymmetric models from @xmath144 , '' jhep * 0712 * ( 2007 ) 026 [ arxiv:0710.3791 ] .
cdf collaboration , `` search for @xmath145 and @xmath146 decays in 3.7 fb@xmath147 of @xmath148 collisions with cdf ii , '' cdf public note * 9892 * ( 2009 ) .
a. g. akeroyd and f. mahmoudi , `` constraints on charged higgs bosons from @xmath149 and @xmath150 , '' jhep * 0904 * ( 2009 ) 121 [ arxiv:0902.2393 ] . m. antonelli _ et al . _
[ flavianet working group on kaon decays ] , `` precision tests of the standard model with leptonic and semileptonic kaon decays , '' arxiv:0801.1817 [ hep - ph ] .
e. komatsu _ et al .
_ [ wmap collaboration ] , `` seven - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation , '' astrophys . j. suppl .
* 192 * ( 2011 ) 18 [ arxiv:1001.4538 ] .
h. baer , r. dermisek , s. rajagopalan and h. summy , `` prospects for hypercharged anomaly mediated susy breaking at the lhc , '' jhep * 0910 * ( 2009 ) 078 [ arxiv:0908.4259 ] . h. baer , e.k . park , x. tata and t.t .
wang , `` collider and dark matter searches in models with mixed modulus - anomaly mediated susy breaking , '' jhep * 0608 * ( 2006 ) 041 [ hep - ph/0604253 ] . for recent reviews ,
see : m. maniatis , `` the next - to - minimal supersymmetric extension of the standard model reviewed , '' int .
j. mod .
phys . a * 25 * ( 2010 ) 3505 [ arxiv:0906.0777 ] ; u. ellwanger , c. hugonie and a.m. teixeira , `` the next - to - minimal supersymmetric standard model , '' phys .
* 496 * ( 2010 ) 1 [ arxiv:0910.1785 ] .
u. ellwanger and c. hugonie , `` nmspec : a fortran code for the sparticle and higgs masses in the nmssm with gut scale boundary conditions , '' comput .
commun .
* 177 * ( 2007 ) 399 [ hep - ph/0612134 ] .
p. gondolo and g. gelmini , `` cosmic abundances of stable particles : improved analysis , '' nucl .
b * 360 * ( 1991 ) 145 ; j. edsjo and p. gondolo , `` neutralino relic density including coannihilations , '' phys . rev .
d * 56 * ( 1997 ) 1879 [ hep - ph/9704361 ] .
p. salati , `` quintessence and the relic density of neutralinos , '' phys .
b * 571 * ( 2003 ) 121 [ astro - ph/0207396 ] .
a. arbey , in preparation , ` http://superiso.in2p3.fr/relic/alterbbn `
. g.b .
gelmini and p. gondolo , `` neutralino with the right cold dark matter abundance in ( almost ) any supersymmetric model , '' phys .
d * 74 * ( 2006 ) 023510 [ hep - ph/0602230 ] .
g. gelmini , p. gondolo , a. soldatenko and c.e .
yaguna , `` the effect of a late decaying scalar on the neutralino relic density , '' phys .
d * 74 * ( 2006 ) 083514 [ hep - ph/0605016 ] .
gelmini and p. gondolo , `` neutralino with the right cold dark matter abundance in ( almost ) any supersymmetric model , '' phys .
d * 74 * ( 2006 ) 023510 [ hep - ph/0602230 ] .
g. gelmini , p. gondolo , a. soldatenko and c.e .
yaguna , `` the effect of a late decaying scalar on the neutralino relic density , '' phys .
d * 74 * ( 2006 ) 083514 [ hep - ph/0605016 ] .
n. okada and o. seto , `` relic density of dark matter in brane world cosmology , '' phys .
d * 70 * ( 2004 ) 083531 [ hep - ph/0407092 ] .
k. jedamzik , `` big bang nucleosynthesis constraints on hadronically and electromagnetically decaying relic neutral particles , '' phys
. rev .
d * 74 * ( 2006 ) 103509 [ hep - ph/0604251 ] .
w. beenakker , m. klasen , m. kramer , t. plehn , m. spira and p.m. zerwas , `` the production of charginos / neutralinos and sleptons at hadron colliders , '' phys .
* 83 * ( 1999 ) 3780 [ erratum - ibid .
* 100 * ( 2008 ) 029901 ] [ hep - ph/9906298 ] .
h. baer and x. tata , `` probing charginos and neutralinos beyond the reach of lep at the tevatron collider , '' phys .
d * 47 * ( 1993 ) 2739 .
h. baer , c.h .
chen , f. paige and x. tata , `` trileptons from chargino - neutralino production at the cern large hadron collider , '' phys .
d * 50 * ( 1994 ) 4508 [ hep - ph/9404212 ] .
|
anomaly mediation is a popular and well motivated supersymmetry breaking scenario .
different possible detailed realisations of this set - up are studied and actively searched for at colliders .
apart from limits coming from flavour , low energy physics and direct collider searches , these models are usually constrained by the requirement of reproducing the observations on dark matter density in the universe .
we reanalyse these bounds and in particular we focus on the dark matter bounds both considering the standard cosmological model and alternative cosmological scenarios .
these scenarios do not change the observable cosmology but relic dark matter density bounds strongly depend on them .
we consider few benchmark points excluded by standard cosmology dark matter bounds and suggest that loosening the dark matter constraints is necessary in order to avoid a too strong ( cosmological ) model dependence in the limits that are obtained for these models .
we also discuss briefly the implications for phenomenology and in particular at the large hadron collider .
cern - ph - th-2011 - 052 + lycen 2011 - 01 * anomaly mediated susy breaking scenarios in the light of cosmology and in the dark ( matter ) * a. arbey@xmath0 , a. deandrea@xmath1 and a. tarhini@xmath1 + _ @xmath2 universit de lyon , france ; universit lyon 1 , cnrs / in2p3 , umr5822 ipnl , f-69622 villeurbanne cedex , france _ + + +
|
in this paper we investigate the computational complexity of some problems in three - dimensional topology and geometry .
we show that the problem of determining a bound on the genus of a knot in a 3-manifold , is * np*-complete . using similar ideas ,
we show that deciding whether a curve in a metrized pl 3-manifold bounds a surface of area less than a given constant @xmath1 is * np*-hard . determining whether a given knot is trivial or not is one of the historically central questions in topology .
the problem of finding an algorithm to determine knot triviality was posed by dehn @xcite .
dehn s investigations into this area led to the formulation of the word and isomorphism problems , which played an important role in the development of the theory of algorithms .
the first algorithm for the unknotting problem was given by haken @xcite .
haken s procedure is based on normal surface theory , a method of representing surfaces introduced by kneser @xcite .
analysis of the computational complexity of this algorithm is more recent .
hass , lagarias and pippenger showed that haken s unknotting algorithm runs in time at most @xmath2 , where the knot @xmath3 is embedded in the 1-skeleton of a triangulated manifold @xmath4 with @xmath5 tetrahedra , and @xmath6 is a constant independent of @xmath4 or @xmath3 @xcite .
it was also shown in @xcite that the unknotting problem is in * np*. the notion of genus was defined by seifert @xcite in 1935 for knots in the 3-sphere , and extends directly to knots in an arbitrary 3-manifold @xmath4 . given a knot @xmath3 ,
consider the class @xmath7 of all orientable spanning surfaces for @xmath3 .
these are surfaces embedded in @xmath4 with a single boundary component that coincides with @xmath3 .
seifert showed that this class is non - empty for any knot @xmath3 in the 3-sphere . for knots in a general manifold , @xmath7 is non - empty when @xmath3 represents a trivial element in the first integer homology group of @xmath4 .
the genus @xmath8 of a knot @xmath3 is the minimum genus of a surface in @xmath7 , or @xmath9 if @xmath10 .
the genus measures one aspect of the degree of `` knottedness '' of a curve .
the unknotting problem is a special case of the more general problem of determining the genus of a knot in a 3-manifold .
given a knot @xmath3 in an orientable 3-manifold and a positive integer @xmath0 , this problem asks for a procedure to determine whether the _ knot genus _ of @xmath3 , the minimal genus of an orientable spanning surface for @xmath3 in a 3-dimensional manifold , is at most @xmath0 .
a knot is trivial , or unknotted , precisely when its genus is zero .
we will show that the problem of determining the genus of a knot in a 3-manifold is * np*-complete .
previous results on this problem were given in @xcite , where it was shown to lie in * pspace * , roughly the class of problems that run in polynomial space .
no lower bounds on the running time were previously known .
we work with 3-manifolds that are triangulated and orientable , and with orientable embedded surfaces .
this is not a significant restriction , since all compact 3-dimensional manifolds admit unique pl structures @xcite .
knot _ in a triangulated 3-manifold @xmath4 is a connected simple ( non self - intersecting ) closed curve in the 1-skeleton of @xmath4 .
any smooth knot in a smooth manifold , or more generally any tame knot , is equivalent to a knot that lies in the 1-skeleton of some triangulation .
we formulate the problem of computing the genus as a language - recognition problem in the usual way , see @xcite . in 1961
schubert @xcite , in an extension of haken s work , showed the decidability of the problem : problem : = 3-manifold knot genus + instance : triangulated 3-dimensional manifold @xmath4 , a knot @xmath3 in the + 1-skeleton of @xmath4 , and a natural number @xmath0 .
+ question : does the knot @xmath3 have @xmath11 ?
the size of an instance is measured by the sum of the number of tetrahedra in @xmath4 . in section [ sectionnph ]
we establish [ thmnph ] 3-manifold knot genus is * np*-hard .
it was established in @xcite that 3-manifold knot genus is in * pspace*. we improve this bound in section [ sectiongnp ] . [ thmgnp ] 3-manifold knot genus is * np*. in combination these two results give : [ 3mkgnpc ] 3-manifold knot genus is * np*-complete .
theorem [ thmnph ] is proved through a connection to one - in - three sat , a known * np*-complete problem that will be reviewed in section 3 .
the theorem carries out a construction that transforms an instance of one - in - three sat to an instance of 3-manifold knot genus . by
`` transform ''
we mean that an instance of one problem is changed to an instance of the second by a procedure that requires time polynomial in the size of the instance . to a boolean expression representing an instance of
one - in - three sat we associate a positive integer @xmath0 and a certain knot in a triangulated , compact 3-manifold .
this knot bounds a surface of genus at most @xmath0 exactly when there is a truth assignment to the boolean expression satisfying the requirements of one - in - three sat . since
one - in - three sat is * np*-hard , this establishes that 3-manifold knot genus is also * np*-hard . in section [ sectiongnp ]
we prove theorem [ thmgnp ] , giving a certificate which demonstrates in polynomial time that a genus @xmath0 knot @xmath3 bounds a surface of genus at most @xmath0 .
the argument in @xcite established that the unknotting problem is * np * using the existence of a normal disk that lies along an extremal ray in the space of normal solutions , called a vertex surface in jaco - tollefson @xcite .
the existence of such an extremal normal surface of minimal genus spanning a knot is not known , so a new technique is needed ( see @xcite for known results here ) .
this is provided in theorem [ orbitcount ] , which gives an algorithm to count the number of orbits of a type of pseudogroup action on a set .
theorem [ orbitcount ] seems likely to have more general applicability . in section [ pseudogroup ]
we describe this algorithm and in section [ sectiongnp ] we apply it to a pseudogroup action that arises in the theory of normal surfaces .
this allows us to determine in polynomial time the number of components in a normal surface described by an integer vector in @xmath12 .
in particular we are able to certify that a normal surface is connected , orientable and has connected boundary .
since calculating the euler characteristic of a normal surface can be done efficiently , establishing orientability and connectedness are the key steps in constructing a certificate of its genus . in section
[ extended ] we extend the orbit counting algorithm to allow the counting of additional integer weight sums associated to each orbit .
this allows for the polynomial time calculation of the genus of all the components of a fundamental normal surface , as well as a count of the number of components .
the genus and the area of a surface are closely connected . in section [ area ]
we extend the methods developed in studying genus to study the problem of determining the smallest area of a spanning surface for a curve in a 3-manifold .
we show that computing an upper bound on the area of a smallest area spanning surface is * np*-hard .
we refer to @xcite for a discussion of complexity classes such as * np * and * pspace * and for a discussion of complexity problems in low - dimensional topology . * remarks : * 1 .
knots are often studied in @xmath13 or @xmath14 rather than in a general manifold .
our methods show that determining knot genus in @xmath13 or @xmath14 , or any fixed manifold , is * np*. it is not clear whether the corresponding problem remains * np*-hard if one restricts consideration to knots in @xmath13 or @xmath14 .
casson has shown that a procedure to determine whether a 3-manifold is homeomorphic to the 3-sphere , following the 3-sphere recognition algorithm described in @xcite and @xcite , runs in time less than @xmath15 , where @xmath16 is a polynomial . in the direction of lower bounds ,
it was shown in @xcite that determining certain values of the jones polynomial of alternating links is * # p*-hard .
we are grateful to the referee for numerous suggestions on the exposition of this paper .
general surfaces in 3-manifolds can wind and twist around the manifold in complicated ways .
kneser described a procedure in which surfaces can be `` pulled taut '' , until they take a simple and rigid position @xcite . in this
_ normal _ position , they have very succinct algebraic descriptions .
we use an approach to normal surfaces in triangulated 3-manifolds based on work of jaco - rubinstein @xcite and jaco - tollefson @xcite .
a _ normal surface _
@xmath17 in a triangulated compact 3-manifold @xmath4 is a @xmath18-surface whose intersection with each tetrahedron in @xmath4 consists of a finite number of disjoint _
elementary disks_. these are properly embedded disks that are isotopic to either triangles or quadrilaterals as shown in figure [ normal ] , by an isotopy preserving each face of the tetrahedron . elementary disks in a normal surface .
] within each tetrahedron of @xmath4 there are four possible triangles and three possible quadrilaterals , up to a _
normal isotopy of @xmath4 _ , an isotopy which leaves each cell of the triangulation of @xmath4 invariant .
w. haken observed that a normal surface is determined up to such isotopies by the number of pieces of each of the seven kinds of elementary disks that occur in each tetrahedron , or a vector in @xmath12 . a normal surface @xmath17 is described by a non - negative integer vector @xmath19 , that gives the _ normal coordinates _ of @xmath17 .
there is a homogeneous rational cone @xmath20 in @xmath21 , called the _
haken normal cone _ , that contains the vectors @xmath22 for all normal surfaces @xmath17 in @xmath4 .
if @xmath23 , then the haken normal cone is specified by linear equations and inequalities of the form @xmath24 @xmath25 the first set of equations expresses _ matching conditions , _ which say that the number of edges on a common triangular face of two adjacent tetrahedra , coming from a collection of elementary disks in each of the tetrahedra , must match . for each triangular face
there are three types of edges ( specified by a pair of edges on the triangle ) , which yield 3 matching conditions per face .
triangular faces in the boundary @xmath26 give no matching equations . the second set
are called the _ positivity
_ conditions .
the cone @xmath20 is _ rational , _ because the above equations have integer coefficients .
we let @xmath27 denote the set of integral vectors in the cone @xmath20 .
an additional set of conditions , the _ quadrilateral conditions , _ is required for an integral vector in the cone to correspond to the normal coordinates of an embedded surface .
this condition states that of the three types of quadrilateral found in each tetrahedron , only one can occur in the vector with non - zero coefficient .
the quadrilateral conditions are required because two distinct types of quadrilateral in a single tetrahedron necessarily intersect , and we are interested in embedded , non - self - intersecting surfaces
. a vector in the haken normal cone that satisfies the quadrilateral conditions corresponds to an embedded normal surface .
this surface is unique up to a normal isotopy .
however it is important to note that the surface corresponding to a given vector in the cone @xmath28 may not be connected .
a normal surface meets each edge of a triangulation in a finite number of points .
the sum of these intersection numbers over all the edges of a triangulation is called the _ weight _ of the normal surface . a _ fundamental normal surface _
is a normal surface @xmath17 such that @xmath29 in the terminology of integer programming , such a vector @xmath22 is an element of the _ minimal hilbert basis _
@xmath30 of @xmath20 , see schrijver ( * ? ? ?
* theorem 16.4 ) . a fundamental normal surface is always connected , but connected normal surfaces need not be fundamental .
vertex minimal solution _ is a special kind of fundamental surface , one that corresponds to a solution of the normal surface equations that lies along an extremal ray of the cone of solutions and is not a multiple of another such extremal solution .
hass , lagarias and pippenger ( * ? ? ?
* lemma 6.1 ) gave a bound for the size of the vectors corresponding to any fundamental surface .
[ bound ] let @xmath4 be a triangulated compact 3-manifold , possibly with boundary , that contains @xmath5 tetrahedra . *
any vertex minimal solution @xmath31 of the haken normal cone @xmath32 in @xmath21 has @xmath33 * any minimal hilbert basis element @xmath31 of the haken fundamental cone @xmath32 has @xmath34 schubert @xcite showed that a surface of smallest genus spanning @xmath3 can be found among the fundamental surfaces .
[ mingenusfund ] there is a minimal genus spanning surface for @xmath3 which is a fundamental normal surface .
similar to the theory of normal surfaces , though somewhat easier , is the theory of normal curves .
these are curves on a surface that intersect each triangle in a collection of _ normal arcs _ , arcs that have endpoints on distinct edges .
normal curves arise as the boundaries of normal surfaces in a manifold with boundary .
since there are three such arcs in each triangle , normal isotopy classes of normal curves in a surface that contains @xmath5 triangles are described by integer vectors in @xmath35 .
in this section we show how to reduce an instance of one - in - three sat to an instance of 3-manifold knot genus .
since one - in - three sat is known to be * np*-hard , this establishes that 3-manifold knot genus is also * np*-hard .
the problem one - in - three sat concerns logical expressions involving collections of literals ( boolean variables or their negations ) gathered in clauses consisting of three literals connected with @xmath36 s .
the logical expression contains a collection of clauses connected with @xmath37 s .
problem : one - in - three sat + instance : a set @xmath38 of variables and a collection @xmath1 of clauses over @xmath38 + such that each clause @xmath39 contains 3 literals .
+ question : s there a truth assignment for @xmath38 such that each clause in @xmath1 + has exactly one true literal ?
schaefer @xcite established that one - in - three sat is * np*-complete . to prove theorem [ thmnph ] ,
establishing that 3-manifold knot genus is * np*-hard , we show that an arbitrary problem in one - in - three sat can be reduced in polynomial time to a problem in 3-manifold knot genus .
see garey and johnson @xcite for a discussion and many examples of such reductions .
let @xmath40 be a set of variables and @xmath41 be a set of clauses in an arbitrary instance of one - in - three sat .
we will describe a knot @xmath3 in a compact 3-dimensional manifold ( with no boundary ) and an integer @xmath0 such that @xmath3 bounds a surface of genus smaller or equal to @xmath0 if and only if @xmath1 is satisfiable so that each clause in @xmath1 contains exactly one true literal .
we construct the 3-manifold @xmath4 in stages .
first we construct a 2-dimensional simplicial complex , then we thicken this complex , replacing triangles with subdivided triangular prisms , getting a triangulated , 3-dimensional manifold with boundary , as indicated in figure [ thicken ] .
finally we use a doubling construction , taking two copies of the manifold with boundary and gluing their boundaries together , to obtain a closed 3-manifold . , shown in cross - section ,
is `` thickened '' to produce a triangulated 3-manifold with boundary . ] to begin , we form a type of singular surface @xmath42 , that we call a _ branching _ surface , by identifying boundary curves of a collection of @xmath43 surfaces with boundary , each forming what we refer to as a _ piece _ of the branching surface .
we construct this collection of surfaces as follows .
let @xmath44 be the number of times that the variable @xmath45 appears in the collection of clauses , and @xmath46 be the number of times that the negation @xmath47 of @xmath45 appears . for @xmath48 ,
let @xmath49 and @xmath50 be genus one surfaces with @xmath51 and @xmath52 boundary curves respectively . also set @xmath53 to be a planar surface with @xmath54 boundary curves .
one of these boundary curves will later become the knot @xmath3 .
the branching surface @xmath42 is constructed by identifying these surfaces along appropriate boundary components as indicated in figure [ branching ] . corresponding to the boolean expression @xmath55 .
the shaded surface @xmath56 indicates the occurrence of @xmath57 in each of the clauses @xmath58 and @xmath59 . ]
branching occurs when more than two boundary curves are identified along a single curve .
we identify pairs of boundary curves by giving a homeomorphism between them .
up to isotopy , this is determined by specifying an orientation on the curves and setting the homeomorphism to be orientation reversing .
we first fix an orientation on each of @xmath60 and @xmath50 .
this induces an orientation on each boundary curve .
all identifications will involve gluing a boundary component of @xmath49 or @xmath50 to a boundary component of @xmath53 , and we require this gluing to be an orientation reversing homeomorphism .
label by @xmath61 the @xmath62 boundary components of @xmath53 .
the boundary component @xmath3 of @xmath53 , which will become our knot , has nothing identified to it . for each @xmath63 , @xmath64 , one boundary curve from the surface @xmath49
is identified to @xmath45 .
the remaining @xmath44 boundary components of @xmath49 are identified with @xmath44 of the curves @xmath65 on @xmath66 , with one component of @xmath67 identified with @xmath68 for each occurrence of the literal @xmath45 in the @xmath69 clause in @xmath1 .
similarly , one curve of @xmath70 is identified to the component of @xmath53 labeled @xmath45 , and the remaining @xmath46 boundary components are glued to @xmath71 , with a component glued to @xmath68 for each occurrence of the literal @xmath47 in the @xmath69 clause of @xmath1 .
a total of three surface boundaries are identified along each of @xmath72 , and exactly four surface boundaries are identified along each of @xmath73 , as in figure [ branching ] .
[ branch ] there is a truth assignment for @xmath38 such that each clause in @xmath1 has exactly one true literal if and only if there is a surface @xmath17 with connected boundary and genus at most @xmath74 and a continuous map @xmath75 such that @xmath76 is a homeomorphism onto @xmath3 .
[ [ proof . ] ] * proof . * +
+ + + + + + + suppose there is a truth assignment for @xmath38 such that each clause in @xmath1 contains exactly one true literal . form a surface @xmath17 inside @xmath42 by taking the union of @xmath53 and either @xmath49 if @xmath45 is true , or @xmath50 if @xmath77 is false .
then exactly two boundary components will be identified along each of the boundary components of @xmath53 other than @xmath3 itself , and @xmath3 becomes the boundary of the resulting embedded surface @xmath17 .
there is a contribution of one to the genus from each of the literals , since @xmath49 and @xmath50 each have genus one , and a contribution of one to the genus from each handle formed when a boundary component of a surface @xmath49 or @xmath50 is glued to @xmath53 along a curve @xmath68 .
the genus of @xmath17 is therefore equal to @xmath74 .
now suppose there is a surface @xmath78 of genus @xmath79 mapped continuously into @xmath42 that has a single boundary component mapped homeomorphically to the boundary curve @xmath3 .
we will show in this case that there is a surface @xmath80 with the same boundary , consisting of certain pieces of @xmath42 identified along their boundaries , having genus precisely @xmath81 , and containing , along each curve @xmath68 , exactly one of the three pieces of surface joining @xmath53 .
the map @xmath82 of @xmath78 into @xmath42 may be quite complicated , winding back and forth across @xmath42 , but by standard transversality arguments we can homotop @xmath82 so that it is a union of homeomorphisms of subsurfaces of @xmath78 mapped homeomorphically to one of the pieces @xmath53 , @xmath49 , @xmath83 forming @xmath42 .
more precisely , we can perturb @xmath78 by a small homotopy so that its intersection with the boundary components of @xmath53 is transverse , and pulls back to a collection of simple closed curves on @xmath78 . if any of these curves bounds a disk in @xmath78 , then the disk is mapped into some subsurface @xmath84 of @xmath42 while the boundary curve is mapped to @xmath85 .
since @xmath84 is not itself a disk , the disk can be homotoped into @xmath85 , and we can therefore homotop @xmath78 in a neighborhood of this disk to remove a component of @xmath86 . after repeating finitely many times , each component of the complement of @xmath87 in @xmath78 has non - positive euler characteristic . the image of @xmath78 in each piece @xmath53 , @xmath49
, @xmath83 has an algebraic degree , which is either even or odd .
this degree equals the number of pre - images in @xmath78 of a generic point in the piece .
the degree of the map from @xmath78 on @xmath53 is odd , since @xmath3 is the boundary of @xmath78 and therefore @xmath78 maps an odd number of times to points near @xmath3 .
the sum of the degrees along each of the pieces meeting a curve @xmath45 or @xmath68 is even , since @xmath78 has no boundary along these curves . in particular , for each @xmath88 , exactly one of @xmath49 , @xmath83 has odd multiplicity in @xmath78 .
form a new surface @xmath89 by taking the union of @xmath53 and each of the pieces @xmath49 , @xmath83 which have odd multiplicity in @xmath78 .
we will show that @xmath90 .
the surface @xmath89 is obtained from @xmath78 by a series of operations that either discard a subsurface with non - positive euler characteristic or replace a subsurface of @xmath78 that maps with odd degree to some piece @xmath91 with a subsurface mapping homeomorphically to @xmath91 .
the euler characteristic of a discarded subsurface of @xmath78 is smaller or equal to that of the subsurface that replaces it , so @xmath92 in either case .
the collection of pieces among @xmath93 , @xmath94 that are in the image of @xmath89 can be attached to @xmath53 along common boundary curves among @xmath95 , forming a connected surface @xmath96 of genus @xmath97 .
finally , a connected surface @xmath80 with boundary @xmath3 is obtained by identifying pairs of curves in @xmath98 that are mapped to the boundary components @xmath71 . each
@xmath99 then has either two or four curves in @xmath98 mapped to it , so there are either one or two identifications made along each @xmath99 .
the choice of which pairs to identify , in case there are four curves mapped to @xmath99 , is not important .
the euler characteristic of @xmath80 is the same as that of @xmath96 and @xmath89 , and therefore greater or equal to that of @xmath78 .
so the genus of @xmath80 is at most that of @xmath78 , genus@xmath100 .
each identification of a pair of surface boundaries along @xmath71 contributes one to the genus of @xmath80 .
there is at least one such identification along each of the @xmath101 curves @xmath102 , though there may be two if each of the three surfaces meeting @xmath53 along @xmath103 has multiplicity one in @xmath89 .
so identification of curves along @xmath71 adds at least @xmath101 to the genus of @xmath80 , and it follows that genus(@xmath104 . since we have seen that genus@xmath100 , equality must hold .
equality holds when exactly one of the three surface pieces meeting @xmath53 along @xmath103 has odd multiplicity for each @xmath105 .
we then assign the value `` true '' to a literal @xmath45 if @xmath49 is used in @xmath80 , and the value `` false '' to @xmath45 if @xmath83 is used in @xmath80 .
this gives a truth assignment to @xmath38 in which each clause in @xmath1 has exactly one true literal . to show that 3-manifold knot genus is * np*-hard we reduce in polynomial time an instance of the * np*-hard problem one - in - three sat to an instance of 3-manifold knot genus . *
proof of theorem [ thmnph ] . * given an instance of one - in - three sat form @xmath42 as in lemma [ branch ] .
then there is a truth assignment for @xmath38 such that each clause in @xmath1 has exactly one true literal if and only if @xmath3 is the boundary of a surface of genus @xmath74 mapped continuously into @xmath42 .
form a 3-manifold @xmath106 by thickening @xmath42 so that it is embedded inside a triangulated 3-dimensional manifold with boundary .
the thickening process replaces each subsurface forming @xmath42 by a product of a surface with an interval , and then glues these surfaces together along portions of their boundaries , as indicated in figure [ thicken ] .
the curve @xmath3 remains on the boundary of @xmath106 , and there is a projection map @xmath107 which fixes @xmath3 .
form a closed manifold @xmath4 by doubling @xmath106 along its boundary , namely by taking two copies of @xmath106 and identifying them along their boundaries by the identity map .
then @xmath4 admits an involution @xmath108 that fixes @xmath3 , and has quotient @xmath106 . to find a triangulation of @xmath106
, we first describe an explicit triangulation of the branching surface @xmath42 . an orientable surface with boundary
has a triangulation with one vertex on each boundary component and no vertices in the interior .
the number of triangles is @xmath109 , where @xmath0 is the genus and @xmath6 is the number of boundary components .
we choose such triangulations for each of the @xmath43 subsurfaces in the branching surface , and we match them together along boundary components to get a triangulation of @xmath42 .
so @xmath42 has a triangulation in which the number of triangles is linearly bounded in @xmath81 .
we thicken the surfaces @xmath49 and @xmath110 , by taking their product with an interval .
we form a cell structure of the thickened surface by dividing the product into prisms , products of a triangle and an interval . for the thickening of @xmath53 , we start by doing the same and then go on to divide each interval into five subintervals .
the top , middle and bottom subintervals are identified with the intervals from the three thickened surfaces meeting each thickened boundary component of @xmath53 corresponding to a curve @xmath99 . only the top and bottom intervals are identified with thickened surfaces from the other boundary components .
we can therefore now glue these thickened surfaces together to get a 3-cell structure on @xmath106 .
each cell in this structure is a prism .
we form a closed 3-manifold @xmath4 by doubling @xmath106 along its boundary , gluing two copies of @xmath106 together along their boundaries to obtain a 3-manifold with no boundary .
finally , we stellar subdivide the cell structure to get a triangulation .
each prism is divided into 14 tetrahedra , by dividing each rectangular face into four triangles by coning to a vertex in the center of each such face , and then coning the 14 triangles of the boundary of the prism to a vertex added to its center .
the number of resulting simplices in @xmath4 is linearly bounded by @xmath81 .
we now check that if @xmath3 bounds a surface of genus @xmath111 with interior in @xmath112 then it bounds a surface of the same genus in @xmath42 .
suppose that @xmath113 is an embedded surface in @xmath4 with boundary @xmath3 . if @xmath113 does not already lie in @xmath106 , then perturb it slightly so that the interior of @xmath113 meets @xmath114 transversely in a finite number of simple closed curves and arcs .
using the involution @xmath108 , reflect the portion of @xmath113 not in @xmath106 into @xmath106 , forming an immersed surface @xmath115 lying in @xmath106 and with the same boundary as @xmath113 .
the interior of @xmath115 remains disjoint from @xmath3 since @xmath3 is fixed by the involution .
the projection @xmath116 is a surface of genus @xmath0 mapped into @xmath42 with boundary @xmath3 .
so if @xmath3 bounds an embedded surface of genus at most @xmath74 in @xmath4 then it bounds a surface of genus at most @xmath74 mapped into @xmath42 .
the converse was shown in the proof of lemma [ branch ] .
it then follows from lemma [ branch ] that @xmath3 bounds an embedded surface of genus at most @xmath74 in @xmath4 if and only if @xmath1 is satisfiable with each clause containing exactly one true literal .
the construction of @xmath117 and @xmath4 described above each requires only a linear number of steps in the size of the instance of one - in - three sat with which we started , so that the reduction requires polynomial time .
in this section we develop a combinatorial procedure that will allow us to count the components of a normal surface .
the procedure computes the number of orbits of a collection of @xmath118 isometries between subintervals of an interval @xmath119 \subset { { \mathbb{z}}}$ ] in time polynomial in @xmath120 . by lining up the intersections of a normal surface with the edges of a triangulation , we obtain such subintervals .
arcs of the normal surface on the faces of the triangulation give rise to correspondences of these intersection points which are subinterval isometries .
we can then apply the algorithm developed here to count the number of components of a normal surface .
assume that we have a set of integers @xmath121 and a collection of bijections , @xmath122 \to [ c_i , d_i]$ ] , @xmath123 , either increasing or decreasing , that are called _
pairings_. if a pairing identifies two intervals @xmath124 $ ] and @xmath125 $ ] by sending @xmath126 to @xmath6 and @xmath127 to @xmath128 , we call it an _ orientation preserving _ pairing , and if it sends @xmath126 to @xmath128 and @xmath127 to @xmath6 , we call it _ orientation reversing_. if @xmath129 we refer to @xmath130 $ ] as the _ domain _ and @xmath131 $ ] as the _ range _ of the pairing
. we work only with integers , and use the term `` connected interval '' to refer to the integers in a connected real interval .
the _ width _ of an interval @xmath130 $ ] with integer endpoints is @xmath132 , the number of integers it contains .
the _ width _ of a pairing is the width of its domain or range , @xmath133 . if the pairing preserves orientation , its _ translation distance @xmath5 _ measures how far it moves points , so @xmath134 .
we can compose two pairings if the range of the first lies in the domain or range of the second .
the collection of pairings generates a pseudogroup , under the operations of composition where defined , inverses , and restriction to subintervals .
the interval @xmath135 $ ] is divided into equivalence classes by the action of the pairings , which are called _
orbits_. we are interested in the orbit structure of the collection of pairings , since with appropriate interpretation an orbit corresponds to a connected component of a normal surface .
we introduce several simplification processes on the set of pairings in order to analyze the structure of the set of orbits .
we introduce some terminology to describe the behavior of pairings .
an interval is called _
static _ if it is in neither the domain nor the range of any pairing , so that its points are identified to no other points by pairings . given a collection of pairings acting on the integers
@xmath135 $ ] , a pairing is said to be _ maximal _ if its range contains both @xmath106 and the range of any other pairing containing @xmath106 . more precisely , define a linear order on pairings using the lexicographical order @xmath136 , so that the maximal pairing has the highest upper endpoint , and among those with that endpoint the widest range , and among those with that range the biggest translation distance ( if orientable ) .
finally , we say that a pairing @xmath137 \to [ c , d]$ ] is _ periodic _ with period @xmath5 if it is orientation preserving with translation distance @xmath5 and @xmath138 , so there is no gap between the domain and range .
the combined interval @xmath139 $ ] is then called a _ periodic interval _ of period @xmath5 .
the following lemma describes the orbits of a periodic pairing .
a periodic pairing has @xmath140 orbits on @xmath139 $ ] .
[ [ proof.-1 ] ] * proof . * + + + + + + + + each point in @xmath124 $ ] greater or equal to @xmath6 lies in the range of @xmath0 and can be mapped to a smaller point in @xmath124 $ ] by a power of @xmath141 .
so each orbit on @xmath139 $ ] has a representative in @xmath142 $ ] . since the congruence class modulo @xmath5 of a point
is preserved by @xmath0 , each of the @xmath5 integers in @xmath142 $ ] lies in a distinct orbit .
these points uniquely represent the @xmath5 orbits .
we now show how to merge two pairings with sufficient overlap into a single pairing with the same orbits .
[ merge1 ] let @xmath143 be a periodic interval with pairing @xmath144 of period @xmath145 .
suppose that there exists an orientation preserving pairing @xmath146 with translation distance @xmath147 and an interval @xmath148 such that @xmath149 has width @xmath145 and @xmath150 .
then the orbits of @xmath151 on @xmath143 are the same as those of a single periodic action on @xmath143 of period @xmath152 .
[ [ proof.-2 ] ] * proof .
* + + + + + + + + let @xmath149 be an interval in @xmath143 of width @xmath145 that is in the domain of @xmath146 and which is paired by @xmath146 to an interval in @xmath143 .
each point in @xmath143 has a unique orbit representative in @xmath149 under the action of @xmath144 on @xmath143 .
the interval @xmath153 lies in @xmath143 by assumption . for @xmath154 let @xmath155 be the unique point in @xmath149 obtained by carrying @xmath156 back to a point in @xmath149 by a power of @xmath144 or @xmath157 .
the effect of @xmath155 on @xmath149 is a shift of @xmath158 .
the orbits of @xmath159 on @xmath149 divide the points of @xmath149 into congruence classes modulo @xmath160 .
neither @xmath144 nor @xmath146 change the congruence class of a point mod @xmath161 , so a subinterval of width @xmath152 in @xmath143 contains exactly one representative of each orbit of @xmath143 under the action of @xmath151 .
the same orbits arise from a periodic action of period @xmath152 .
the following is a special case of lemma [ merge1 ] that applies when both @xmath144 and @xmath146 are periodic pairings .
[ merge2 ] let @xmath143 and @xmath162 be overlapping periodic intervals , associated with pairings @xmath163 having periods @xmath145 and @xmath147 .
suppose that width(@xmath164 .
then the orbits of @xmath151 on @xmath165 are the same as those of a single periodic pairing on @xmath165 of period @xmath152 .
[ [ proof.-3 ] ] * proof . * + + + + + + + + the leftmost interval @xmath149 of width @xmath145 in @xmath166 is translated by @xmath146 a distance of @xmath147 to the right , to an interval @xmath167 which lies in @xmath166 .
lemma [ merge1 ] then states that the action of @xmath151 on @xmath143 is the same as those of a single periodic pairing on @xmath143 of period @xmath152 . by symmetry
the same result holds for the action of @xmath151 on @xmath162 .
thus the orbits of @xmath151 on @xmath165 are the same as those of a single periodic pairing on @xmath165 of period @xmath152 .
we state a consequence in a form which will be convenient for our applications .
[ merge3 ] let @xmath163 be periodic pairings with periods @xmath168 and let @xmath169 be intervals with @xmath170 and @xmath171 .
suppose that @xmath172 is contained in the union @xmath173 .
then the hypothesis of lemma [ merge2 ] is satisfied and the orbits of @xmath174 are the same as those of a single periodic pairing of period @xmath160 , acting on the union of the periodic intervals of @xmath163 .
[ [ proof.-4 ] ] * proof . *
+ + + + + + + + let @xmath175 , j= [ a , b]$ ] and @xmath176 $ ] .
we have @xmath177 and @xmath178 .
since @xmath179 is a connected interval , we have that @xmath180 .
note that @xmath181 and @xmath182 . then @xmath183 .
the width of @xmath172 is @xmath184 and we have seen above that @xmath185 .
so the width of @xmath172 is at least @xmath186 and lemma [ merge2 ] implies the conclusion of the lemma .
the orbit counting algorithm applies a series of modifications to a collection of pairings .
we now describe these modifications .
* periodic merger + the _ periodic merger _ operation replaces @xmath144 and @xmath146 by a single periodic action on @xmath165 of period @xmath152 , as in lemma [ merge2 ] . *
contraction + the operation of _ contraction _ is performed on a static interval @xmath187 $ ] .
we eliminate this interval , replace @xmath119 $ ] by @xmath188 $ ] , and alter each @xmath189 by replacing any point @xmath190 in a domain or range which lies entirely to the right of @xmath191 by @xmath192 .
( this operation will lead to a decrease of @xmath193 in the number of orbits , since the eliminated points are each unique representatives of an orbit . ) * trimming + the _ trimming _ operation simplifies an orientation reversing pairing whose domain and range overlap .
suppose that @xmath137 \to [ c , d]$ ] is a pairing with @xmath194 and @xmath195 .
define a new pairing @xmath196 $ ] by restricting the domain and range of @xmath0 , and say that @xmath197 is obtained from @xmath0 by _
the domain and range of a trimmed pairing are disjoint .
* truncation + if an interval lies in the domain and range of exactly one pairing , then the interval can be `` peeled off '' without changing the orbit structure , in a way we now describe .
the algorithm applies this operation to strip off points from the right of the interval @xmath119 $ ] .
+ when there is a pairing @xmath137 \to [ c , n]$ ] and a value @xmath198 with @xmath199 , such that all points in the interval @xmath200 $ ] are in the range of no pairing other than @xmath0 , then we can perform an operation called _ truncation _ of @xmath0 .
truncation shortens the interval @xmath201 $ ] to the interval @xmath202 $ ] , and similarly shortens the domain and range of @xmath0 .
if @xmath0 is orientation preserving , pairings other than @xmath0 are unchanged , while @xmath0 is eliminated entirely if @xmath203 , or replaced by a shortened pairing @xmath204 \to [ c , n']$ ] if @xmath205 .
we can perform this operation even if the interval @xmath206 $ ] intersects both the range and the domain of @xmath0 .
+ if @xmath0 is orientation reversing , truncation is applied only when @xmath0 has disjoint domain and range ( i.e. after trimming ) , and @xmath200 $ ] is contained in the range .
suppose @xmath207 \to [ c , n]$ ] is a pairing with @xmath208 and @xmath209 , that @xmath200 $ ] is disjoint from the domains and ranges of all pairings other then @xmath0 and that @xmath210 .
if @xmath211 then we eliminate @xmath0 .
otherwise replace @xmath0 by a shortened orientation reversing pairing , formed by restricting its domain to @xmath212 $ ] and its range to @xmath213 $ ] . *
transmission + transmission is an operation in which two pairings are composed .
a pairing @xmath144 used to shift down the domain and range of a second pairing @xmath146 as much as possible .
this operation will allow us to shift pairings leftwards from the right end of the interval @xmath202 $ ] and subsequently apply truncation .
+ if @xmath144 is orientation reversing and has overlapping domain and range , then as a first step in transmission we trim @xmath144 .
now consider a pairing @xmath144 , either orientation preserving or orientation reversing , and a second pairing @xmath146 whose range is contained in the range of @xmath144 .
if the domain of @xmath146 is not contained in the range of @xmath144 , form the composite map @xmath214 , where @xmath215 if @xmath144 is orientation reversing and otherwise @xmath216 is the largest integer such that @xmath217)$ ] is contained in the range of @xmath144 .
the domain of @xmath218 is the same as that of @xmath146 in this case . if the domain of @xmath146 is also contained in the range of @xmath144 then form the composite map @xmath219 ) \to g_1^{-r}([c_2 , d_2])$ ] , where @xmath220 is as above , @xmath221 if @xmath144 is orientation reversing , and otherwise @xmath222 is the largest integer such that @xmath223 ) $ ] is contained in the range of @xmath144 .
the domain of @xmath218 is then that of @xmath146 shifted left by @xmath224 .
the process of replacing @xmath146 by @xmath225 is called a _ transmission of @xmath146 by @xmath144_. we now construct a sequence of pseudogroups of pairings , terminating with the trivial pseudogroup acting on the empty set . at each stage , each orbit is associated to a unique orbit in the previous set of pairings , though the number of orbits may decrease .
a counter is kept at each stage that records the total decrease in the number of orbits .
since the final pseudogroup has no orbits , the final value of this counter gives the initial number of orbits .
[ orbit.same ] a contraction decreases the number of orbits by the width of the contracted interval . altering a collection of pairings by any of the operations of periodic merger , trimming
, truncation and transmission preserves the number of orbits .
[ [ proof.-5 ] ] * proof . *
+ + + + + + + + we refer to the collection of pairings @xmath226 as @xmath227 and to the new collection of pairings produced by one of the operations as @xmath228 .
the effect of a contraction is to shorten the interval @xmath119 $ ] by removing points which are fixed by the entire collection of pairings .
the number of orbits removed equals to the width of the contracted interval .
a periodic merger joins two periodic intervals @xmath143 and @xmath229 and their pairings into one .
lemma [ merge2 ] shows that the orbits in @xmath165 of @xmath151 are the same as those produced by a single periodic pairing @xmath197 .
suppose that @xmath190 and @xmath230 are points in @xmath119 $ ] in the same orbit of the action of @xmath231 .
then @xmath232 , where @xmath233 is some finite word in the elements of @xmath227 and their inverses .
wherever a @xmath144 or @xmath146 occurs in this word we can replace it by a power of @xmath197 , since @xmath197 translates by an amount that divides the translation distance of @xmath144 and @xmath146 .
so @xmath190 and @xmath230 are in the same orbit of @xmath234 .
conversely suppose that that @xmath190 and @xmath230 are in the same orbit under @xmath228
. then @xmath235 , where @xmath236 is some finite word in @xmath237 .
lemma [ merge2 ] implies that wherever a @xmath197 occurs in @xmath236 we can replace it by some word in @xmath144 and @xmath146 , since the orbits of @xmath197 and @xmath151 coincide on the periodic interval of @xmath197 .
so periodic mergers preserve orbits .
suppose next that @xmath238 \to [ c_i , d_i]$ ] is an orientation reversing pairing with @xmath239 , and that @xmath240 $ ] is obtained by trimming @xmath241 .
if @xmath190 and @xmath230 are in the same orbit of @xmath231 , then there is a sequence of points @xmath242 where each @xmath243 is the image of @xmath244 under a pairing in @xmath227 .
we can replace an occurrence of @xmath241 or @xmath245 by @xmath197 if the point @xmath244 is smaller than @xmath246 and by @xmath247 otherwise .
so @xmath190 and @xmath230 remain in the same orbit under the action of @xmath228 .
conversely if @xmath190 and @xmath230 are in the same orbit under @xmath228 then replacing each occurrence of @xmath197 by @xmath0 and of @xmath247 by @xmath141 gives a word in @xmath227 taking @xmath190 to @xmath230 .
we conclude that trimming preserves orbits .
next consider the effect of a truncation .
suppose that all points in @xmath200 $ ] are in the range of exactly one pairing @xmath137 \to [ c , n]$ ] , and we truncate , shortening to an interval @xmath248 $ ] and either eliminating @xmath0 or shortening it to @xmath204 \to [ c , n']$ ] in the orientation preserving case , or to @xmath249 \to [ c , n']$ ] in the orientation reversing case .
suppose that @xmath250 where @xmath233 is a reduced product of pairings ( a product that does not contain a subproduct of the form @xmath251 ) .
we can assume that @xmath252 , as otherwise we can replace @xmath233 by its inverse .
let @xmath236 be obtained from @xmath233 by replacing all occurrences of @xmath0 with @xmath197 .
the successive images of @xmath190 under the subwords of @xmath233 can never enter and leave the interval @xmath200 $ ] , since they can only enter it under the action of some positive power of @xmath0 , and only leave it under a negative power of @xmath0 .
these do nt occur in succession in a reduced product .
the image of a point @xmath253 under @xmath189 is unchanged unless @xmath254 , and @xmath255 .
so the image of @xmath190 under @xmath233 is the same as its image under @xmath236 unless @xmath256 . in that case
@xmath233 is of the form @xmath257 where @xmath258 , @xmath216 , and @xmath259 .
so two points in @xmath248 $ ] are in the same orbit under @xmath228 if and only if they are in the same orbit under @xmath227 .
but each point in @xmath200 $ ] is in the same @xmath227 orbit as a point in @xmath202 $ ] .
it follows that the number of orbits is unchanged by truncation .
finally suppose that a transmission of @xmath189 by @xmath241 replaces the pairing @xmath189 in @xmath227 by @xmath260 .
if @xmath250 where @xmath233 is a word in @xmath227 , then @xmath261 where @xmath236 is obtained by replacing every occurrence of @xmath189 with @xmath262 . similarly if @xmath263
where @xmath236 is a word in @xmath228 , then @xmath250 where @xmath233 is obtained by replacing every occurrence of @xmath264 with @xmath265 .
so transmissions preserve the collection of orbits and do nt change their number .
we now describe an algorithm which uses these operations to count the number of orbits of a pseudogroup of pairings acting on @xmath119 $ ] .
we initially set a counter for the number of orbits to zero .
a pairing @xmath241 is said to be _ maximal _ if @xmath266 and @xmath267 $ ] contains the range of any other pairing with an endpoint at @xmath106 .
let @xmath268 and let @xmath122 \to [ c_i , d_i],\ 1 \le i \le k$ ] be a collection of pairings between subintervals of @xmath17 .
the algorithm will contract @xmath17 and reduce the number of pairings .
it keeps a running count of the number of orbits detected in an integer referred to as the orbit counter .
denote by @xmath198 the current size of the interval as we proceed .
the algorithm repeats the following steps , reducing @xmath198 and @xmath118 , until there are no points remaining . + * orbit counting algorithm : * 1 .
delete any pairings that are restrictions of the identity .
2 . make any possible contractions and , if any exist , increment the orbit counter by the sum of the number of points deleted by the contractions .
if the number of pairings remaining is zero , output the number of orbits and stop .
3 . trim all orientation reversing pairings whose domain and range overlap .
4 . search for pairs of periodic pairings @xmath241 and @xmath189 whose domains and ranges satisfy the condition of lemma [ merge2 ] . if any such pair exists then perform a merger as in lemma [ merge2 ] , replacing @xmath241 and @xmath189 by a single periodic pairing , with translation distance @xmath269 .
the new pairing acts on the union of the domains and ranges of @xmath241 and @xmath189 .
repeat until no mergers can be performed .
find a maximal @xmath241 .
for each @xmath270 whose range is contained in @xmath271 $ ] , transmit @xmath189 by @xmath241 .
6 . find the smallest value of @xmath6 such that the interval @xmath213 $ ] intersects the range of exactly one pairing .
truncate the pairing whose range contains the interval @xmath213 $ ] .
[ orbitcount ] the orbit counting algorithm gives the number of orbits of the action of the pairings @xmath272 on @xmath273 $ ] in time bounded by a polynomial in @xmath274 .
[ [ proof.-6 ] ] * proof . *
+ + + + + + + + we first check that the orbit counting algorithm correctly counts the number of orbits . in step
( 1 ) , deleting a pairing which is the identity on its domain does not change the number of orbits . in step ( 2 ) , contracting a static interval removes a number of points that are unique orbit representatives , and the count of these is added to the running total kept in the orbit counter . if all the points have been removed , then there are no more orbits to count and the orbit count is complete .
lemma [ orbit.same ] shows that the operations of transmission , trimming , truncation and merger occurring in steps ( 3),(5 ) and ( 6 ) do not change the number of orbits of the collection of pairings .
mergers carried out in step ( 4 ) occur when the conditions of lemma [ merge2 ] are satisfied , and lemma [ merge2 ] shows that they preserve the orbit structure of the collection of pairings .
a merger reduces the number of pairings by one , replacing two pairings @xmath241 and @xmath189 by a single periodic pairing acting on the union of their periodic intervals . in each cycle through these steps the interval width decreases by at least one , either in step ( 2 ) or in step(6 ) .
it follows that the algorithm terminates , yielding a count of the number of orbits after a number of steps bounded by the width @xmath106 of the interval .
we will obtain a much better bound .
to do so , we define a complexity which decreases as we iterate the above steps .
recall that @xmath275 and @xmath118 is the number of pairings .
the complexity @xmath84 is defined to be @xmath276 the process of executing steps ( 1)-(7 ) in turn is called a _
cycle_. we will show that when we run through @xmath277 cycles , @xmath84 is reduced by a factor of at least two . see the remark at the end of this section for a geometric interpretation of this complexity .
call an interval @xmath278 $ ] a _ @xmath279-close _
interval if it is the domain or range of a pairing of subintervals of @xmath280 $ ] and if @xmath281 , or equivalently , if @xmath282 . being @xmath279-close corresponds to being relatively close to @xmath279 .
it means that there is no room for another interval of the same size to the right of @xmath230 .
the value of @xmath279 is initially set to @xmath106 , and as the algorithm proceeds it is reset to the current interval width @xmath198 each time the number of pairings @xmath118 is decreased .
the number of pairings decreases when a merger occurs , or when a pairing is truncated to zero width , or when a pairing is transmitted to become trivial ( a restriction of the identity ) and then eliminated .
[ overlap ] the union of two @xmath279-close intervals of equal width is a connected interval .
[ [ proof.-7 ] ] * proof . *
+ + + + + + + + suppose that @xmath130 $ ] and @xmath131 $ ] are equal width @xmath279-close intervals with @xmath283 .
then @xmath284 , @xmath285 and @xmath286 , so @xmath287 implying that @xmath288 and @xmath130 \cup [ c , d ] $ ] is connected .
[ transmit ] suppose that the domain @xmath130 \subset [ 1,z]$ ] of a pairing @xmath0 is not @xmath279-close . then the image @xmath289 $ ] of an interval @xmath278 $ ] under @xmath290 is not @xmath279-close .
[ [ proof.-8 ] ] * proof . *
+ + + + + + + + suppose that @xmath289 $ ] is the image of @xmath278 $ ] under @xmath290 , where @xmath291 \to [ c , d]$ ] .
by assumption we have that @xmath292 .
the interval @xmath289 $ ] is contained in @xmath130 $ ] so @xmath293 .
then @xmath294 and @xmath289 $ ] is not @xmath279-close .
[ 5pairs ] after a series of five cycles either the number of @xmath279-close intervals decreases or the number of pairings decreases . [ [ proof.-9 ] ] * proof .
* + + + + + + + + in each cycle at least one maximal pairing is truncated .
suppose that during a series of five cycles , five successive maximal pairings @xmath295 occur in turn in the orbit counting algorithm .
the initial maximal pairing @xmath144 is truncated in the first cycle .
eventually @xmath296 stops being maximal , and it is then transmitted by the new maximal pairing @xmath297 in the next cycle .
the cycle @xmath146 is then itself truncated until no longer maximal and it in turn is then transmitted by @xmath298 and so on .
when @xmath299 transmits @xmath300 , the pairing of @xmath301 $ ] with @xmath302 $ ] is replaced by a pairing of @xmath303 $ ] with @xmath304 $ ] and one of the following three cases occurs . 1 .
the range @xmath305 $ ] of @xmath241 is transmitted to a non @xmath279-close interval by @xmath299 .
the range @xmath305 $ ] is transmitted to a @xmath279-close interval @xmath306 $ ] by @xmath299 , and the domain @xmath307 $ ] of the transmitted pairing is @xmath279-close .
the range @xmath305 $ ] is transmitted to a @xmath279-close interval @xmath306 $ ] by @xmath299 , and the domain @xmath307 $ ] of the transmitted pairing is not @xmath279-close .
we first consider the case where each of @xmath308 is orientation preserving .
setting @xmath279 to the initial interval width @xmath198 and noting that @xmath296 is initially maximal , we see that the range of @xmath144 is initially @xmath279-close , with @xmath309 .
truncation reduces the range to of @xmath144 to @xmath310 $ ] , at which point the pairing @xmath297 becomes maximal . we can assume the range @xmath310 $ ] of @xmath144 is still @xmath279-close , or we are done . in the next cycle
the interval @xmath310 $ ] is transmitted by @xmath297 to an interval @xmath311 $ ] .
if case ( 1 ) applies , this new interval is not @xmath279-close and the number of @xmath279-close intervals has decreased .
if case ( 2 ) , the three intervals @xmath312 , [ c_1 ' , d_1']$ ] and @xmath313 $ ] are all @xmath279-close and of equal width , and therefore satisfy the hypothesis of claim [ overlap ] .
lemma [ merge3 ] then implies that two pairings can be merged , and the number of pairings decreases during the execution of step ( 4 ) in the next cycle .
if case ( 3 ) occurs , the maximal pairing @xmath297 is truncated in the second cycle .
we can assume its domain remains @xmath279-close , or we are done .
when the next pairing @xmath298 becomes maximal and transmits @xmath146 in the third cycle , we must fall into one of the first two cases .
so the number of @xmath279-close intervals is decreased in the third cycle or the number of pairings decreases in a merger during a subsequent application of step ( 4 ) in the fourth cycle .
now we consider the possibility that one of @xmath308 is orientation reversing .
the domain and range of a trimmed orientation reversing pairing are disjoint , and hence any interval transmitted by an orientation reversing pairing is not @xmath279-close . if any of @xmath314 are orientation reversing then it transmits the previously maximal pairing s range to an interval that is not @xmath279-close and the number of @xmath279-close pairings decreases .
if non is orientation reversing then there is a sequence of three successive orientation preserving pairings and the previous argument applies .
[ div2 ] suppose that @xmath279 is set to the current interval size @xmath198 and that after a series of truncations in which the maximal pairings are successively @xmath315 , no @xmath279-close intervals remain .
then the complexity @xmath84 is reduced by a factor of at least two .
[ [ proof.-10 ] ] * proof . * + + + + + + + + truncation of a maximal pairing @xmath241 results in its range @xmath316 $ ] , of width @xmath317 , being reduced to a shorter interval @xmath318 $ ] of width @xmath319 .
this reduces the complexity @xmath84 by a factor of @xmath320 the maximal pairing changes after a truncation if the range of @xmath321 has truncated sufficiently so that it is contained in the range of @xmath299 , so that @xmath322 and @xmath323 .
define @xmath324 by @xmath325 and define @xmath326 differentiation shows that the function @xmath327 is increasing with @xmath190 for @xmath328 , since @xmath329 .
so @xmath330 after a series of truncations of the pairings @xmath331 , the complexity @xmath84 is multiplied by a factor of @xmath332 now @xmath333 , since @xmath144 was the first maximal pairing truncated , so @xmath334 . by assumption , there are no @xmath279-close intervals remaining after @xmath335 is truncated , so that @xmath336 $ ] is not @xmath279-close .
we then have by the definition of @xmath279-close that @xmath337 it follows that @xmath84 is multiplied by a factor smaller than 1/2 .
* proof of theorem [ orbitcount ] : * the operations involved in one cycle consist of comparisons , additions , subtractions and computing greatest common divisors of the @xmath118 pairings of @xmath227 .
the number of these operations occurring in each cycle is linear in @xmath338 .
pairings are described by pairs of intervals , whose boundary points are integers of size at most @xmath106 .
so the running time of each cycle is polynomial in @xmath339 .
if the number of pairings @xmath118 decreases during a cycle , then a pairing has been eliminated because its width has truncated to zero , because it has been transmitted and become the identity on its domain , or because two pairings have merged .
when two pairings @xmath163 merge , the value of @xmath118 decreases by one . the product @xmath340 which occurs in @xmath84
is replaced by the width of the new pairing , which is at most @xmath341 .
therefore @xmath342 decreases by a factor of at least two after a merger .
since each pairing has width at most @xmath106 , the initial complexity is bounded above by @xmath343 . by claim [ 5pairs ] , each time we run through five cycles either the number of pairings decreases or the number of @xmath279-close intervals decreases .
there are at most @xmath118 @xmath279-close intervals initially , so after @xmath344 cycles either there is a reduction in the number of pairings or there are no @xmath279-close intervals remaining . in the second case
the complexity @xmath84 has decreased by a factor of at least two , by claim [ div2 ] .
so @xmath84 decreases by a factor of at least two after @xmath344 cycles , and the complexity reduces to zero after at most @xmath345 successive series of @xmath344 cycles , or after at most @xmath346 cycles .
each cycle runs in time polynomial in @xmath120 , so the total running time is also polynomial in @xmath120 .
we now apply theorem [ orbitcount ] to count the number of components of a normal curve or normal surface .
an obvious algorithm to count components proceeds by marking vertices connected by common edges until all vertices in a component are reached .
this procedure takes time linear in the number of edges of the curve .
this is equal to the sum of the normal coordinates ( called @xmath347 below ) , but we can achieve an exponential improvement .
we first look at normal curves .
[ normalcurvecount ] let @xmath113 be a surface with a triangulation @xmath348 containing @xmath5 triangles and let @xmath349 be a normal curve in @xmath113 with normal coordinates summing to @xmath347 .
there is a procedure for counting the number of components of @xmath349 that runs in time polynomial in @xmath350 .
[ [ proof.-11 ] ] * proof . *
+ + + + + + + + the 1-skeleton of @xmath348 contains @xmath351 edges , where @xmath352 .
fix once and for all an ordering of these edges .
a normal curve @xmath349 intersects each edge of @xmath348 in a finite number of points .
set @xmath106 to be the weight @xmath347 of @xmath349 , the sum of the number of intersection points of @xmath349 with all the edges of @xmath348 .
label the intersections of @xmath349 and the first edge of the 1-skeleton by the integers @xmath353 , the intersections of @xmath349 and the @xmath354th edge by @xmath355 , and the intersection of @xmath349 and the @xmath356 edge of @xmath348 by @xmath357 . then @xmath358 .
each triangular face of @xmath348 has three sets of arcs pairing points of @xmath119 $ ] , with one set running between each pair of edges of the face .
to each set of arcs we associate a pairing between the intervals at either end of the arcs , as in figure [ pairings ] .
all the pairings are orientation reversing in this example . in general
some will be orientation preserving , as the edge orientations on any triangle can be arbitrary .
normal arcs on a triangular face give three pairings of ordered intervals : @xmath359 \to [ n_3 , n_3+r_{12}-1]$ ] , @xmath360 \to [ n_5 , n_5+r_{23}-1]$ ] , and @xmath361 \to [ n_1 , n_1+r_{13}-1]$ ] the normal curve coordinates associated to this face are @xmath362 and @xmath363 . in a normal surface , @xmath362 and @xmath363
are each a sum of two normal coordinates . ]
the number of connected components of @xmath349 is the same as the number of orbits of an action of a collection of pairings on @xmath364 , where two points are paired if they are connected by an edge of @xmath365 lying in a triangle .
this is precisely the number returned by the orbit counting algorithm .
the number of pairings is at most @xmath366 and the interval size is @xmath347 . applying theorem [ orbitcount ]
, we can determine the number of components of the normal curve @xmath349 in time polynomial in @xmath350 .
a similar argument applies to normal surfaces .
[ normalcount ] let @xmath4 be a 3-manifold with a triangulation @xmath348 containing @xmath5 tetrahedra and let @xmath113 be a normal surface in @xmath4 whose normal coordinates sum to @xmath347 .
there is a procedure for counting the number of components of @xmath113 which runs in time polynomial in @xmath350 .
[ [ proof.-12 ] ] * proof . *
+ + + + + + + + the 2-skeleton of @xmath348 contains at most @xmath367 faces and the 1-skeleton contains @xmath368 edges .
set @xmath106 to equal the weight @xmath347 of @xmath113 , the total number of points in which it intersects the 1-skeleton . order the edges of @xmath369 in an arbitrary way and label the intersections of @xmath113 and the @xmath354th edge of @xmath369 by @xmath355 .
again we have @xmath358 .
to a pair of edges on a triangular face of the 2-skeleton we associate a pairing between the intervals at either end of the corresponding set of arcs .
there are at most three pairings for each face , and the number of faces is at most @xmath367 , so the number of pairings is bounded above by @xmath370 .
these pairings are determined by the normal coordinates of @xmath113 . for a normal surface ,
the number of connected components of @xmath113 is the same as the number of components of @xmath371 , since every component of @xmath113 intersects the 1-skeleton of @xmath348 , and if two points on @xmath372 can be joined by a path then that path can be homotoped into the 2-skeleton of the triangulation .
so the the number of components equals the number of orbits of @xmath373 under pairings that identify two points connected by an edge of @xmath371 contained in a face of the 2-skeleton .
this number is precisely what is computed by the orbit counting algorithm .
therefore we can determine the number of components of the normal surface @xmath113 in time polynomial in @xmath350 .
* remark : * there is a motivating geometrical construction behind the combinatorics of the orbit counting algorithm which we informally explain
. we can associate to each pairing @xmath122 \to [ c_i , d_i]$ ] a pair of `` transmission towers '' in the upper half - plane , with one tower being the vertical line segment from @xmath374 to @xmath375 and the other the vertical segment from @xmath376 to @xmath377 .
these towers capture the information contained in the pairings .
points in the domain and range of a pairing beam up to the tower leftward at a 45 degree angle , then beam to the paired tower ( either straight across , if the pairing preserves orientation , or crossing if not ) then down again .
assign a `` cost '' to each transmission tower equal to the hyperbolic length in the upper half space model from @xmath378 to the top of the tower .
this cost equals @xmath379 , and the sum of all these costs is essentially the logarithm of the complexity @xmath84 used above .
the counting algorithm starts with a euclidean line emerging from the right endpoint of a @xmath279-close interval and going upward at a 45 degree angle .
in hyperbolic terms , this is an equidistant curve from the geodesic from the right endpoint to the point at infinity .
initially , we may assume that this sweep line hits the top of at least one transmission tower .
we make the highest of these the @xmath279-close tower , and use it to beam all the other towers to its paired interval .
if the two domains overlap , we use the highest power of the transmission that can be applied .
eventually the triangle between the current tower and the sweep line is vacant .
we sweep leftward with this equidistant curve .
as long as the sweep line hits only one tower , we truncate it .
consider a second equidistant curve @xmath380 through a point @xmath190 on @xmath381 , with slope -1/2 .
the hyperbolic distance between the slope -1 and slope -1/2 lines is a constant equal to about 0.49 .
call the region between these two lines a _
zone_. if two paired towers have their tops in a single zone , their domains overlap . if no towers are completely truncated away between the time when the sweep line hits a point @xmath190 and when there are no towers intersecting the zone above @xmath380 , each equidistant curve between the two given ones hits a transmission tower based at a point that will be removed by truncation .
so in this time either the number of towers has decreased or the sum of all the tower costs has decreased by at least 0.49 .
if we merge transmission towers when possible , a complexity based on the cost decreases sufficiently fast to give a polynomial time algorithm .
the calculations in theorem [ orbitcount ] implemented this geometric picture .
in this section we establish that 3-manifold knot genus is * np*. * proof of theorem [ thmgnp ] : * we begin with a simplicial complex consisting of @xmath5 tetrahedra whose faces are identified in pairs , and a collection @xmath3 of edges in the 1-skeleton of this complex . while there are alternate formats in which a knot and a 3-manifold may be presented , all reasonable ones appear to be transformable to one another in polynomial time . in time
polynomial in @xmath5 we can check that the link of each vertex is connected and has euler characteristic two , which means that it is a sphere , and that the link of each edge is a connected curve .
these are the necessary conditions to ensure that the underlying space of the complex is a 3-manifold @xmath4 .
similarly we can check in time polynomial in @xmath5 that the edges of @xmath3 form a simple closed curve in @xmath4 , and that this curve represents a trivial element of the first homology group of @xmath4 with integer coefficients .
we then form the second barycentric subdivision of the triangulation of @xmath4 , replacing each tetrahedron by 576 tetrahedra . removing all closed tetrahedra that meet @xmath3 results in a 3-manifold @xmath382 with a single torus boundary component , the `` peripheral torus '' that surrounds the knot @xmath3 .
the knot @xmath3 bounds a surface of genus @xmath0 in @xmath4 if and only if there is a surface in @xmath382 of genus @xmath0 with a single boundary component that is an essential curve on @xmath383 .
we restate our problem in the triangulated manifold @xmath382 as the question of whether there exists in @xmath382 an orientable surface of genus @xmath0 with a single essential boundary component on @xmath383 . by schubert
@xcite , if such a surface exists , then such a surface exists among the fundamental normal surfaces in @xmath382 .
the certificate consists of an integer vector @xmath384 in @xmath385 giving the normal surface coordinates of this surface .
recall that not all vectors in @xmath385 correspond to normal surfaces .
it is necessary that @xmath384 satisfies the matching , positivity and quadrilateral conditions .
these conditions can be checked in time which is linear in @xmath386 .
so we can verify that there is a normal surface @xmath113 with @xmath387 . to verify that @xmath388 is essential on the peripheral torus , we include in the certificate a non - trivial cycle in the 1-skeleton of @xmath383 that intersects @xmath388 in an odd number of points .
such a cycle can be found in the 1-skeleton of @xmath383 , since curves embedded in the 1-skeleton generate its first homology .
odd intersections with such a cycle implies that @xmath388 is non - separating on @xmath383 , and in particular does not bound a disk on @xmath383 , ensuring that @xmath388 corresponds to a longitude curve parallel to @xmath3 in @xmath389 . using the orbit counting algorithm
, we can count the number of components of the normal surface @xmath113 and verify that @xmath113 is connected in time polynomial in @xmath350 , where @xmath347 is the weight of @xmath113 the number of points in which @xmath113 meets the 1-skeleton of @xmath382 .
we apply corollary [ normalcount ] to verify that @xmath388 is connected in time bounded by a polynomial in @xmath350 .
( this last step can be avoided .
the number of essential boundary components must be odd , since @xmath388 represents a non - trivial element in @xmath390 , and an even number of them can be removed by joining adjacent pairs of curves with annuli on @xmath383 .
this gives a surface with one boundary component and the same euler characteristic .
inessential boundary curves can be capped while increasing euler characteristic , which gives lower genus . )
to check that @xmath113 is orientable , we take the vector @xmath391 that doubles each coordinate of the normal surface @xmath113 , and apply the orbit counting algorithm to determine if the corresponding normal surface @xmath392 is connected .
since @xmath4 is orientable , @xmath392 is connected if @xmath113 is connected and non - orientable , and has two components if @xmath113 is connected and orientable .
thus we can verify if @xmath113 is orientable in time which is linear in @xmath350 .
the euler characteristic @xmath393 is determined by the number of vertices , edges and faces of @xmath113 , which are computable from its normal coordinates in time which is linear in @xmath5 .
following @xcite , we let @xmath394 be the number of tetrahedra containing edge @xmath395 and set @xmath396 if the edge @xmath395 meets the @xmath69 normal disk . then
the normal surface @xmath113 with coordinates @xmath397 has @xmath398 where @xmath399 and @xmath400 .
the values of @xmath401 are determined by the triangulation and are independent of @xmath113 .
they can be computed in time polynomial in @xmath5 . since @xmath113 is a connected orientable surface with one boundary component the genus of @xmath113
is @xmath402 .
thus we can determine the genus of @xmath113 in time which is linear in @xmath386 .
theorem [ bound ] implies that the normal coordinates of this surface are at most @xmath403 .
there are @xmath404 normal coordinates , and each represents a triangle or quadrilateral , so that the total number of intersections with the 1-skeleton satisfies @xmath405 .
in particular , @xmath406 is bounded above by a polynomial in @xmath5 . so the fact that @xmath113 is a spanning surface for @xmath3 can be verified in time polynomial in @xmath5 .
in this section we develop a generalized version of the orbit counting algorithm , that counts not only the number of orbits of a collection of isometries between subintervals of an interval , but also more general quantities which are useful in applications .
for example , we can use the extended algorithm to answer the following question : given a normal surface and a triangulation , how many times does each component of the surface intersect a fixed edged of the triangulation ?
the extended algorithm allows one to effectively compute the normal coordinates and the euler characteristic of each connected component of the surface , hence the genus , even when there are exponentially many components . to carry out such computations ,
we extend the previous analysis to pairings of weighted intervals , in which each point of the interval has associated to it a vector in @xmath407 .
we are interested in the sum of these vectors over an orbit , the orbit weights .
consider again a pseudogroup of interval isometries acting on @xmath408 $ ] .
we will assume that there is given as input a nonnegative weight function @xmath409 \to { { \mathbb{z}}}_+d $ ] , associating to each element of @xmath408 $ ] a vector in @xmath410 satisfying the following condition : the weight at successive points @xmath354 and @xmath411 changes at most @xmath412 times . in our application , @xmath118 will be the initial number of pairings , and @xmath412 gives an upper bound on how many times an endpoint of a domain or range of a pairing is reached as one moves across @xmath408 $ ] .
the algorithm proceeds as before , while maintaining data on the orbit weights .
we keep track of the orbit weights by maintaining two lists of weighted subintervals .
the first @xmath413,z_1),\dots,([p_m , q_m],z_m)\},\ 1 \le p_i \le q_i < p_{i+1}\ ] ] records the current weight values at each point in @xmath119 $ ] , with @xmath414 initially equal to @xmath106 .
this list is updated as the algorithm proceeds .
points in the interval @xmath415 $ ] have constant weight @xmath416 , and there are at most @xmath412 such intervals .
a second list of @xmath5 subintervals @xmath417,v_1),\dots,([r_t , s_t],v_t)\ } , \ 1 \le r_i \le s_i < r_{i+1}\ ] ] consists of a collection of intervals @xmath418 $ ] paired with a vector @xmath419 .
this pair represents @xmath420 orbits , one for each point in the interval @xmath418 $ ] , and to each of these orbits is assigned the orbit weight @xmath421 .
initially empty , at the algorithm s conclusion @xmath422 records the total number of orbits @xmath423 , along with the orbit weight @xmath421 assigned to each of the @xmath420 orbits in the interval @xmath424 $ ] .
we define an additional operation , called _ transferring weights by a pairing @xmath0_. suppose that @xmath137\to [ c , d]$ ] is a pairing and that @xmath425 $ ] carries @xmath97 different weights , given by the list @xmath426,v_1),\dots,([r_n , s_n = d],v_n)\ } , \
1 \le r_i \le s_i < r_{i+1 } .\ ] ] the weight function can be split into constant functions on @xmath97 subintervals of @xmath425 $ ] , where @xmath427 .
the transfer operation sets the weights on @xmath425 $ ] to zero and keeps the orbit weights the same by translating the weight vectors of @xmath425 $ ] to smaller orbit representatives , as below : case 1 : @xmath0 is orientation preserving and @xmath209 .
set the weights on @xmath425 $ ] to zero and for each @xmath428 , add @xmath429 to @xmath430)$ ] .
case 2 : @xmath0 is orientation preserving and @xmath431 .
then @xmath137 \to [ c , d]$ ] is a periodic pairing of period @xmath432 .
we set the weights of points in @xmath433 $ ] to zero and adjust weights in @xmath434 $ ] to preserve the orbit weight .
the points in @xmath433 $ ] have weights given by the intersection of @xmath433 $ ] with intervals in @xmath435 .
these weights are described by @xmath426,v_1),\dots,([r_n , s_n = d],v_n)\ } , \ 1 \le r_i \le s_i < r_{i+1 } .\ ] ] for each interval @xmath436,v_j),\ 1 \le j \le n$ ] , of width @xmath437 and constant weight @xmath429 , add @xmath438v_j $ ] to the weight of each point in @xmath439 $ ] , and add an additional @xmath429 to the weight of each point in @xmath439 $ ] that is congruent mod@xmath440 to a point in @xmath441t , s_j]$ ] , if any such point exists .
case 3 : @xmath0 is orientation reversing .
we first trim @xmath0 .
this does not affect the orbits or the orbit weights .
for a trimmed , orientation reversing pairing @xmath137 \to [ c , d]$ ] , set the weights on @xmath425 $ ] to zero and for each @xmath428 add @xmath429 to @xmath430)$ ] .
[ orbitweights.same ] the operation of transfer sets the weights on @xmath433 $ ] to zero and preserves the orbit weights of a collection of pairings . the number of distinct weights taken by the weight function on @xmath119 $ ] increases by at most four following a transfer operation .
[ [ proof.-13 ] ] * proof . * + + + + + + + + in cases ( 1 ) and ( 3 ) , the decrease in the weight function at one point @xmath190 in an orbit is exactly offset by an equal increase at the point @xmath442 in the same orbit . in case ( 2 ) , each point in @xmath443 $ ] is in the orbit of a unique point in @xmath444 = [ a , c-1]$ ] under the iterates of @xmath0 . the total weight of an orbit within a periodic pairing is transferred to the orbit representative in this initial subinterval by applying powers of @xmath0 .
adding weight @xmath429 to an orbit representative in @xmath439 $ ] of each point in @xmath445 while setting the weight at that point to zero , preserves the orbit weight .
the resulting weights on @xmath439 $ ] are gotten by adding appropriate multiples of @xmath429 , with the factor being the number of orbits of a point in @xmath439 $ ] that lie in @xmath446 $ ] . the number of orbits under @xmath0 of a point in @xmath434 $ ] that lie in @xmath446 $ ] is @xmath447 $ ] or one more than this for points
whose orbit hits the last @xmath448 points of @xmath446 $ ] .
it follows that the transfer operation in case ( 2 ) preserves the number of orbits and the orbit weight .
setting the weights on @xmath131 $ ] to zero can cause at most two new points where a weight change occurs .
the transferred weights from a constant weight interval @xmath449 $ ] result in a net increase of at most two pairs of successive points where the weight changes , at the preimages of its two endpoints . in
cases ( 1 ) and ( 3 ) , transferred weights from an interval with non - constant weights results in an increase in the number of weight changes in the domain of @xmath0 .
the increase in the number of weight changes in the domain is exactly canceled by the decrease in the number of weight changes in the range of @xmath0 , except possibly for two extra weight changes at the boundary points @xmath450 of the domain . in case
( 2 ) the same holds , but with @xmath439 $ ] replacing the domain . in each case
the number of constant weight intervals @xmath451 is increased by at most four during a transfer operation .
we now describe the modified algorithm .
again @xmath198 represents the current interval length , and we set @xmath451 to be the current number of constant weight intervals .
+ * weighted orbit counting algorithm : * let @xmath452 \to [ c_i , d_i],\ 1 \le i \le k \}$ ] be a collection of pairings between subintervals of @xmath453 , and let @xmath413,z_1),\dots,([p_m , q_m],z_m)\ } , \
1 \le p_i \le q_i < p_{i+1 } \le n\ ] ] be a list representing a collection of weights on @xmath119 $ ] , with the weight on @xmath454 $ ] equal to @xmath455 .
initialize a second weight list @xmath422 to be empty .
the algorithm proceeds as before , reducing the interval size @xmath106 until it reaches zero , but this time keeping track of orbit weights by maintaining the lists @xmath456 . 1 .
search through the pairings and delete any pairings which are restrictions of the identity . leave the weight lists @xmath456 unchanged .
2 . search for and contract static intervals .
if the interval @xmath187 $ ] is contracted , has constant weight @xmath253 , and is not contained in a larger contracted interval with the same weight @xmath253 , add an interval @xmath457 , z)$ ] of width @xmath193 to the end of @xmath422 , with associated weight @xmath253 . alter @xmath435 by replacing @xmath202 $ ] by @xmath458 $ ] , and altering each @xmath241 by replacing any point @xmath190 in the domain or range of @xmath241 with @xmath459 by @xmath192 .
replace the weight function by a new weight function @xmath460 , which at points @xmath459 satisfies @xmath461 and agrees with @xmath462 at points @xmath463 .
3 . trim all orientation reversing pairings whose domain and range overlap . leave the weight lists @xmath456 unchanged .
4 . search for pairs of periodic pairings @xmath241 and @xmath189 whose domains and ranges satisfy the condition of lemma [ merge2 ] . if any such pair exists then perform a merger as in lemma [ merge2 ] , replacing @xmath241 and @xmath189 by a single periodic pairing , with translation distance @xmath269 , acting on the union of the domains and ranges of @xmath241 and @xmath189 .
leave the weight lists @xmath456 unchanged .
repeat until not mergers can be performed .
5 . find a maximal @xmath241 .
for each @xmath189 with @xmath464 , if the range of @xmath189 is contained in @xmath271 $ ] , transmit @xmath189 by @xmath241 .
leave the weight lists @xmath456 unchanged .
6 . find the smallest value of @xmath6 such that the interval @xmath213 $ ] intersects the range of at most one pairing @xmath241 , with @xmath122 \to [ c_i , n']$ ] .
transfer the weights on @xmath271 $ ] by @xmath241 , and then truncate the pairing @xmath241 .
if the interval size @xmath198 has decreased to zero , output the list @xmath422 and stop .
otherwise start again with step ( 1 ) . for a @xmath465-valued function on
@xmath466 $ ] whose values are given by the list @xmath467,z_1),\dots,([p_m , q_m],z_m)\ } , \
1 \le p_i \le q_i < p_{i+1}$ ] , define the _ total weight _ of @xmath435 to be @xmath468 .
[ weightedcount ] suppose there is a pseudogroup generated by @xmath118 pairings with @xmath465-valued weights , @xmath272 on @xmath119 $ ] such that there is a partition of @xmath119 $ ] into @xmath101 disjoint subintervals in which the weights are constant , and such that the total weight is at most @xmath469 .
then the weighted orbit counting algorithm outputs a list with one point for each orbit and corresponding orbit weights , and runs in time polynomial in @xmath470 .
[ [ proof.-14 ] ] * proof . *
+ + + + + + + + we will check that the running time of the algorithm is larger than that of the previous unweighted version by a factor which is a polynomial in @xmath471 .
the proof that the algorithm terminates is the same as that given for theorem [ orbitcount ] .
there is some extra overhead involved in keeping track of weights that modifies the calculation of the running time .
we indicate these additional calculations below .
we now check that at each step in the algorithm the orbit weight is unchanged for any orbit remaining in @xmath435 , and that eliminated orbits have their orbit weights correctly recorded in @xmath422 . as we run through the steps of the algorithm , steps ( 1),(3),(4),(5 ) and ( 7 )
preserve the orbit structure , the weight function and the interval @xmath202 $ ] , so neither of the lists @xmath456 are changed .
the number of constant weight sub - intervals is also unchanged .
step ( 2 ) , contraction , does change the orbit structure , and also shortens @xmath202 $ ] . in step ( 2 )
the procedure adds the eliminated orbits and their weights to @xmath422 .
the number of constant weight intervals is not increased , and may be decreased .
maintaining the two lists requires at most @xmath472 additional steps . in step ( 6 ) , truncation
, points are eliminated from the end of the interval @xmath202 $ ] .
however since we first transfer the weights of these points , the eliminated points all have weight zero and the weight of an orbit is unaffected .
the number of steps involved in resetting the weights in @xmath435 for a transfer operation is given by a polynomial in @xmath473 .
since @xmath451 increases by at most four at each of the polynomially many steps of the algorithm , @xmath451 is bounded by a polynomial in @xmath474 .
combining the running time of each of the steps , whose number is given by a polynomial in @xmath475 , gives a polynomial in @xmath470 for the total running time .
[ eulercount ] let @xmath4 be a 3-manifold with a triangulation @xmath348 containing @xmath5 tetrahedra and let @xmath113 be a normal surface in @xmath4 of total weight @xmath347 .
there is a procedure for counting the number of components of @xmath113 and determining the topology of each component which runs in time polynomial in @xmath350 .
[ [ proof.-15 ] ] * proof . *
+ + + + + + + + we begin as in corollary [ normalcount ] by assigning an integer in @xmath119 $ ] to each point of intersection between the normal surface and an edge of the triangulation , where @xmath106 is the total number of intersections of @xmath113 with the 1-skeleton , and again associate three pairings to each face of the triangulation , one to each pair of edges in the face .
the number of pairings that results is bounded above by @xmath370 .
we next define a weight function @xmath476 which assigns integer weights @xmath477 to each point in @xmath119 $ ] .
initially @xmath478 is set to zero for all @xmath63 at all points @xmath479 $ ] .
a tetrahedron can have as many as five distinct elementary disk types with non - zero coefficients , four triangles and one quadrilateral .
if the @xmath69 elementary disk type occurs , then fix one of the edges that it meets , and add @xmath480 to the @xmath69 component of the weight vector at each of the indices that the @xmath69 elementary disk meets on that edge .
the orbit weights are then the normal coordinates of the components of the normal surface @xmath113 .
each point in the output list @xmath422 corresponds to a component of the normal surface with normal coordinates given by the corresponding weight in @xmath481 .
theorem [ weightedcount ] tells us that the list @xmath422 is computed in time polynomial in @xmath470 .
we now bound these constants in terms of @xmath5 .
since each edge of a tetrahedron meets at most three disk types in that tetrahedron , each edge of a tetrahedron can contribute at most six points at which the weight vector changes .
given six edges to each tetrahedron , we have @xmath482 . as before we have a bound for the number of pairings @xmath483 and the number of normal coordinates is given by @xmath484 .
the total weight bounds the normal coordinates , and with @xmath485 we get a bound on the running time that is polynomial in @xmath350 .
let @xmath4 be a 3-manifold with a triangulation @xmath348 containing @xmath5 tetrahedra and let @xmath113 be a fundamental normal surface in @xmath4 .
there is a procedure for counting the number of components of @xmath113 and the topology of each component which runs in time polynomial in @xmath5 .
[ [ proof.-16 ] ] * proof . * + + + + + + + + theorem [ bound ] gives a bound for the normal coordinates of @xmath113 of @xmath403 .
recall that there are @xmath404 normal coordinates , and each represents a triangle or quadrilateral , so the total number of intersections with the 1-skeleton satisfies @xmath486 . in particular
, @xmath487 is bounded above by a polynomial in @xmath5 . plugging this in for @xmath347 in corollary [ eulercount ] we get a bound for the running time which is a polynomial in @xmath5 .
in this section we examine the complexity of the problem of determining the smallest area of a spanning surface for a curve in a 3-dimensional manifold .
such an area calculation problem seems at first to be ill suited to a complexity analysis , since it has real solutions depending on a choice of riemannian metric .
we recast the area calculation problem into a discretized form where its complexity can be analyzed .
given a curve in a suitably discretized riemannian 3-manifold , we ask whether it bounds a surface of area less than @xmath1 , where @xmath1 is an integer . to describe a metric on a 3-manifold with a finite amount of data
, we restrict to piecewise flat metrics , and manifolds constructed from collections of flat tetrahedra and triangular prisms whose faces are identified by isometries .
the curvature of such pl metrics can be defined as a limit of smooth curvatures , and is concentrated along their edges and vertices .
a particular manifold in this class is described by a decomposition into tetrahedra or triangular prisms with a rational ( or integer ) length assigned to each edge .
the metric on this tetrahedron or prism is then taken as the metric on the euclidean tetrahedron or prism with those edge lengths . in the case of a prism
we also set the angles of quadrilateral faces to be right angles .
prisms are allowed in this construction in order to form metrics with rational lengths on spaces that are products . identified 2-dimensional faces
are required to be isometric .
we do not require that the total angle around an edge is @xmath488 , nor do we make any metric conditions at a vertex .
this type of metric is described by a finite set of data , and can be used to approximate riemannian metrics on a manifold . up
to scaling , we can take all the edge lengths to be integers .
we call these objects _ metrized pl 3-manifolds_. a curve is given as a collection of edges in the 1-skeleton of @xmath4
. we will show that given an integer @xmath1 , determining whether the smallest spanning surface for a curve in such a 3-manifold has area less than @xmath1 is * np*-hard .
_ problem : _ minimal - spanning - area + _ instance : _ a 3-dimensional metrized pl manifold @xmath4 , a 1-dimensional + curve @xmath3 in the 1-skeleton of @xmath4 , and a natural number @xmath1 .
+ _ question : _ does the curve bound a surface of area @xmath489 ?
the size of an instance is given by the number of bits needed to describe all the edge lengths and @xmath1 .
[ nphard ] minimal - spanning - area is * np*-hard .
[ [ proof.-17 ] ] * proof . * + + + + + + + + we reduce in polynomial time an instance of the * np*-hard problem one - in - three sat to an instance of minimal - spanning - area .
this shows that minimal - spanning - area is at least as hard , up to polynomial time reduction , as one - in - three sat . as a first step
, we set up a 2-dimensional version of minimal - spanning - area .
we then construct a 3-manifold by a thickening process , with the property that a minimizing surface must remain within the 2-complex .
given a boolean expression representing an instance of one - in - three sat , we construct a triangulated metrized 2-complex and an integer @xmath1 .
this complex contains a curve @xmath3 with the property that the expression admits a satisfying assignment if and only if @xmath3 bounds a surface of area less than @xmath1 .
this metrized complex is shown in figure [ metrized ] for the expression @xmath490 . .
the picture is not to scale .
the shaded prisms are constructed to each have area one , while the rest of the surface has total area less than 1/2 . ]
the branching surface is similar to the one used in the proof of theorem [ thmnph ] , but carries the additional structure of a metrized triangulation , whose triangles have flat metrics of prescribed edge length .
the metrized triangles are constructed so that near each of the @xmath101 boundary components corresponding to clauses of the boolean expression there are three triangulated disks of area close to one , one on each of the three handles coming into the punctured sphere near the boundary component .
these disks are shaded in figure [ metrized ] .
each of these shaded disks is chosen to have area between 1 and @xmath491 .
the surface is constructed so that the union of all triangles in the rest of the surface has total area less than 1/2 .
we saw in theorem [ thmnph ] that a spanning surface which has minimal genus goes over each of the shaded disks at most once and goes over exactly one shaded disk for each of the @xmath101 clauses .
it follows that such a surface has total area @xmath492 .
furthermore , a satisfying assignment for one - in - three sat leads to an embedded spanning surface , with the satisfying values of the variables selecting branches of the surface , and such a spanning surface has area less than @xmath493 .
so an instance of one - in - three sat can be reduced to an instance of minimal - spanning - area for this 2-complex . to pass to a 3-manifold , thicken each triangle in the branching surface to a triangular prism , triangulated as in the proof of theorem [ thmnph ] , and with a product metric .
this produces a 3-manifold @xmath4 which is a thickened up version of the 2-complex .
projection to the branching surface is area non - increasing , and area decreasing for a surface with boundary on the branching surface but not contained in it .
therefore a least area surface spanning @xmath3 must lie on the branching surface .
a closed manifold @xmath494 with a piecewise - smooth metric can be obtained by a doubling construction as in theorem [ thmnph ] .
the doubling involution is an isometry , so that reflecting a surface meeting @xmath495 into @xmath4 does not increase area .
it follows that the embedded spanning surface on the branching surface is a least area surface in @xmath494 .
among many unresolved questions are : \2 . is determining the genus of a knot in a 3-manifold * np * ? this amounts to showing that finding a lower bound to the knot s genus is an * np * problem , in contrast to the upper bound we have investigated .
recall that the genus of a knot is the least possible genus of all spanning surfaces .
we have shown that certifying that the genus is at most @xmath0 is * np * , but left open the possibility that the genus may be smaller than @xmath0 .
if the answer to this question is yes , then we can certify that a non - trivial knot has positive genus , and it would follow that unknotting is both * np * and * conp*. 99 m. dehn , `` uber die topologie des dreidimensional raumes '' , _ math .
annalen _ , 69 ( 1910 ) , 137168 .
garey and d.s .
johnson , `` computers and intractability .
a guide to the theory of * np*-completeness '' , w. h. freeman and co. , san francisco , 1979 .
w. haken , `` theorie der normalflchen : ein isotopiekriterium fr den kreisknoten '' , _ acta math .
_ , 105 ( 1961 ) 245375 . j. hass , `` algorithms for recognizing knots and 3-manifolds '' , _ chaos , solitons and fractals
_ , 9 ( 1998 ) 569581
. j. hass , j. c. lagarias and n. pippenger , `` the computational complexity of knot and link problems '' , _ journal of the acm _ , 46 ( 1999 ) 185211
. j. hass and j. c. lagarias , `` the number of reidemeister moves needed for unknotting '' , _ j. amer
_ 14 ( 2001 ) , no .
2 , 399428 g. hemion , _ the classification of knots and @xmath496-dimensional spaces _ , oxford university press , 1992 . j. hempel , _
3-manifolds _ , princeton university press , princeton , nj 1976 .
w. jaco , _ lectures on three - manifold topology _ , cbms regional conference series in mathematics , 43 ams , providence , ri , 1980 .
w. jaco and u. oertel , `` an algorithm to decide if a @xmath496-manifold is a haken manifold '' , _ topology _ , 23 ( 1984 ) 195209 .
w. jaco and j. l. tollefson , `` algorithms for the complete decomposition of a closed @xmath496-manifold '' , _
illinois j. math .
_ , 39 ( 1995 ) 358406 .
w. jaco and j. h. rubinstein , `` pl equivariant surgery and invariant decompositions of @xmath496-manifolds '' , _ advances in math .
_ , 73 ( 1989 ) 149191 . f. jaeger , d. l. vertigan and d. j. a. welsh , `` on the computational complexity of the jones and tutte polynomials '' , _ math .
cambridge phil .
_ , 108 ( 1990 ) 3553
. h. kneser , `` geschlossene flchen in dreidimensionalen mannigfaltigkeiten '' , _ jahresbericht math .
_ , 28 ( 1929 ) 248260 .
e. e. moise , `` affine structures in @xmath496-manifolds , v : the triangulation theorem and hauptvermutung '' , _ ann .
_ , 56 ( 1952 ) 96114 .
papadimitriou , computational complexity .
addison - wesley publishing company , reading , ma , 1994 .
m. o. rabin , `` recursive unsolvability of group - theoretic problems '' , _ ann .
_ , 67 ( 1958 ) 172194 .
rubinstein , `` an algorithm to recognize the @xmath496-sphere '' , _ proceedings of the international congress of mathematicians _ , vol . 1 , 2 ( zurich , 1994 ) , birkhauser , basel , 601611 .
schaefer , t.j .
`` the complexity of satisfiability problems '' , _ proc 10th ann .
symp . on theory of computing _ ,
acm , ny , ( 1978 ) 216 - 226 .
a. schrijver , _ theory of linear and integer programming _
, john wiley and sons , 1986 .
h. schubert , `` bestimmung der primfaktorzerlegung von verkettungen '' , _ math .
_ , 76 ( 1961 ) 116148 .
a. seb , `` hilbert bases , caratheodory s theorem and combinatorial optimization '' , in : r. kannan and w. r. pulleyblank ( eds . ) , _ integer programming and combinatorial optimization _ ,
university of waterloo press , 1990 , pp .
. h. seifert , `` ber das geschlecht von knoten '' , _ math .
annalen _ , 110 ( 1935 ) 571592 .
a. thompson , thin position and the recognition problem for @xmath14 .
lett . 1 ( 1994 )
, 613630 .
d. j. a. welsh , _ complexity : knots , colourings and counting _ , cambridge university press , 1993 .
d. j. a. welsh , `` the complexity of knots '' , _ ann .
discr . math .
_ , 55 ( 1993 ) 159173 .
d. j. a. welsh , `` knots and braids '' , _ contemp .
_ , 147 ( 1993 ) 109123 .
|
we show that the problem of deciding whether a polygonal knot in a closed three - dimensional manifold bounds a surface of genus at most @xmath0 , is * np*-complete . we also show that the problem of deciding whether a curve in a pl manifold bounds a surface of area less than a given constant @xmath1 is * np*-hard .
|
hints of direct detection of dark matter ( dm ) currently exist from two experiments .
there is a long - standing observation of an annual modulation in the signal observed by dama @xcite , whose statistical significance is beyond question .
last year the cogent experiment reported excess events in their lowest electron energy bins @xcite , followed more recently by a @xmath2 detection of annual modulation in the signal @xcite . under the simplest assumptions about the nature of the dark matter interactions ,
these two observations appear to be incompatible with each other @xcite and with upper limits obtained by other experiments , especially cdms @xcite , xenon10 @xcite and xenon100 @xcite .
channeling of recoiling ions along the crystal planes in the detectors has been suggested as one loophole for reconciling the conflicts , but this has been argued to be too small an effect by ref .
uncertainties in quenching factors can also be used to help reconcile the two positive detections @xcite .
there are alternative microphysical ways of ameliorating the tensions that have been explored in recent papers @xcite-@xcite .
one important modification is to allow for isospin - violating interactions of the dark matter with nucleons , since the coherent spin - independent matrix element is a sum over the interactions with protons and neutrons : _
n ~(f_p z + f_n ( a - z))^2 _ n where @xmath3 are the number of nucleons and protons in the nucleus , respectively , and @xmath4 are the respective cross sections for dm scattering on neutrons and the nucleus . if isospin is conserved then @xmath5 , but if @xmath6 , the limits placed on @xmath7 by the xenon experiments are greatly relaxed @xcite .
moreover , several groups have indicated that inelastic scattering @xcite of dm with a small mass splitting @xmath8 kev can have a beneficial effect either for the the agreement between dama and cogent @xcite , the conflict between cogent and cdms @xcite , the goodness of fit to the cogent modulated signal alone @xcite , or marginally the overall goodness of fit to all data @xcite . in this work
, we look for the simplest and most natural kind of models that could accommodate these two generalizations , isospin violation and inelasticity , with an emphasis on hidden sector models with higgs or gauge kinetic portals to the standard model .
( for previous model building efforts which do not focus on these aspects , see references @xcite-@xcite . ) a similar study to the present one was done in ref .
@xcite , but considering more elaborate models than we examine here .
our inelastic model also has elements in common with that of ref .
@xcite , although the latter incorporated neither isospin violation nor inelasticity .
( see @xcite for another recent isospin - violating model . ) in ref .
@xcite , isospin violating couplings of a single vector were obtained through a combination of kinetic mixing and mass mixing with the @xmath9 .
we will argue that interference between two new vector mediators is an elegant way of getting isospin violation if one demands that only inelastic scattering takes place .
it is well known that dirac states coupling to vectors become off - diagonal in the mass eigenstate basis if small majorana masses are introduced @xcite .
thus it is not challenging to account for the inelastic nature of the couplings . on the other hand , ref .
@xcite finds that inelasticity improves the global fit to the data only moderately , so that one might also contemplate models with no dm mass splitting and purely elastic scattering .
the overall fit is in fact not very good , indicating either that some of the data have inconsistencies or that the dark matter interpretation is not correct . in this work
we will assume the former , in which case one might be motivated to consider inelasticity as a secondary criterion , which might or might not survive as the data improve . accordingly , we consider both elastic and inelastic models in the following and leave it to the reader to judge how strongly the latter is preferred by the data .
elastic scattering naturally arises in alternative models where the isospin violation comes about by interference between vector and scalar exchange .
we will construct our models for isospin - violating dark matter starting with this simpler possibility , showing why it does not naturally accommodate inelastic couplings .
a natural way to induce scalar - mediated interactions between the dark sector and the standard model ( sm ) is to introduce a new higgs field @xmath10 that is a singlet under the sm gauge group , and communicates to the sm through the interaction @xmath11 .
if @xmath10 gets a vev then the mass eigenstates are admixtures of @xmath10 and @xmath12 with mixing angle @xmath13 , where @xmath14 stand for the mass eigenvalues ( we assume small mixing , @xmath15 ) and @xmath16 is the sm higgs vev . if the dm @xmath17 has a yukawa coupling @xmath18 , then @xmath10 mediates interactions with sm fermions @xmath19 with the strength @xmath20 , assuming that @xmath21 . here
@xmath22 is the sm yukawa coupling of the higgs to fermion @xmath19 . in this scenario , @xmath10 couples to sm matter proportionally to @xmath12 .
although these couplings violate isospin , it is a very small violation when it comes to scattering on nuclei , because nucleons get almost none of their mass from the valence quarks .
the couplings of the higgs to nucleons are dominated by the sea - quark and gluon content , which are the same for neutrons and protons ( see for example @xcite ) .
therefore we get no significant isospin violation from couplings mediated by the higgs portal : it has @xmath5 .
if in addition we invoke a gauge kinetic mixing portal , @xmath23 ( where @xmath24 and @xmath9 are the respective field strengths of the sm hypercharge and the hidden sector u(1)@xmath25 , and @xmath26 is the weinberg angle ) , then interference between the new vector and the new higgs does lead to tunable isospin violation .
this is because the vector only mixes significantly with the photon , leading to @xmath27 .
of course , this is not the ratio we need for dama and cogent .
but interference between the vector and scalar exchange allows for any desired value .
if @xmath28 is the new u(1)@xmath25 gauge coupling to the dm , then [ frat ] where @xmath29 is the coupling of @xmath12 to the nucleon : @xmath30 in terms of the nucleon mass @xmath31 and the higgs vev @xmath32 @xcite . the main objection to this scenario arises if we want the couplings to dm to also be inelastic .
suppose there are two mass eigenstates @xmath33 with masses @xmath34 split by the small amount @xmath35 .
let us first write down an effective potential : v = _ |_m _ _ + 12 m^2_z z^2 + 12 m^2_^2 + g|_+ z _ - + y |_+ _ - [ scalarv ] ( it is also understood that @xmath9 couples to the electromagnetic current with strength @xmath36 and @xmath10 to the sm fermions @xmath19 with strength @xmath37 . )
the mass splitting can arise by spontaneous symmetry breaking through couplings of the form @xmath38 where @xmath39 are the weyl components in the original lagrangian , and @xmath10 carries twice the u(1)@xmath25 charge of @xmath17 .
the troublesome question is how the scalar interactions came to be purely off - diagonal in the mass eigenbasis , given that @xmath10 has a vev which is needed in order to mix with the sm higgs @xmath12 .
to arrive at ( [ scalarv ] ) , there must have been a bare mass term @xmath40 ( here expressed in the mass eigenbasis ) that was exactly canceled by @xmath41 . in the absence of this tuning
, @xmath10 will have diagonal plus off - diagonal couplings , and this presents a complication for getting the desired level of isospin violation , since the interference between @xmath10 and @xmath9 only occurs in the inelastic channel .
this is because the majorana vector couplings are purely off - diagonal ( see the following subsection ) .
we do not contemplate this finely - tuned situation any further .
however if instead of couplings of the form @xmath42 in which @xmath10 must carry a compensating charge , we have the interactions |(m+y ) in terms of a dirac fermion @xmath17 , then @xmath17 remains dirac after @xmath10 gets its vev , and there is no mass splitting .
@xmath10 can interfere with @xmath9 through the purely elastic couplings . even though we must invoke an additional singlet higgs @xmath43 to break the u(1)@xmath25 symmetry , since now @xmath10 is neutral under u(1)@xmath25 , this is still an economical model , whose consequences we will consider . for simplicity
we take @xmath10 to be real .
we do not consider the case of a pseudoscalar ( @xmath44 ) coupling because this leads to a nuclear scattering amplitude that is suppressed by the dm velocity , making it more difficult to interfere with the vector exchange contribution to get the desired isospin violation ( since the latter has no such velocity suppression ) .
in contrast to the couplings of a scalar to dm , the purely off - diagonal coupling of the gauge field is natural if @xmath33 are majorana fermions that originated from a dirac particle before spontaneous symmetry breaking @xcite .
consider the interactions ( again in the model where @xmath10 carries two units of the @xmath17 charge ) v = 12 |_m _
+ ^ * ( |_ p_l _ + | _ p_r _ ) + h.c .
[ lagrangian ] where now @xmath45 , @xmath46 denote majorana - dirac spinors constructed from the weyl spinors , here renamed @xmath47 to avoid confusion .
when @xmath10 gets a vev , the mass matrix becomes ( cc & m + m & ) which is diagonalized by @xmath48 with mass eigenvalues @xmath49 , where @xmath50 .
if the u(1)@xmath25 interaction was orginally vector - like , then it becomes exactly off - diagonal in the mass basis because there is no vector current for a single majorana state .
this is a strong motivation to prefer vector mediators if we aim for both isospin violation and inelastic off - diagonal couplings .
the previous discussion motivates us to build a model in which the interfering scalar current is replaced by another vector current .
the new vector need only couple to isospin differently from the kinetically mixed u(1 ) that has @xmath27 .
coupling to @xmath51 is attractive from the point of view of anomaly cancellation , but such couplings are very strongly constrained because of the leptonic interactions ( see for example @xcite ) .
the simplest possibility that avoids these constraints is coupling to @xmath52 alone .
u(1)@xmath53 is anomalous and it also has mixed anomalies with the sm gauge groups , that can be canceled by adding the appropriate exotic heavy particles @xcite-@xcite .
we will not discuss the implications of these new particles further here , although they can provide complementary collider signatures to test the model .
our addition of a single vector - like dm particle coupling to @xmath52 does not spoil the anomaly cancellation achieved in these models .
we refer to the u(1)@xmath53 gauge boson as @xmath54 and for simplicity assume that it couples with equal strength @xmath55 to the dm and to the sm baryons
. it also couples with equal strength to protons and neutrons , just like the singlet higgs of the previous section .
therefore it is clear that ( [ frat ] ) is replaced by [ frat2 ] this model is almost complete , but we have accounted for the breaking of only one linear combination of the two new u(1)s through the vev of @xmath10 .
notice that @xmath10 must have charges @xmath56 under @xmath57 in order for ( [ lagrangian ] ) to be gauge invariant . to completely break the symmetry
we need another field @xmath43 with different charges .
this means that the mass eigenstates for the gauge bosons are generally admixtures of the original fields , and that both will therefore kinetically mix with the sm hypercharge .
we need to clarify the relation between the couplings appearing in ( [ frat2 ] ) and the original lagrangian parameters .
a simple way to ensure that the above relations are approximately correct despite mixing of the new gauge bosons is to assume that @xmath58 , by assigning @xmath43 the charges @xmath59 such that @xmath60 . in that case
the mixing is suppressed by the large mass difference and @xmath61 , then the gauge boson mass matrix in the basis @xmath62 is @xmath63 and we can take m_b g_b m_z 2 g [ masses ] the mixing angle is approximately @xmath64 .
we will find that its value scales proportionally to the gauge kinetic mixing parameter @xmath65 , such that @xmath66 .
( this relation follows from eq .
( [ mbzconst ] ) below and the relic density constraint , fig .
[ fig : relicden](a ) . )
.[tab1 ] best - fit values of the dm - neutron elastic scattering cross section , dm mass , mass splitting , and isospin violation , from ref . , appropriate to the given theoretical model . [ cols="^,^,^,^,^",options="header " , ]
we must show that values for the parameters exist that can give the right cross section for cogent and dama , and the right relic density for the dark matter . and
@xmath67 being the origin of the relic density , which has been widely discussed in the recent literature ( see for example @xcite ) .
for this work we will assume the asymmetry vanishes .
] for the effective elastic cross section of dm on the neutron , the dm mass and mass splitting , and level of isospin violation , we will consider the three cases shown in table [ tab1 ] , which is corresponds to the best - fit values found by ref .
@xcite for the cases of endothermic @xmath68 , and elastic scatterings , respectively .
these are the ones appropriate to our two models .
we will show that the exothermic reactions , @xmath69 , are possible when @xmath70 , whereas otherwise the excited state is depleted by @xmath71 downscatterings in the early universe .
the theoretical cross sections for dm - neutron scattering in our models , in the elastic limit , are _
n = _ n^2\ { rl g_b^4/m_b^4 , & b _ + ( y y_n_s)^2/m_^4 , & scalar + } [ sigma_n ] where @xmath72 is the reduced mass for the dm - nucleon system and @xmath73 is the dm velocity . by equating the @xmath7 values in table [ tab1 ] to those in eq .
( [ sigma_n ] ) and using eqs .
( [ frat],[frat2 ] ) , we obtain & = & 232 gev , = -(79.9 gev)^2(b_)[mbzconst ] + m _ & = & 346 gev , = -(118.9 gev)^2 ( scalar ) [ mphizconst ] to get the correct mass spectrum for the model with @xmath54 , we can set the bare dirac mass directly to @xmath74 gev , and choose @xmath75 kev . recall that only @xmath9 gets its mass primarily from @xmath10 , in this model , so if @xmath65 is sufficiently small , it is possible to have @xmath76 gev or less .
the yukawa coupling still needs to be quite small in that case , @xmath77 .
however this is only 4 times smaller than the electron yukawa coupling in the standard model , so it is not unreasonable .
alternatively , for the purely elastic models with @xmath10 exchange , we need to set @xmath78 7.5 gev .
next we consider the relic density . starting with the @xmath54 vector exchange model ,
there are several possible annihilations into gauge bosons : @xmath79 , @xmath80 , @xmath81 .
in addition there are coannihilations into quarks , @xmath82 mediated by @xmath52 in the @xmath83-channel , and also @xmath84 into all charged sm fermions @xmath19 except the kinematically inaccessible top , mediated by the @xmath9 .
averaging over all the possibilities , we find the annihilation cross section _
ann v= 132 m^2 s(g,g_b,,x,x_b ) [ sigv ] where @xmath85 is a dimensionless function of the couplings and the mass ratios @xmath86 , @xmath87 , given by s & = & 12 g^4 f_1(x,x ) + g_b^2 g^2 f_2(x,x_b ) + 12 g_b^4 f_1(x_b , x_b ) + & + & g_b^4 _ i = u , d , s , c , b n_c , i f_3(x_b , x_i ) + ( ge)^2 _
i = e,,,u , d , s , c , b n_c , i q_i^2 f_3(x,x_i ) [ xeq ] here @xmath88 for sm fermion @xmath89 , with charge @xmath90 and number of colors @xmath91 , and the functions @xmath92 are given in the appendix .
for the the scalar exchange model , all the analogous processes to the previous case are present , with @xmath54 replaced by @xmath10 .
the @xmath93 contribution is @xmath94-wave suppressed , as is @xmath95 by @xmath10 exchange , so we neglect them .
the @xmath96 and @xmath9-mediated @xmath97 contributions are the same as in ( [ xeq ] ) .
we find that ( [ xeq ] ) is replaced by s 12 g^4 f_1(x,x ) + y^2
g^2 f_4(x,x _ ) + ( ge)^2 _ i = e,,,u , d , s , c , b n_c , i q_i^2 f_3(x,x_i ) [ xeq2 ] where @xmath98 and @xmath99 is defined in the appendix . (
a ) : contours in the @xmath100-@xmath101 plane that give the observed relic density in the vector exchange model , for gauge kinetic mixing parameter @xmath102 , @xmath103 , and @xmath104 .
the ellipse highlights the region where the gauge boson mixing angle @xmath105 is small as needed for consistency .
( b)-(d ) : analogous contours in the @xmath106-@xmath101 plane for the scalar exchange model , for @xmath102 ( b ) , @xmath103 ( c ) , @xmath104 ( d ) , and several values of the higgs mixing angle @xmath107 . ] for each of the models , we equate ( [ sigv ] ) to the standard value of the cross section for the observed relic density , @xmath108 @xmath109/s .
this gives the constraint s = 1.710 ^ -5 [ xconst ] to determine the parameters satisfying ( [ xconst ] ) in the vector exchange model , we assume several choices for @xmath65 that can be compatible with laboratory bounds on kinetic mixing ( see next section ) , and use ( [ mbzconst ] ) to eliminate @xmath110 in favor of @xmath111 .
for the scalar exchange model , we similarly use ( [ mphizconst ] ) to eliminate @xmath112 in favor of @xmath113 .
this case has the free parameter of the higgs mixing angle @xmath107 to be varied in addition to @xmath65 . for @xmath114 5 gev @xmath107
is constrained to be less than @xmath115 from the width of the @xmath116 boson due to decays @xmath117 and from @xmath52 meson decays @xmath118 . @xcite .
we thus obtain , for each value of @xmath65 , the contour in the @xmath101-@xmath119 plane corresponding to the observed relic density for the vector exchange model , shown in fig .
[ fig : relicden](a ) .
similarly in the scalar model , for each pair @xmath120 , we find a contour in the @xmath101-@xmath113 plane , shown in fig . [
fig : relicden](b)-(d ) .
it is clear from fig .
[ fig : relicden ] that @xmath101 and @xmath106 tend to be @xmath121 gev and both scale as @xmath122 .
only the vector @xmath54 can remain somewhat heavier as we now explain .
recall that we previously made a simplifying technical assumption , @xmath123 , to ensure small mixing @xmath105 between the two gauge bosons .
this assumption might be relaxed somewhat , but at the risk of increasing the highly constrained couplings of @xmath52 to leptons due to the gauge kinetic mixing . to the extent
that @xmath105 is small , only the vertical part of the contours where @xmath124 gev is relevant , giving a sharp prediction for the @xmath54 mass , if @xmath125 is known .
since the determination of @xmath125 could well be uncertain by @xmath1261 gev , we find an uncertainty of @xmath127 gev in @xmath119 by varying @xmath125 . in this vertical branch ,
the @xmath128 annihilation cross section is dominated by the @xmath129 contributions in ( [ xeq ] ) . in the vector exchange model where we have a small dm mass splitting ,
the process @xmath130 mediated by the @xmath9 ( and the @xmath52 , although we find the former dominates ) efficiently depletes the @xmath131 population in the early universe over part of the allowed parameter space .
the downscattering cross section for a similar model was calculated in @xcite , which adapts to the present case as _ v ( g^2m_z^2 + 2 m + g_b^2m_b^2 + 2m)^2 m^2 4 these interactions freeze out at temperature @xmath132 given by @xmath133 where @xmath134 is the dm number density and @xmath135 is the hubble constant .
if the dm remained in kinetic equilibrium with the sm down to @xmath132 then the relative abundance of @xmath131 to @xmath136 would be suppressed by @xmath137 .
however it is the kinetic temperature of the dm which is important here , and if kinetic decoupling occurs at a temperature @xmath138 , then the suppression is more severe , @xmath139 ( see @xcite for a discussion of this issue . ) the kinetic equilibrium is controlled by the scattering of dm on electrons , whose cross section is approximately _
e v ( gem_z^2 + 2m)^2 m_e^2 we find that this goes out of equilibrium at @xmath140 mev .
using this methodology , we estimate that the relative abundance of @xmath131 is unsuppressed for @xmath70 .
the suppression turns on exponentially fast as a function of @xmath65 , with @xmath141 giving a relative abundance of @xmath142 , while @xmath143 leads to almost no suppression , @xmath144 .
therefore it is possible to realize the exothermic dark matter scenario suggested in ref .
@xcite over some part of the allowed parameter space . in these cases
the scatterings will be an average over the endothermic and exothermic ones since both states are equally populated .
we note that ref .
@xcite finds a moderate preference for exothermic reactions in their global fits .
to recapitulate , we have investigated two hidden sector dark matter models that violate isospin in an optimal manner for reconciling the cogent and dama signals with constraints from xenon10 and xenon100 .
the vector exchange model has two majorana mass eigenstates @xmath33 ( or a single pseudo - dirac particle ) with masses @xmath145 gev @xmath146 kev , and two light gauge bosons @xmath9 and @xmath52 with masses @xmath147 gev , @xmath148 gev , and the couplings boson current @xmath149 given by @xmath150 that come from the mixing of @xmath9 with weak hypercharge ; see for example @xcite .
larger contributions to the kinetic mixing of the @xmath52 with @xmath116 can be generated from sm loops below the scale of baryon symmetry breaking @xcite . ]
|_+(g + g_b)_- + g_b j_b^b^+ j^_em(z_+ b _ ) here @xmath151 is the gauge kinetic mixing parameter for the @xmath9 , @xmath152 is the mixing angle between @xmath52 and @xmath9 , and @xmath153 are the respective electromagnetic and baryon number currents of the standard model .
the couplings are adjusted to give the optimal isospin violation @xmath154 @xcite via eq .
( [ frat2 ] ) .
there must also be diagonal yukawa interactions @xmath155 to a singlet higgs @xmath10 whose vev leads to the small mass splitting , but @xmath156 is much smaller than the gauge couplings ( @xmath157 ) and therefore we have neglected these interactions . the dirac mass @xmath74 gev appearing in the original lagrangian is protected by chiral symmetry , and so does not introduce any new hierarchy problem .
the small scale @xmath158 gev on the other hand is unexplained unless one invokes supersymmetry in the hidden sector @xcite or some other uv completion .
the scalar exchange model has exactly dirac dark matter with @xmath159 gev ( hence only elastic scattering ) and a real singlet @xmath10 that mixes with the sm through the higgs portal @xmath11 .
like the previous model these also have the kinetically mixed @xmath9 vector .
the interactions are given by |(g + y)+ _ s _ i y_i |f_i f_i + j^_emz _ where @xmath92 are the sm fermions with their yukawa couplings @xmath160 .
we find that @xmath161 gev to satisfy the relic density constraint .
if @xmath113 happens to be close to this upper limit , it could be discoverable at the lhc , while for the very light cases @xmath114 1 gev , indirect discovery could come from rare decays such as @xmath162 followed by @xmath163 @xcite .
one of the most exciting aspects of these models is that they predict new low - energy interactions mediated by the light @xmath9 with a strength relevant for detection in beam - dump experiments @xcite such as apex @xcite and the mainz microtron @xcite and the low - energy @xmath164-@xmath165 collider experiment kloe @xcite .
the still - open window of parameter space in the @xmath65-@xmath101 plane corresponds roughly to that which we have identified in this paper as being compatible with the @xmath33 relic density . in our model ,
@xmath101 is only bounded from above , depending on the value of @xmath65 , as shown in fig .
[ fig : relicden ] : @xmath166 gev in the vector exchange model . on the other hand ,
the gauge boson @xmath52 of baryon number is extremely hard to detect due to its very weak coupling and relatively large mass .
for example , constraints from new contributions to @xmath167 decay into quarks are easily satisfied @xcite .
the tevatron sets limits on @xmath168 from the nonobservation of @xmath169 where @xmath170 is a single jet @xcite , which is also satisfied by our model .
@xcite pointed out that the kinetic mixing of @xmath52 leads to weak tevatron constraints from the drell - yan production of lepton pairs .
it may be interesting to update these constraints since @xcite was written before the upgrade of the tevatron
. the best indirect confirmation of its presence will be the discovery of an exotic extra family of quarks with baryon number @xmath171 @xcite . in the simplest such models
@xcite , this fourth generation gets its mass through the usual couplings to the higgs , requiring its mass to be at the electroweak scale and limited by large yukawa couplings leading to a landau pole near the tev scale .
but ref .
@xcite shows that this limitation can be removed using vector - like quarks ( from the point of view of the sm su(2 ) gauge symmetry ) and giving mass to them through the vev of the field which breaks baryon number , @xmath43 in our model .
if the coupling @xmath172 in ( [ masses ] ) is sufficiently small , for example @xmath173 , then @xmath174 can be at the tev scale .
astrophysical constraints are rapidly closing in on light dark matter models .
if @xmath175 gev dm annihilates predominantly into @xmath176 with the standard relic density cross section , it is ruled out by its distortions of the cmb @xcite .
the @xmath177 channel is still open since a large fraction of the muon energy is converted to neutrinos which have no effect on the cmb .
other channels have an intermediate effect between these two extremes @xcite . in our model with two vector bosons , the annihilation is primarily into @xmath52 s followed by decay into light quarks , which appear to be still be allowed . on the other hand , models that produce _ too many _ neutrinos are constrained by superkamiokande limits on @xmath178 from the sun @xcite . even in
optimally isospin violating models such as we have considered here , annihilation of @xmath175 gev dm with the required cross section on nucleons for cogent / dama is ruled out for the @xmath179 channel and marginally allowed for @xmath180 and lighter quarks as in our vector model .
these limits can be improved in the future using data from icecube / deepcore @xcite .
finally , annihilations of light dm in dwarf satellite galaxies of the milky way that produce too many gamma rays in association with charged particles have recently been severely constrained by fermi data @xcite ( see also @xcite ) .
again , the @xmath179 channel is excluded but annihilation into light quarks is still allowed . after the first version of this paper was posted , ref .
@xcite appeared , which considers a similar class of models .
* note added : * after completing this work we became aware of refs.@xcite showing that pamela antiproton constraints are in conflict with a @xmath52 vector boson of mass greater than @xmath181 , which would favor the lower-@xmath119 parts of the contours of fig .
in addition , we discovered that the gauge boson mixing angle effects in the present model can not be ignored even when @xmath105 is small , due to the occurrence of @xmath182 in the amplitudes , which scales as @xmath183 .
this can be overcome by introducing an additional contribution to @xmath119 .
we intend to address these issues in a forthcoming publication .
we thank b. grinstein , c. carone , r. essig , g. gelmini , s.koushiappas , j. kumar , j. lavalle , g. moore , m. pospelov , i.shoemaker , l. strigari , a. strumia , t. volansky , y. zhang for helpful correspondence or discussions .
jc thanks the aspen center for physics for stimulating interactions during the completion of this work .
this work was supported in part by the natural sciences and engineering research council of canada .
the functions of mass ratios appearing in the annihilation cross sections are as follows .
they were computed using feyncalc @xcite .
f_1(x_1,x_2 ) & = & ( 1-x_2 ^ 2)^3/2(1 - 12
x_1 ^ 2)^2 ( 1-x_2 ) , + f_2(x_1,x_2 ) & = & ( 1 - 12(x_1 ^ 2+x_2 ^ 2 ) + ( x_1 ^ 2-x_2 ^ 2)^2)^3/2 ( 1 - 14(x_1 ^ 2+x_2 ^ 2))^2(2-x_1-x_2 ) , + f_3(x_1,x_2 ) & = & ( 1 + 12 x_2 ^ 2)(1-x_2 ^ 2)^1/2(1 - 14 x_1 ^ 2)^2(1-x_2 ) r. bernabei _ et al . _
[ dama collaboration ] , eur .
j. c * 56 * ( 2008 ) 333 [ arxiv:0804.2741 [ astro - ph ] ] . c. e. aalseth _
[ cogent collaboration ] , phys .
lett . * 106 * , 131301 ( 2011 ) [ arxiv:1002.4703 [ astro-ph.co ] ] . c. e. aalseth _
et al . _ , arxiv:1106.0650 [ astro-ph.co ] . c. savage , g. gelmini , p. gondolo , k. freese , phys
. rev . * d83 * , 055002 ( 2011 ) .
[ arxiv:1006.0972 [ astro-ph.co ] ] .
z. ahmed _ et al .
_ [ the cdms - ii collaboration ] , science * 327 * , 1619 ( 2010 ) [ arxiv:0912.3592 [ astro-ph.co ] ] .
z. ahmed _ et al .
_ [ cdms - ii collaboration ] , phys .
* 106 * , 131302 ( 2011 ) .
[ arxiv:1011.2482 [ astro-ph.co ] ] .
j. angle _ et al . _
[ xenon10 collaboration ] , phys .
d * 80 * , 115005 ( 2009 ) [ arxiv:0910.3698 [ astro-ph.co ] ] .
e. aprile _ et al .
_ [ xenon100 collaboration ] , arxiv:1104.2549 [ astro-ph.co ] . n. bozorgnia , g. b. gelmini , p. gondolo , jcap * 1011 * , 019 ( 2010 ) .
[ arxiv:1006.3110 [ astro-ph.co ] ] ; jcap * 1011 * , 028 ( 2010 ) .
[ arxiv:1008.3676 [ astro-ph.co ] ] .
p. sandick , j. diemand , k. freese and d. spolyar , arxiv:1108.3820 [ astro-ph.co ] . h. an and f. gao , arxiv:1108.3943 [ hep - ph ] . q. -h . cao , i. low , g. shaughnessy , phys . lett . * b691 * , 73 - 76 ( 2010 ) .
[ arxiv:0912.4510 [ hep - ph ] ] .
|
motivated by recent attempts to reconcile hints of direct dark matter detection by the cogent and dama experiments , we construct simple particle physics models that can accommodate the constraints .
we point out challenges for building reasonable models and identify the most promising scenarios for getting isospin violation and inelasticity , as indicated by some phenomenological studies . if inelastic scattering is demanded , we need two new light gauge bosons , one of which kinetically mixes with the standard model hypercharge and has mass @xmath0 gev , and another which couples to baryon number and has mass @xmath1 gev .
their interference gives the desired amount of isospin violation .
the dark matter is nearly dirac , but with small majorana masses induced by spontaneous symmetry breaking , so that the gauge boson couplings become exactly off - diagonal in the mass basis , and the small mass splitting needed for inelasticity is simultaneously produced .
if only elastic scattering is demanded , then an alternative model , with interference between the kinetically mixed gauge boson and a hidden sector scalar higgs , is adequate to give the required isospin violation . in both cases ,
the light kinetically mixed gauge boson is in the range of interest for currently running fixed target experiments .
|
one of the important open problems in the physics of the quantum hall effect ( qhe ) @xcite@xcite is the complete understanding of the hierarchical hall plateaus , whose filling fractions fall beyond the laughlin sequence @xmath1 @xcite .
there are two kinds of theoretical descriptions available at present : the wave - function constructions and the effective conformal field theories ( cft ) in @xmath2 dimensions .
the first approach has culminated in the jain theory of the composite - fermion correspondence between the integer hall states with @xmath3 and the hierarchical states with @xmath4 , @xmath5 , such as @xmath6 @xcite .
the existence of the composite - fermion excitations has been confirmed by many experiments @xcite ; the corresponding ansatz wave - functions have been tested in numerical simulations of few electron systems @xcite@xcite@xcite .
these have been mostly done on the spatial geometry of the sphere and have firmly established that the jain states describe the bulk excitations of quantum hall fluids . on the other hand ,
these incompressible fluids have characteristic edge excitations @xcite , which can not be seen on the sphere geometry .
these excitations are the relevant low - energy degrees of freedom in the conduction experiments , and their basic properties , like the fractional charge , have been already measured in the simplest ( yet non - trivial ) case of the @xmath7 laughlin hall state @xcite @xcite @xcite .
the edge excitations are naturally described by the conformal field theories dimensions @xcite . ]
@xcite , because their low - energy dynamics is effectively one - dimensional , being localized on the boundary of the sample within a width of the order of the magnetic length @xmath8 @xcite@xcite .
the conformal field theories are a powerful tool because they can be solved explicitly in a non - perturbative framework @xcite , and predict universal data like the filling fraction , the fractional charge and quantum statistics of the edge excitations ; moreover , they directly describe the relevant experimental regime of a large number of electrons .
however , these effective descriptions can not be easily related to the microscopic dynamics of the electrons .
actually , based on simple arguments and symmetry principles , two classes of cfts have been proposed for the edge excitations of the hierarchical plateaus .
the first is given by the @xmath9-component abelian theories @xmath10 , which are generalizations of the successful one - component theory for the laughlin plateaus @xcite@xcite@xcite . in the literature ,
these conformal theories have been naturally associated with the jain approach , because the composite - fermion correspondence implies the existence of several effective landau levels , which may have independent edges of the laughlin type .
the second class of conformal field theories encompasses the @xmath0 minimal models @xcite , which exploit the @xmath0 symmetry of the incompressible fluids under area - preserving diffeomorphisms of the plane @xcite@xcite@xcite .
there is a one - to - one correspondence between the hierarchical hall plateaus and the @xmath0 minimal theories .
these theories can be obtained by projecting out some states of the multi - component abelian theory , those which are not fluctuations of an elementary incompressible fluid .
this projection implies different properties for the edge excitations in the two theories , which are qualitatively and quantitatively important @xcite . in this paper
, we study numerically the spectrum of a finite systems of @xmath11 and @xmath12 electrons on a disk geometry ; we diagonalize exactly the hamiltonian with the haldane short - range interaction in the first landau level @xcite .
the low - lying states occur in branches which are separated by gaps ; each branch contains one state of lowest angular momentum , the `` bottom '' state , followed by several higher angular momentum levels with close energies ( see fig.([fig1 ] ) ) .
this pattern is well understood for the @xmath7 laughlin hall fluid : the bottom state is the ground state , which is an exact eigenstate of the haldane interaction ; the higher levels are the degenerate edge excitations .
the plot of density profiles shows that the ground state is rather flat , a characteristic of incompressible fluids @xcite , and that the edge excitations are infinitesimal density deformations .
we use the density plots to gain a qualitative understanding of the spectrum : we find that the other bottom states can be either quasi - particle excitations ( oscillating density profile ) , or new incompressible fluids with lower filling fractions , such as @xmath13 ( flat density profile ) .
furthermore , we overlap the low - lying states with the jain ansatz states .
the jain composite - fermion theory on the disk geometry does predict the branches in the spectrum : they correspond to effective landau levels which are filled selectively at the boundary .
these states are usually denoted by @xmath14 $ ] , and correspond to filling the @xmath15-th level with @xmath16 electrons , with @xmath17 and @xmath18 @xcite .
our numerical analysis shows that each branch in the spectrum is well described by one of these states , corresponding to a given set of fillings @xmath19 , and by the corresponding particle - hole excitations .
we then find the following results .
smooth density profiles identify unambiguously the branches corresponding to the @xmath13 and @xmath20 hierarchical plateaus , for each value of @xmath21 ; the rest are quasi - particles branches over the @xmath7 laughlin and the hierarchical plateaus ; previous numerical analyses
@xcite@xcite@xcite only charted the energy spectrum and were naturally led to misinterpret the branches .
next , we observe that , within each branch , the low - lying states can be divided into families of close density profiles : the edge excitations correspond to infinitesimal deformations of the bottom state , while families with different density profiles are other bulk excitations or magneto - phonons , which are not well separated in energy for these small values of @xmath21 .
this decomposition in families is first understood in the case of a quasi - particle over the laughlin state ( @xmath22 ) : we find a precise correspondence between the exact spectrum and the conformal field theory by using the jain composite - fermion theory , as it follows .
we begin by computing the overlap matrix between the exact low - lying states and the jain states of the branch ( for example , suppose that the quasi - particle corresponds to the filling @xmath23 $ ] for eight electrons ) .
these jain states are particle - hole excitations of two effective landau levels , and can be mapped one - to - one to the states of the two - component abelian conformal theory . as a consequence , each low - lying exact state in the branch
has a label of this conformal theory ( up to @xmath24 finite - size errors ) .
we then find that the edge excitations of the laughlin quasi - particle match certain conformal states which are symmetric with respect to the two abelian components , i.e. the two effective landau levels .
more precisely , these states are obtained by the projection @xmath25 relating the two - component and the one - component abelian theories .
therefore , the laughlin quasi - particles are indeed described by the one - component abelian theory , like the well - understood quasi - holes this is expected in a relativistic theory . moreover ,
the quasi - particle edge excitations amount to a specific sub - set of the larger jain spectrum .
having understood the pattern of the edge excitations of the laughlin quasi - particles , we proceed to analyse other branches in the spectrum , whose edge excitations are less understood . for eight electrons , we find that the edge spectrum of the @xmath13 hierarchical hall state and its first quasi - particle match the predictions of the @xmath0 minimal conformal theory .
this theory realizes the weaker projection @xmath26 @xcite , which again eliminates the antisymmetric excitations between the two abelian components , but keeps all the symmetric ones @xcite ; there remains a non - trivial sector of neutral excitations , as explained in section iic .
therefore , the numerical study support the description of the hierarchical plateaus by the @xmath0 minimal conformal theory .
the electrons form an irreducible , minimal incompressible fluid , which is in many respects analogous to the original laughlin fluid @xcite . the same pattern of families of edge excitations in the low - lying spectrum is also found for @xmath27 and @xmath12 electrons . for @xmath27 ,
the finite - size span for the angular momentum of edge excitations , @xmath28 , is too small to see any difference between the one - component abelian ( laughlin states ) and the @xmath0 minimal theories ( hierarchical states ) ; namely , the results are consistent with our picture but are not very significative . for @xmath29
, we find that the @xmath13 hall plateau clearly displays the edge spectrum predicted by the @xmath0 minimal theory , as in the @xmath30 case ; on the other hand , the @xmath31 hierarchical state and its quasi - particles are less neat .
anyhow , these data can not be consistently interpreted by the alternative multi - component abelian theory .
the paper is organized as follows . in section @xmath32
, we recall the main results of the jain theory and the predictions of the two classes of cfts , which are then used to analyse the data in section @xmath33 ( @xmath30 ) and section @xmath34 ( @xmath35 ) .
finally , we end up with some conclusions .
we consider a system of @xmath36 electrons in the first landau level on a disk geometry ; we use open boundary conditions , namely , we truncate the hilbert space of single - particle angular momentum states at a value that is very large compared to the highest occupied level in the laughlin state . the interaction among the electrons
is given by the haldane short - range potential , which selects the @xmath7 laughlin incompressible ground state .
of course , we are interested in studying the @xmath13 incompressible ground state , and other hierarchical states , for which there is no analogous model interaction .
an overview of the @xmath30 spectrum is shown in fig.([fig1 ] ) as a function of the angular momentum @xmath37 : one clearly sees the branches of low - lying states .
no confining potential is present , but this can be easily added afterwards : a quadratic potential increases the energy of each state by an amount proportional to @xmath37 ; thus , the bottom state of each branch can be made the ground state by suitably tuning the strength of the potential .
if this ground state has an approximately constant density , its filling fraction is @xmath38 , up to finite - size corrections of relative order @xmath24 .
let us finally quote the previous analysis on the disk geometry , which were useful in setting up our work @xcite@xcite@xcite . according to the composite - fermion theory , there is a correspondence between @xmath9 filled effective landau levels ( @xmath39 ) and the hierarchical hall states @xcite : ^*=m = m pm 1 , m=1,2 ,
, p=2,4, .
[ cfnu]this correspondence is made explicit by the ansatz wave - functions _ ( z_1, ,z_n ) = p
( _ i < j^n ( z_i - z_j ) ^2 _ ^*(z_1, ,z_n ) ) , [ jwf]where @xmath40 is the slater determinant of the filled landau levels and @xmath41 is the projector into the first landau level .
the composite - fermion correspondence also describes the excited states by inserting in ( [ jwf ] ) the slater determinants for the @xmath39 electron transitions . on the disk geometry
, the landau levels can be filled selectively by making electron droplets with different edge shapes : these `` bottom '' states are denoted by @xmath42 $ ] and correspond to the slater determinants of @xmath16 electrons filling the lowest angular momentum states of the @xmath15-th landau level @xcite .
each bottom state has an independent branch of low - lying states corresponding to the particle - hole excitations of the effective levels .
the jain correspondence ( [ jwf ] ) applied to these bottom states yields @xmath43 states which are characterized by the following angular momenta and approximate energies @xcite : j_[n_1, ,n_k ] & = & n(n-1 ) + n_1(n_1 - 1 ) 2 + n_2(n_2- 3 ) 2 + n_3(n_3 - 5 ) 2 + , + e^(0)_[n_1, ,n_k ] & = & e_d ( n_2 + 2 n_3 + 3 n_4 + ) .
[ jed]in the last equation , @xmath44 is the effective landau level gap , which has been recently interpreted as the energy for creating a `` defect '' in the electron fluid . ]
@xcite . in this work ,
the jain ansatz was also modified to be completely written in the first landau level ; nevertheless , in this paper we shall use the original proposal ( [ jwf ] ) and perform the projection @xmath41 numerically .
the jain bottom states ( [ jed ] ) for eight electrons are summarized in table ( [ tab1 ] ) : the @xmath45 $ ] state is identified as the laughlin @xmath7 incompressible state
. the rest of states could be candidates for the @xmath13 ground state , because they have @xmath46 up to @xmath47 finite - size corrections ; actually , in the thermodynamic limit @xmath48 , any state @xmath49 $ ] , with @xmath50 fixed , would work .
the jain bottom states match rather well the numerical spectrum in fig.([fig1 ] ) : for almost all predicted @xmath37 values in table ( [ tab1 ] ) , there is a bottom state of a branch of low - lying states ; its energy is rather well approximated by the jain formula ( [ jed ] ) .
on the other hand , a disagreement is seen for the smaller @xmath37 states @xmath51 $ ] and @xmath52 $ ] , which have @xmath53 but are almost degenerate with the @xmath54 states ; moreover , these two states seem to describe the same branch of levels we shall discuss these points later on .
it is rather well established that the low - energy excitations of a droplet of quantum incompressible fluid reside on the boundary and can be described by a @xmath2-dimensional conformal field theory @xcite .
this can be simply understood in the case of one filled landau level ( see fig.([fig2 ] ) ) , which can be considered as the fermi sea ( in configuration space ) of a one - dimensional system with one chirality only @xcite .
its low - lying excitations are particle - hole transitions near the fermi surface , which is actually the physical edge of the ( circular ) electron droplet . by a well - known procedure
, one can take the thermodynamic limit @xmath48 and approximate the fermi sea by a dirac vacuum ; moreover , the energy of the low - lying excitations can be linearized around the fermi level , @xmath55 , where @xmath56 is the momentum , @xmath57 the fermi velocity and @xmath58 the size of the disk .
these are the edge excitations of the @xmath59 quantum hall state ; in a finite system , their wave number is limited by @xmath60 , i.e. their angular momentum must be @xmath61 @xcite .
besides the edge excitations , there are quasi - hole excitations which amount to moving one electron from deep inside the fermi sea to the edge ( @xmath62 ) , and , conversely , for the quasi - particles . in the thermodynamic limit ,
the relativistic one - dimensional effective fermion is identified as the charged , chiral weyl fermion ; the conformal symmetry of this theory is described by the virasoro algebra with central charge @xmath63 @xcite .
this effective conformal field theory of the @xmath59 plateau can be generalized to the laughlin states by the well - known bosonization procedure : one rewrites the weyl fermion in terms of a bosonic field , changes its compactification radius and obtains a general @xmath63 abelian conformal theory , whose one - dimensional current satisfies the abelian current algebra @xmath64 @xcite ; an equivalent name is the chiral luttinger liquid @xcite .
this description of the edge excitations of the laughlin states is well established both theoretically and experimentally @xcite@xcite@xcite .
an important property for the following discussion is that the hilbert spaces of the edge excitations of the integer and laughlin hall states , @xmath1 , are all isomorphic @xcite ; thus , these excitations can be still visualized as the particle - hole transitions in fig.([fig2 ] ) .
it is rather simple to count them for small values of @xmath65 , and obtain the multiplicities reported in the first line of table ( [ tab2 ] ) @xcite ; let us stress that the cft description of edge excitations is valid in the linear range @xmath28 @xcite .
a natural generalization of the above picture is to consider the case of two filled landau levels : the corresponding edge excitations are described by the two - component abelian cft , which is simply the tensor product of two one - component theories .
each landau level has its own fermi surface and its particle - hole excitations ; their total number is obtained by adding the excitations of the two levels which have a given @xmath66 ( see the second row of table ( [ tab2 ] ) ) . according to the jain composite - fermion correspondence ( [ cfnu ] ) , it is rather natural to use this two - component abelian theory for describing the edge excitations of the hierarchical states ( [ jwf ] ) with @xmath67 . as in the one - component case
, these @xmath68 values can be described in the conformal theory by bosonizing the two weyl fermions and by suitably changing their compactification radii @xcite . a characteristic feature of the resulting conformal theories is to realize an extension of the abelian current algebra symmetry @xmath69 @xcite@xcite
; this will be useful in the following discussion .
as in the one - component case , the hilbert spaces of the edge excitations above the @xmath70 hall plateau and all the @xmath71 hierarchical states ( [ jwf ] ) are isomorphic .
another theory of the hierarchical edge excitations has been independently proposed in the ref.@xcite : it corresponds to the one - component abelian cft for the simplest laughlin fluids , but differs from the multi - component theory .
it is based on the physical picture of the incompressible fluid @xcite , which possesses the dynamical symmetry under the area - preserving diffeomorphisms of the spatial coordinates @xcite @xcite .
this symmetry has been promoted to a building principle for the cfts describing its edge excitations @xcite : actually , it implies that the conformal fields should carry a representation of the @xmath0 algebra , which is a generalization of the abelian current algebra @xcite . among the conformal theories with @xmath0 algebra @xcite ,
a particular class of models has been found , the @xmath0 _ minimal models _
@xcite , whose filling fractions are in one - to - one correspondence with the hierarchical values ( [ cfnu ] ) .
this is already a strong indication that these models are experimentally relevant .
the minimal models are characterized by being a reduction of the previous multi - component abelian cfts , in the sense that some excitations are projected out , as explained hereafter .
this projection implies @xcite : ( i ) a reduced number of edge excitations above the ground state , as given in the third line of table ( [ tab2 ] ) ; ( ii ) only one independent abelian charge for the quasi - particles , which is hence identified with the electric charge ; ( iii ) the existence of neutral quasi - particle excitations characterized by a non - abelian quantum number of the @xmath72 lie algebra , where @xmath9 is the number of would - be components .
this projection has been recently made explicit by a hamiltonian formulation of the minimal incompressible models @xcite , which will be briefly summarized in the first non - trivial case of two components ( @xmath73 ) .
let us start from a closer look into the two - component abelian theory with symmetry @xmath74 ; for @xmath70 , this is described by two weyl fermions @xmath75 , where @xmath76 is the angular variable on the disk and @xmath77 denote the upper and lower levels , respectively .
their excitations can be labelled by the total abelian charge @xmath78 and the @xmath79 isospin charge @xmath80 , which are defined as follows : j_0 & = & _ 0 ^ 2 d2 ( _ 1_1 + _
2_2 ) , + j_0 ^ 3 & = & _ 0 ^ 2 d2 12 ( _ 1_1 - _ 2_2 ) , + j_0^+ & = & _ 0 ^ 2 d2 _
1_2 , j_0 ^ - = _ 0 ^ 2 d2 _
[ jdef]one can check that @xmath81 satisfy the @xmath79 algebra by using the fermionic canonical commutation relations generate the current algebra @xmath82 which contains this @xmath79 as a subalgebra @xcite . ] .
the abelian and iso - spin charges measure the edge excitations in the two layers symmetrically and anti - symmetrically , respectively .
note , however , that the quasi - particles in these abelian theories carry the iso - spin quantum number in such a way that it is linearly additive as another electric charge ; namely , there are no non - abelian effects @xcite .
these @xmath79 generators can be defined for all the hierarchical ground states @xmath83 , but are not realized in terms of fermions ; we can nevertheless continue to use the more intuitive fermionic language , owing to the aforementioned isomorphism between hilbert spaces .
after these preliminaries , we are ready to define the @xmath0 minimal theory ( for @xmath84 ) : this is obtained from the two - component abelian theory by imposing the constraint @xcite , j^-_0 = 0 .
[ hamred ] it can be shown that the zero modes @xmath85 commute with the virasoro generators ; thus , the constraint ( [ hamred ] ) does not spoil the conformal invariance and defines a new conformal theory with the same central charge .
the effect of the constraint is the following : the operator @xmath86 moves electrons down and holes up between the two layers ( with a minus sign in the latter case ) , while keeping their ( normal - ordered ) angular momentum constant ; it relates the edge excitations in the two layers and actually vanishes on their symmetric linear combinations . therefore , the condition ( [ hamred ] ) projects out the edge excitations which are antisymmetric with respect to the two levels .
the ground state is unique and symmetric , then it satisfies the constraint : namely , the two cfts share the same ground state .
let us see some examples of allowed edge excitations in the minimal theory .
we first need to clarify the angular - momentum labels in cft .
the weyl fermions define a second - quantized relativistic theory , which describes excitations above the ground state ; therefore , the latter must be specified in order to relate this description to the numerical data .
for example , one can suppose that the @xmath30 bottom state @xmath87 $ ] with @xmath88 is the @xmath13 incompressible ground state ( see fig.([fig1 ] ) ) .
this fixes the fermi surface for both layers and sets the reference value for the angular momentum of excitations : @xmath89 and @xmath90 .
the usual moding of conformal fields thus corresponds to the normal - ordered angular momentum , which is equal to zero for the charges in ( [ jdef ] ) .
the ground state @xmath87 $ ] and its edge excitations are identified by vanishing eigenvalues of both @xmath78 and @xmath91 ; the other jain branches in the spectrum , corresponding to @xmath23 $ ] , etc .
, in table([tab1 ] ) , are described by the cft as excitations with @xmath92 , etc , respectively .
let us now introduce the fermionic fock space of the two weyl fermions ( [ jdef ] ) , which also describe the particle - hole excitations of the two effective landau levels , `` up '' and `` down '' , in the jain construction : the fermionic second - quantized operators are , respectively , @xmath93 , @xmath94 and @xmath95 , @xmath96 ( @xmath97 ) ; they act on the ground state @xmath98 .
there are two abelian edge excitations at the first excited level @xmath99 , which can be written : 1 ; = ( d_1^d_0 u_1^u_0 ) .
[ delta1]the constraint ( [ hamred ] ) can be written explicitly : @xmath100 .
we find that the symmetric combination @xmath101 satisfies this constraint and the antisymmetric does not : therefore , the @xmath0 minimal conformal theory only contains the symmetric excitations , as we anticipated . at the next level @xmath102
, there are five abelian edge states : 2 ; a & = & d_1^d_0 u_1^u_0 , + 2 ; b&= & ( d_2^d_0 u_2^u_0 ) , + 2 ; c&= & ( d_1^d_-1 u_1^u_-1 ) . [ deltwo]the antisymmetric combinations @xmath103 and @xmath104 do not satisfy the constraint ( [ hamred ] ) and are not present in the minimal cft . note that the counting of these states is in agreement with table ( [ tab1 ] ) . in the semiclassical picture developed in ref.@xcite , the incompressible hall fluid is identified with a fermi sea and its area - preserving deformations are the particle - hole excitations ( fig.([fig2 ] ) ) .
the two - level structure introduces one additional degree of freedom : the antisymmetric excitations are tangential to the fermi surface and do not correspond to deformations of the incompressible fluid ; therefore , they need not to be included in a minimal theory .
this is the physical meaning of the condition ( [ hamred ] ) .
a precise derivation of this constraint can be obtained by analysing the irreducible representations of the @xmath0 algebra .
a general property of cfts is that they can be constructed by assembling the representations of their infinite - dimensional symmetry algebra .
the @xmath0 minimal theories were first obtained in such a way @xcite , by using the special , degenerate @xmath0 representations , which are equivalent to those of the algebra @xmath105 , in particular @xmath106 for @xmath71 @xcite .
the constraint ( [ hamred ] ) precisely realizes the projection of conformal theories @xmath107 @xcite , which is a simplified case of the general mechanism of hamiltonian reduction @xcite .
note that the projection only keeps one state for each @xmath79 multiplet in the abelian spectrum , i.e. the highest - weight state ; these states form a non - trivial set of neutral edge excitations , which are characterized by their virasoro weight .
the edge excitations of the other , @xmath108 part are generated by the modes of the symmetric current in ( [ jdef ] ) , j_-k = _
j=-^ d^_k - jd_j + u^_k - j u_j , k=1,2 , [ udiag ] these excitations are symmetric with respect to the two landau levels , but are not the most general ones : for example , the state @xmath109 in ( [ deltwo ] ) can not be obtained by applying ( [ udiag ] ) on the ground state .
the symmetric edge excitations obtained by ( [ udiag ] ) realize the reduction of the two - component abelian theory to the one - component theory @xmath25 , which also entails a change of central charge @xmath110 .
finally , we remark that the constraint ( [ hamred ] ) can be enforced dynamically by modifying the hamiltonian of the two - component abelian theory : h h + j^+_0 j^-_0 .
[ ham]the added term is diagonal in the two - component abelian hilbert space , and is relevant in the renormalization - group sense , because @xmath111 has dimension of a mass .
it increases the energy of the states which do not satisfy ( [ hamred ] ) and in the limit @xmath112 , it performs the projection leading to the @xmath0 minimal model . in general
, we may also consider the non - conformal theory with @xmath113 , which interpolates between the two - component and the minimal , i.e. irreducible , incompressible fluids .
a more complete discussion of these matters can be found in ref.@xcite .
we now proceed to analyse each branch of levels in the @xmath30 spectrum and interpret it according to the theories of the previous section .
the exact eigenstates are denoted by @xmath114 , where @xmath37 is the angular momentum and @xmath115 the ordering by increasing values of the energy .
the complete set of our numerical data is accessible on - line . * branch [ 8 ] : the laughlin incompressible hall fluid at @xmath7 * the density profile @xmath116 of the laughlin ground state @xmath117 is drawn in fig.([fig3 ] ) : we see that this droplet of incompressible fluid is fairly flat in the interior ; moreover , its value at the origin . ] @xmath118 is close to the expected value of @xmath119 in the thermodynamic limit .
the edge excitations are recognized as very small deformations of the shape of the ground state density ; their energies above the ground state vanish for the haldane interaction , as shown in the inset of fig.([fig3 ] ) .
having identified the edge states , we can count their number : we find the multiplicities @xmath120 for @xmath121 , in agreement with the predictions of the one - component abelian cft in table([tab2 ] ) .
the matching of the exact edge excitations with the the corresponding jain ( laughlin ) wave functions is rather obvious in this case ; the total overlap with each exact numerical state is one by construction .
* branches [ 7,1 ] , [ 6,2 ] : the quasi - particles over the laughlin state * fig.([fig30 ] ) shows the density profiles of all the bottom states found in the @xmath30 spectrum : these are all more oscillating than the laughlin ground state .
the states @xmath122 and @xmath123 are clearly quasi - particle excitations with the characteristic bump of size @xmath124 : therefore , they can not be interpreted as other incompressible ground states with @xmath125 .
let us analyse the @xmath23 $ ] branch @xmath126 in more detail : the density profiles and energies of its low - lying states are drawn in fig.([fig4 ] ) .
one distinguishes two families of states with opposite oscillations , which have multiplicities @xmath127 and @xmath128 , respectively . the first group can be interpreted as a quasi - particle on top of the laughlin @xmath7 ground state : actually , the multiplicities of its edge excitations agree with the predictions of the one - component abelian conformal theory ( table ( [ tab2 ] ) ) . the second group can be interpreted as another bulk excitation , which might acquire a larger gap in the thermodynamic limit , where the cft description of the low - lying states should become exact .
although this limit can not be inferred from the simulation , the inset of magnified energy levels in fig.([fig4 ] ) clearly shows that these states have a higher energy than those of the previous group .
note that the sum of the degeneracies of the two groups considered above match the predictions of the two - component abelian cft in table ( [ tab2 ] ) , i.e. @xmath129 . by ignoring the information arising from the density plots , one might be tempted to interpret all these low - lying states as edge excitations ; then , the corresponding bottom state could not be a laughlin quasi - particle , because the edge multiplicities would not match ; it would rather be interpreted as the @xmath13 incompressible fluid @xcite
. however , the @xmath130 branch , i.e. @xmath131 $ ] , being qualitatively similar , should be interpreted in the same way , while it is necessarily a laughlin quasi - particle , thus arriving to a contradiction .
we conclude that the new criterion of density plots is essential for distinguishing laughlin quasi - particles from new hierarchical plateaus .
we can actually observe the formation of a hierarchical hall fluid @xcite : as soon as sizable number @xmath132 of laughlin quasi - particles are present , they condense into a new incompressible fluid , whose density shape is again smooth ( see next paragraph ) .
after having established that @xmath126 is a laughlin quasi - particle , we should understand the origin of the extra family @xmath128 of low - lying states .
this is explained by the jain theory , which we now analyse in detail ; we are going to illustrate the following chain of relations : ccc low - lying exact states & jain states & two - component abelian cft + & & + edge excitations & & one - component abelian cft [ rela ] the jain wave functions corresponding to the @xmath23 $ ] branch are obtained by plugging in ( [ jwf ] ) the slater determinants for the particle - hole excitations of the first and second levels filled with @xmath133 and @xmath32 electrons , respectively .
these states are , by construction , in one - to - one correspondence with those of the two - component abelian conformal theory ; thus , we use the same notation for them , in particular the basis of section iic , equations ( [ delta1 ] ) and ( [ deltwo ] ) .
note , however , that the jain states are not orthogonal ; therefore , we diagonalize them by the gram - schmidt method ( by also computing the overlaps among themselves ) . for each angular momentum value @xmath134 , we have computed the matrix of overlaps between these jain states and the low - lying spectrum ( see table ( [ tab20 ] ) ) . the total square overlap with each state is very large ( about 0.96 ) , and the matrix determinant is large enough to match these states one - to - one .
we conclude that the jain theory describes very well all the low - lying states .
nonetheless , this does not imply that they all should be edge excitations , as shown by the previous arguments of the density plot and of the consistency with the laughlin theory .
let us recall that , in principle , the jain theory and the conformal field theory give rather different descriptions of the spectrum : the former applies directly to all excitations for finite @xmath21 , while the latter only accounts for the edge excitations in the large @xmath21 limit , which requires some imagination .
the analysis of the overlap matrices in table ( [ tab20 ] ) let us identify which jain states correspond to the edge excitations of the laughlin quasi - particle in fig.([fig5 ] ) . the edge state @xmath135 clearly matches the symmetric state @xmath136 in equation ( [ delta1 ] ) .
the five states at @xmath137 can be divided in two groups : ( @xmath138 , @xmath139 , @xmath140 ) have larger overlaps with the symmetric states in ( [ deltwo ] ) , and ( @xmath141 , @xmath142 ) with the antisymmetric states .
the precision is lower than for @xmath143 , but it is nonetheless remarkable that the signal is not completely washed out by the finite - size effects . this pattern of overlaps can be understood as a relation among conformal field theories by applying the analysis of section iic : the edge excitations of the laughlin quasi - particle , @xmath135 , @xmath138 and @xmath139 , have large overlaps with the symmetric states of the two - component abelian theory which are generated by the @xmath108 current ( [ udiag ] ) , namely @xmath136 , @xmath144 and @xmath145 .
therefore , the laughlin quasi - particle is indeed described by the one - component abelian conformal theory ; the new result is that this is obtained in the jain spectrum by the projection @xmath25 described in the section iic .
similar families of symmetric laughlin edge excitations with multiplicities ( 1,1,2 ) are actually found in any branch and for any number of electrons : this is perhaps the most important result of this paper .
these `` standard '' families are enlarged by further edge excitations in the case of hierarchical hall states , as discussed hereafter .
* branches [ 5,3 ] , [ 5,2,1 ] : the @xmath13 hierarchical hall state and its quasi - particle * fig.([fig5 ] ) shows the family of low - lying states starting at @xmath88 , which is identified with the @xmath87 $ ] branch .
the bottom state @xmath146 has a rather flat profile and can be considered as a new incompressible ground state ; the value of @xmath147 is close to its thermodynamic limit of @xmath148 ( at variance with the quasi - particle state @xmath149 ) .
moreover , a family of edge excitations is clearly associated to @xmath146 , which have multiplicities @xmath150 , in agreement with the predictions of the @xmath0 minimal cft ( third row of table ( [ tab2 ] ) ) .
next , we check whether these edge excitations match the states of the minimal theory , by analysing the overlaps with the jain states ( see table ( [ tab3 ] ) ) .
we find that the ground state @xmath151 is well - approximated by the jain bottom state @xmath152\rangle$ ] .
for @xmath143 , the unique edge state @xmath153 is again identified with the symmetric jain state @xmath154 . at @xmath102 ,
the exact edge states @xmath155 , with @xmath156 have large overlaps with the symmetric jain states @xmath157 , @xmath158 and @xmath159 , and @xmath24 projection on the antisymmetric jain states .
therefore , the edge excitations match all the possible symmetric jain states ; according to section iic , these are singled out by the projection @xmath160 , leading to the @xmath0 minimal theory . putting all these informations together , we conclude that this branch can be definitely interpreted as the @xmath13 incompressible fluid for @xmath30 electrons , and that its edge excitations are described by the @xmath0 minimal conformal theory @xcite within the finite - size accuracy .
let us add one remark .
the third state @xmath161 at @xmath137 which makes the difference between the ubiquitous laughlin edge spectrum and the hierarchical one , has an alternative description in the jain theory as the bottom state of the @xmath162 $ ] branch .
here we first encounter the phenomenon of superposition of jain branches , which is a sort of degeneracy in the description of the spectrum ; it is an interesting dynamical problem in the jain theory , which has not been addressed so far .
it can be rephrased as an `` interaction among defects '' in the regime of compact electron droplets @xcite ; it might also be associated to the mechanism of the `` condensation of quasi - particles '' of the original hierarchical scheme @xcite ( or maybe not ) . as a matter of fact , this dynamical phenomena does not concern the cft interpretation , which is an effective description of the edge excitations above a given ground state
; there is just a practical consequence that the one - to - one identification of states in ( [ rela ] ) might be lost in case of mixing of two competing jain branches .
the overlaps of the @xmath162 $ ] jain branch with the @xmath163 low - lying states are not very good , although the bottom state itself has a good overlap .
this branch is rather atypical , and can not be considered as a candidate incompressible state because it is almost gapless with the @xmath87 $ ] branch .
we now proceed to analyse the branch @xmath164 , which is plotted in fig.([fig6 ] ) .
its natural interpretation is of a quasi - particle over the @xmath13 ground state , due to the oscillating density profile .
the @xmath0 minimal theory predicts the same pattern of edge excitations as for the @xmath13 ground state , i.e. ( 1,1,3 ) @xcite ; this is actually observed in fig.([fig6 ] ) , and add further support to whole picture put forward in this paper .
a slight peculiarity is that the third state at @xmath137 is higher in energy than other non - edge states but the shape is more important in our analysis .
next , we proceed to the identification of these states by the overlap analysis ( see table ( [ tab40 ] ) ) .
the jain theory presents a degenerate description for this branch , namely @xmath51 $ ] and @xmath52 $ ] ( see table ( [ tab1 ] ) ) .
the overlaps show that the latter wins , i.e. the states of the @xmath51 $ ] branch have less than @xmath24 projection on the low - lying states ; apparently , the electron tends to `` pile up '' in the effective landau levels .
the empirical rule taken from this and other cases of two competing jain branches is that one of them describes well the data , while the other definitely does not .
the jain states of the @xmath52 $ ] branch should be related to a three - component abelian conformal theory ; this is the first step in extending the chain of relations ( [ rela ] ) to this branch . for @xmath143 , there are three abelian edge states , which can be written , in analogy with eq.([hamred ] ) : 1 ; a & = & 1 ( u^_1 u_0 + c^_1 c_0 + d^_1 d_0 ) , + 1 ; b & = & 1 ( u^_1 u_0 - c^_1 c_0 ) , + 1 ; c & = & 1 ( u^_1 u_0 + c^_1 c_0 - 2 d^_1 d_0 ) .
[ three]the ground state is @xmath165\rangle$ ] and the @xmath166 are fermionic fock operators for the upper , central and lower ( effective ) landau levels . in this case
, we expect that the edge excitations are described by a reduction from the three - component abelian theory to the @xmath84 @xmath0 minimal theory ( analogous to the @xmath110 projection ( [ rela ] ) for the laughlin quasi - particles ) .
it can be shown @xcite that this reduction implies that the unique edge excitation at @xmath143 is the completely symmetric state @xmath167 .
this is in agreement with the overlaps reported in table ( [ tab30 ] ) .
next , the overlap analysis can not be carried over to the @xmath137 edge excitations , owing to the finite - size limitations : the jain state @xmath52 $ ] possesses only one electron in the highest effective landau level , and is very far from the thermodynamic limit of three fermi surfaces which is implicit in the cft description ; in particular , one of the @xmath137 particle - hole excitations in the higher level is missing .
therefore , we can not analyse the projection of states in this branch ; nevertheless , the shape and number of edge excitations are in agreement with the predictions of the @xmath0 minimal theory for a quasi - particle over the @xmath13 state .
fig . ( [ fig7 ] ) shows the exact spectrum as a function of the angular momentum and table ( [ tab4 ] ) reports the bottom states of the jain branches with @xmath168 .
the observed structure in branches is similar to that of the @xmath30 case : the laughlin branch starts at @xmath169 , and its ground state is identified with the @xmath170 $ ] jain state . in decreasing order of @xmath37 , we identify as @xmath171 $ ] the state appearing at @xmath172 and as @xmath173 $ ] the one at @xmath174 . for the state appearing at @xmath175
there are two possible jain states , @xmath176 $ ] and @xmath177 $ ] , which have the same energy @xmath178 .
figure ( [ fig70 ] ) shows the density profiles of all the bottom states : these give the first hints for identifying the incompressible ground states .
the profiles should be compared with the data of average density @xmath179 and angular momentum for idealized flat droplets , which are reported in table ( [ tab40 ] ) .
one finds that the bottom states with @xmath172 and @xmath175 are candidates for the incompressible hall states with @xmath13 and @xmath180 , respectively .
let us analyse them in turn .
the profiles of the low - lying states of the @xmath172 branch are shown in fig.([fig71 ] ) : the edge states are @xmath181 , @xmath182 and @xmath183 , i.e. the observed multiplicities are @xmath127 . the analysis of the overlaps with the jain wave functions is similar to the previous cases : the results are that the bottom state @xmath184 is identified with @xmath185\rangle$ ] , as expected ; for @xmath143 , the edge state @xmath186 is the symmetric combination of abelian edge states @xmath136 ( see eq.([delta1 ] ) ) ; instead , the non - edge eigenstate is @xmath187 .
the overlap analysis can not be extended to @xmath137 , because the @xmath171 $ ] ground state has only one electron in the second effective landau level , and its particle - hole excitations can not match the two - component abelian cft . within the @xmath143 analysis , we can not distinguish the minimal @xmath84 edge ( multiplicities @xmath188 ) from the @xmath63 abelian edge @xmath189 associated with the laughlin quasi - particle excitations . in conclusion
, this branch can be interpreted either as the @xmath13 incompressible hall state described by the @xmath0 minimal theory ( but one edge excitation is missing ) , or as a laughlin quasi - particle ( but its profile is exceptionally flat ) .
the density profiles of the @xmath175 branch are presented in fig .
( [ fig8 ] ) : the edge excitations are @xmath190 , @xmath191 and @xmath192 , i.e. again laughlin - like multiplicities .
their interpretation in conformal field theory begins by identifying the ground state @xmath193 with one of the two possible jain bottom states : the overlaps in table ( [ tab5 ] ) shows that the exact state is well described by the three - level state @xmath177 $ ] , rather than the `` simpler '' @xmath176 $ ] one ; this is the `` piling - up '' of the electrons , which we have already encountered .
actually , the computation of the energies of the two jain bottom states shows that the @xmath176 $ ] branch is separated from the low - lying @xmath177 $ ] branch by a gap of order @xmath44 ; in this case , the interaction among defects is of the same order as the energy @xmath178 for non - interacting defects @xcite .
next , we analyse the low - lying states of this branch : the jain excitations of the @xmath177 $ ] bottom state should be compared with the three - component abelian cft , as we have already done for the branch @xmath52 $ ] of @xmath30 electrons .
again , we must limit ourselves to the excitations with @xmath143 , due to the finite sizes of the would - be fermi seas .
the results for the overlaps in table ( [ tab5 ] ) are very similar to those in table ( [ tab30 ] ) for the quasi - particle over the @xmath13 state : the unique edge state @xmath194 is clearly identified as the symmetric excitation @xmath167 .
the density shape of the @xmath175 bottom state suggests its interpretation as the @xmath31 hierarchical hall state for @xmath27 electrons .
then , we expect that its edge excitations are described by the @xmath195 @xmath0 minimal conformal theory , whose multiplicities are again @xmath150 ( see the last row of table ( [ tab2 ] ) ) @xcite ; actually , one can show that the excitations of the @xmath195 and @xmath84 minimal theories only differ for @xmath196 , and that both agree with those of the simpler laughlin theory for @xmath143 . in conclusion ,
the @xmath27 exact edge states are consistently described by the @xmath0 minimal conformal theory @xcite , within the finite - size limitations : we recall that the span for edge excitations in cft is @xmath197 @xcite .
another consistent interpretation for all the branches is given by the laughlin one - component theory ; possibly , the condensation of quasi - particles leading to the hierarchical hall fluids can not take place in such a small system . nevertheless , it is important that the @xmath27 and the @xmath30 data can be consistently interpreted . the spectrum of energies as a function of the angular momentum for @xmath29 is presented in fig .
( [ fig10 ] ) and the branches of jain states are given in table ( [ tab6 ] ) ; the laughlin ground state @xmath198 $ ] and the @xmath199 $ ] branch are not presented in fig .
( [ fig10 ] ) , which focusses on the @xmath200 region .
the comparison between the exact branches and the jain predictions shows the dynamical phenomenon already seen before : for any pair of jain branches which have degenerate energy @xmath178 ( table ( [ tab6 ] ) ) , only one is realized in the spectrum , and correctly describes the low - lying states ( as shown by the overlap analysis ) .
actually , @xmath201 matches @xmath202 $ ] , but @xmath203 $ ] is not observed ; similarly @xmath204 matches @xmath205 $ ] , @xmath206 is @xmath207 $ ] and @xmath208 is @xmath209 $ ] , while @xmath210 $ ] , @xmath211 $ ] and @xmath212 $ ] are not observed ; these are new examples of the `` piling up '' effect .
the analysis of the density profiles of all the bottom states in fig.([fig100 ] ) shows that the branches @xmath213 and @xmath214 are quasi - particles , with the same qualitative features of the analogous @xmath30 cases .
the bottom states of the @xmath201 and @xmath215 branches are not growing in the bulk : although rather oscillating , they are the possible candidates for the @xmath13 and @xmath31 incompressible hall states with @xmath29 electrons , respectively ( see the data in table ( [ tab40 ] ) ) . the by - now standard analysis of the edge excitations is first performed on the @xmath201 branch ;
the plots are shown in fig.([fig101 ] ) and the overlaps with the jain states are reported in table ( [ tab60 ] ) .
the usual laughlin edge excitations ( @xmath216 , @xmath217,@xmath218 ) are clearly seen in the plots and they definitely overlap on the symmetric jain states @xmath136 , @xmath219 , @xmath144 and @xmath159 . according to the @xmath0 minimal conformal theory , the interpretation of this branch as the @xmath13 hierarchical hall state requires the identification of a third edge excitation with @xmath137 .
this can reasonably be @xmath220 : its density profile is a non - infinitesimal , but in - phase , deformation of the bottom state , and its projection is large on the symmetric jain states ( see table ( [ tab60 ] ) ) . we conclude that the @xmath0 minimal theory consistently describes this branch as the @xmath13 hierarchical state : this is the third definite evidence for this theory .
the next branch @xmath204 , i.e. @xmath205 $ ] , is plotted in fig.([fig102 ] ) : the profile of the bottom state grows in the bulk and this fits the natural expectation that this branch is a quasi - particle over the @xmath13 state .
the families of low - lying states are actually very similar to those of the analogous @xmath30 branch ( @xmath164 ) in fig.([fig6 ] ) .
the multiplicities of the edge excitations should again be ( 1,1,3 ) as for the @xmath13 ground state ( they are observed in the corresponding @xmath30 branch ) ; instead the actual computing yields ( 1,1,2 ) presumably , the missing state is higher in energy than @xmath221 .
let us now discuss the @xmath206 branch , which is drawn in fig.([fig11 ] ) ; the corresponding overlaps with the jain branch @xmath207 $ ] are reported in table ( [ tab7 ] ) .
the density profile of the bottom state is similar to that of the @xmath201 state , i.e. the @xmath13 hall state ; thus , @xmath222 can be interpreted as the @xmath31 hierarchical state for @xmath29 .
the overlap matrix is rather standard for a three - level jain branch and identifies the cft labels of the @xmath143 low - lying states ; on the other hand , the @xmath137 excitations can not be analysed for the usual reason that there is a single electron in the highest effective landau level .
the edge excitations are identified by the density profiles and their multiplicities are found to be ( 1,2,4 ) : this is in disagreement with the @xmath195 @xmath0 minimal conformal theory , which predicts the values ( 1,1,3 ) ( see table ( [ tab2 ] ) ) .
it is possible that there is an accidental degeneracy with another family ( 0,1,1 - 2 ) of excitations .
this interpretation is supported by the analysis of the next branch @xmath208 , which is shown in fig .
( [ fig12 ] ) .
the bottom state is identified as a quasi - particle over the @xmath31 state , as expected ; the multiplicities ( 1,1 , ) of its edge excitations are again in agreement with the @xmath0 minimal theory . in conclusion
, the @xmath29 spectrum presents the general features already encountered for @xmath30 and @xmath133 ; the edge excitations of its hierarchical states can be interpreted within the @xmath0 minimal conformal theory ( and the laughlin quasi - particles by the one - component abelian theory , of course ) .
however , we should remark that the finite - size effects are not smaller than in the @xmath30 spectrum ; this is contrary to the expectation of a smooth thermodynamic limit towards the conformal field theory .
in this paper we have presented a comprehensive analysis of the low - lying spectrum of the electrons in the quantum hall effect , in the regime of hierarchical hall states @xmath200 .
we have gone beyond previous works along these lines : regarding the studies of the composite - fermion correspondence @xcite @xcite , we have done the first detailed analysis of the edge state structure , which was overlooked by the studies on the spherical geometry @xcite@xcite .
we have shown that the jain composite - fermion theory describes very well the low - lying spectrum ; but we also revealed the dynamical mechanism which takes place when two jain states are allowed : the electrons tend to `` pile up '' , if they are let to freely fill the effective landau levels . moreover , we improved and reviewed previous analyses on the disk geometry @xcite@xcite@xcite .
we introduced two new criteria for the analysis of the exact states : i ) the plot of their density profiles for distinguishing quasi - particle excitations from new incompressible ground states , and for identifying the real edge excitations within the low - lying states ; ii ) the interpretation of their overlaps with the jain states in the language of conformal field theory , with concrete relations among the states and projections thereof .
we have presented a consistent analysis of the low - lying spectrum .
we have shown that the edge excitations form specific subset of the low - lying states , by applying the previous criteria and by checking that the laughlin quasi - particles are described by the well - understood one - component abelian conformal theory . while all the low - lying states are nicely described by the jain theory , i.e. by the multi - component abelian conformal theory @xcite@xcite@xcite
, the real edge excitations of the hierarchical hall states match the predictions of the @xmath0 minimal theory @xcite ( and those of the laughlin quasi - particles naturally match the one - component abelian theory ) .
although the numerical data show some blurs and some finite - size limitations , the general picture seems firmly established ; we found four neat positive evidences out of the six hierarchical states with @xmath30 and @xmath12 electrons . in conclusion , we hope that this work will stimulate further analyses of the hierarchical hall states .
a. c. and g. r. z. would like to thank the c.e.r.n .
theory division and the theory group at l.a.p.p . , annecy , for hospitality .
a. c. also thanks the theory group of the centro atmico bariloche for hospitality and acknowledges the partial support of the european community program fmrx - ct96 - 0012 .
g. r. z. is grateful to i.n.f.n .
sezione di firenze for hospitality .
the work of g. r. z. is supported by a grant of the antorchas foundation ( argentina ) . for a review see : r. a. prange , s. m. girvin , _ the quantum hall effect _ , springer verlag , new york ( 1990 ) . for a review see : s. das sarma and a. pinczuk , _ perspectives in quantum hall effects _ , wiley , new york ( 1996 ) .
r. b. laughlin , ( 1983 ) 1395 ; for a review see : r. b. laughlin , _ elementary theory : the incompressible quantum fluid _ , in @xcite
. j. k. jain , ( 1989 ) 199 ; ( 1990 ) 7653 ; for reviews see : j. k. jain , _ adv . in phys .
_ * 41 * ( 1992 ) 105 , and _ composite fermions _ , in @xcite .
for a review see : h. l. stormer and d. c. tsui , _ composite fermions in the fractional quantum hall effect _ , in @xcite for a review see : e. fradkin and a. lopez , ( 1993 ) 67 . x. g. wu and j. k. jain , ( 1995 ) 1752 .
j. k. jain and r. k. kamilla , ( 1997 ) 2621 . for a review ,
see : x. g. wen , ( 1992 ) 1711 , _ adv . in phys .
_ * 44 * ( 1995 ) 405 .
r. c. ashoori , h. l. stormer , l. n. pfeiffer , k. w. baldwin and k. west , ( 1992 ) 3894 ; f. p. milliken , c. p. umbach and r. a. webb , _ solid state commun .
_ * 97 * ( 1996 ) 309 ; p. fendley , a. w. w. ludwig and h. saleur , ( 1995 ) 8934 ; for a review , see : c. l. kane and m. p. a. fisher , _ edge - state transport _ , in @xcite .
v. j. goldman and b. su , _ science _ * 267 * ( 1995 ) 1010 ; r. de - picciotto et al .
, cond - mat/9707289 ; l. saminadayar et al .
, cond - mat/9706307 . j. frhlich and a. zee , ( 1991 ) 517 ; x .- g .
wen and a. zee , ( 1993 ) 2290 .
j. frhlich and e. thiran , _ j. stat .
* 76 * ( 1994 ) 209 ; j. frlich , t. kerler , u. m. studer and e. thiran , ( 1995 ) 670 . a. a. belavin , a. m. polyakov and a. b. zamolodchikov , ( 1984 ) 333 ; for a review see : p. ginsparg , _ applied conformal field theory _ , in _ fields , strings and critical phenomena _
, les houches school 1988 , e. brezin and j. zinn - justin eds .
, north - holland , amsterdam ( 1990 ) . m. stone , ( ny ) * 207 * ( 1991 ) 38 .
a. cappelli , g. v. dunne , c. a. trugenberger and g. r. zemba , ( 1993 ) 531 .
n. read , ( 1990 ) 1502 .
a. cappelli , c. a. trugenberger and g. r. zemba , ( 1995 ) 470 ; for a review , see : ( 1996 ) 112 .
s. iso , d. karabali and b. sakita , ( 1992 ) 700 , ( 1992 ) 143 .
a. cappelli , c. a. trugenberger and g. r. zemba , ( 1993 ) 465 , ( 1993 ) 100 ; for a review , see : a.cappelli , g.v.dunne , c.a.trugenberger and g.r.zemba , ( 1993 ) 21 .
a. cappelli , c. a. trugenberger and g. r. zemba , ( 1994 ) 1902 .
f. d. m. haldane , _ the hierarchy of fractional states and numerical studies _ , in @xcite .
g. dev and j. k. jain , _ phys .
_ * 45 b * ( 1992 ) 1223 .
m. kasner and w. apel , _ phys . rev . _ * 48 b * ( 1993 ) 11435 ; _ ann
. physik _ * 3 * ( 1994 ) 433 .
a. cappelli and g. r. zemba , _ hamiltonian formulation for the minimal models of the incompressible quantum hall fluids _ , to appear . v. kac and a. radul , _ comm .
* 157 * ( 1993 ) 429 ; h. awata , m. fukuma , y. matsuo and s. odake , _ prog .
( supp . ) _ * 118 * ( 1995 ) 343 ; e. frenkel , v. kac , a. radul and w. wang , ( 1995 ) 337
. a. m. polyakov , ( 1990 ) 833 ; m. bershadsky and h. ooguri , ( 1989 ) 49 .
f. d. m. haldane ( 1983 ) 605 ; b. i. halperin , 91984 ) 1583 ; m. greiter , ( 1994 ) 48 .
.bottom states of the jain theory for @xmath30 , ordered by the decreasing angular momentum @xmath223 , and the corresponding approximated energies @xmath178 , in units of @xmath44 .
[ cols="^,^,^",options="header " , ]
|
we present a detailed analysis of the exact numerical spectrum of up to ten interacting electrons in the first landau level on the disk geometry .
we study the edge excitations of the hierarchical plateaus and check the predictions of two relevant conformal field theories : the multi - component abelian theory and the @xmath0 minimal theory of the incompressible fluids .
we introduce two new criteria for identifying the edge excitations within the low - lying states : the plot of their density profiles and the study of their overlaps with the jain wave functions in a meaningful basis .
we find that the exact bulk and edge excitations are very well reproduced by the jain states ; these , in turn , can be described by the multi - component abelian conformal theory .
most notably , we observe that the edge excitations form sub - families of the low - lying states with a definite pattern , which is explained by the @xmath0 minimal conformal theory . actually , the two conformal theories are related by a projection mechanism whose effects are observed in the spectrum .
therefore , the edge excitations of the hierarchical hall states are consistently described by the @xmath0 minimal theory , within the finite - size limitations .
200 1 2_1
|
in the mixed state , the magnetization of superconductors is a combination of two different contributions , @xmath6 and @xmath7 .
@xmath6 is the equilibrium ( or reversible ) magnetization@xcite and @xmath7 is the irreversible magnetization .
the former is caused by the equilibrium surface current .
the latter arises from the surface ( bean - livingston ) barrier effect,@xcite as well as the bulk pinning due to the interaction between vortices and various defects within the superconductor .
the surface barrier originates from the competition between two forces , ( a ) an attractive interaction between a vortex and its image vortex and ( b ) a repulsive interaction between a vortex and the surface shielding current . for high-@xmath8 cuprate superconductors , the irreversible magnetization at low temperatures
is dominated by the bulk pinning .
however , the role of the surface barrier effect becomes significant as the temperature increases.@xcite recently , superconductivity in a non - cuprate binary compound mgb@xmath9 was discovered by akimitsu _
et al._@xcite this material is known to be type ii superconductor with the ginzburg - landau parameter @xmath10 26 and @xmath11 40 k,@xcite and various experimental studies@xcite have been carried out to elucidate the fundamental properties of this new superconductor . in the mixed state ,
the magnetic behavior of mgb@xmath9 has been known to resemble that of the conventional superconductors such as nb - ti and nbsn@xmath12 .
et al_.@xcite showed that the parameter @xmath13 is linear in @xmath14 over a wide range of temperature as in nb@xmath12sn , where the @xmath15 is the magnetic hysteresis in @xmath2 .
they also reported the proportionality of the irreversible field @xmath16 and the upper - critical field @xmath17 , which are usually independent in high-@xmath8 cuprates . in this work , we measured magnetization @xmath2 of mgb@xmath9 superconductor as a function of the external magnetic field to elucidate its pinning properties in detail .
we found that the @xmath2 curves for various temperatures can be describe by the exponential critical state model.@xcite from this analysis , we present evidence of the significant role of bulk pinning in this system even up to @xmath18 , which is contrary to the case of high-@xmath8 cuprates .
a commercially available powder of mgb@xmath0 ( alfa aesar)@xcite was used to make a pellet .
high - pressure heat treatment was performed with a 12-mm cubic multi - anvil press.@xcite the pellet was put into a au capsule in a high - pressure cell . a _
d_-type thermocouple was inserted near the au capsule to monitor the temperature .
it took about 2 hours to pressurize the cell to 3 gpa .
after the pressurization , the heating power was increased linearly and then maintained constant for 2 hours .
the sample was sintered at a temperature of @xmath19c and then quenched to room temperature .
the sample weighed about 130 mg and was about 4.5 mm in diameter and 3.3 mm in height .
the magnetization curves were measured by using a squid magnetometer ( quantum design , mpms-_xl _ ) .
figure [ fig1 ] shows the temperature dependence of the zero - field - cooled magnetization measured at @xmath20 oe . from this figure
, we found that the superconducting transition temperature @xmath8 and the transition width @xmath21 were about 37 k and 1 k , respectively.@xcite figure [ fig2 ] shows the magnetization curves , @xmath2 , of mgb@xmath9 , which were measured in the temperature region of 5 k @xmath22 33 k.@xcite one notable feature is the symmetry in the increasing and the decreasing field branches , _
i.e. _ , @xmath23 , where @xmath24 and @xmath25 are the magnetizations in the increasing and the decreasing field branches , respectively .
such a feature can be commonly found in previous reports.@xcite this means that the contribution of the equilibrium magnetization and the surface pinning is negligible compared to that of the bulk pinning .
the irreversible magnetization can be described by various critical - state models .
the bean model@xcite has been used to calculate the critical current density of superconducting materials .
the model assumes that the slope @xmath26 is constant and field independent , where @xmath27 denotes the local magnetic induction inside a sample .
thus , the critical current density ( or irreversible magnetization ) should also be field independent , which is contrary to most experimental results .
other critical - state models , such as the exponential and the watson models,@xcite which take into account the field dependence of the critical current density , can be used to describe the irreversible magnetization properly . in the frame of the watson model , the critical current density , @xmath28 ,
is given by @xmath29 where @xmath30 and @xmath31 are adjustable parameters which depend on the material .
the exponential model proposes that the critical current density has the form @xmath32 where @xmath30 and @xmath31 are again adjustable parameters as in the watson model .
according to ampere s law , the field gradient inside a sample is given by @xmath33 where sgn(@xmath34 ) is the sign function and @xmath35 is the speed of light . in cylindrical coordinates ,
we obtain an average magnetic induction @xmath36 of a sample with a radius @xmath37 @xmath38 if the surface barrier effect is ignored , the boundary condition for @xmath27 is @xmath39 , where @xmath14 is the external magnetic field .
figure [ fig3](a ) shows our attempt to fit @xmath2 at @xmath40 k by using eq .
( [ avginduction ] ) with the exponential critical - state model with @xmath41 emu/@xmath42 and @xmath43 t. for the theoretical description of the @xmath2 , we can choose an arbitrary number for a sample size , @xmath37 , within the constraint that the multiplier @xmath44 is a constant .
as one can see , the data are well described by the critical - state model without considering the contribution of the reversible magnetization and the surface barrier effect . as stated before
, this implies that , in the mixed state in the mgb@xmath9 superconductor , the magnetization mainly comes from the contribution of the bulk pinning effect .
a fit using the watson model was also attempted , but was poor for all adjustments of the parameters .
the dashed line of fig .
[ fig3](a ) represents the average critical current density @xmath45 calculated from the decreasing - field branch of the theoretical magnetization curve assuming the grain size @xmath46 @xmath47 m .
we note that the shapes of the @xmath2 curves shown in fig .
[ fig2 ] are remarkably similar to each other .
this suggests that the vortex pinning mechanism in mgb@xmath9 does not change even as the temperature is varied up to near @xmath8 .
more concrete evidence for this can be found from the scaling analysis of the @xmath2 curves .
for the scaling of the @xmath2 curves , we define two phenomenological parameters as shown in the inset of fig .
[ fig3](b ) .
@xmath3 means the field where the magnetization in the increasing field branch reaches it s maximum value @xmath4 .
we divided the @xmath48 and the @xmath14 in each curve of fig .
[ fig2 ] by @xmath49 and @xmath3 , respectively .
the result is shown in fig .
[ fig3](b ) . without any exception ,
all the curves collapse on a single universal curve .
the solid line in the figure denotes the exponential critical - state model .
this result is consistent with the scaling of the pinning force , @xmath50 , in a temperature range of @xmath51 reported by larbalestier _
et al_.@xcite in the case of high-@xmath8 cuprates , such scaling behavior of the @xmath2 curves is established in a limited low - temperature region .
this implies that the fundamental mechanism determining the magnetic hysteresis at low temperatures changes as the temperature is increased toward @xmath8 . for bi@xmath9sr@xmath9cacu@xmath9o@xmath52 ( bi-2212 ) , while the bulk pinning is dominant at low temperatures , the contribution of the surface or geometrical barrier effects to the magnetization becomes more important as the temperature is increased.@xcite thus , universal scaling of @xmath2 is not seen in bi-2212 .
figure [ fig4 ] shows the temperature dependence of the scaling parameters @xmath3 and @xmath4 used in the above analysis .
the right axis for @xmath3 in the figure was corrected by the demagnetization factor @xmath53 .
the value of @xmath54 was obtained from the low - field susceptibility curve in fig .
[ fig1 ] assuming 100 % magnetic screening .
it was obvious that the @xmath55 and @xmath56 scaled lineally with temperature as indicated by solid lines .
the linearity in @xmath56 requires a linear temperature dependence of the irreversible field @xmath57 where the magnetic hysteresis disappears .
this is because the normalized hysteresis curves @xmath2 for @xmath58 collapsed into a single universal curve as we showed .
in fact , a nearly linear @xmath16 was shown from the kramer analysis of magnetization.@xcite this feature differentiates mgb@xmath9 from other cuprate high-@xmath8 materials with @xmath59 .
in summary , we measured the magnetization curves @xmath2 for the newly discovered metallic mgb@xmath9 superconductor , which has a @xmath60 k , in the region 5 k @xmath61 33 k and @xmath625 t @xmath63 5 t. the magnetic hysteresis in our experimental region was well described by the exponential critical - state model , without considering the reversible magnetization and the surface barrier effect .
also , we found that all the magnetization curves collapsed onto a single universal curve when the field and the magnetization were normalized by the characteristic scaling factors @xmath3 and @xmath4 , respectively .
these results lead us to the conclusion that the irreversible magnetization of mgb@xmath9 is dominated by bulk pinning and that the pinning mechanism does not change even when the temperature is varied up to @xmath18 .
, of mgb@xmath9 for @xmath20 oe .
this curve reveals the superconducting transition temperature @xmath8 and the transition width @xmath21 to be about 37 k and 1 k , respectively.,title="fig:",width=264,height=188 ] + , at @xmath64 k. the solid line represents the theoretical curve for the exponential critical - state model .
the line denotes the absolute value of the average critical current density @xmath45 calculated from the decreasing - field branch of the theoretical magnetization curve at @xmath64 k. ( b ) scaling of the magnetization curves , @xmath2,in the temperature region 5 k @xmath61 33 k. the inset illustrates the definitions of @xmath55 and @xmath65 ( see text).,title="fig:",width=294,height=340 ] + d. c. larbalestier , m. o. rikel , l. d. cooley , a. a. polyanskii , j. y. jiang , s. patnaik , x. y. cai , d. m. feldmann , a. gurevich , a. a. squitieri , m. t. naus , c. b. eom , e. e. hellstrom , r. j. cava , k. a. regan , n. rogado , m. a. hayward , t. he , j. s. slusky , p. khalifah , k. inumaru , and m. haas , nature * 410 * , 343 ( 2001 ) . at @xmath67
k , around @xmath68 in the decreasing field branch , the magnetization data were considerably noisy as shown in fig .
the same feature was observed by dou _
et al . _ , ( cond - mat/0102320 ( 2001 ) ) .
they interpreted the phenomenon in terms of the flux jumping effect .
|
we report the magnetic - field dependence of the irreversible magnetization of the recently discovered binary superconductor mgb@xmath0 . for the temperature region of @xmath1
, the contribution of the bulk pinning to the magnetization overwhelms that of the surface pinning .
this was evident from the fact that the magnetization curves , @xmath2 , were well described by the critical - state model without considering the reversible magnetization and the surface pinning effect .
it was also found that the @xmath2 curves at various temperatures scaled when the field and the magnetization were normalized by the characteristic scaling factors @xmath3 and @xmath4 , respectively .
this feature suggests that the pinning mechanism determining the hysteresis in @xmath2 is unique below @xmath5 . 2
|
cataclysmic variable stars ( cvs ) are close binaries consisting of a white dwarf primary that accretes matter from a secondary star via roche lobe overflow .
the secondary is more extended than the white dwarf , and usually resembles a low - mass main - sequence star .
cvs have a rich phenomenology and have attracted a great deal of observational and theoretical interest .
@xcite gives a dated but useful review of these objects . because the secondary fills its roche lobe , its mean density is closely constrained by the orbital period , with the secondary s mean density being larger at shorter periods @xcite . on the main sequence ,
the mean density rises toward lower masses , so if the mass - radius relation of the secondary is similar to that of main - sequence stars , the orbital period gives a rough proxy measurement of the secondary s mass .
short period cvs therefore tend to have low - mass secondaries , which tend to have correspondingly cool temperatures and low visible luminosities .
@xcite updated the long - known correlation between system s orbital period and the secondary s spectral type .
he found that at a given orbital period , secondaries tended to be near the expected main - sequence surface temperature , or somewhat cooler .
there are exceptions a few systems are anomalously warm , with k - type secondaries in orbits of only a few hours or less .
these secondaries appear to have undergone substantial hydrogen burning during previous evolution , which increased the mean molecular weight and hence changed the mass - luminosity - temperature relation .
i recently identified an example css 1340 , which has @xmath4 = 2.4 hr ; @xcite details this discovery and includes references and further background on these objects . here ,
i report another , less extreme sytem , 13cl .
this previously uncatalogued dwarf nova was discovered in outburst by the asas - sn surveyassassin / index.shtml ] on 2013 august 29 .
table [ tab : properties ] gives further details on this object .
the outburst history of 13cl is not well - determined .
it appears consistently , with little obvious variability , on digitized sky survey images served at space telescope science institute .
this area of the sky ( at @xmath5 deg ) is not covered by the catalina surveys data release 2 @xcite .
all the observations are from mdm observatory , on kitt peak , arizona .
spectra were obtained on 2013 sept .
11 , 13 , 14 , 15 , and 17 ut , and photometric time series were taken on sept . 16 , 17 , and 19 ut . the system appeared to have returned to quiescence from its recent outburst .
the spectra are from the 2.4 m hiltner telescope and modular spectrograph . the instrument setup ,
observing protocols , and reduction procedures were essentially identical to those described in @xcite . the mean spectrum ( fig .
[ fig : specs ] , top panel ) shows the absorption lines of a late - type star together with the broad emission lines of hydrogen and hei commonly seen in dwarf novae at minimum light .
i estimated the spectral type and light fraction of the secondary by scaling and subtracting spectra of stars classified by @xcite and looking for the best cancellation of late - type features .
this constrained the spectral type to be k4 @xmath0 2 subclasses , with the secondary contributing roughly half the system s light ; the lower trace in the upper panel of fig .
[ fig : specs ] shows the result of subtracting a scaled k - star from the observed spectrum .
the sdss magnitudes in table [ tab : properties ] are broadly consistent with this classification .
the @xmath6 color suggests a type near k5 @xcite , with bluer passbands giving earlier types , consistent with an increasing contribution from a blue accretion disk toward shorter wavelengths . to measure the ( absorption ) radial velocity of the k star ,
i cross - correlated the spectra against a velocity - compensated sum of many late - type spectra , using the _ rvsao _ package @xcite .
the cross - correlation covered from 5100 to 6500 , excluding 5800 to 5950 to avoid hei @xmath75876 and nad , which can have an interstellar component .
table [ tab : velocities ] gives the velocities .
a period search of the resulting timeseries with a ` residual - gram ' method @xcite revealed an unambiguous sinusoidal modulation at just under 5 cycle d@xmath3 ( fig .
[ fig : folded ] ) .
the h@xmath8 velocities , measured with a convolution routine , were much noisier than the absorption velocities , but they do show significant modulation consistent with the well - defined absorption period .
table [ tab : fits ] gives parameters sinusoidal fits to the velocities .
the velocity modulation of the absorption lines is clearly visible in the lower panel of fig .
[ fig : specs ] , which shows a phase - averaged greyscale representation of the spectra .
the time - series photometry is from an andor ikon du-937n camera mounted on the mdm 1.3 m mcgraw - hill telescope , as described in @xcite .
individual exposures were 30 s , in the @xmath9 filter . in each frame , the program star , a comparison star , and two check stars were measured ; the main comparison star was 54 arcsec west and 17 arcsec south of the program star .
conditions were generally clear but not photometric , and exposures in which the comparison star s instrumental magnitude more than @xmath10 mag fainter than average were excluded from the analysis . in the sdss dr10 ( table [ tab : properties ] ) , the program object has @xmath11 and @xmath12 , while the main comparison star has @xmath13 and @xmath14 . using transformations derived by robert lupton , these correspond to @xmath15 and @xmath16 respectively .
[ fig : tsphot ] shows the time series for the three nights , with the zero - point adjusted to approximate @xmath9 magnitude , and the lower panel of fig . [
fig : folded ] shows the magnitude plotted against orbital phase .
the photometry shows variation at twice the orbital period , consistent with ellipsoidal modulation .
superposed on this is a sharp dip at inferior conjunction of the secondary , which is apparently a grazing eclipse .
the eclipse is clearly visible on the first and second nights , but only just discernible on the third .
table [ tab : eclipses ] gives the timings of the three eclipses . _ period and ephemeris . _ the radial velocities and the grazing eclipses give independent estimates of the period , which agree within their uncertainties . adopting the weighted mean period yields @xmath17 where @xmath18 is an integer .
_ masses . _
the grazing eclipse requires an orbital inclination @xmath19 not too far from edge - on , and the amplitude of the absorption line velocity variation @xmath20 should reflect the secondary star s center - of - mass motion with reasonable fidelity .
one is less confident of the dynamical interpretation of the h@xmath8 velocities ( see , e.g. , @xcite ) ; however , the phase at which they cross their mean velocity is @xmath21 in the eclipse ephemeris , consistent with the antiphased motion expected if they coincide with the white dwarf .
if we take the velocities as reliable for dynamics , we find @xmath22 . for @xmath23 degrees ,
the mass function then implies @xmath24 for the white dwarf , and roughly half that for the secondary .
as expected , the secondary is undermassive for its spectral type ; a recent tabulation by pecaut and mamajekemamajek / eem_dwarf_ubvijhk_colors_teff.txt ] ( based on @xcite ) estimates 0.7 m@xmath25 for typical main - sequence field star at k6 , the late ( hence low - mass ) end of our spectral type range .
_ distance .
_ the secondary star s spectral type constrains its surface brightness , and the roche lobe geometry combined with the orbital period lets us estimate the secondary s radius , with only a weak dependence on its mass .
the radius and surface brightness yield an absolute magnitude , which together with the observed brightness of the secondary ( and an estimated interstellar extinction ) yields a distance .
all the quantities entering the calculation are uncertain , and some are correlated ( e.g. , the light fraction attributed to the secondary decreases at cooler spectral types , because the metal lines used in estimating the fraction mostly grow stronger toward cooler types ) .
i therefore used a monte carlo procedure to estimate the distance and its uncertainty , which explicitly included the correlation between the secondary s spectral type and its contribution to the light . for the k@xmath26 secondary ,
i used magnitudes centered around @xmath27 , and ranging between extremes of 19.55 ( minimum contribution at k6 ) and 18.10 ( maximum contribution at k2 ) .
the secondary s mass ( which enters weakly ) was taken to be uniformly distributed in the range @xmath28 .
the @xmath9-band surface brightness at each spectral type was computed using a polynomial relation tabulated by @xcite .
i assumed @xmath29 , guided by the extinction map of @xcite , which gives a total extinction of @xmath30 at this location .
the monte carlo procedure yielded a median distance of 1180 pc , with a 68-percent confidence range from 990 to 1400 pc .
_ ellipsoidal variation .
_ in favorable cases the light variations of the secondary star can yield useful constraints on system parameters , especially the inclination . to exploit this
, @xcite developed a code to model the light variation of a roche - lobe filling star .
13cl , unfortunately , is rather too faint and messy for fruitful light - curve fitting , but nonetheless the lower panel of fig .
[ fig : folded ] includes a representative calculation m@xmath25 , @xmath31 m@xmath25 , @xmath23 deg , @xmath32 k for the m4 secondary ( following @xcite ) , and a contribution from the disk equivalent to an observed @xmath33 , which includes @xmath34 mag of extinction .
the model is scaled here to a distance of 1100 pc to match the observational data ; the distance is nearly identical to that found earlier by an essentially equivalent method . ] which is useful for comparison .
several features stand out : 1 .
the sharp dip around phase zero departs significantly from the secondary light curve , indicating that it is very likely an eclipse .
the maxima are unequal , with the second hump ( phase 0.75 ) higher than the first . at minimum light dwarf novae
often show a pre - eclipse brightening as the ` hot spot ' where the mass - transfer stream strikes the disk rotates into view ( @xcite present several examples of this ) .
while this has the correct sense to explain the inequality , the extra light does not persist until eclipse as it does in other systems , so it seems more likely that the secondary s leading and trailing hemispheres differ in mean brightness .
the curve shown has extra light added ( assumed constant through the orbit ) to approximate the contributions from the rest of the system ( accretion disk and so on ) .
this tends to flatten the light curve and push the system to greater computed distances .
the amplitude of the computed curve appears to be smaller than that of the data , suggesting that the extra light may have been overestimated ; however , it is difficult to justify a smaller contribution on the basis of the spectral decomposition ( fig .
[ fig : specs ] ) .
although the data are noisy and the detailed match imperfect , the modulation s period and phase confirms that it is mostly ellipsoidal .
the secondary of 13cl is warmer than expected . at @xmath35 h , the temperature discrepancy is not as dramatic as in shorter period systems such as ei psc @xcite , qz ser @xcite , sdss j170213.26 + 322954.1 @xcite , and 1340 @xcite , but it is significant . the recent generation of synoptic sky surveys , most notably the catalina surveys @xcite , asas - sn , and the master survey @xcite have dramatically accelerated the rate of discovery of new dwarf novae ( see , e.g. , @xcite ) .
the variability criteria favor the discovery of large - amplitude outbursts @xcite . in most cvs with periods
less than a few hours , the secondaries are so faint that the systems can become very faint between outbursts , but in systems similar to 13cl the anomalously bright secondary imposes a ` floor ' on the total brightness , which can make the amplitude rather modest . in 13cl , the measured amplitude is less than 3 mag , though it is unlikely to have been caught exactly at maximum . at such relatively small amplitudes ,
discovery is somewhat more difficult . also , the known examples of warm - secondary dwarf novae have relatively long inter - outburst times , or have only a single observed outburst . for these reasons ,
it is likely that a substantial number of these warm - secondary systems await discovery .
i gratefully acknowledge support from nsf grant and ast-1008217 .
i also thank the mdm observatory staff for their cheerful and excellent support , and the referee for a prompt and useful report .
this paper uses data from the sdss .
funding for sdss - iii has been provided by the alfred p. sloan foundation , the participating institutions , the national science foundation , and the u.s .
department of energy office of science .
the sdss - iii web site is http://www.sdss3.org/ ; further acknowledgments can be found there .
sdss - iii is managed by the astrophysical research consortium for the participating institutions of the sdss - iii collaboration .
barrett , p. e. , & bridgman , w. t. 2000 , in astronomical society of the pacific conference proceedings , vol 216 , astronomical data analysis software and systems ix , ed .
n. manset , c. veillet , & d. crabtree , 67 lll @xmath36 & 21@xmath37 38@xmath38 [email protected] & sdss + @xmath40 & + 26@xmath41 & + outburst date & 2013 - 08 - 29.35 & asas - sn + outburst magnitude & 15.66 & + u & @xmath42 & sdss + g & @xmath43 & + r & @xmath44 & + i & @xmath45 & + z & @xmath46 & + v & 17.87 & + j & @xmath47 & 2mass + h & @xmath48 & + k & @xmath49 & + [ tab : properties ] lrrrr 56546.9338 & @xmath50 & @xmath51 & @xmath52 & @xmath53 + 56546.9408 & @xmath54 & @xmath55 & @xmath56 & @xmath57 + 56548.9009 & @xmath58 & @xmath59 & @xmath60 & @xmath61 + 56548.9093 & @xmath62 & @xmath63 & @xmath64 & @xmath65 + 56549.7790 & @xmath54 & @xmath63 & @xmath66 & @xmath67 + 56549.7875 & @xmath68 & @xmath69 & @xmath70 & @xmath71 + 56549.7960 & @xmath72 & @xmath73 & @xmath74 & @xmath75 + 56549.8087 & @xmath76 & @xmath77 & @xmath78 & @xmath79 + 56549.8172 & @xmath80 & @xmath55 & @xmath81 & @xmath82 + 56549.8257 & @xmath83 & @xmath84 & @xmath85 & @xmath86 + 56549.8341 & @xmath87 & @xmath77 & @xmath88 & @xmath89 + 56549.8426 & @xmath90 & @xmath69 & @xmath91 & @xmath92 + 56549.8511 & @xmath93 & @xmath84 & @xmath94 & @xmath95 + 56549.8595 & @xmath96 & @xmath63 & @xmath97 & @xmath98 + 56549.8680 & @xmath76 & @xmath99 & @xmath100 & @xmath82 + 56549.8765 & @xmath101 & @xmath102 & @xmath103 & @xmath86 + 56549.8850 & @xmath104 & @xmath105 & @xmath106 & @xmath107 + 56549.8935 & @xmath108 & @xmath109 & @xmath110 & @xmath86 + 56549.9020 & @xmath111 & @xmath63 & @xmath112 & @xmath65 + 56549.9105 & @xmath113 & @xmath77 & @xmath114 & @xmath115 + 56549.9189 & @xmath116 & @xmath63 & @xmath117 & @xmath118 + 56549.9274 & @xmath119 & @xmath120 & @xmath121 & @xmath122 + 56549.9360 & @xmath123 & @xmath77 & @xmath117 & @xmath124 + 56550.7524 & @xmath125 & @xmath84 & @xmath126 & @xmath127 + 56550.7609 & @xmath128 & @xmath55 & @xmath129 & @xmath130 + 56550.7694 & @xmath91 & @xmath73 & @xmath131 & @xmath132 + 56550.7778 & @xmath133 & @xmath59 & @xmath134 & @xmath135 + 56550.7863 & @xmath136 & @xmath73 & @xmath137 & @xmath130 + 56550.7948 & @xmath138 & @xmath63 & @xmath139 & @xmath140 + 56552.8044 & @xmath141 & @xmath55 & @xmath66 & @xmath132 + [ tab : velocities ] lrrrrrr absorption & 56549.7839(11 ) & 0.20246(20 ) & 246(9 ) & @xmath142 & 30 & 20 + & 56549.7841(12 ) & [ 0.20219 ] & 248(9 ) & @xmath143 & 30 & 22 + h@xmath8 emission & 56549.888(5 ) & 0.2026(9 ) & 105(17 ) & @xmath144 & 30 & 40 + & 56549.888(5 ) & [ 0.20219 ] & 106(17 ) & @xmath145 & 30 & 40 + [ tab : fits ]
|
the spectrum of the recently discovered cataclysmic variable star ( cv ) 13cl shows that a secondary star with spectral type k4 ( @xmath0 2 subclasses ) contributes roughly half the optical light .
the radial velocities of the secondary are modulated on an orbital period @xmath1 hr with a velocity semiamplitude @xmath2 km s@xmath3 , and the light curve shows ellipsoidal variations and an apparent grazing eclipse . at this orbital period ,
the secondary stars in most cvs are substantially cooler , with spectral types near m3 .
13cl therefore joins the small group of cvs with anomalously warm secondary stars , which apparently form when the onset of mass transfer occurs after the secondary has undergone significant nuclear evolution .
|
molecular quantum systems are described by an unbounded self - adjoint operator , the schrdinger operator @xmath0 acting on the hilbert space of complex - valued square - integrable functions @xmath1 . here
, @xmath2 is a small positive parameter related to the inverse square root of the average nuclear mass , and @xmath3 is the nuclear potential function resulting from the born oppenheimer approximation , see @xcite .
the relevant time scales of nuclear quantum motion are of the order @xmath4 .
the time evolution of an initial wave function @xmath5 is governed by the time - dependent linear schrdinger equation @xmath6 with the appropriate @xmath7-scaling of the time - derivative or
equivalently by the action of the unitary evolution operator @xmath8 , since @xmath9 even though the schrdinger equation is a _ linear _ partial differential equation , the numerical simulation of physical quantities derived from the wave function @xmath10 is notoriously difficult for two reasons : the dimension @xmath11 of the nuclear configuration space is large . if a molecule consists of @xmath12 nuclei , then @xmath13 .
just for a single water molecule , for example , we have @xmath14 .
moreover , nuclear quantum motion is highly oscillatory . solutions typically oscillate with frequencies of the order @xmath4 in time and space , while @xmath7 ranges between @xmath15 and @xmath16 , depending on the molecular system under consideration . for the hydrogen molecule h@xmath17 , for example , one has @xmath18 , while @xmath19 for iodine monobromide ibr . as an answer to these challenges ,
chemical physicists have developed approximations involving the _ nonlinear _ time evolution of classical mechanics .
one uses the hamiltonian function @xmath20 the associated hamiltonian system @xmath21 and the corresponding hamiltonian flow @xmath22 as the classical counterparts to the schrdinger operator @xmath23 , the time - dependent schrdinger equation , and the unitary evolution operator @xmath24 .
this quantum - classical correspondence is most elegantly elaborated by using wigner functions and weyl quantization .
the wigner function @xmath25 of a square integrable function @xmath26 is a real - valued , square integrable , continuous function on phase space @xmath27 obtained by an inverse fourier transform of the autocorrelation function of @xmath28 , see the short summary in [ app : wigner ] .
the wigner function can be thought of as a probability density on phase space @xmath27 , though in general it might attain negative values .
it has been introduced by e. wigner in @xcite when developing the thermodynamics of quantum mechanical systems .
crucial properties of the wigner function are the orthogonality relation @xmath29 and the asymptotically classical time evolution @xmath30 for the solution @xmath10 of the schrdinger equation .
the wigner phase space method of e. heller @xcite and the statistical quasiclassical method of h. lee and m. scully @xcite , for example , use these properties for computing the transition probabilities from a given state @xmath31 to @xmath10 according to @xmath32 viewing the wigner function @xmath33 of a square integrable function @xmath28 as a tempered distribution , that is , as a continuous linear mapping from the schwartz functions @xmath34 to the real numbers , one arrives at the identity@xmath35 where @xmath36 denotes the bounded linear operator on @xmath1 obtained by the weyl quantization of the function @xmath37 .
the weyl operator @xmath36 is a pseudo - differential operator , a generalized partial differential operator , which treats differentiation and multiplication by functions on the same footing , see the appendix . with an appropriate handling of operator domains
, weyl quantization also applies for unbounded linear operators , and the schrdinger operator @xmath38 , for example , is the weyl quantized classical hamiltonian function @xmath39 defined in .
often , the weyl operator @xmath36 is also called the quantum observable assocaited with the classical observable @xmath37 . besides the convenient relation , weyl quantization enjoys the beautiful property that the trace of the product of weyl operators can be expressed as the integral @xmath40 provided that the classical observables @xmath37 and @xmath41 satisfy appropriate regularity and growth conditions , see e.g. ( * ? ? ?
* proposition 9.2 ) or ( * ? ? ?
* proposition 284 ) . from the weyl point of view
, the asymptotic time evolution of the wigner function can equivalently be formulated as @xmath42 this quantum - classical approximation of time evolved quantum observables is known as egorov s theorem and has first been formulated in @xcite within the hrmander theory of pseudo - differential operators , see also @xcite for refined error estimates in the context of semiclassical microlocal analysis .
the computational power of classically propagating quantum observables has been recognized by w. miller and h. wang @xcite , who approximate the time - dependent correlation function of @xmath36 and @xmath43 according to @xmath44 in the chemical physics literature this approximation is referred to as the linearized semiclassical initial value representation ( lsc - ivr ) , and it is derived from fourier integral operator representations of the unitary evolution operator @xmath45 , see also @xcite for a recent review .
it seems that the link of lsc - ivr to egorov s theorem has not been noticed so far . in a study of the one - dimensional morse oscillator ,
h. lee and m. scully @xcite have proposed to improve the second order approximation for the wigner function of @xmath10 by @xmath46 with @xmath47 where the derivative of the effective potential @xmath48 is given by @xmath49 for the correcting factor @xmath50 they mention the three options @xmath51 and they use the term wigner trajectories for the solutions of the ordinary differential equation @xmath52 , @xmath53 , see also @xcite . later
, a. donoso and c. martens @xcite have defined their entangled classical trajectories by a fourth variant @xmath54 .
trajectories generated by @xmath55 and @xmath56 and generalisations thereof have also been incorporated for the numerical simulation of time - dependent correlation functions of one - dimensional systems by j. liu and w. willer @xcite .
our aim here is also a higher order approximation of the dynamics .
however , we will work with the rigorous mathematical framework of egorov s theorem and construct phase space trajectories without any entanglement for arbitrary dimensions @xmath57 .
the key element for proving egorov s theorem is an asympotic expansion of the commutator @xmath58 \;\sim\ ; \sum_{k\in 2\n } \left(\frac{\eps}{2\rmi}\right)^{k } \op(\{h , a\}_{k+1}).\ ] ] where @xmath59 denotes a generalization of the usual poisson bracket involving derivatives of the functions @xmath39 and @xmath37 up to order @xmath60 , see . in the context of response theory ,
m. kryvohuz and j. cao @xcite use this expansion up to the fourth term for a systematic improvement of linear response computations over long times .
the commutator expansion also reveals , that for hamiltonians @xmath39 , wich are polynomials of degree less or equal than two , the remainder of egorov s theorem vanishes .
this exact egorov result is utilized by h. waalkens , r. schubert and s. wiggins for their quantum normal form algorithm in dynamical transition state theory @xcite . for general hamiltonian functions @xmath39 ,
the expansion implies a higher order version of egorov s theorem , @xmath61 with leading order term @xmath62 and corrections @xmath63it is remarkable that the corrections are only built from derivatives of the observable @xmath37 , the hamiltonian function @xmath39 and the flow @xmath64 , which has also been emphasized by m. pulvirenti in @xcite , when deriving higher order estimates for the wigner function of the wave function @xmath10 .
since the higher order egorov expansion is built of even powers of the parameter @xmath7 , the first correction @xmath65provides the fourth order estimate @xmath66 the aim of this paper is the exemplary analysis of this correction and its discretization for the numerical computation of expectation values . in a first step ,
we reformulate it as @xmath67 where @xmath68 is an explicitly defined linear mapping from the space of @xmath69-tensors to the real numbers , which depends on derivatives of the hamiltonian function @xmath39 and the flow @xmath64 , but not on the observable @xmath37 . in spirit , this formulation of @xmath70 is close to the wigner trajectories generated by the first correction factor @xmath71 . in the next step ,
we derive a first order ordinary differential system for the components of @xmath72 , @xmath73 , and @xmath74 .
the vectorization of this equation results in @xmath75 where the building blocks of the matrix @xmath76 and the vector @xmath77 are components of the tensors @xmath78 , @xmath79 , and @xmath80 evaluated along the flow @xmath64 . in the final step
we derive a fourth order splitting scheme for the discretization of the ordinary differential equation , which is then applied for numerical test calculations .
this paper is organized as follows .
develops egorov s theorem to the next order with respect to the parameter @xmath7 and formulates this correction as a first order ordinary differential equation .
discusses a discretization of this corrected approximation for the computation of expectation values . provides numerical experiments for a two - dimensional torsional quantum system , which confirm the theoretical considerations .
the appendix summarizes basic properties of wigner functions and weyl operators .
let @xmath81 be a smooth function of subquadratic growth .
that is , for all @xmath82 with @xmath83 there exists @xmath84 with @xmath85 let @xmath34 be a schwartz function .
then , the following formal considerations can be turned into a proof according to ( * ? ? ?
* thorme iv.10 ) or ( * ? ? ?
* theorem 1.2 ) .
we look for an approximate classical observable @xmath86 such that @xmath87 we require @xmath88 at time @xmath89 , rewrite the difference according to @xmath90 - \frac{\rmd}{\rmd t}\,\op(a_{\rm appr}(t-\tau ) ) \right ) \rme^{-\rmi h\tau/\eps}\rmd\tau,\end{aligned}\ ] ] and observe that the commutator @xmath91 = \op(h)\op(a_{\rm appr}(t))-\op(a_{\rm appr}(t))\op(h)\ ] ] plays a crucial rule . multiplying this commutator with @xmath92
, we obtain an asymptotic expansion in even powers of @xmath7 ,
@xmath93 \sim \sum_{k\in2\n } \left(\frac{\eps}{2i}\right)^k \op(\{h , a_{\rm appr}(t)\}_{k+1}),\ ] ] where @xmath94 denotes the @xmath69th generalized poisson bracket of two smooth functions @xmath95 , see [ app : wigner ] .
since the first generalized poisson bracket coincides with the usual poission bracket , @xmath96 , we have @xmath97 = \op(\{h,\op(a_{\rm appr}(t)\ } ) + o(\eps^2).\ ] ] let @xmath98 be the flow associated with the classical hamilton function @xmath39 , and set @xmath99 .
then , @xmath100 and we obtain egorov s theorem , @xmath101 the asymptotic commutator expansion allows to systematically derive higher order corrections to egorov s theorem . for constructing the first correction , we set @xmath102 .
the computation @xmath97 = \op(\{h , a_{\rm appr}(t)\ } ) - \frac{\eps^2}{4}\ , \op(\{h , a_0(t)\}_3)+ o(\eps^4)\ ] ] suggests to choose @xmath103 indeed , we compute the time derivative , @xmath104where the last equation uses that the classical flow as a symplectic transformation of phase space preserves the poisson bracket.we therefore obtain @xmath105 consequently , @xmath106 [ rem : inv ] if @xmath34 is an observable invariant along the hamiltonian flow , that is , @xmath107 for all @xmath108 , and additionally @xmath109 , then the corrected approximation is also invariant , @xmath110for all @xmath108 .
in particular , mass ( @xmath111 ) and energy ( @xmath112 ) are conserved .
our next aim is to reformulate the correction as @xmath113 where the components of the time - dependent tensors @xmath114 , @xmath115 , and @xmath116 satisfy a first order system of coupled ordinary differential equations , that is independent of the observable @xmath37 and can be efficiently solved alongside the hamiltonian flow @xmath64 . here and
in the following , we use einstein s summation convention for notational brevity .
let @xmath117 we observe that we can write @xmath118 where the 3-tensor @xmath119 is defined by @xmath120 and @xmath121 a closer look at this 3-tensor reveals that it is symmetric : [ lem_symm_jd3h ] let @xmath122 and @xmath123 be a @xmath69-tensor .
then , @xmath124 in particular , the 3-tensor @xmath119 defined in is symmetric . for every @xmath125 @xmath126 let @xmath127 . for @xmath128 , we set @xmath129 , @xmath130 and otherwise @xmath131 , @xmath132 .
then , @xmath133 hence , a permutation of @xmath134 results in a permutation of @xmath135 and @xmath136 .
thus , @xmath137 is symmetric if and only if @xmath138 is symmetric .
we compute the third derivative of @xmath139 by the chain rule , where we denote the derivatives of @xmath140 by @xmath141,@xmath142 since the 3-tensor @xmath119 is symmetric and @xmath143 , we obtain the reformulation @xmath144 with @xmath145 \circ \phi^{t-\tau } \rmd\tau , \\
\gamma^t_{ij } & = \int_{0}^t \left [ \left ( d^2\phi^\tau \right)_{ikl } \left ( d\phi^\tau \right)_{jm } \left(j\widetilde{d^3h}\right)_{mlk } \right ] \circ \phi^{t-\tau } \rmd\tau , \\ \xi^t_{i } & = \int_{0}^t \left [ \left ( d^3\phi^\tau \right)_{ijkl } \left(j\widetilde{d^3h}\right)_{lkj } \right ] \circ \phi^{t-\tau } \rmd\tau . } \ ] ] next , we compute the time derivatives of integrals , which are of the form observed in the defining equations of the time - dependent tensors @xmath114 , @xmath115 , and @xmath116 .
[ lem_main_pde ] let @xmath146 and @xmath147 smooth functions . then , @xmath148 we start with @xmath149for introducing the @xmath150-derivative in the integral , we compute @xmath151 therefore , @xmath152 now we are ready to formulate and prove our first main result , the explicit system of ordinary differential equations describing the second order correction to egorov s theorem . in m.
zworski s recent monograph , the higher order terms are referred to `` as difficult to compute '' , see @xcite .
our result shows nonetheless , that the computation is feasible .
[ ode_thm_odes ] let @xmath81 be a smooth function of subquadratic growth , @xmath108 , and @xmath98 the hamiltonian flow associated with @xmath39 .
let @xmath34 be a schwartz function .
then , there exists a constant @xmath153 such that for all @xmath2 @xmath154 where @xmath155 and @xmath156 and the functions @xmath157 , @xmath158 , @xmath159 , @xmath160 , solve the ordinary differential system @xmath161 with @xmath162 and @xmath163 where for a matrix @xmath164 and @xmath69-tensor @xmath165 , @xmath166 is given by @xmath167 and @xmath168 is defined by ( [ def_jd3tilde ] ) .
the initial values @xmath169 are clear , since @xmath157 , @xmath158 and @xmath159 are defined in as an integral from @xmath170 to @xmath171 , over a smooth integrand .
next , we study the time derivative of @xmath157 . for this , we write the hamiltonian system as @xmath172 and apply the differential operator @xmath173 to obtain the jacobi stability equation @xmath174 then , by lemma [ lem_main_pde ] , @xmath175 \circ \phi^{t-\tau } \rmd\tau \\\fl&&+ \left [ \left ( d\phi^0 \right)_{il } \left ( d\phi^0 \right)_{jm } \left(d\phi^0 \right)_{kn } \left(j\widetilde{d^3h}\right)_{lmn } \right ] \circ \phi^t \\\fl&= & \int_0^t \left [ \left ( j\cdot d^2h \circ \phi^\tau \right)_{i\nu } \left ( d\phi^\tau \right)_{\nu l } \left ( d\phi^\tau \right)_{jm } \left(d\phi^\tau \right)_{kn } \left(j\widetilde{d^3h}\right)_{lmn } \right ] \circ \phi^{t-\tau } \rmd\tau \\\fl&&+ \int_0^t \left [ \left ( d\phi^\tau \right)_{il } \left ( j\cdot d^2h \circ \phi^\tau \right)_{j\nu } \left ( d\phi^\tau \right)_{\nu m } \left(d\phi^\tau \right)_{kn } \left(j\widetilde{d^3h}\right)_{lmn } \right ] \circ \phi^{t-\tau } \rmd\tau \\\fl&&+ \int_0^t \left [ \left ( d\phi^\tau \right)_{il } \left ( d\phi^\tau \right)_{jm } \left ( j\cdot d^2h \circ \phi^\tau \right)_{k\nu } \left(d\phi^\tau \right)_{\nu n } \left(j\widetilde{d^3h}\right)_{lmn } \right ] \circ \phi^{t-\tau } \rmd\tau
\\\fl&&+ \left(j\widetilde{d^3h}\right)_{ijk } \circ \phi^t .
\end{aligned}\ ] ] so we can finish the proof for @xmath176 by using that @xmath177 .
the proofs for @xmath178 and @xmath179 are analogous : we differentiate the jacobi stability equation to obtain @xmath180 moreover , we use @xmath181 \left(j\widetilde{d^3h}\right)_{mlk } , \end{aligned}\ ] ] @xmath182 \left(j\widetilde{d^3h}\right)_{lkj}. \end{aligned}\ ] ] as well as @xmath183 and @xmath184 . for the numerical simulation of the ordinary differential system , we vectorize the tensors .
recall , that for matrices @xmath185 and @xmath186 , the kronecker product @xmath187 is defined as @xmath188 that is , @xmath189 the vectorization of a k - tensor @xmath190 is defined as @xmath191 that is , @xmath192 if @xmath193 the following observation allows the vectorization of the products occuring on the right hand side of our differential equation .
[ prop_vectorisation ] let @xmath194 a matrix , @xmath195 an n - tensor , and @xmath196 .
define the n - tensor @xmath197 as @xmath198 then , @xmath199 we define the matrix @xmath200 .
that is , @xmath201 for @xmath202 and @xmath203 .
then , @xmath204 for @xmath202 .
now we reformulate the results of theorem [ ode_thm_odes ] in vectorized form .
[ thm_vec_odes ] consider the time - dependent tensors @xmath205 and @xmath206 , @xmath207 , together with the hamiltonian flow @xmath64 of theorem [ ode_thm_odes ] .
let the functions @xmath157 , @xmath158 and @xmath159 solve the ordinary differential system @xmath208 then , @xmath209 with @xmath210 and @xmath211 and @xmath212 applying lemma [ prop_vectorisation ] to the @xmath213 part of the ordinary differential equation , we obtain @xmath214 the proofs for @xmath215 and @xmath116 are analogous .
we now analyse the vectorized system of ordinary differential equations for the special case , that the hamilton function is given by @xmath216 then , @xmath217 moreover , @xmath218 to exploit this zero pattern , we reorder our system of differential equations .
we set @xmath219 for @xmath220 .
these terms are recollected according to @xmath221 and @xmath222 this reordering enhances the zero pattern of the right hand side of the ordinary differential equation .
[ theo : ham_schro ] let @xmath3 be a smooth function of subquadratic growth and @xmath81 , @xmath223 .
let @xmath224 and @xmath225 be the solution of the ordinary differential equation given in corollary [ thm_vec_odes ] with initial values @xmath226 we set @xmath227 then , @xmath228 with @xmath229 , @xmath230 and @xmath231 , where @xmath232 and @xmath233 and @xmath234 denoting @xmath235 and @xmath236 the matrices @xmath237 are given by @xmath238 in the following , we denote @xmath239 for tensors @xmath240 and @xmath241 . by the specific form of the matrix @xmath205 given in , we obtain @xmath242 therefore , @xmath243 since @xmath244 in the same way , we obtain @xmath245 since @xmath246 finally , we compute @xmath247 using that @xmath248 reordering the above differential equations , we obtain the claimed result .
we adopt the following general scheme which has been previously developed for discretizing egorov s theorem @xcite : the classical hamiltonian flow @xmath64 is discretized by a symplectic order @xmath249 method @xmath250 with sufficiently small step size @xmath251 .
the initial wigner function is split into its positive and negative part @xmath252 and is sampled by sufficiently many phase space points @xmath253.then , expectation values for various observables @xmath34 with respect to the solution @xmath10 of the schrdinger equation @xmath254 are approximated according to @xmath255 the computational work of this algorithm lies in the sampling of the initial wigner function and the classical evolution of the sample points .
then , expectation values are computed by a final phase space summation .
if the initial wave function @xmath256 is a gaussian wave packet , then the wigner function @xmath257 is positive , see ( * ? ? ?
* theorem 1.102 ) , and the approximation reads as @xmath258 with @xmath259 sampled according to @xmath257 . if @xmath256 is a superposition of gaussian wave packets , then the wigner function @xmath257 is a sum of phase space gaussians plus oscillatory cross terms , such that stratified sampling can be applied , see @xcite . here , we add the second order correction @xmath260 to @xmath155 and discretize @xmath261 where we split the phase space integral into two parts .
this splitting has an impact on the computing time , since the prefactor @xmath262 allows to discretize the second summand rather coarsely without diminishing the overall accuracy of the approximation .
we approximate the two integrals via @xmath263 such that @xmath264 .
the two integrals depend on the wigner function of the initial wave function . here
, we consider a gaussian wave packet centered at @xmath265 , @xmath266 in this case the wigner function can be calculated analytically as a phase space gaussian with mean @xmath267 and covariance matrix @xmath268 ( see , for example @xcite ) , namely @xmath269 let @xmath270 be the function to be integrated .
due to the high dimensionality of the problem , we use quasi - monte carlo quadrature . that is , @xmath271 with quadrature nodes @xmath272 of low star discrepancy with respect to the multivariate normal distribution .
then , the koksma - hlawka inequality yields a constant @xmath273 such that @xmath274 where @xmath275,see , for example , @xcite . to compute @xmath155 we have to discretize the hamiltonian equation
@xmath276 now let @xmath277 and @xmath278 be the flows of the following differential equations @xmath279 these flows can be computed exactly by @xmath280 and @xmath281 defines a symplectic second order splitting scheme for @xmath251 , the so - called strang splitting , see @xcite .
by suitable compositions of this scheme , one can construct symplectic splitting schemes of arbitrary order , see e.g. @xcite ) . for computing the correction term @xmath70 , we write the ordinary differential equation of theorem [ theo : ham_schro ] as @xmath282 with @xmath283 the zero pattern of the matrices @xmath284 allows to compute explicit flow maps .
indeed , let @xmath285 , @xmath286 , @xmath287 be the flows of the differential equations @xmath288 then , @xmath289 for @xmath251 , we obtain a first order splitting scheme for equation by @xmath290 see , @xcite .
its adjoint is @xmath291 , and we can create a second order symmetric splitting method by @xmath292 a corresponding fourth order splitting is obtained by @xmath293 see @xcite , and we obtain @xmath294 as @xmath295 . by construction , the first @xmath296 components of @xmath297 , and @xmath298 , respectively , define symplectic maps on phase space . at this point
we have built up all the ingredients for computing an approximation to expectation values , which is fourth order accurate with respect to the semiclassical parameter @xmath7 .
let @xmath34 be a schwartz function . according to the discussion in section [ sec_discrete_a2 ] , we use time splitting schemes of order eight and order four , for the approximation of @xmath299 and @xmath70 , respectively .
that is , @xmath300 as @xmath295 , with @xmath301 and @xmath302 where @xmath303 is a symplectic eighth order splitting for the hamiltonian flow @xmath140 , while @xmath304 and the tensors @xmath305 consist of the appropriate components of the fourth order splitting @xmath306 . using the quasi - monte carlo estimate ( [ approx_error_phase_sampl ] )
, we obtain @xmath307 for @xmath308 and @xmath309 . choosing the time steps and the number of sampling points such that @xmath310 we finally obtain the desired asymptotic approximation @xmath311 we note , that the number of sampling points @xmath312 can be chosen much smaller than @xmath313 and that the step size @xmath314 can be chosen about the same size as @xmath315 . ignoring the logarithmic term in the quasi - monte carlo estimate
, we deduce as a rule of thumb : @xmath316 hence , although the system of differential equations for the approximation of the correction @xmath70 is more intricate than the one for @xmath299 , for moderate dimensions @xmath317 , the computation of the correction is less costly .
for our numerical experiments , we consider the time dependent schrdinger equation @xmath318 in two dimensions with torsional potential .
the time interval is @xmath319 $ ] , and the initial data @xmath256 is chosen as a single gaussian wave packet ( [ intro_wave_pack ] ) with center @xmath320 .
we validate the approach developed in section [ discretization ] for different observables , namely the position and momentum operators given by @xmath321 as well as the potential , kinetic and total energy operators defined by @xmath322 since the computation of the leading order egorov term @xmath155 has already been elaborated in @xcite , we will mainly focus on the correction term @xmath70 .
the main goal of our numerical experiments is to show that our suggested algorithm can reach order four accuracy with respect to @xmath7 , does this in an efficient way , and thus is feasible in a moderately high dimensional setting . as a reference
, we use grid - based solutions to the schrdinger equation computed by a strang splitting scheme with fourier collocation , see ( * ? ? ?
* appendix ) , and derive the needed expectation values .
the parameters we use for computing the reference solution are given in table [ tab_ref_data ] .
[ tab_ref_data ] @ l l l l l l l @xmath323 & # time steps & domain & space grid + @xmath16 & @xmath324 & @xmath325\times[-3,3]$ ] & @xmath326 + @xmath327 & @xmath328 & @xmath325\times[-3,3]$ ] & @xmath326 + @xmath329 & @xmath328 & @xmath325\times[-3,3]$ ] & @xmath326 + @xmath330 & @xmath331 & @xmath325\times[-3,3]$ ] & @xmath326 + figure [ fig_vals_all ] presents the expectation values computed by the full approximation @xmath332 for the case @xmath333 .
we note that the total energy is preserved , see also remark [ rem : inv ] .
figure [ fig_errors_all ] compares the expectation values computed from the reference solution with those from the egorov approximation @xmath334 as well as the corrected approximation ( [ eq_approx_all ] ) .
we observe that the error of the corrected approach is of the order @xmath335 , while the error of the uncorrected approximation ( [ eq_approx_a0 ] ) is only of the order @xmath262 .
+ to validate that the corrected algorithm is of the order of four in @xmath7 , we compute the maximal and mean deviation from our reference solution for @xmath336 , where the number of halton points @xmath313 and @xmath312 and the size of the time steps @xmath315 and @xmath314 are given in table [ tab_diff_eps ] . in table
[ tab_diff_eps ] the computing times of the egorov approximation scheme ( [ eq_approx_a0 ] ) and the correction term @xmath337 are presented as well . for a given accuracy we can observe that , especially for small semiclassical parameters , ( [ eq_approx_a2 ] )
can be computed much faster than ( [ eq_approx_a0 ] ) . in figure [ fig_diff_eps ] the expected fourth order of our approximation
can be seen .
[ tab_diff_eps ] @ l l l l l l l @xmath7 & @xmath313 & @xmath315 & @xmath338 & @xmath312 & @xmath314 & @xmath339 + @xmath16 & @xmath340 & @xmath16 & @xmath341 sec & @xmath342 & @xmath343 & @xmath344 sec + @xmath327 & @xmath345 & @xmath16 & @xmath346 sec & @xmath347 & @xmath348 & @xmath349 sec + @xmath329 & @xmath350 & @xmath16 & @xmath351 min & @xmath352 & @xmath353 & @xmath344 min + @xmath330 & @xmath354 & @xmath16 & @xmath355 min & @xmath352 & @xmath353 & @xmath344 min + for the experiments shown in figure [ fig_err_n ] and figure [ fig_err_tau ] the values of the leading order approximation ( [ eq_approx_a0 ] ) are computed with @xmath356 sampling points and a time step size of @xmath357 . in figure
[ fig_err_n ] , we examine the dependency of the absolute error of the expectation values from the number of particles @xmath312 . to do so we compare expectation values coming from ( [ eq_approx_all ] ) with those from our reference and compute the maximal as well as the mean error over the time interval @xmath319 $ ] . for the computation of ( [ eq_approx_a2 ] )
we apply our fourth order splitting scheme with a time step size of @xmath358 .
we can see that the mean as well as the maximal error decrease with an order slightly worse than @xmath359 until they reach a lower bound of the order @xmath335 . in figure [ fig_err_tau ] ,
the mean and maximal error over time of the approximated expectation values depending on the step size @xmath314 is displayed . here
the number of sampling points in chosen as @xmath360 .
as expected , we observe that the mean as well as the maximum error decrease with an order of @xmath361 until they reach values of the order @xmath335 .
we thank one of the anonymous referees for pointing us to the literature on wigner trajectories .
this research was supported by the german research foundation ( dfg ) , collaborative research center sfb - tr 109 .
let @xmath2 be a small positive parameter .
the @xmath7-scaled wigner function @xmath362 of a square integrable function @xmath26 is defined as @xmath363 see for example @xcite or @xcite . using plancherel s theorem it can easily be shown that @xmath364 implies @xmath365 the real - valued wigner function @xmath33 can therefore be interpreted as a phase space description of the wave function @xmath28 .
in contrast to classical phase space distributions , the wigner function may also attain negative values .
the wigner function is in close relation to the @xmath7-scaled weyl quantization @xmath36 of schwartz functions @xmath34 , since @xmath366 with @xmath367 for @xmath26 .
weyl quantization also extends to unbounded phase space functions @xmath37 , if operator domains are defined suitably .
for example , @xmath368 .
the composition of weyl operators is a weyl operator @xmath369 , and the symbol @xmath370 has an asymptoic expansion in powers of @xmath7 , @xmath371 where the @xmath69th generalized poission bracket is defined as @xmath372 see ( * ? ? ?
* appendix ) .
we note that the first generalized poisson bracket coincides with the usual poission bracket , @xmath373 .
moreover , the antisymmetry of the possion bracket is also satisfied by its generalizations , if @xmath69 is odd .
[ prop_sym_poisson ] let @xmath374 be smooth functions and @xmath122 . then ,
@xmath375 by interchanging the multi - indices @xmath376 and @xmath377 in the formula for @xmath378 we get @xmath379 since @xmath380 , we conclude that for even @xmath69 , @xmath381 is even if and only if @xmath382 is even and for odd @xmath69 , @xmath381 is even if and only if @xmath382 is odd , which finishes the proof .
in consequence , the commutator of two weyl operators has an asymptotic expansion in odd powers of @xmath7 , @xmath383 = \op(a)\op(b)-\op(b)\op(a ) \sim 2 \sum_{k\in2\n+1 } \left(\frac{\eps}{2i}\right)^k \op(\{a , b\}_k).\ ] ]
99 bouzouina a and robert d 2002 uniform semiclassical estimates for the propagation of quantum observables _ duke math .
j. _ * 111 * 223252 brown r and heller e 1981 classical trajectory approach to photodissociation : the wigner method 186188 donoso a and martens c 2001 quantum tunneling using entangled classical trajectories 223202 dimassi m and sjstrand j 1999 _ spectral asymptotics in the semi - classical limit ( lms lecture note series no .
( cambridge : cambridge university press ) egorov y 1969 on canonical transformations of pseudodifferential operators _ uspekhi mat .
* 24 * 235236 folland g 1989 _ harmonic analysis in phase space _
( princeton : princeton university press ) de gosson m 2011 _ symplectic methods in harmonic analysis and in mathematical physics _
( basel : birkhuser ) heller e 1976 wigner phase space method : analysis for semiclassical applications 12891298 kryvohuz m and cao j 2005 quantum - classical correspondence in response theory 180405 lasser c and rblitz s 2010 computing expectation values for molecular quantum dynamics _
siam j. sci .
comput . _ * 32 * 14651483 lee h 1992 wigner trajectories of a gaussian wave packet perturbed by a weak potential _ found
* 22 * 9951010 lee h 1995 theory and application of the quantum phase - space distrubtion functions _ phys .
rep . _ * 259 * 147211 lee h and scully m 1980 a new approach to molecular collisions : statistical quasiclassical method 22382242 lee h and scully m 1982 wigner phase - space description of a morse oscillator 46044610 liu j and miller w 2007 real time correlation function in a single phase space integral beyond the linearized semiclassical initial value representation 234110 martinez a 2002 _ an introduction to semiclassical and microlocal analysis _
( _ universitext _ ) ( new york : springer ) miller w 1974 quantum mechanical transition state theory and a new semiclassical model for reaction rate constants 18231834 pulvirenti m 2006 semiclassical expansion of wigner functions 052103 robert d 1987 _ autour de lapproximation semi - classique _
( _ progress in mathematics _
vol 68 ) ( boston : birkhuser ) spohn h and teufel s 2001 adiabatic decoupling and time - dependent born - oppenheimer theory _ commun .
phys . _ * 224 * 113172 taylor m 1991 _ pseudodifferential operators and nonlinear pde _ ( _ progress in mathematics _ vol 100 ) ( boston : birkhuser ) thoss m and wang h 2004 semiclassical description of molecular dynamics based on initial - value representation methods _ annu .
chem . _ * 55 * 299332 waalkens h , schubert r and wiggins s 2008 wigner s dynamical transition state theory in phase space : classical and quantum r1r118 wang h , sun x and miller h 1998 semiclassical approximations for the calculation of thermal rate constants for chemical reactions in complex molecular systems 97269736 wigner e 1932 on the quantum correction for thermodynamic equilibrium 749759 zworski m 2012 _ semiclassical analysis _ ( _ graduate studies in mathematics _ vol 138 ) ( providence : ams ) yoshida h 1990 _ construction of higher order symplectic integrators _ _ physics letters a _ * 150 * no.5,6,7 262268 hairer e and lubich c and wanner g 2006 _ geometric numerical integration .
structure - preserving algorithms for ordinary differential equations , second edition _ ( _ springer series in computational mathematics _
vol 31 ) ( berlin : springer - verlag )
|
over decades , the time evolution of wigner functions along classical hamiltonian flows has been used for approximating key signatures of molecular quantum systems .
such approximations are for example the wigner phase space method , the linearized semiclassical initial value representation , or the statistical quasiclassical method .
the mathematical backbone of these approximations is egorov s theorem . in this paper , we reformulate the well - known second order correction to egorov s theorem as a system of ordinary differential equations and derive an algorithm with improved asymptotic accuracy for the computation of expectation values . for models with
easily evaluated higher order derivatives of the classical hamiltonian , the new algorithm s corrections are computationally less expensive than the leading order wigner method .
numerical test calculations for a two - dimensional torsional system confirm the theoretical accuracy and efficiency of the new method . [ section ] [ section ] [ section ] [ section ] [ section ] [ section ]
|
the problem of formulating supersymmetric theories on lattices has a long history going back to the earliest days of lattice gauge theory .
however , after initial efforts failed to produce useful supersymmetric lattice actions the topic languished for many years . indeed a folklore developed that supersymmetry and the lattice were mutually incompatible . however , recently , the problem has been re - examined using new tools and ideas such as topological twisting , orbifold projection and deconstruction and a class of lattice models have been constructed which maintain one or more supersymmetries exactly at non - zero lattice spacing .
while in low dimensions there are many continuum supersymmetric theories that can be discretized this way , in four dimensions there appears to a unique solution to the constraints @xmath1 super yang - mills .
the availability of a supersymmetric lattice construction for this theory is clearly very exciting from the point of view of exploring the connection between gauge theories and string / gravitational theories .
but even without this connection to string theory it is clearly of great importance to be able to give a non - perturbative formulation of a supersymmetric theory via a lattice path integral in the same way that one can formally define qcd as a limit of lattice qcd as the lattice spacing goes to zero and the box size to infinity . from a practical point of view
one can also hope that some of the technology of lattice field theory such as strong coupling expansions and monte carlo simulation can be brought to bear on such supersymmetric theories . in this talk
i will outline some of the key ingredients that go into these constructions , the kinds of applications that have been considered so far and highlight the remaining difficulties .
first , let me explain why discretization of supersymmetric theories resisted solution for so long .
the central problem is that naive discretizations of continuum supersymmetric theories break supersymmetry completely and radiative effects lead to a profusion of relevant supersymmetry breaking counterterms in the renormalized lattice action .
the coefficients to these counterterms must then be carefully fine tuned as the lattice spacing is sent to zero in order to arrive at a supersymmetric theory in the continuum limit . in most cases this is both unnatural and practically impossible particularly if the theory contains scalar fields .
of course , one might have expected problems the supersymmetry algebra is an extension of the poincar algebra which is explicitly broken on the lattice . specifically , there are no infinitesimal translation generators on a discrete spacetime so that the algebra @xmath2 is already broken at the classical level .
equivalently it is a straightforward exercise to show that a naive supersymmetry variation of a naively discretized supersymmetric theory fails to yield zero as a consequence of the failure of the leibniz rule when applied to lattice difference operators . in the last five years or
so this problem has been revisited using new theoretical tools and ideas and a set of lattice models have been constructed which retain exactly some of the continuum supersymmetry at non - zero lattice spacing .
the basic idea is to maintain a particular subalgebra of the full supersymmetry algebra in the lattice theory .
the hope is that this exact symmetry will constrain the effective lattice action and protect the theory from dangerous susy violating counterterms .
two approaches have been pursued to produce such supersymmetric actions ; one based on ideas drawn from the field of topological field theory @xcite and another pioneered by david b. kaplan and collaborators using ideas of orbifolding and deconstruction @xcite .
remarkably these two seemingly independent approaches lead to the same lattice theories
see @xcite and the recent reviews @xcite .
this convergence of two seemingly completely different approaches leads one to suspect that the final lattice theories may represent essentially unique solutions to the simultaneous requirements of locality , gauge invariance and at least one exact supersymmetry .
we will only have time to discuss the approach via topological twisting in this talk .
perhaps the simplest way to understand how this subalgebra emerges is to reformulate the target theory in terms of `` twisted fields '' . the basic idea of twisting goes back to witten in his seminal paper on topological field theory @xcite but
actually had been anticipated in earlier work on staggered fermions @xcite . in our context
the idea is decompose the fields of the theory in terms of representations not of the original ( euclidean ) rotational symmetry @xmath3 but a twisted rotational symmetry which is the diagonal subgroup of this symmetry and an @xmath4 subgroup of the r - symmetry of the theory .
so(d)^=diag(so_lorentz(d)so_r(d ) ) to be explicit consider the case where the total number of supersymmetries is @xmath5 . in this case
i can treat the supercharges of the twisted theory as a @xmath6 matrix @xmath7 .
this matrix can be expanded on products of gamma matrices q = qi+q__+q___nu+ the @xmath8 antisymmetric tensor components that arise in this basis are the twisted supercharges and satisfy a corresponding supersymmetry algebra following from the original algebra @xmath9 the presence of the nilpotent scalar supercharge @xmath10 is most important ; it is the algebra of this charge that we can hope to translate to the lattice .
the second piece of the algebra expresses the fact that the momentum is the @xmath10-variation of something which makes plausible the statement that the energy - momentum tensor and hence the entire action can be written in @xmath10-exact form there is an additional @xmath10-closed term needed ] .
notice that an action written in such a @xmath10-exact form is trivially invariant under the scalar supersymmetry provided the latter remains nilpotent under discretization .
the rewriting of the supercharges in terms of twisted variables can be repeated for the fermions of the theory and yields a set of antisymmetric tensors @xmath11 which for the case of @xmath5 matches the number of components of a real khler - dirac field .
this repackaging of the fermions of the theory into a khler - dirac field is at the heart of how the discrete theory avoids fermion doubling as was shown by becher , joos and rabin in the early days of lattice gauge theory @xcite .
it is important to recognize that the transformation to twisted variables corresponds to a simple change of variables in flat space
one more suitable to discretization .
a true topological field theory only results when the scalar charge is treated as a true brst charge and attention is restricted to states annihilated by this charge . in the language of the supersymmetric parent theory such a restriction corresponds to a projection to the vacua of the theory .
it is _ not _ employed in these lattice constructions .
this theory satisfies our requirements for supersymmetric latticization ; its r - symmetry possesses an @xmath12 subgroup corresponding to rotations of the its two degenerate majorana fermions into each other .
explicitly the theory can be written in twisted form as s= q(_f_+-d)[2daction]the degrees of freedom are just the twisted fermions @xmath13 previously described and a complex gauge field @xmath14 .
the latter is built from the usual gauge field and the two scalars present in the untwisted theory @xmath15 with corresponding complexified field strength @xmath16 .
notice that the original scalar fields transform as vectors under the original r - symmetry and hence become vectors under the twisted rotation group while the gauge fields are singlets under the r - symmetry and so remain vectors under twisted rotations .
this structure makes possible the appearance of a complex gauge field in the twisted theory .
notice though , that the theory is only invariant under the usual @xmath17 gauge symmetry and not its complexified cousin .
the nilpotent transformations associated with @xmath10 are given explicitly by @xmath18 performing the @xmath10-variation and integrating out the auxiliary field @xmath19 yields s=(-_f_+ [ _ , d_]^2- _ d_- _ _ ) [ twist_action ] to untwist the theory and verify that indeed in flat space it just corresponds to the usual theory one can do a further integration by parts to produce s=(-f^2_+2b_d_d_b_-[b_,b_]^2+l_f)where @xmath20 is the usual yang - mills term .
it is now clear that the imaginary parts of the gauge fields @xmath21 can now be given an interpretation as the scalar fields of the usual formulation .
similarly one can build spinors out of the twisted fermions and write the action in the manifestly dirac form l_f= ( cc_12 & ) ( cc - d_2-ib_2&d_1+ib_1 + d_1-ib_1&d_2-ib_2 ) ( c_1 + _ 2 )
the prescription for discretization is somewhat natural .
( complex ) gauge fields are represented as complexified wilson gauge links @xmath22 living on links of a lattice which for the moment we can think of as hypercubic .
these transform in the usual way under @xmath17 lattice gauge transformations u_(x)g(x)u_(x)g^(x)supersymmetric invariance then implies that @xmath23 live on the same links and transform identically .
the scalar fermion @xmath24 is clearly most naturally associated with a site and transforms accordingly ( x)g(x)(x)g^(x)the field @xmath25 is slightly more difficult . naturally as a 2-form it should be associated with a plaquette . in practice
we introduce diagonal links running through the center of the plaquette and choose @xmath25 to lie _ with opposite orientation _ along those diagonal links .
this choice of orientation will be necessary to ensure gauge invariance .
figure 1 . shows the resultant lattice theory ( with @xmath26 and @xmath27 and @xmath28 ) [ fig1 ] to complete the discretization we need to describe how continuum derivatives are to be replaced by difference operators . a natural technology for accomplishing this in the case of adjoint fields
was developed many years ago and yields expressions for the derivative operator applied to arbitrary lattice p - forms @xcite . in the case
discussed here we need just two derivatives given by the expressions @xmath29 the lattice field strength is then given by the gauged forward difference @xmath30 and is automatically antisymmetric in its indices .
furthermore it transforms like a lattice 2-form and yields a gauge invariant loop on the lattice when contracted with @xmath25 .
similarly the covariant backward difference appearing in @xmath31 transforms as a 0-form or site field and hence can be contracted with the site field @xmath32 to yield a gauge invariant expression .
this use of forward and backward difference operators guarantees that the solutions of the theory map one - to - one with the solutions of the continuum theory and hence fermion doubling problems are evaded @xcite . indeed ,
by introducing a lattice with half the lattice spacing one can map this khler - dirac fermion action into the action for staggered fermions @xcite .
notice that , unlike the case of qcd , there is no rooting problem in this supersymmetric construction since the additional fermion degeneracy is already required by the continuum theory . many other examples of supersymmetric lattices exist .
figure 2 . shows two such lattices arising in the case of eight supercharges a two dimensional triangular lattice and a generalized hypercubic lattice ( including body and face links ) in three dimensions .
notice that in all cases almost all fields live on links with the exception of a small number of fermion site fields the number of those corrresponding to the number of exact supersymmetries preserved in the lattice theory .
furthermore , in all cases the number of fermions exactly fills out multiples of a basic khler - dirac field in the corresponding number of dimensions .
supersymmetric lattices , title="fig:",scaledwidth=45.0% ] [ fig2 ]
in four dimensions the constraint that the target theory possess 16 supercharges singles out a single theory for which this construction can be undertaken
@xmath1 sym .
the continuum twist of @xmath1 that is the starting point of the twisted lattice construction was first written down by marcus in 1995 @xcite although it now plays a important role in the geometric - langlands program and is hence sometimes called the gl - twist @xcite .
this four dimensional twisted theory is most compactly expressed as the dimensional reduction of a five dimensional theory in which the ten ( one gauge field and six scalars ) bosonic fields are realized as the components of a complexified five dimensional gauge field while the 16 twisted fermions naturally span one of the two khler - dirac fields needed in five dimensions .
remarkably , the action of this theory contains a @xmath10-exact piece of precisely the same form as the two dimensional theory given in eqn .
[ 2daction ] provided one extends the field labels to run now from one to five .
in addition the marcus twist requires a new @xmath10-closed term which was not possible in the two dimensional theory .
mnpqr_qr_p_mn[closed]the supersymmetric invariance of this term then relies on the bianchi identity @xmath33 .
the four dimensional lattice that emerges from examining the moduli space of the resulting discrete theory is called @xmath34 and is constructed from the set of five basis vectors @xmath35 pointing out from the center of a four dimensional equilateral simplex out to its vertices together with their inverses @xmath36 .
it is the four dimensional analog of the 2d triangular lattice .
complexified wilson gauge link variables @xmath37
are placed on these links together with their @xmath10-superpartners @xmath38 .
another 10 fermions are associated with the diagonal links @xmath39 with @xmath40 .
finally , the exact scalar supersymmetry implies the existence of a single fermion for every lattice site .
the lattice action corresponds to a discretization of the marcus twist on this @xmath34 lattice and can be represented as a set of traced closed bosonic and fermionic loops .
it is invariant under the exact @xmath10 scalar susy , lattice gauge transformations and a global permutation symmetry @xmath41 and can be proven free of fermion doubling problems as discussed before . the @xmath10-exact part of the lattice action is again given by eqn .
[ twist_action ] where the indices @xmath42 now correspond to the indices labeling the five basis vectors of @xmath34 .
while the supersymmetric invariance of this @xmath10-exact term is manifest in the lattice theory it is not clear how to discretize the continuum @xmath10-closed term .
remarkably , it is possible to discretize eqn . [ closed ] in such a way that it is indeed exactly invariant under the twisted supersymmetry ! s_closed=-_*x * _ mnpqr_qr(*x*+_m+_n+_p ) ^(-)_p_mn(*x*+_p)and can be seen to be supersymmetric since the lattice field strength satisfies an exact bianchi identity @xcite .
the lattice theory may be simulated using standard algorithms and techniques drawn from lattice qcd .
the first step is to integrate out the twisted fermions which results in a pfaffian of the twisted fermion kinetic operator @xmath43 . assuming that the phase of this operator can be neglected this pfaffian
can be represented by the integral m(u,)=df df^ e^-f^(m^m)^fwhere the _ bosonic _ pseudofermion field @xmath44 carries the same labels as its fermion counterpart .
the fractional inverse power is then approximated to whatever accuracy is desired by a partial fraction expansion whose coefficients are determined using the remez algorithm to minimize the error over some interval in the spectrum .
a standard rhmc algorithm can then be used to simulate the full partition function of the resultant theory .
one of the first things to check is the of course the exact supersymmetry .
table 1 . compares the monte carlo measurements of the bosonic action against the exact value computed assuming the exact supersymmetry ( the bosonic action can be derived using an exact @xmath10 ward identity ) .
results are shown for @xmath45 and several lattice couplings @xmath46 for both the two dimensional model with 4 supercharges and @xmath1 super yang - mills in four dimensions with 16 supercharges @xcite .
cc .tests of exact susy - bosonic action for @xmath45 theory at various lattice couplings [ cols="^,^,^",options="header " , ] + @xmath47&@xmath48 [ tab1 ] these results confirm rather convincingly that the lattice theory does indeed possess exact supersymmetry .
next we turn to the moduli space . in the continuum and on the lattice
both the 4 and 16 supercharge theories possesses an infinite set of classical vacua corresponding to constant commuting @xmath49 matrices .
furthermore , since the complex fields are in principle unbounded from above one might worry that the partition function would be unbounded . in the usual formulation
the usual statement is that the scalar fields can run off to infinity along these flat directions . in practice
we find that the partition function and low moments of the scalar field distribution are well defined in these theories .
figures 3 . and 4 .
show the distribution of eigenvalues of the scalar fields ( given by @xmath50 ) for the 4 and 16 supercharge @xmath45 models . while one observes long power law tails particularly in the 4 supercharge case
the fields remain localized around the origin in the moduli space , the partition function is finite and no instability is seen in the simulations . for @xmath51 su(2 )
theory , title="fig:",scaledwidth=50.0% ] [ fig3 ] for @xmath52 su(2 ) theory , title="fig:",scaledwidth=50.0% ] [ fig4 ] it is interesting to examine the eigenvalues of the fermion operator accumulated over the same set of monte carlo configurations used in the moduli space analysis .
these are shown in figures 5 .
and 6 . for 4 and 16
supercharges respectively . @xmath45 model on @xmath53 lattice ,
title="fig:",scaledwidth=50.0% ] [ fig5 ] @xmath45 model on @xmath53 lattice , title="fig:",scaledwidth=50.0% ] [ fig6 ] notice these picture show only the region close to the origin
the full spectrum extends to larger values and shows a clustering along the imaginary ( vertical ) axis corresponding to eigenvalues of the derivative operator appearing in the fermion kinetic terms .
nevertheless , these pictures reveal one strong difference between the 4 and 16 supercharge cases ; the relative scarcity of eigenvalues close to the origin in the 16 supercharge case .
the absence of a zero eigenvalue in the fermion spectrum indicates by supersymmetry a corresponding absence of boson zero modes those associated with the classical flat directions .
in the context of the monte carlo this is consistent with the observation that the scalars do not penetrate far down the classical flat directions but rather correspond to a wavefunction which localizes close to the origin of moduli space .
in contrast the scalars in the 4 supercharge theory can be found at relatively large distance from the origin ( the tail of the eigenvalue distribution falls only as a single inverse power of the eigenvalue and thus the variance of the eigenvalues actually diverges logarithmically ) . at such large distances
there is a near bosonic zero corresponding to translations along the flat directions and corresponding a near zero fermion eigenvalue as is seen in the spectrum .
indeed , the presence of a nearly massless fermion mode in the spectrum is a necessary condition for supersymmetry breaking where it would play the role of a goldstino .
an order parameter for supersymmetry breaking is the witten index .
the measured value of @xmath54 is close to zero in this model ( it can be measured via the phase of the pfaffian in the phase quenched theory we simulate ) .
thus we conclude that our simulations yield tantalizing evidence for dynamical supersymmetry breaking in the two dimensional model .
no such breaking apparently occurs for the four dimensional theory in line with expectations .
one of the key issues that still remains to be explored is the question of how much residual fine tuning will be required to achieve a continuum limit in which full supersymmetry is restored .
this is controlled by the flows in all relevant operators which could be induced in the effective action as a result of quantum corrections .
we have used the exact lattice symmetries together with power counting to enumerate the possible set of such lattice operators .
only four terms appear in this counter term analysis and three of these correspond to wavefunction renormalizations of kinetic terms already present in the bare lattice action .
there is one additional term of the form q(u__)which leads to supersymmetric mass terms for the fermions and scalars .
however , such a term would lead to a lifting of the classical moduli space which does not happen in the continuum theory .
we have computed the effective action to one loop in the lattice theory and find that it vanishes as a consequence of exact @xmath10 supersymmetry just as in the continuum .
indeed , this result can be generalized to any finite order of perturbation theory using the @xmath10-exact property of the lattice action and indicates that such a dangerous radiative correction is _ not _ induced to all orders of perturbation theory @xcite .
this is a strong and useful result as it indicates at worst a logarithmic tuning of the other terms in the action will be needed .
having eliminated all such mass terms the remaining question concerning the restoration of full supersymmetry then rests on whether the ratios of the coefficients to the renormalized kinetic operators flow away from their classical values as the lattice spacing is decreased .
a one loop calculation is in progress which should shed light on this issue .
beyond this issue it would be very interesting to use monte carlo simulation to test aspects of the ads / cft conjecture .
parallel code has now been developed to study @xmath1 super yang - mills and a program of numerical investigations of this theory is ongoing . at finite temperature
this theory and its dimensional reductions and deformations should be dual to a variety of black hole solution in supergravity see the recent work reported in @xcite and complementary numerical work using non - lattice methods reported in @xcite .
one would hope that the results of such simulations could be useful in the quest to understand how aspects of the quantum geometry can be understood in terms of the dual gauge theory .
s.c would like to thanks joel giedt , anosh joseph , mithat unsal and toby wiseman for useful discussions .
numerical results shown in this talk were obtained using usqcd resources at fermilab .
smc is supported in part by doe under grant number de - fg02 - 85er40237 .
9 a. dadda , i. kanamori , n. kawamoto , k. nagata , twisted superspace on a lattice , nucl .
phys . b707 ( 2005 ) 100144 .
a. dadda , i. kanamori , n. kawamoto , k. nagata , exact extended supersymmetry on a lattice : twisted n = 2 super yang - mills in two dimensions , phys .
b633 ( 2006 ) 645652 .
a. dadda , i. kanamori , n. kawamoto , k. nagata , exact extended supersymmetry on a lattice : twisted n=4 super yang - mills in three dimensions , nucl .
b798 ( 2008 ) 168183 .
a. g. cohen , d. b. kaplan , e. katz , m. unsal , supersymmetry on a euclidean spacetime lattice .
i : a target theory with four supercharges , jhep 08 ( 2003 ) 024 .
a. g. cohen , d. b. kaplan , e. katz , m. unsal , supersymmetry on a euclidean spacetime lattice .
ii : target theories with eight supercharges , jhep 12 ( 2003 ) 031 .
d. b. kaplan , m. unsal , a euclidean lattice construction of supersymmetric yang- mills theories with sixteen supercharges , jhep 09 ( 2005 ) 042 .
s. catterall , lattice supersymmetry and topological field theory , jhep 05 ( 2003 ) 038 .
f. sugino , a lattice formulation of super yang - mills theories with exact supersymmetry , jhep 01 ( 2004 ) 015 .
s. catterall , s. ghadab , lattice sigma models with exact supersymmetry , jhep 05 ( 2004 ) 044 .
s. catterall , a geometrical approach to n = 2 super yang - mills theory on the two dimensional lattice , jhep 11 ( 2004 ) 006 .
h. damgaard , s. matsuura , lattice supersymmetry : equivalence between the link approach and orbifolding , jhep 09 ( 2007 ) 097 .
h. damgaard , s. matsuura , classification of supersymmetric lattice gauge theories by orbifolding http://arxiv.org/abs/0704.2696 [ hep - lat ] s. catterall , from twisted supersymmetry to orbifold lattices , jhep 01 ( 2008 ) 048 .
m. unsal , twisted supersymmetric gauge theories and orbifold lattices , jhep * 0610 * , 089 ( 2006 ) [ arxiv : hep - th/0603046 ] .
j. giedt , progress in four - dimensional lattice supersymmetry http://arxiv.org/abs/0903.2443 simon catterall , david b. kaplan and mithat nsal , exact lattice supersymmetry phys.rept.484:71-130,2009 .
e. witten , topological quantum field theory , commun .
phys . 117 ( 1988 ) 353 .
s. elitzur , e. rabinovici , a. schwimmer , supersymmetric models on the lattice , phys .
b119 ( 1982 ) 165 . j. m. rabin , homology theory of lattice fermion doubling , nucl .
b201 ( 1982 ) 315 .
p. becher , h. joos , the dirac - kahler equation and fermions on the lattice , zeit .
c15 ( 1982 ) 343 .
h. aratyn , m. goto , a. h. zimmerman , a lattice gauge theory for fields in the adjoint representation , nuovo cim .
a84 ( 1984 ) 255 .
tom banks , y. dothan , d. horn , geometric fermions .
n. marcus , the other topological twisting of n=4 yang - mills , nucl .
b452 ( 1995 ) 331345 .
a. kapustin , e. witten , electric - magnetic duality and the geometric langlands program http://arxiv.org/abs/hep-th/0604151 s. catterall , first results from simulations of supersymmetric lattices , jhep 01 ( 2009 ) 040 .
s. catterall , e. dzeimkowski , j. giedt and a. joseph , in preparation . extracting black hole physics from the lattice .
simon catterall , toby wiseman arxiv:0909.4947 [ hep - th ] + black hole thermodynamics from simulations of lattice yang - mills theory .
simon catterall , toby wiseman , phys.rev.d78:041502,2008 . + towards lattice simulation of the gauge theory duals to black holes and hot strings .
simon catterall , toby wiseman , jhep 0712:104,2007 .
first results from lattice simulation of the pwmm .
simon catterall , greg van anders , arxiv:1003.4952 [ hep - th ] k. n. anagnostopoulos , m. hanada , j. nishimura and s. takeuchi , _ monte carlo studies of supersymmetric matrix quantum mechanics with sixteen supercharges at finite temperature _ , _ phys .
* 100 * ( 2008 ) 021601 [ arxiv:0707.4454[hep - th ] ] .
m. hanada , a. miwa , j. nishimura and s. takeuchi , _ schwarzschild radius from monte carlo calculation of the wilson loop in supersymmetric matrix quantum mechanics _
, _ phys .
* 102 * ( 2009 ) 181602 [ arxiv:0811.2081[hep - th ] ] .
m. hanada , y. hyakutake , j. nishimura and s. takeuchi , _ higher derivative corrections to black hole thermodynamics from supersymmetric matrix quantum mechanics _
, _ phys . rev .
* 102 * ( 2009 ) 191602 [ arxiv:0811.3102[hep - th ] ] .
g. ishiki , s. w. kim , j. nishimura and a. tsuchiya , _ deconfinement phase transition in @xmath55=4 super yang - mills theory on @xmath56 from supersymmetric matrix quantum mechanics _ , _ phys .
* 102 * ( 2009 ) 111601 [ arxiv:0810.2884[hep - th ] ] ; g. ishiki , s. w. kim , j. nishimura and a. tsuchiya , _ testing a novel large - n reduction for n=4 super yang - mills theory on @xmath57 _ , _ jhep _ * 09 * ( 2009 ) 029 [ arxiv:0907.1488[hep - th ] ] .
m. hanada , j. nishimura , y. sekino and t. yoneya , _ monte carlo studies of matrix theory correlation functions _ , _ phys .
* 104 * ( 2010 ) 151601 [ arxiv:0911.1623[hep - th ] ] .
|
discretization of supersymmetric theories is an old problem in lattice field theory .
it has resisted solution until quite recently when new ideas drawn from orbifold constructions and topological field theory have been brought to bear on the question .
the result has been the creation of a new class of lattice gauge theory in which the lattice action is invariant under one or more supersymmetries .
the resultant theories are local and free of doublers and in the case of yang - mills theories also possess exact gauge invariance . in principle
they form the basis for a truly non - perturbative definition of the continuum supersymmetric field theory . in this talk
these ideas are reviewed with particular emphasis being placed on @xmath0 super yang - mills theory .
|
the propagation of an electron beam in a two - dimensional degenerate electron gas ( 2deg ) can be used to study the role of electron - electron ( e - e ) collisions in reduced dimensions , see e.g. refs .
of main interest are `` high - energy '' beams with a typical electron energy that exceeds considerably the thermal energy of the equilibrium electrons within the 2deg , but remains significantly below the fermi energy .
a special combination of ac- and dc - measurement techniques allows for an energy selective investigation of electrons traversing a defined distance within a 2deg @xcite .
some of these experiments were analyzed in terms of the @xmath0-approximation @xcite , where it is supposed that each collision scatters an electron out of the beam .
however , this approach is obviously insufficient for the case of electron - electron ( ee ) scattering in a 2deg , where small angle scattering events dominate , and scattered electrons may remain within the region of the electron beam @xcite . for an adequate theoretical description of ee - scattering
it is necessary to take into account the incoming terms in the collision operator @xmath1 ( see the second term on the right of the equation for @xmath1 ) of the dynamic integro - differential linearized boltzmann equation : @xmath2 here , @xmath3 is the distribution function of non - equilibrium electrons , @xmath4 determines the location of a non - equilibrium electron , and @xmath5 and @xmath6 are the electron velocity and momentum , respectively ( quasi - classical approximation ) .
@xmath7 is the lorentz force acting on the electron .
the function @xmath8 determines the probability of non - equilibrium electrons ( @xmath9 ) or holes ( @xmath10 ) to be found in state @xmath6 after scattering from state @xmath11 .
a hole in this case describes the absence of an equilibrium electron below the fermi surface . in the general case ,
the solution of the kinetic boltzmann equation ( [ e - kin ] ) is a very difficult task , both in analytical and numerical calculations .
but it is well known from the kinetics of gases that two regimes exist which allow for an essential simplification of the description .
first , in the few - collision case the probability is small for an electron to be scattered over a characteristic length scale @xmath12 ( e.g. , the distance between the injector to the detector , see fig.[g - exp ] below ) , which is much smaller than the electron - electron mean - free path , @xmath13 , i.e. @xmath14 .
therefore , using perturbation theory it is possible to solve the boltzmann equation by iteration of the collision operator ( quasiballistic approximation ) .
second , in the case of frequent collisions ( @xmath15 ) a hydrodynamic situation arises .
the influence of frequent collisions establishes a quasi - equilibrium distribution which is characterized by hydrodynamic variables , such as the average velocity or temperature which depend only smoothly on @xmath4 .
these regimes were studied in our previous publications @xcite . for obvious reasons
a treatment of the intermediate regime ( @xmath16 ) presents considerable difficulties . in this paper
we show that it is possible to study the intermediate regime successfully if the scattering probability decreases substantially with each collision .
the occurrence of a decreasing scattering probability can be expressed by an additional small parameter @xmath17 which is defined as the ratio of the scattering probability of a nonequilibrium electron after a collision to the probability before this event .
such situations can be found in a wide range of problems where non - equilibrium quasi - particles exhibiting some excess energy , above thermal equilibrium , collide with equilibrium quasi - particles and lose a considerable amount of this energy due to this event .
usually , the probability for inelastic scattering decreases drastically with decreasing energy of the quasi - particles , because the statistical weight decreases for both the possibility for finding a collision partner and for finding empty final states .
electron - electron , electron - phonon or phonon - phonon collisions at temperatures below the debye temperature can be taken as an example .
our approach is based on the following concept : ( i ) a few collisions are treated exactly , while ( ii ) subsequent collisions are neglected , or considered perturbationally .
we will refer to this approach as the `` modified one - collision approximation '' ( moc ) , which accounts for the fact that in most cases a restriction to an one collision approximation is sufficient to explain the general features of the scattering experiments . in the present work ,
a full account of the moc approximation is given and the results are applied to analyze experimental data obtained from electron beam experiments .
it is shown , that applicability of the moc approximation to the scattering of high - energy electrons substantially extends the description of spectroscopic data from electron - electron collisions experiments @xcite .
the non - monotonic dependence of the signal of scattered electrons on the beam energy can be explained within the moc approximation .
this observation can be interpreted as a crossover from the ballistic to multi - collision ( hydrodynamic ) flow regime .
let us consider non - equilibrium electrons at a characteristic energy @xmath18 , measured from the fermi level @xmath19 , and larger than the temperature @xmath20 of the equilibrium electron gas , @xmath21 .
in a first stage of relaxation these highly non - equilibrium electrons are scattered by equilibrium electrons thus generating a new non - equilibrium system but with already much lower excess energy .
this stage of relaxation was named the `` pretemperature '' stage in ref.@xcite .
this `` pretemperature '' stage ends when the quasi - particles energy becomes comparable to the bath temperature @xmath20 .
the most important fact for the `` pretemperature '' stage is the following : if the energy @xmath22 , then the probability of secondary collision is an order of magnitude lower than for the first one @xcite .
indeed , the energy @xmath23 of a non - equilibrium electron after collision with an equilibrium electron will be distributed between three states in phase space : @xmath24 @xmath25 for @xmath26 , we have @xmath27 and @xmath28 due to pauli principle , i.e. the state @xmath11 is below the fermi level while the final states @xmath29 and @xmath30 are well above .
the energy of the initial electron is redistributed between three final states . on average ,
the energy of each state after the first collision can be expressed as @xmath31 , with @xmath32 ( note , the numerical value of @xmath33 depends on the specific type of problem considered)@xcite . since the energy dependent mean - free path @xmath34 , this means that @xmath35 of non - equilibrium quasi - particles will increase by about an order of magnitude , which implies that @xmath36 .
this may lead to a situation where the probability for a secondary collision is low while the probability for the first collision is close to unity in a system of size @xmath12 : @xmath37 . in this case , it is a natural choice to consider the first collision exactly while the secondary collision is treated by perturbation theory , which leads to an extension of the validity of the one - collision approximation up to @xmath38 . in the following , we use this approach for the case of propagation of nonequilibrium electrons in an electron beam injected into a 2deg via a quantum point - contact . for steady state conditions , the operator @xmath39 on the left side of eq .
( [ e - kin ] ) must be replaced by @xmath40 . for further consideration it
is convenient to use cartesian coordinates .
the x - axis is along the beam direction , while the origin is located at the center of the electron beam injector ( a quantum point - contact in an actual experiment ) .
( [ e - kin ] ) will be solved by integrating along the trajectory .
we define @xmath41 as the time of motion of the particle along a given trajectory .
@xmath42 is the ( lagrange ) coordinate of the particle at a time @xmath41 . at the time of observation ( taken as the origin of the time counting )
, the particle is located at the point @xmath4 with a momentum @xmath6 .
the momentum of the particle at a time @xmath41 is @xmath43 .
the particle trajectory @xmath44 is uniquely defined for given applied fields .
the solution of the equation @xmath45 defines the start time @xmath46 of the electron injection into the system at @xmath47 .
the solution of eq .
( [ e - kin ] ) is given as @xmath48 @xmath49 f^{(n)}({\bf r } , { \bf p'})\nonumber\\ \equiv \int_{t_0}^0 dt ( \hat{\nu } - \nu ) f^{(n)}.\end{gathered}\ ] ] in this expression each @xmath50-th term of the series is obtained by @xmath50 subsequent actions of the collision operator on the initial or unpertubated function @xmath51 which describes the ballistic motion of the particle : @xmath52 here , @xmath53 is an injected distribution where all the injected electrons have a characteristic energy @xmath54 .
the quasi - particles , which appear as the result of a collision , are described by the action of the incoming part @xmath55 of the collision operator @xmath56 on @xmath57 . in other words , @xmath58 correspond to a particle that has been scattered once , while @xmath59 describes a subsequent second scattering event , and so forth .
correspondingly , the typical energies of these states differ : @xmath60 is characterized by the energy @xmath54 , for @xmath58 the characteristic energy will be @xmath61 and for @xmath62 it will be given by @xmath63 etc .
consequently we find that @xmath64 and that @xmath65 . therefore , as a main approximation in the parameters @xmath66 , @xmath67 , it is possible to neglect all terms in the series of eq .
( [ e - mocriad ] ) that include @xmath55 or @xmath68 to an order higher than one , where @xmath68 is placed to the left of @xmath55 . after making this approximation ,
the summation of the remaining terms can be carried out easily .
thus , the main idea consists in a partial summation of the infinite series to a first or subsequent approximation , as required in order to reach sufficient accuracy for an adequate description of the experiment . for a stationary beam in a magnetic field , a green s function of the `` moc-1 '' ( i.e. the solution of eq .
( [ e - kin ] ) for @xmath69 , that takes into account the first collision exactly ) has the following form : @xmath70 here @xmath71 is the distribution of the ballistic part of the electrons , i.e. , the solution of eq .
( [ e - kin ] ) in the absence of the integral operator @xmath55 .
@xmath72 is the time of collision and @xmath46 is the starting time .
@xmath73 and @xmath74 are the momenta before and after collision , respectively , while @xmath75 .
the values of @xmath41 , @xmath46 , @xmath76 , @xmath77 , @xmath78 , and @xmath6 are determined through the equations of motion by the initial momentum @xmath79 and location @xmath80 .
the equations of motion are reduced to @xmath81 .
the presence of walls in the experimental setup , see fig .
[ g - exp ] , excludes solutions with a full cyclotron orbit .
note that the moc approximation is applicable in principle under conditions , where a pre - temperature stage occurs for the non - equilibrium quasi - particles .
in other words , it holds when the following hierarchy of mean - free paths is given : @xmath82 where @xmath83 is the energy after the @xmath84-th collision and @xmath85 ; for @xmath86 the inequality corresponds to the quasi - ballistic regime , where perturbation theory is applicable .
if @xmath12 corresponds to the @xmath50-th segment of this hierarchy , i.e. @xmath87 , it is necessary to construct a `` moc-@xmath50 '' approximation by treating the first @xmath50 collisions exactly ; which leads to relative corrections of the order of @xmath88 .
the approach breaks down when the quasi - particle energy is comparable to @xmath20 . as an example
, the exact result for the first two collisions ( `` moc-2 '' ) , for the spatially homogeneous case and @xmath89 , reads : @xmath90\hat{\nu } \big\}f_{0},\hspace{1mm}\\ \nu_{0}=\nu(\varepsilon_{0}).\end{gathered}\ ] ] this formula is valid for times @xmath91 .
thus , just a few steps lead to a high accuracy .
it turns out , that for the experiments discussed below , just one step is enough to produce satisfactory results .
in order to investigate the electron - electron scattering properties in a 2deg , experiments are needed which allow to distinguish between scattered and non - scattered particles , and reveal the details of scattered signal .
therefore , we used the experimental setup of the electron spectroscopy method proposed in ref .
@xcite ( see fig . [ g - exp ] ) .
the sample is based on a conventional si - modulation doped ga - as(al , ga)as heterojunction with a carrier density of @xmath92 @xmath93 and an impurity mean - free path of @xmath94 @xmath95 m .
the scheme of the experiment is shown in fig .
[ g - exp ] .
schottky gates define quantum point - contacts in the 2deg and are adjusted such that the injector and detector quantum point - contact ( qpc ) are carrying only one conducting mode .
the distance between the injector and detector qpc is @xmath96 @xmath95 m .
ohmic contacts are indicated by crossed squares in fig .
[ g - exp ] .
the sample was cooled down to @xmath97 k. an ac - modulated dc - bias voltage applied between ohmic contacts 1 and 2 was used to inject the electrons into the 2deg area between injector and detector .
the amplitude of the ac modulation @xmath98 was much smaller than the dc component , @xmath99 .
the ac - component of the detector signal , i.e. the voltage that occurs between contacts 3 and 4 ( @xmath100 ) is measured with a lock - in technique . with this technique only the signal of electrons with energy @xmath101 is measured , while the whole e - beam consists of electrons with energies in the range of @xmath102 .
( note that , throughout this paper the electron charge @xmath103 is set equal to one @xmath104 . )
mv , ( dash )
@xmath105 mv , ( dot )
@xmath106 mv , ( dash - dot )
@xmath107 mv.,width=302,height=6 ] the magnetic field dependence of the detector signal @xmath108 has been studied for a wide range of energies @xmath101 @xmath109 @xmath110 mv .
an example is given in fig .
[ g - expb ] for different injection energies @xmath101 .
this figure display the evolution of an mono - energetic electron beam with increasing injection energy@xmath111 , which reflects the energy - dependent electron - electron scattering . from the half - hight width of low energy curve ( 1 ) ( @xmath112 mv )
it is possible to extract the opening angle of the collimated electron beam : @xmath113 , which is identical for injector and detector when the qpcs are equally adjusted to one conducting mode ( @xmath114 ) .
the contribution of scattered particles to the detector signal @xmath115 can be easily extracted following the approach of ref .
@xcite : @xmath116 here , @xmath117 is the energy dependent e - e scattering length according to ref .
@xcite , where @xmath118 is the thomas - fermi screening wave vector ; @xmath119 is the cyclotron radius ; @xmath120 is the fermi - velocity .
the second term on the right - hand side of this expression corresponds to @xmath71 defined in eq .
( [ e - moca ] ) , @xmath121 , represents the signal that would be observed in the absence of scattering .
a good approximation for this value can be obtained from the signal measured for the lowest excitation voltage @xmath122 at @xmath123 mv . in this cases ,
@xmath124 and electron - electron collisions can be ignored .
note that for injection energies @xmath125 mv @xmath126 the above equation for @xmath127 yields a large inaccuracy .
however , this can be ignored because @xmath127 becomes so small that the subtraction of the ballistic contribution to the signal leads only to corrections of less than one percent @xcite .
in fig . [ g - vsmx ] the maximum value of the magnetic field dependent signal due to the scattered electrons , @xmath128 is presented as a function of the injection energy ( @xmath101 ) .
this dependence shows a distinct non - monotonic behavior , i.e. a rapid increase at small @xmath129 , a distinct maximum at @xmath130 mv and an abrupt decrease for larger injection energies up to @xmath131 mv , followed by a weak increase .
the presence of a maximum and the subsequent decrease in the energy dependence of @xmath132 can not be explained in the context of a simple perturbation theory applied to the collision integral , because in the regime where the maximum is observed the electron - electron mean - free path is already smaller than the distance between the injector and detector qpc .
perturbation theory can only be applied to describe the increase of the signal for small energies @xmath101 .
the presence of a strong decrease appears at first sight to be unexpected , because in the multi - collision regime that is entered upon increasing @xmath101 , a decrease of the ac - signal from scattered particles is not anticipated . in this multi - collision regime
the ballistic beam is already absent but the motion of the electrons can be represented by a hydrodynamic drift . however , the behavior shown in fig .
[ g - vsmx ] can be explained in the context of the modified one - collisional approximation described above by taking into account the specific properties of e - e scattering in two - dimensions @xcite .
a calculation shows , that by taking into account the first collision exactly , eq . ( [ e - moca ] ) leads to the following expression for the detector signal due to scattered particles @xmath133 where @xmath134 is the scattering angle and @xmath135 is the angular distribution function of e - e scattering , @xmath136 , averaged over an interval @xmath137 ; the weighting coefficient @xmath138 is @xmath139 for @xmath140 and @xmath141 for @xmath142 .
( it is sufficient to consider only the case where the cyclotron radius @xmath119 is smaller than @xmath12 . )
it should be noted that eq .
( [ e - specform ] ) is valid for all ratios of @xmath137 and the angular width of @xmath143 .
the expression @xmath144 in the exponent of eq .
( [ e - specform ] ) represents the length of the electron trajectory from the injector to the point @xmath145 where the scattering takes place ( see fig . [ g - exp ] ) . from eq .
( [ e - specform ] ) it follows , that the signal due to scattered particles is directly proportional , on the one hand , to the e - e scattering probability , or the frequency of the collisions @xmath68 and , on the other hand , it is proportional to the probability of an electron to reach the scattering point @xmath145 ballistically .
as has been shown earlier in @xcite , the pre - exponential factor increases with increasing @xmath101 at low injection energies ( mainly due to the increase of the collision frequency @xmath146 ) . the existence of a maximum in fig .
[ g - vsmx ] can be attributed to a competition between an increase of the scattering probability ( the pre - exponential factor ) and a decrease of ballistic transport into the neighborhood of the scattering point @xmath145 ( the exponent ) with @xmath101 .
this explanation is supported by the excellent quantitative agreement between the experimental data and a theoretical fit based on eq .
( [ e - specform ] ) , where the angle @xmath134 is replaced by the experimental value @xmath147 , see fig .
[ g - vsmx ] .
the good fit also justifies the use of the one - collision moc approximations indicating that higher orders collisions can be neglected . for small energies ,
the pre - exponential factor determines the behaviour of the signal due to scattered particles , which also can be described by standard perturbation theory applied to the collisions integral . around the maximum
the probability for secondary collisions is still small and the signal is still defined by the scattering that takes place in the neighborhood of the point @xmath145 - however , the probability for a first collision on the way between the injector and the detector becomes larger .
this underlines the decrease of the signal with increasing @xmath101 . as mentioned above this regime
can not be described by ordinary perturbation theory , requing the moc approximation . at this point
we would like to note that the non - monotonic behavior of the ac - current component which has been discussed above can not be explained by nonlinearities due to heating effects which are expected to appear at approximately the same values of @xmath101 in dc - measurements @xcite .
it can be shown , that such an effect gives only a small correction of the order of @xmath148 to the ac - component of the signal , where @xmath149 is the temperature due to the heating effect . finally , we may only speculate about the origin of the weak increase of the detector signal which is observed for the highest injection energies @xmath101 . as the injection energy approaches @xmath150 , the potential drop across the injector qpc becomes so large that the second subband of the qpc may become populated .
this enhances the injector conductance , and , consequently , the detected signal . for these energies
( [ e - specform ] ) is no longer valid ( see ref .
@xcite ) . a second reason could be transition to the two - collision ( or multi - collision ) regime as the mean - free path is less then @xmath151 at these energies .
in summary , we have shown in this paper that it is possible to successfully investigate the evolution of a high - energy non - equilibrium quasi - particles distribution with the help of the `` modified one - collision approximation '' .
this method was used to explain novel measurements of the non - monotonic energy dependence of the signal of scattered electrons in a 2d system .
we have obtained an excellent quantitative agreement between the experimental data and a theoretical calculation . in conclusion
we would like to emphasize that the moc approximation may be of interest for a wide range of problems independent of the nature of the scattering particles ; the only restriction is that the scattering probability should strongly decrease after one or several collisions .
along with the case of electron - electron scattering , this is a typical situation for the majority of inelastic scattering processes , such as electron - phonon , phonon - phonon etc .
let us discuss as an example , electron - phonon scattering in the case when the electron energy is high enough @xmath152 , but is low in comparison with the debye temperature ( the electron energy is measured from the fermi level ) . in this case
a phonon with an energy of the order of @xmath23 will be emitted as a result of the electron - phonon scattering event and the corresponding mean - free path is @xmath153 .
in contrast to electron - electron scattering the energy is redistributed here between two states according to the momentum and energy conservation laws , @xmath154 and @xmath155 . here
@xmath156 and @xmath157 are the quasimomentum and energy before the collision , and after the collision they are @xmath158 , and @xmath157 , respectively ; @xmath159 is quasimomentum of the emitted phonon and @xmath160 its energy .
since , on average , the energy is redistributed equally between the scattering product @xmath161 , we also have here the small parameter @xmath162 .
the case of three - particles high - energy phonon - phonon scattering processes differs only by the energy dependence of the mean - free path .
furthermore , the inelasticity , on which the approach developed in this paper is based , is likely to be relevant not only to the energy redistribution but also to some other quantities ; this includes , for example , the spin in cases where the probability of scattering decreases significantly after a spin - flip scattering event .
this case may be realized for electron scattering from spin - polarized impurities ( there are no spin subbands in this case ) and if the exchange interaction depends significantly on the mutual orientation of the electron spin and the impurity spin .
let us neglect the interaction between electrons of one of the spin directions ( spin - up , for example ) and the impurities .
then the spin relaxation process is completed just after the first spin - flip scattering of the spin - down electrons .
thus , it is straightforward to describe this process in the framework of the modified one - collision approximation .
the method we proposed in this work can be very effective also for point - contact spectroscopy @xcite .
the reason is the following .
the current - voltage characteristic of the electric circuit containing a microconstriction is determined by the back - scattering due to inelastic electron scattering processes just in the region of the microconstriction .
if @xmath163 ( l is the length of the constriction ) the current - voltage characteristic is determined only by one , or few , first collisions , the probability of which is not small and , thus , it can not be taken into account using a perturbation theory ( as it was done in the study described in ref.@xcite for the case when @xmath164 ) .
the same also holds for the nonlinear thermoelectric phenomenon in point - contacts @xcite .
the influence of the first collisions on the point - contact spectrum may be calculated exactly using equations of the general form of eq.(6 ) because in point - contact spectroscopy the excess energy of the electron is high , @xmath152 .
the research described in this publication was made possible in part by award no.up2-2430-kh-02 of the u.s .
civilian research & development foundation for the independent states of the former soviet union ( crdf)(r.n.g , a.n.k , a.i.k .
, e.n.b . , and u.l . ) , by the us department of energy grant fg05 - 86er-45234 ( e.n.b . and u.l . ) , and by the deutsche forschungsgemeinschaft sfb 410 ( h.b . and l.m .
molenkamp , m.j .
brugmans , h. van houten and c.t .
foxon , semicond .
. technol .
* 7 * , b228 , ( 1992 ) .
th.schpers , m.krger , j.appenzeller , a.frster , b.lengeler , and h.lth , appl .
66 * , 3603 ( 1995 ) .
cumming and j.h .
davis , appl .
. lett . * 69 * , 3363 ( 1996 ) .
gurzhi , a.n .
kalinenko , a.i .
kopeliovich , phys .
lett . , * 74 * , 3872 , ( 1995 ) ; phys .
b , * 52 * , 4744 ( 1995 ) .
h. predel , h. buhmann , r.n .
gurzhi , a.n .
kalinenko , a.i .
kopeliovich , l.w .
molenkamp , and a.v .
yanovsky , physical review * b * , * 62 * , 2057 , ( 2000 ) .
h. buhmann , r.n .
gurzhi , a.n .
kalinenko , a.i .
kopeliovich , l.w .
molenkamp , h. predel , a.v .
yanovsky , physica e , * 6 * , 310 , ( 2000 ) . i.k .
yanson , zh.eksp.theor.fiz . * 66 * , 1035 ( 1974)[sov.phys.jetp * 39 * , 506 ( 1074 ) ] .
kulik , a.n .
omelyanchuk , and r.i.shekhter , fiz .
nizk . temp .
, * 3 * , 1543 , ( 1977 ) [ sov . j. low temp . phys .
* 3 * 789 , ( 1977 ) ] .
bogachek , i.o .
kulik , and a.g.shkorbatov , fiz .
nizk . temp .
, * 11 * , 1189 , ( 1985 ) [ sov .
j. low temp .
phys . * 11 * 656 , ( 1985 ) ] .
|
a theory is developed for the evolution of the non - equilibrium distribution of quasiparticles when the scattering rate decreases due to particle collisions .
we propose a `` modified one - collision approximation '' which is most effective for high - energy quasiparticle distributions .
this method is used to explain novel measurements of the non - monotonic energy dependence of the signal of scattered electrons in a 2d system .
the observed effect is related to a crossover from the ballistic to the hydrodynamic regime of electron flow .
|
the process of galaxy formation involves the interplay of many non - linear phenomena that span a wide range of length and time - scales . a galaxy like our milky way , for example , forms from a region that initially extends to roughly one comoving mpc , yet its angular momentum is determined by the mass distribution within tens of comoving mpc . star formation ( sf ) , on the other hand , takes place in the densest cores of giant molecular clouds ( gmcs ) , on scales of the order of 0.1 pc .
the challenge in simulations of galaxy formation is to capture this vast dynamic range , while simultaneously accounting for the different physical processes that intervene on relevant scales .
this is usually achieved with ad hoc sub - grid models that attempt to emulate the most important small - scale phenomena . in particular ,
one of the biggest uncertainties in simulations of galaxy formation is the means by which gas is converted into stars ( see * ? ? ? * for a recent review ) .
the standard approach to this problem , motivated by observations , is to adopt a schmidt - like law @xcite , often coupled to conditions on the local gas properties .
however , there are several issues with this method .
first , its parameters are poorly constrained and are usually fine - tuned to match the observed kennicutt - schmidt ( ks ) relation @xcite .
secondly , there is a growing body of evidence that the local star formation rate ( sfr ) correlates more tightly with the observed density of molecular hydrogen than with that of the total gas density ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , though there is yet no consensus as to whether this reflects a causal relation .
in particular , numerical simulations of isolated molecular clouds suggest that the presence of molecules does not boost the ability of the gas to cool and form stars @xcite .
the tight spatial correlation between h@xmath1 and young stars may then be due to the fact that they are both formed in high density regions where gas is effectively shielded from the interstellar radiation field . despite the ongoing debate , there are strong motivations for including a treatment of molecular hydrogen in cosmological simulations of galaxy formation .
observations of h@xmath1 proxies ( such as co luminosity ) , for example , have progressed tremendously over the past decade ( see * ? ? ? * for a recent review ) , underlining the need for robust theoretical templates to aid in the design of observational campaigns and the interpretation of their results .
furthermore , numerical simulations constitute a unique tool to test the impact of h@xmath1-regulated sf on the global structure of galaxies , provided their h@xmath1 content can be reliably determined .
tracking h@xmath1 , however , requires solving a challenging network of rate equations which are coupled to a radiative - transfer computation for h@xmath1-dissociating photons .
given that the spatial resolution of current simulations is comparable in size to gmcs , these calculations must be done at the sub - grid level and include a description of gas structure on the unresolved scales ( e.g. a clumping factor for the gas density ) .
recently , several authors have incorporated simple algorithms to track molecular hydrogen in hydrodynamical simulations of galaxy formation .
for instance , @xcite monitored the h@xmath1 distribution in dwarf - sized galaxies within a fixed dark matter ( dm ) potential and showed that the resulting molecular mass depends strongly on the metallicity of the interstellar medium ( ism ) .
similar conclusions were drawn by ( * ? ? ?
* see also feldmann et al .
( 2011 ) ) , who followed the evolution of the h@xmath1 content for 100 myr in a cosmologically simulated galaxy at @xmath2 .
these authors showed that it is only possible to form fully shielded molecular clouds when the gas metallicity is high ( i.e. @xmath3 z@xmath4 ) , and argued that h@xmath1-regulated sf can act as an effective feedback mechanism , delaying sf in the low - metallicity progenitors of a galaxy .
the implications of these results for galaxy formation in low - mass haloes were studied further by @xcite , who suggested the possible existence of a large population of low - mass , gas - rich galaxies that never reached the critical column density required for the h@xmath1/hi transition and are thus devoid of stars .
their work , however , was based on an analytic model for h@xmath1 formation that assumes chemical equilibrium between its formation and destruction rates @xcite .
none the less , @xcite showed that this model agrees well with a time - dependent solution to the chemical network provided the local metallicity of the gas is above @xmath5 z@xmath4 , lending support to these conclusions .
@xcite modelled the non - equilibrium abundance of h@xmath1 in a dwarf galaxy that was simulated down to redshift @xmath6 , connecting sf explicitly to the local h@xmath1 content of the gas .
these authors showed that , compared to simulations rooted on the schmidt law , molecule - based sf produces a galaxy which is more gas rich , has bluer stellar populations and a clumpier ism . on the other hand , strong stellar feedback , when included ,
tends to mitigate these differences by regulating the formation and destruction rates of gmcs @xcite . in this paper
, we introduce a new time - dependent sub - grid model for tracking the non - equilibrium abundance of h@xmath1 in cosmological simulations of galaxy assembly .
our approach builds upon the work of @xcite and @xcite by including additional information on the unresolved distribution of gas temperatures and densities .
in particular , our model : ( i ) explicitly accounts for the distribution of sub - grid densities , as determined by observations and numerical simulations of turbulent gmcs ; ( ii ) invokes a gas temperature - density relation that was determined from detailed numerical studies of the ism @xcite ; and ( iii ) consistently takes into account that denser , unresolved clumps have larger optical depths . as an application
, we employ the model in a high - resolution simulation that follows the formation of a milky way - sized galaxy down to @xmath0 . in order to explore the interplay between sf , h@xmath1 abundance and galactic structure ,
we re - simulate the same volume using different algorithms for computing the density of molecular hydrogen and the local sfr .
this paper is organized as follows . in section [ sec : mol_hydrogen_formation ] , we introduce our model for tracking the non - equilibrium h@xmath1 abundance and compare it with other commonly adopted prescriptions that have been discussed in the literature . in sections [ sec : code ] and [ sec : analysis ] , we describe our suite of simulations and present our main results .
finally , we summarize our main conclusions and then critically discuss some of our assumptions in section [ sec : discussion ] .
the abundance of molecular hydrogen in the metal - rich ism is mainly regulated by the competition between its formation due to the catalytic action of dust grains and its photo - dissociation via the two - step solomon process . in the ground state ,
h@xmath1 absorbs electromagnetic radiation in two densely packed series of lines ( the lyman band characterized by photon energies @xmath7 ev or wavelengths @xmath8 and the werner band
@xmath9 ev , @xmath10 ) .
radiative decay from the excited states leads to dissociation in approximately 15 per cent of the cases .
direct photo - dissociation would require photons with energy @xmath11 ev , but these are principally absorbed by hydrogen atoms as they lie above the hydrogen photo - ionization threshold ( 13.6 ev , 912 )
. as a result , only photons between @xmath12 can photo - dissociate molecular hydrogen .
lyman - werner ( lw ) photons are copiously emitted by ob stars , but intervening h@xmath1 and dust effectively shield the densest regions of the ism .
this results in an h@xmath1 abundance that increases rapidly towards regions in which the medium becomes optically thick to lw radiation .
an exact treatment of these effects is challenging : it requires three - dimensional radiative transfer calculations capable of resolving length and time - scales orders of magnitude shorter than those associated with galaxy evolution . nevertheless , it is possible to follow the formation of molecular complexes in a phenomenological way , using approximate sub - grid treatments of the most crucial physical processes involved . in this work ,
we use three different mathematical models that attempt to approximate these effects .
the simplest ( kmt - eq ) is fully analytical ; it returns the equilibrium h@xmath1 fraction in terms of quantities that can be determined locally in a simulation @xcite .
the model is based on a spherical molecular complex immersed in an isotropic bath of lw photons .
assuming that the ism is in a two - phase equilibrium between a cold and a warm neutral medium , it allows both the intensity of the radiation field and the resulting h@xmath1 fraction to be expressed in terms of the local gas column density and metallicity @xcite .
the second model ( kmt - uv ) employs the same analytical relation between the equilibrium h@xmath1 fraction and the radiation density in the lw band .
however , in this case , the latter is calculated numerically within the simulations themselves
. this is achieved by propagating photons from stellar particles assuming that the ism is optically thin within a characteristic length scale and optically thick beyond that ( see appendix [ sec : uv ] for further details ) .
note , however , that the formation of molecular hydrogen on dust grains is a very inefficient process . in the metal poor ism , for example , the h@xmath1 formation
time - scale approaches a hubble time .
this calls into question the appropriateness of the equilibrium - based models , which _ instantaneously _ populate each simulation volume element with a given fraction of h@xmath1 .
this has motivated several authors @xcite to develop more sophisticated algorithms capable of tracking the non - equilibrium h@xmath1 fraction in high - resolution simulations .
inspired by these efforts , we have developed a novel dynamical model ( dyn ) for calculating the abundance of molecular hydrogen .
this model is described in detail in the following subsection .
the evolution of the h@xmath1 number density is described by the following system of equations : @xmath13 - 2 \ ,
\dfrac{{\rm d}n_{\rm h_2}}{{\rm d}t } \nonumber\ , , \\ & \label{eq_closure_h } n_{\textsc{hi}\,}+n_{\textsc{hii}}+2n_{\rm h_2 } = n_{\rm h}\,\,\,.\end{aligned}\ ] ] here @xmath14 , @xmath15 and @xmath16 are , respectively , the number densities of neutral , ionized and molecular hydrogen ; @xmath17 is the electron number density ; @xmath18 is the @xmath19 collisional destruction rate due to interaction with species @xmath20 ; @xmath21 is the unshielded interstellar uv - radiation flux in habing units ( see appendix [ sec : uv ] ) ; @xmath22 is the h@xmath1 photo - dissociation rate for @xmath23 ; @xmath24 is the photo - ionization rate , the ionization cross - section @xmath25 and the corresponding optical depth @xmath26 , @xmath27 , where @xmath28 and @xmath29 denotes the planck constant . ] of @xmath30 ; @xmath31 and @xmath32 are the hydrogen recombination and collisional destruction rates ; @xmath33 is the optical depth of dust in the lw bands ( conventionally evaluated at 1000 ) , where @xmath34 is the total hydrogen column density .
this assumes that the dust abundance scales linearly with the gas metallicity and the dust - to - gas mass ratio is equal to the value measured in the milky way ( mw ) , i.e. 10@xmath35 @xmath36z@xmath4 .
the parametrization of the h@xmath1 self - shielding function , @xmath37 ( approximated for the plane parallel case ) and the h@xmath1 formation rate on dust grains , @xmath38 , are given in table [ table : pararameters ] , and adopt a dust temperature @xmath39 k. [ cols="<,^,^,^",options="header " , ] [ sec : stellar_disc ] the evolution of the h@xmath1 ( or total gas ) profile in our simulations determines the structure of the stellar disc . in order to better assess the differences between our runs , we study the radial distribution of stars at @xmath0 within the galaxy in bins of stellar age .
we choose the bins so that , in the std run , the stellar mass produced in each is equal to one fifth of the galaxy s total stellar mass at @xmath0 ( corresponding to age bins of [ 3.35,1.80 ] , [ 1.80,1.39 ] , [ 1.39,0.98 ] , [ 0.98,0.48 ] , [ 0.48,0 ] gyr ) .
the results for the std ( left - hand panels ) and dyn ( right - hand panels ) runs are plotted in the top row of figure [ fig : rad_distr_stars ] ( qualitatively similar results are obtained when using the dyn - sne model instead of dyn ) .
one can see that the dyn model shows significant differences in the age distribution of stars , especially at early formation epochs ( represented with darker shades of blue ) and in the innermost @xmath40 kpc .
these regions are populated by younger stars in dyn than in the std case .
for instance , the mean age of the stellar particles located at @xmath41 kpc is 1.1 gyr in the std simulation , but only 0.6 gyr in dyn .
the total stellar mass within 1 kpc from the galaxy centre at @xmath0 is also different .
in fact , we find values of @xmath42 m@xmath4 for the std run and @xmath43 m@xmath4 for the dyn one .
( for all the other models see table [ table : stars_1kpc ] . ) to explain the origin of this difference , we separate the stars that formed in the central galaxy of the most massive progenitor halo ( second row in figure [ fig : rad_distr_stars ] ) from those that formed in other haloes ( third row ) . in the std model ,
nearly 66 per cent of the stars in the first quintile of the age distribution ( @xmath44 ) does not from in the most massive progenitor . on the other hand , in the dyn simulation ,
only 8 per cent of the final stellar mass is created at such early times mostly in the central galaxy of the most massive progenitor ( 41 per cent ) and in other haloes ( 59 per cent ) .
cumulative distribution of the dm ( thin lines ) and stellar mass ( connected points ) of substructures within the virial radius of the most massive halo at @xmath0 for the std ( red ) , dyn ( black ) , dyn - sne ( magenta ) , kmt - uv ( blue ) and kmt - eq ( green ) runs .
the total number of substructures ( with @xmath45 and hosting at least 100 stellar particles ) is indicated in parenthesis . ]
a visual impression can be formed from figure [ fig : snapshot_z_5 ] where we compare the spatial distribution of dm and stars extracted from the two simulations in the region surrounding the most massive progenitor at @xmath46 . at this redshift ,
in the std run , 99 per cent of the haloes with @xmath47 m@xmath4 ( i.e. above the mass threshold where sf is suppressed by the cosmic reionization of hydrogen ) contain some stars while only 33 per cent do in the dyn simulation .
this striking difference is due to the fact that most low - mass haloes ( with @xmath48 m@xmath4 ) in the dyn model can not efficiently form molecular hydrogen at early times .
in fact , their dust content and central densities are too low to promote efficient h@xmath1 formation , resulting in much lower sfrs .
we find that their characteristic h@xmath1 formation time , @xmath49 , is several orders of magnitude larger than the age of the universe .
therefore , many low - mass haloes stay gas rich but dark ( similar conclusions have been reached using the kmt - eq model by * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
this phenomenon has strong implications for the assembly of the galaxy in our simulations . in the std run
, the main - progenitor galaxy is built up by a sequence of mergers between objects with extended stellar components .
on the other hand , in the dyn simulation , the same mergers provide gas - rich haloes with a few compact stellar distributions .
once a well defined central galaxy is in place , the merger history in the two models is nearly identical .
however , the different rules with which sf and feedback take place produce a clumpier stellar pattern in the dyn model while the std run more often develops grand design spiral arms in between major merging events . off - centre mergers between substructures
produce large clumps in both models ( see figure 8) . in particular , a massive ( @xmath50 m@xmath4 ) gas - rich condensation forms at @xmath51 is the dyn model and orbits within a few kpc from the galactic centre afterwards .
a first encounter between this object and the central stellar clump of the main galaxy takes place at @xmath52 .
this event heats the distribution of stellar orbits in the core of the dyn galaxy and scatters the old stars ( that were previously concentrated in the middle ) over a more extended area ( see the bottom right panel in figure [ fig : rad_distr_stars ] ) .
sf in the central regions of the galaxy is also boosted at this time ( see the right - hand panel in the second row of figure [ fig : rad_distr_stars ] ) . in both models ,
the late history of the galaxy is characterized by a close encounter with a satellite at @xmath53 which mildly distorts the gas and stellar distributions .
finally , we focus our attention on the distribution of satellite galaxies at @xmath0 .
figure [ fig : mass_function2 ] shows , for all models , the cumulative number of substructures that are found within the virial radius of the main halo as a function of their dm and stellar masses . while the number of dm clumps is virtually identical in all runs , the std model produces a larger number of bright satellites than any other run ( approximately 30 per cent more than in the dyn simulation and 10 to 30 per cent more in the other models ) .
this is a direct consequence of the suppression of sf in metal - poor , low - mass haloes that we have already discussed above .
we conclude that molecule - regulated sf can act as an effective feedback mechanism that may alleviate the satellite problem of the cdm scenario ( e.g. * ? ? ?
* ; * ? ? ?
further study will be necessary to determine whether this conclusion also applies to low redshift galaxies .
we have presented a sub - grid model for tracking the non - equilibrium abundance of molecular hydrogen in cosmological simulations of galaxy formation .
the novelty of the model is that it phenomenologically accounts for the distribution of unresolved sub - grid densities determined from observations and simulations of the turbulent ism . in this sense
, it improves upon previous time - dependent schemes ( e.g. * ? ? ?
* ; * ? ? ?
* ) in which the h@xmath1 formation rate on dust grains is amplified by a fixed clumping factor . we have implemented our model in the code in order to run simulations that track the evolution of the h@xmath1 content of a massive galaxy at @xmath0 , and to study the imprint of h@xmath1-regulated sf . the resulting h@xmath1 fractions in different environments spanning a vast range of gas metallicities and interstellar uv fields are consistent with observations of the milky way galaxy , the lmc and the smc . in order to better understand what determines the properties of a galaxy , we ran a suite of simulations of the same dm halo , each with a different prescription for computing the h@xmath1 distribution . in the runs where h@xmath1 is calculated explicitly ,
sf was regulated by the local h@xmath1 abundance , while , for another ( std ) , we adopted the traditional schmidt law based on the total gas density .
our main findings can be summarized as follows : 1 .
all simulations produce a galaxy which is in broad agreement with several high - redshift observational data sets . however , the different models for star and h@xmath1 formation result in important differences in the stellar and gas masses of the galaxies ( see table [ table : par_r200 ] ) .
in particular , if sf is molecule - regulated and the h@xmath1 abundances are computed with a detailed treatment of photo - dissociation including a simplified radiative - transfer scheme : 1 .
the galaxy produces less stars and is in better agreement with the observed stellar - to - halo mass relation with respect to the std model ( which , however , can not be ruled out as we did not study the effect of increasing the spatial resolution of the simulations ) .
2 . the galaxy harbours a larger gas reservoir and its gas fraction better matches observations of high redshift galaxies . 3 .
early sf is inhibited in metal - poor haloes with mass @xmath48 m@xmath4 in which gas and dust densities are too low to trigger efficient conversion of hi into h@xmath1 . as a result ,
the main galaxy assembles by accreting many gas - rich substructures and , consequently , hosts a younger stellar population and harbours a larger cold gas reservoir than in the std case .
also , the number of satellites of the main galaxy at @xmath0 is reduced by 30 per cent at all stellar masses compared with the std simulation .
regardless of the h@xmath1 model : 1 .
the main galaxy in our simulations ( with the exception of the kmt - eq model ) has similar spatial distribution and total mass of molecular hydrogen at @xmath0 .
this is mainly due to the fact that the average metallicity of the gas in its dominant progenitors is already @xmath54 z@xmath4 at @xmath55 . as a result ,
h@xmath1 formation is rapid , mitigating any subtle differences independent of whether stars form from atomic or molecular gas .
the molecular mass in a cell scales linearly with that of the metals ( above a model - dependent threshold density ) .
this is a consequence of assuming that dust traces the metals in the simulations .
if h@xmath1 destruction by sne is substantial , sf is suppressed by nearly 20 per cent at all times relative to an identical model in which this destruction channel is neglected .
contrary to the assumption that gas is fully molecular in high - redshift galaxies ( commonly used to interpret co observations , e.g. @xcite ) , the atomic gas fraction in our simulated galaxy is comparable to the molecular contribution , independent of the h@xmath1 formation model .
6 . using the std model to form stars , but
` painting on ' h@xmath1 in post - processing using the kmt - uv prescription , gives a reasonable estimate of the total h@xmath1 mass of the galaxy .
however , as already mentioned above , the resulting galaxy contains more stars ( in particular old stars ) than in all of the molecule - regulated schemes .
although our non - equilibrium h@xmath1 model represents a significant improvement over previous methods , many challenges remain .
for instance , we have assumed that the sub - grid density pdf of gmcs can be accurately described by a lognormal distribution .
this is based on several observations of molecular clouds which , in some cases , show high - density tails in star forming regions ( e.g. * ? ? ?
* ; * ? ? ?
several complex physical phenomena like energy injection , turbulence , gravity and external compression influence the density structure of molecular clouds . yet
, numerical studies of the ism have shown that the lognormal model is a good approximation when an isothermal gas flow is supersonically turbulent ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
power - law tails in the high - density regime form under the presence of self - gravity ( which generates dense cores and super - critical filaments ) .
non - isothermal turbulence can also increase the occurrence of dense clumps . none the less ,
these uncertainties are likely sub - dominant to those associated with modelling the effects of sna feedback on gmcs , which must be tackled with small - scale simulations of the ism . in addition
, we have set the dispersion of the lognormal density distribution to be @xmath56 , consistent with a constant clumping factor @xmath57 .
this choice was motivated by theoretical work that relates local density enhancements to the three - dimensional rms mach number , @xmath58 , with values of @xmath59 ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
moreover , the same value for the clumping factor has been adopted in the literature to best match the h@xmath1 content in observations and simulations ( e.g. * ? ? ?
* ; * ? ? ?
however , observations of gmcs have revealed substantial variations in the mach number @xcite . in future implementations
, the realism of our model can be improved by adjusting the clumping factor ( as well as the pdf of the sub - grid density ) in cells with different mean densities and temperatures . from the technical point of view , this is straightforward to do : the difficulty lies in linking the mean properties of a cell to the sub - grid parameters that regulate the density pdf .
one intriguing possibility could be to implement a simplified description of supersonic turbulence along the lines of that proposed by @xcite .
we plan to return to these issues in future work .
we wish to thank our referee for a constructive report that has improved this paper .
we also thank romain teyssier and tom abel for helpful discussions and acknowledge support from mark labadens for the software pymses and from oliver hahn for the music code .
we are also grateful to andreas schruba for kindly providing us with data from the heracles survey .
this work was supported by the deutsche forschungsgemeinschaft ( dfg ) through the project sfb 956 _ conditions and impact of star formation _ , sub - project c4 .
mt was supported through a stipend from the international max planck research school ( imprs ) for astronomy and astrophysics in bonn .
we acknowledge that the results of this research have been achieved using the prace-2ip project ( fp7 ri-283493 ) resources hector based in the uk at the uk national supercomputing service and the abel computing cluster based in norway at the university of oslo .
t. , anninos p. , norman m. l. , zhang y. , 1998 , apj , 508 , 518 t. , anninos p. , zhang y. , norman m. l. , 1997 , new astron .
, 2 , 181 o. , kravtsov a. v. , leitner s. n. , gnedin n. y. , 2013 , apj , 770 , 25 behroozi , p. s. , conroy , c. , & wechsler , r. h. 2010 , apj , 717 , 379 f. , leroy a. , walter f. , brinks e. , de blok w. j. g. , madore b. , thornley m. d. , 2008 , aj , 136 , 2846 n. , et al .
2010 , apj , 718 , 1001 brook , c. b. et al . 2011 , mnras , 415 , 1051 brook , c. b. , stinson , g. , gibson , b. k. , wadsley , j. , & quinn , t. 2012 , mnras , 424 , 1275 j. r. , hollenbach d. j. , 1983 , apj , 265 , 223 c. l. , walter f. , 2013 , ara&a , 51 , 105 s. , spaans m. , 2004 , apj , 611 , 40 s. , spaans m. , 2009 , a&a , 496 , 365 ceverino , d. , & klypin , a. 2009 , apj , 695 , 292 c. , quinn t. , governato f. , stilp a. , shen s. , wadsley j. , 2012 , mnras , 425 , 3058 e. , et al . , 2010 , apj , 714 , l118 c. l. , et al . , 2013 , preprint ( arxiv:1312.3223 ) b. t. , bertoldi f. , 1996 , apj , 468 , 269 c. , klessen r. s. , 2013 , apj , 763 , 51 c. , klessen r. s. , schmidt w. , 2009 , apj , 692 , 364 r. , gnedin n. y. , kravtsov a. v. , 2011 , apj , 732 , 115 feldmann , r. , gnedin , n. y. , & kravtsov , a. v. 2012 , apj , 758 , 127 r. , et al . , 2010 , mnras , 407 , 2091 s. p.
d. , knebe a. , gibson b. k. , 2004 , mnras , 351 , 399 k. , shull j. m. , tumlinson j. , danforth c. , 2006 , apj , 636 , 891 s. c. o. , clark p. c. , 2012 , mnras , 426 , 377 s. c. o. , mac low m .-
, 2007a , apjs , 169 , 239 s. c. o. , mac low m .-
, 2007b , apj , 659 , 1317 n. y. , kravtsov a. v. , 2011 , apj , 728 , 88 n. y. , tassis k. , kravtsov a. v. , 2009 , apj , 697 , 55 gnedin , n. y. , & kravtsov , a. v. 2010 , apj , 714 , 287 guedes , j. , callegari , s. , madau , p. , & mayer , l. 2011 , apj , 742 , 76 f. , madau p. , 2012 ,
apj , 746 , 125 o. , abel t. , 2011 , mnras , 415 , 2101 p. f. , kere d. , murray n. , quataert e. , hernquist l. , 2012 , mnras , 427 , 968 jaacks , j. , thompson , r. , & nagamine , k. 2013 , apj , 766 , 94 j. , beuther h. , henning t. , plume r. , 2009 , a&a , 508 , l35 jr .
r. c. , 1989 , apj , 344 , 685 jr .
r. c. , 1998 , apj , 498 , 541 jr .
r. c. , calzetti d. , walter f. , helou g. , hollenbach d. j. , armus l. , bendo g. , 2007 , apj , 671 , 333 s. r. , knebe a. , 2009 , apjs , 182 , 608 e. , et al .
, 2011 , apjs , 192 , 18 p. , 2001 ,
mnras , 322 , 231 krumholz , m. r. 2014 , preprint ( arxiv:1402.0867 ) m. r. , gnedin n. y. , 2011 , apj , 729 , 36 m. r. , mckee c. f. , tumlinson j. , 2008 , apj , 689 , 865 m. r. , mckee c. f. , tumlinson j. , 2009 , apj , 693 , 216 m. , krumholz m. r. , madau p. , smith b. d. , wise j. , 2012 , apj , 749 , 36 kuhlen , m. , madau , p. , & krumholz , m. r. 2013 , apj , 776 , 34 klypin , a. , kravtsov , a. v. , valenzuela , o. , & prada , f. 1999 , apj , 522 , 82 c. , schaerer d. , goldader j. d. , gonzlez delgado r. m. , robert c. , kune d. f. , de mello d. f. , devost d. , heckman t. m. , 1999 , apjs , 123 , 3 a. , bolatto a. , stanimirovic s. , mizuno n. , israel f. , bot c. , 2007 , apj , 658 , 1027 a. k. , walter f. , brinks e. , bigiel f. , de blok w. j. g. , madore b. , thornley m. d. , 2008 , aj , 136 , 2782 c. f. , krumholz m. r. , 2010 , apj , 709 , 308 magnelli , b. , et al . 2013 , a&a , 553 , a132 m. , glover s. c. o. , federrath c. , klessen r. s. , 2012 , mnras , 421 , 2531 moore , b. , ghigna , s. , governato , f. , lake , g. , quinn , t. , stadel , j. , tozzi , p. , 1999 ,
apjl , 524 , l19 moster , b. p. , naab , t. , & white , s. d. m. 2013 , mnras , 428 , 3121 munshi , f. , et al .
2013 , apj , 766 , 56 munshi , f. , christensen , c. , quinn , t. r. , goveranto , f. , wadsley , j. , loebman , s. , shen , s. , 2014 , apj , 781 , l14
e. c. , stone j. m. , gammie c. f. , 2001 , apj , 546 , 980 p. , nordlund a. , jones b. j. t. , 1997 , mnras , 288 , 145 f. i. , papadopoulos p. p. , van der werf p. , 2006 ,
apj , 645 , 1024 d. j. , federrath c. , brunt c. m. , 2011 , apjl , 727 , l21 robertson , b. e. , & kravtsov , a. v. 2008 , apj , 680 , 1083 h. , pittard j. m. , 2013 , mnras , 431 , 1337 b. d. , bohlin r. c. , drake j. f. , budich w. , 1977 , apj , 216 , 291 c. , et al . , 2012 , mnras , 423 , 1726 m. , 1959 , apj , 129 , 243 n. , andr p. , knyves v. , bontemps s. , motte f. , federrath c. , ward - thompson d. , arzoumanian d. , 2013 , apjl , 766 , l17 schruba , a. et al .
2011 , aj , 142 , 37 shull , j. m. , anderson , k. , tumlinson , j. , & fuse science team 2004 , baas , 36 , 772 g. , seth a. , katz n. , wadsley j. , governato f. , quinn t. , 2006 , mnras , 373 , 1074 sternberg , a. , le petit , f. , roueff , e. , & le bourlot , j. 2014 , apj , 790 , 10 l. j. , et al . , 2010 , nature , 463 , 781 y. , umemura m. , 1998 , apj , 502 , 59 r. , 2002 , a&a , 385 , 337 r. , chapon d. , bournaud f. , 2010 , apj , 720 , l149 r. , pontzen a. , dubois y. , read j. i. , 2013 , mnras , 429 , 3068 thompson , r. , nagamine , k. , jaacks , j. , & choi , j .- h . 2014 , apj , 780 , 145 a. g. g. m. , hollenbach d. , 1985 , apj , 291 , 722 j. k. , klein r. i. , mckee c. f. , holliman ii j. h. , howell l. h. , greenough j. a. , 1997 , apjl , 489 , l179 j. , et al . , 2002 , apj , 566 , 857 e. , 1994 , apj , 423 , 681 j. h. , turk m. j. , norman m. l. , abel t. , 2012 , apj , 745 , 50 a. m. , gawiser e. , prochaska j. x. , 2005 , ara&a , 43 , 861 m. g. , tielens a. g. g. m. , hollenbach d. , kaufman m. j. , 2008 , apj , 680 , 384 zemp , m. , gnedin , o. y. , gnedin , n. y. , & kravtsov , a. v. 2012 , apj , 748 , 54
each star particle in our simulations represents a simple stellar population .
we use the starburst99 templates @xcite with a @xcite initial mass function ( imf ) to compute the luminosity in the lw band , @xmath60 , as a function of stellar age , @xmath61 .
we then propagate this radiation up to a maximum distance assuming that the gas is optically thin . for each spatial location
, we only consider sources of radiation that lie within an oct , i.e. a collection of 8 cells at the highest level of refinement .
therefore the corresponding uv flux is : where @xmath63 is the time elapsed in the simulation , @xmath64 is the size of a resolution element , @xmath65 is the speed of light , @xmath66 and @xmath67 erg @xmath68 @xmath69 .
the factor @xmath70 gives the correct average flux at the centre of a cell when stars are randomly distributed ( with uniform density ) within an oct .
this coefficient has been measured with a monte carlo method . in our simulations , @xmath71 pc and is comparable with the typical size of gmcs .
|
we introduce a sub - grid model for the non - equilibrium abundance of molecular hydrogen in cosmological simulations of galaxy formation .
we improve upon previous work by accounting for the unresolved structure of molecular clouds in a phenomenological way which combines both observational and numerical results on the properties of the turbulent interstellar medium .
we apply the model to a cosmological simulation of the formation of a milky way - sized galaxy at @xmath0 , and compare the results to those obtained using other popular prescriptions that compute the equilibrium abundance of h@xmath1 . in these runs
we introduce an explicit link between star formation and the local h@xmath1 abundance , and perform an additional simulation in which star formation is linked directly to the density of cold gas . in better agreement with observations ,
we find that the simulated galaxy produces less stars and harbours a larger gas reservoir when star formation is regulated by molecular hydrogen . in this case
, the galaxy is composed of a younger stellar population as early star formation is inhibited in small , metal poor dark - matter haloes which can not efficiently produce h@xmath1 .
the number of luminous satellites orbiting within the virial radius of the galaxy at @xmath0 is reduced by 10 - 30 per cent in models with h@xmath1-regulated star formation .
[ firstpage ] methods : numerical - ism : molecules - galaxies : evolution - galaxies : formation
|
low - dimensional spin - gapped quantum systems are of current interest since they show interesting ground states and a variety of unconventional low - lying excitations . examples of such systems include spin-1 haldane chains @xcite , spin-@xmath2 even - leg ladders@xcite and spin-@xmath2 alternating chains @xcite .
the discovery of appropriate compounds like y@xmath0banio@xmath1@xcite ( haldane chain ) , sr@xmath7cu@xmath8o@xmath9 @xcite or srcu@xmath0o@xmath10 @xcite ( ladder systems ) has brought new insight into the study of these systems .
there is a long list of various spin-@xmath2 alternating chain systems which are being intensively studied in connection to their magnetic excitations , to mention a few of them kcucl@xmath10 @xcite , tlcucl@xmath10 @xcite , cu(no@xmath10)@xmath0.@xmath11d@xmath0o @xcite .
if frustration is also present new features appear in the magnetization spectrum of the low - dimensional quantum systems as in cugeo@xmath10 @xcite , ( vo)@xmath0p@xmath0o@xmath12 ( vopo ) @xcite , srcu@xmath0(bo@xmath10)@xmath0 @xcite or in the recently synthesized cu@xmath0te@xmath0o@xmath1br@xmath0 @xcite .
an important family of low - dimensional compounds are the layered vanadates av@xmath0o@xmath1 ( a = ca , mg , na , li , cs)@xcite . while cav@xmath0o@xmath1 and mgv@xmath0o@xmath1 contain only magnetic v@xmath13 ions and behave like spin-@xmath2 ladders with spin - gaps of the order of 600k @xcite and 20k @xcite respectively , @xmath3-nav@xmath0o@xmath1 , @xmath4-liv@xmath0o@xmath1 and csv@xmath0o@xmath1 are mixed - valence systems ( v@xmath14 on average ) with important charge and spin fluctuations .
this last property makes this family specially interesting since it allows to study the influence of charge - ordering and the corresponding crystallographic distortions on the magnetic interactions .
the quarter - filled ladder compound @xmath3-nav@xmath0o@xmath1 @xcite is a highly discussed material and has become a model substance for the study of spin , charge and orbital coupling .
it behaves like a spin-@xmath2 heisenberg system at high temperatures and at 34k undergoes a charge ordering ( 2v@xmath14 @xmath15 v@xmath13 + v@xmath16 ) transition@xcite accompanied with a lattice distortion @xcite and the opening of a spin gap@xcite .
@xmath4-liv@xmath0o@xmath1 is a charge ordered system with no spin gap opening at low temperatures @xcite .
recent investigations based on an _ ab initio _ analysis of the electronic structure@xcite show that charge fluctuations in this system are still significant and the experimental observations @xcite can be well described by a spin-@xmath2 asymmetric quarter - filled ladder model . such proposal has also found support by a numerical analysis of optical conductivity experiments@xcite .
the third member of this vanadate family , csv@xmath0o@xmath1 , has been given much less attention .
existing susceptibility measurements of isobe _
et al._@xcite could be rather well explained by considering this material as a system of isolated dimers with a spin - gap of @xmath17 .
measurements other than that are unknown to us except for older susceptibility measurements @xcite which somewhat disagree with isobe _
s data and different values of the exchange integral @xmath18 and the gyromagnetic ratio @xmath19 are predicted .
there are various reasons why we believe that this system deserves attention : ( i ) the question of how the electronic and magnetic properties of this material compare with the properties of the other two members of the alkali - vanadate family , _
i.e. _ @xmath3-nav@xmath5o@xmath20 and @xmath4-liv@xmath5o@xmath20 ( ii ) this system crystallizes in a similar monoclinic structure as the most discussed vopo and many other materials _ e.g. _
kcucl@xmath10 , tlcucl@xmath10 showing an alternating chain behavior of the magnetic interactions .
it is then worthwhile to find out whether csv@xmath0o@xmath1 has also a similar behavior of the magnetic interactions as its isostructural members .
( iii ) as will be shown below , a re - examination of the susceptibility measurements reveals that a dimer model is not the only model that can provide reasonable fitting of the susceptibility data . in view of the above
, we have performed a microscopic study of this system by considering density - functional - theory - based ( dft ) calculations .
the microscopic study of the electronic behavior of this compound reveals that the inter - dimer coupling in this material in addition to the intra - dimer interaction is also significant and should nt be neglected .
further , an analysis of the susceptibility data shows that getting estimates of the coupling constants and the gyromagnetic factor based merely on a fitting procedure is very subtle and with such an analysis one can only test the consistency with the assumed model , but can not prove the uniqueness of that model .
results on that will be presented in the next sections .
we shall also discuss the similarities and differences to the other members of this vanadate family , @xmath3-nav@xmath5o@xmath20 and @xmath4-liv@xmath5o@xmath20 as well as to isostructural compounds like vopo .
the crystal structure of csv@xmath0o@xmath1 is somewhat different from that of @xmath3-nav@xmath0o@xmath1 and @xmath4-liv@xmath0o@xmath1 . @xmath3-nav@xmath0o@xmath1 and @xmath4-liv@xmath0o@xmath1 crystallize in the orthorhombic space group @xmath21 ( @xmath22 ) and @xmath23 ( @xmath24 ) respectively .
they consist of a double chain - structure of edge - sharing distorted square vo@xmath1 pyramids running along the orthorhombic @xmath25 axis , which are linked together via common corners to form layers .
these layers are stacked upon each other along @xmath26 .
na / li atoms are located between these layers .
crystallizes@xcite , on the other hand , in the monoclinic space group @xmath27 ( @xmath28 ) .
csv@xmath5o@xmath20 has also a layered structure with cs ions in between the layers .
these layers , which are somewhat tilted ( _ i.e. _ not strictly in _ yz _ plane ) , and are stacked upon each other along @xmath29 , are made of two types of vanadium atoms v1 and v2 .
the v1 atoms are in a distorted square pyramidal environment of oxygens and form pairs by sharing the edge of the pyramids .
these pairs are separated from one another by the v2 vanadiums in a tetrahedral co - ordination of oxygens as shown in fig .
[ struct_yz ] . within a pair of square pyramids ,
the two apical o atoms are pointing in opposite directions and the remaining o atoms of the pyramids are shared with tetrahedra .
the v1-v1 distance across the shared edge of the square - pyramids - pair is 3.073 @xmath30 and the v1-v2 distance across the shared corner of a pyramid and a tetrahedron is 3.352 @xmath30 .
since the average v2-o distance within the tetrahedron d=1.718 @xmath30 is shorter than the average v1-o distance within the square pyramid d=1.882 @xmath30 , one can conclude from a bond - valence analysis @xcite that the v1 sites are in the oxidation v@xmath13 and the v2 sites are in the oxidation v@xmath16 .
as the intra - pair v@xmath13-v@xmath13 distance is 3.073 @xmath30 while the inter - pair distance along the @xmath26-axis is 5.501 @xmath30 , this compound has been assumed to be a system of isolated dimers .
in order to analyze the electronic behavior of csv@xmath0o@xmath1 , we carried out dft calculations within the local density approximation ( lda ) by employing the full - potential linearized augmented plane wave code wien97 @xcite ( lapw ) and by linear - muffin - tin orbital ( lmto)@xcite based on the stuttgart tblmto-47 code .
both calculations agree well within the allowed error - bars of the various different approximations involved in two different methods . in both approaches we treated the exchange - correlation part by using the generalized gradient approximation @xcite . in fig .
[ cs_spag ] the energy bands for csv@xmath0o@xmath1 along the path @xmath31 are shown .
since the unit cell of csv@xmath0o@xmath1 contains four formula units , i.e. four v@xmath13 and four v@xmath16 , there are four bands around the fermi level which correspond to v@xmath13 - 3d orbitals .
these bands are separated by an energy gap of about 2 ev from the lower o - p valence bands and an energy gap of about 0.5 ev from the upper conduction bands .
the conduction bands up to 4ev are of v-3d nature .
there are a few points to be noted here : ( i ) the energy range on the csv@xmath0o@xmath1 bands is of the same order of magnitude as it is in @xmath3-nav@xmath0o@xmath1@xcite and @xmath4-liv@xmath0o@xmath1@xcite .
( ii ) the contribution of v@xmath16 - 3d is mostly in the upper conduction bands , starting from about 0.75 ev above the fermi level and have a very small contribution at the fermi level indicating that , contrary to @xmath3-nav@xmath0o@xmath1 and @xmath4-liv@xmath0o@xmath1 , here the system is close to complete charge ordering .
( iii ) there is no dispersion along the @xmath29 axis while along the @xmath26 axis ( chain direction ) the dispersion is non - negligible and of about 300mev . along the @xmath25 axis
the lda - bands show a small dispersion which corresponds to inter - chain interactions . in order to investigate the dimension - dependent behavior of csv@xmath0o@xmath1
, we show in fig .
[ partial_dos](a ) the partial density of states ( dos ) of the v@xmath13 3d - orbitals for csv@xmath0o@xmath1 . for comparison
we have also plotted in fig .
[ partial_dos](b ) the partial density of states of v1 - 3d orbitals for @xmath4-liv@xmath0o@xmath1@xcite and in fig .
[ partial_dos](c ) that of v-3d orbitals for @xmath3-nav@xmath0o@xmath1 .
note that , while the @xmath4-liv@xmath0o@xmath1 and @xmath3-nav@xmath0o@xmath1 dos show the characteristic quasi-1d van hove singularities near the band edges around the fermi level , this is not the case for the cs - compound .
the dos for csv@xmath0o@xmath1 shows more a 2d - behavior than 1d .
( iv ) in the global frame of reference , the character of the bands around the fermi level turn out to be that of d@xmath32 with small admixture from d@xmath33 character . rotation to local co - ordinate system with the @xmath26-axis aligned along the _
( v - apical o ) _
direction for individual pyramids transform this to d@xmath34 character , in agreement with the crystal field analysis of a v@xmath13 ion in the pyramidal environment@xcite .
the band - structure analysis leads us to the conclusion that inter - dimer coupling is non - negligible .
if the system would have been a pure dimer system we would have expected no dispersion in any direction .
the existence of a non - negligible dispersion tells us that a model of unlinked dimers is too simple and there must be some inter - dimer coupling to explain the band picture . from the dos analysis
we learn that inter - dimer interactions in both @xmath26 and @xmath25 direction are important for the electronic properties .
the possible role of the v@xmath16o@xmath35 tetrahedra in this context will be investigated in the next section .
in order to understand the nature of the ground state in csv@xmath0o@xmath1 we need to determine the appropriate microscopic model which explains the low - energy physics of this compound . we have therefore applied the so - called _ downfolding method _ offered by the new and generalized version of the lmto @xcite method , together with a tight - binding ( tb ) analysis on the band - structure results .
the downfolding technique consists of integrating out the high energy degrees of freedom so as to describe the details of lda energy bands close to the fermi energy in terms of few - orbital effective hamiltonians . from this analysis
, we can extract the effective hopping matrix elements @xmath36 between vanadium ions in csv@xmath0o@xmath1 ( by fourier transform of the downfolded hamiltonian @xmath37 ) which reproduce the behavior of the lda bands close to the fermi level .
since , as seen in the previous section , the contribution of the v@xmath16 - 3d orbitals around the fermi level is very small , and the biggest contribution comes from the v@xmath13 - 3@xmath38 orbital , a one - orbital effective tight - binding hamiltonian should give a good description of the low energy physics .
the tight - binding hamiltonian therefore , can be written as : @xmath39 where @xmath40 and @xmath41 denote a pair of v@xmath13 ions labeled each one by @xmath42 where @xmath43 denotes the unit cell and @xmath44 v@xmath13 ions in the unit cell . in fig .
[ hoppings ] all the considered hoppings @xmath45 , @xmath46 , @xmath47 and t@xmath1 have been drawn .
the subscript @xmath48 , @xmath49 , @xmath50 and @xmath51 indicate respectively that the hopping integral is between first nearest neighbors ( n.n . ) , second n.n .
, third n.n and fifth n.n .
the hopping term @xmath52 , which connects two v@xmath13 ions belonging to two different layers along @xmath29 , has been neglected due to its vanishingly small value .
note here that no hopping terms have been considered along the @xmath29 direction , since as we have learnt from the band picture , there is no dispersion along this direction and that means that the interlayer coupling along @xmath29 must be negligible . in fig .
[ tb_fit ] we show a comparison of the _ downfolding_-tb bands to the lda bands .
we observe that not only the hopping corresponding to the intra - dimer interaction @xmath45 is important but also the inter - dimer matrix element @xmath47 along @xmath26 is considerable . a model with only @xmath45 and @xmath47 , i.e. alternating chain , shows a dispersion relation of the form : @xmath53 which is symmetric with respect to e=0 ( i.e. to @xmath54 ) and does nt account for the band - splittings in the paths @xmath55 and @xmath56 .
these features are described by the inter - chain hopping parameters along the @xmath25 direction , @xmath46 and @xmath57 , though they are much smaller than @xmath45 and @xmath47 .
note that the tb - results show an artificial band - crossing along the path @xmath58 and @xmath59 .
we have analyzed this result and seen that the inclusion of longer - ranged interactions lifts this artificial crossing .
concerning the nature of the hopping integrals , we observe that the values of the dominant hopping parameter @xmath60 ( which agrees with the overall band - width of the v@xmath13-d@xmath34 derived low - energy bands in csv@xmath0o@xmath1 ) is smaller than the dominant hopping parameter for @xmath3-nav@xmath0o@xmath1@xcite @xmath61 and @xmath4-liv@xmath0o@xmath1@xcite @xmath62 .
this difference can be explained by the nature of the path .
while @xmath45 in csv@xmath0o@xmath1 corresponds to a @xmath63 path in between two _ edge_-sharing square pyramids , @xmath64 in @xmath3-nav@xmath0o@xmath1 and @xmath4-liv@xmath0o@xmath1 corresponds to a @xmath63 path between two _ corner_-sharing square pyramids .
this feature has important implications in the strength of the exchange interaction as has been pointed out in ref .
also note that the same kind of _ edge_-sharing pyramids are also present in @xmath3-nav@xmath0o@xmath1 and @xmath4-liv@xmath0o@xmath1@xcite with effective hopping values smaller than @xmath64 and similar to @xmath45 in csv@xmath0o@xmath1 . about the nature of the rest of the hopping integrals in csv@xmath0o@xmath1 , we observe that @xmath46 , @xmath47 and @xmath57 correspond to v@xmath13-v@xmath13 paths through the bridging v@xmath16o@xmath35 tetrahedra with different lengths and angles .
et al._@xcite studied , based on geometrical considerations , the various super - exchange paths for the case of vanadyl phosphates . in particular they considered the vopo compound .
vopo and csv@xmath0o@xmath1 have a similar monoclinic structure if one identifies the p@xmath16o@xmath35 tetrahedra and the v@xmath13o@xmath1 square pyramids in vopo with the v@xmath16o@xmath35 tetrahedra and the v@xmath13o@xmath1 square pyramids in csv@xmath0o@xmath1 respectively . as argued by beltran _
et al . _ , based on the formation of coherent molecular electron orbitals , the coupling between two v@xmath13 through bridging oxygens for separated pyramids is weaker than those via the pathways involving v@xmath16 ions . furthermore , double bridging modes , involving two v@xmath16 ions , as in @xmath46 and @xmath47 ( see fig . [ hoppings ] ) can be stronger than single bridging modes involving one v@xmath16 ion as in @xmath57 .
these interactions can be as strong as paths of the type @xmath45 , i.e. v@xmath13-o - v@xmath13 where the square pyramids are linked to each other by sharing an edge .
geometry considerations then lead to a relative order of importance among the various bridging modes , which for csv@xmath0o@xmath1 is observed to be @xmath65 .
we note here that while in csv@xmath0o@xmath1 we have seen that the dimers are formed by the edge - sharing pyramids ( @xmath45 ) with the predominant inter - dimer interaction formed by v@xmath13-v@xmath13 interaction bridged by v@xmath16o@xmath66 ( @xmath47 ) , in vopo these roles are interchanged in the sense that dimers are formed by v@xmath13o@xmath1 square pyramids bridged by p@xmath16o@xmath35 groups@xcite ( @xmath47 ) and the inter - dimer interaction is formed by neighboring v@xmath13o@xmath20 pyramids ( @xmath45 ) .
this is due to the different type of distortion in the square pyramids as well as different angle and distance relations between these two compounds . using the relation @xmath67 , @xmath68 being the on - site coulomb electron - electron interaction and considering u@xmath69 2.8 ev @xcite
the dft calculation gives an estimate of the intra - dimer exchange coupling , @xmath70 .
to summarize , the conclusions that we draw from the above analysis is that _ ab initio _ calculations support an alternating chain behavior for csv@xmath0o@xmath1 along @xmath26 with weaker inter - chain interactions along @xmath25 .
in order to test the results obtained from the _ ab initio _ calculation we only have available two sets of susceptibility data @xcite as a function of temperature which do nt completely agree quantitatively with each other .
we consider here the data by isobe _
et al._@xcite these authors proposed that their experimental data are well fitted by the susceptibility of a spin-@xmath2 dimer system with @xmath71 given by : @xmath72 with a g - factor of 1.8 , @xmath73 and @xmath74 .
we have re - analyzed the data by considering the temperature dependence of the susceptibility for a spin-@xmath2 alternating heisenberg chain with : @xmath75 with @xmath76 and @xmath77 ( antiferromagnetic couplings ) obtained by the stochastic quantum monte carlo method @xcite .
we define the parameter @xmath78 which measures the ratio of the inter and intra - dimer coupling .
@xmath79 corresponds to the dimer model .
we introduce also an additional parameter @xmath80 so that @xmath81 , @xmath82 and therefore @xmath83 . in fig .
[ susceptibility](a ) we show the comparison of the experimental data to the dimer - model susceptibility given by eq .
[ dimer_susc ] .
we find , in fact , that the best fit corresponds to @xmath73 , @xmath74 and a value of @xmath84 , somewhat smaller than the one proposed in ref .
@xcite . figs .
[ susceptibility ] ( b ) and ( c ) show the comparison of the experimental data to the spin-@xmath2 alternating heisenberg chain model for @xmath85 and @xmath86 respectively .
we observe that values of @xmath87 up to @xmath88 give still very good fits to the susceptibility data by choosing appropriate @xmath19 and @xmath18 parameters .
this analysis makes plausible our _ ab initio _ results , namely the fact that the inter - dimer interactions can be also significant@xcite .
the important conclusion that we gain from this susceptibility analysis is that it is a very hard task to determine the behavior of a compound only from susceptibility fits .
extra information is needed so as to minimize the possible options .
for instance , electron spin -resonance ( esr ) experiments could complement the susceptibility measurements by delivering the value of the gyromagnetic factor .
there are in the literature other examples ( _ e.g. _ vopo ) where the only examination of the susceptibility led to a wrong interpretation of the behavior of the system@xcite .
we have presented _ ab initio _ band - structure calculations for csv@xmath0o@xmath1 , a low - dimensional spin - gapped system and member of the alkali - vanadate family av@xmath0o@xmath1 ( a = na , li , cs ) which has a similar monoclinic structure as the most studied vopo . by means of a _ downfolding_-tight - binding analysis ,
we observe that the behavior of this system can be best described by a spin-@xmath2 alternating chain model with weak interchain interactions .
this proposal has been complemented by an analysis of available susceptibility data for this compound .
a re - examination of the temperature dependence of the experimental susceptibility leads us to the conclusion that an alternating chain model can also provide a good fitting by choosing appropriate @xmath19 and @xmath18 values .
we have discussed the structural , electronic and magnetic similarities and differences between csv@xmath0o@xmath1 and @xmath3-nav@xmath0o@xmath1 , @xmath4-liv@xmath0o@xmath1 and vopo . with this work
we would like to draw the attention on this compound which shows interesting features , especially since it has a lot of points of comparison to the most studied vopo .
we think that new experimental studies like esr , ins , raman scattering and the study of a possible field - induced magnetic ordering under a very strong magnetic field would be helpful to unveil the behavior of this material .
see for instance j. ldecke , a. jobst , s. van smaalen , e. morre , c. geibel and h. -g .
krane , phys .
rev . lett . * 82 * , 3633 ( 1999 ) .
j. l. de boer , a. m. meetsma , j. baas and t. t. m. palstra , phys . rev . lett . * 84 * , 3962 ( 2000 ) .
p. blaha , k. schwarz and j. luitz , wien97 , _ a full potential linearized augmented plane wave package for calculating crystal properties _ ,
( karlheinz schwarz , techn . univ .
wien , vienna 1999 ) .
isbn 3 - 9501031 - 0 - 4 .
[ updated version of p. blaha , k. schwarz , p. sorantin , and s. b. trickey , comp .
. commun . * 59 * , 399 ( 1990 ) ] .
|
we have studied the electronic structure of the spin - gapped system csv@xmath0o@xmath1 by means of an _ ab initio _ calculation .
our analysis and a re - examination of the susceptibility data indicate that the behavior of this system is much closer to that of an alternating spin-@xmath2 antiferromagnetic chain with significant inter - dimer coupling and weaker inter - chain couplings than that of isolated dimers as was initially proposed .
comparison to the vanadate family members @xmath3-nav@xmath0o@xmath1 , @xmath4-liv@xmath0o@xmath1 and isostructural compounds like ( vo)@xmath5p@xmath5o@xmath6 is discussed .
pacs numbers : 75.30.gw , 75.10.jm , 78.30.-j
|
for nearly integrable hamiltonian systems , the set of kam tori has a relatively large lebesgue measure in phase space . for systems with two degrees of freedom
, it implies the dynamical stability : all orbits are stable , the variation of actions stays small for all the time as each 2-dimensional kam torus separates the 3-dimensional energy level .
however , this is a special property of lower - dimensional space , kam torus of @xmath5-dimension does not separate @xmath6-dimensional energy level if @xmath7 .
it is conceivable that the complement of all @xmath5-dimensional invariant tori forms dense and connected set in phase space .
this would mean that by arbitrary small changes of the initial states one would find orbits along which the action variables ultimately escape . the underlying phenomenon is now called arnold diffusion " . * conjecture * ( @xcite ) : _ the typical case in a higher - dimensional problems is topological instability : through an arbitrarily small neighborhood of any point there passes a phase trajectories along which the slow variables drift away from the initial value by a quantity of order 1 .
_ since the celebrated example of arnold @xcite was published half a century ago , there are many works for the study of this problem . in recent years
, it has become clear that diffusion is a typical phenomenon in so called _ a priori _ unstable systems , refer to @xcite .
the _ a priori _ unstable condition guarantees the existence of normally hyperbolic cylinder , from which one derives certain regularity of the barrier functions .
the genericity of the diffusion is obtained by using the regularity @xcite .
there are also many works for the study of the problem , for instance , see @xcite .
general perturbation of integrable hamiltonian is usually called _ a priori _ stable system .
a bit away from strong complete resonance in such systems , some pieces of normally hyperbolic cylinder still exist and the method for _ a priori _ unstable system can also be applied @xcite . for _ a priori _ stable systems with three degrees of freedom , a notable difficulty occurs at the point of double resonance , around which the cylinder for prescribed single resonance may disappear .
the averaged system has two homoclinic orbits associated with different classes in @xmath8 . as the energy decreases , the periodic orbit on the cylinder approaches these two homoclinic orbits simultaneously .
thus , the transition chain in @xmath9 for the single resonance may break . to solve this difficulty ,
mather suggested a path in @xmath9 to cross double resonance , along which one moves the cohomology class in the channel determined by the prescribed homology class and switches it to the channel determined by one of these two classes when it is getting close to the double resonance @xcite . in this paper ,
the path we choose to construct transition chain is different from that suggested by mather .
we find an annulus surrounding the flat of double resonance in @xmath9 , which has a foliation of circles .
each of these circles is actually a transition chains , possibly , of incomplete intersection .
although the annulus is not so thick , each single resonance path extends into .
it allows us to use one of these circles connecting one single resonance path to another . in this way
, we find a path of transition chain along which the diffusion orbits are constructed by variational method .
we consider nearly integrable hamiltonian systems with 3 degrees of freedom : @xmath10 where @xmath11 is assumed to strictly convex , namely , the hessian matrix @xmath12 is positive definite .
it is also assumed that @xmath13 , both @xmath11 and @xmath14 are @xmath15-function with @xmath16 . for @xmath17 , let @xmath18 denote the energy level set , @xmath19 denote a ball in @xmath20 such that @xmath21 .
let @xmath22 denote a sphere and a ball with radius @xmath23 respectively : @xmath24 if and only @xmath25 and @xmath26 if and only @xmath27 .
they inherit the topology from @xmath28 . for perturbation
@xmath14 independent of @xmath29 ( classical mechanical system ) we use the same notation @xmath30 to denote a sphere and a ball with radius @xmath23 let @xmath31 be a set residual in @xmath32 , each @xmath33 is associated with a set @xmath34 residual in the interval @xmath35 $ ] with @xmath36 .
a set @xmath37 is said cusp - residual in @xmath38 if @xmath39 let @xmath40 denote the hamiltonian flow determined by @xmath41 .
given an initial value @xmath42 , @xmath43 generates an orbit of the hamiltonian flow @xmath44 .
an orbit @xmath44 is said to visit @xmath45 if there exists @xmath46 such that @xmath47 a ball centered at @xmath48 with radius @xmath49 .
[ mainthm ] given any two balls @xmath50 and finitely many small balls @xmath51 @xmath52 , where @xmath53 with @xmath17 and @xmath54 is small , there exists a cusp - residual set @xmath55 such that for each @xmath56 , the hamiltonian flow @xmath40 admits orbits which , on the way between passing through @xmath57 and @xmath58 , visit the balls @xmath59 in any prescribed order .
this is the main result of the paper .
it is generic not only in usual sense , but also in the sense of ma , namely , it is a typical phenomenon when the system is perturbed by potential .
the same result for time - periodic systems can be proved by using the same method .
the statement of the result is : for typical time - periodic perturbations of integrable hamiltonian with 2-degrees of freedom , the hamiltonian flow admits orbits passing through any prescribed two balls @xmath57 and @xmath58 in the phase space and finitely many small balls @xmath60 @xmath52 in the action variable space .
the proof is easier from technical point of view , one can see it in the proof .
similar result was announced by mather earlier @xcite .
recently , two other groups ( kaloshin and zhang , marco @xcite ) announced , using different approach , a similar result for time - periodic systems with two degrees of freedom . the result obtained here is stronger than what was formulated in @xcite . by dropping the requirement that orbit passes two prescribed balls in the phase space and using the same construction
, one can get an orbit that visits these balls @xmath51 @xmath52 infinitely many times with any prescribed order , as it was announced in @xcite . indeed , as finitely many balls are given , there exists a path with finite length passing through finitely many resonance layers and connecting any two of these balls directly .
it does not damage the cusp - residual property .
we use variational method to prove the result , which is based on mather theory .
this theory is established for tonelli lagrangian .
let @xmath61 be a closed manifold . a @xmath62-function @xmath63 : @xmath64 is called tonelli lagrangian if it satisfies the following conditions : positive definiteness . for each @xmath65 ,
the lagrangian function is strictly convex in velocity : the hessian @xmath66 is positive definite .
super - linear growth .
we assume that @xmath63 has fiber - wise superlinear growth : for each @xmath65 , we have @xmath67 as @xmath68 .
all solutions of the lagrangian equations are well defined for the whole @xmath46 . for autonomous systems ,
the completeness is automatically satisfied , since each orbit entirely stays in certain compact energy level set .
let @xmath69 denote a closed 1-form @xmath70 evaluated at @xmath71 , with its first co - homology class @xmath72=c\in h^1(m,\mathbb{r})$ ] .
we introduce a lagrange multiplier @xmath73 . without danger of confusion ,
we call it closed 1-form also . for each @xmath74 curve @xmath75 : @xmath76 with period @xmath77 , there is unique probability measure @xmath78 on @xmath79 so that the following holds @xmath80 for each @xmath81 , where we use the notation @xmath82
. let @xmath83 the set @xmath84 of holonomic probability measures is the closure of @xmath85 in the vector space of continuous linear functionals .
one can see that @xmath84 is convex . for each @xmath86 the action @xmath87 is defined as follows @xmath88 it is proved in @xcite that for each co - homology class @xmath89 there exists at least one invariant probability measure @xmath90 minimizing the action over @xmath84 @xmath91 called @xmath89-minimal measure .
let @xmath92 be the set of @xmath89-minimal measures , the mather set @xmath93 is defined as @xmath94 the @xmath95-function is defined as @xmath96 , it is convex , finite everywhere with super - linear growth .
its legendre transformation @xmath97 is called @xmath98-function @xmath99 it is also convex , finite everywhere with super - linear growth ( see @xcite ) . note
that @xmath100 for each exact 1-form @xmath101 and each @xmath102 .
therefore , for each measure @xmath103 one can define its rotation vector @xmath104 such that @xmath105,\omega(\mu)\rangle=\int\lambda d\mu,\ ] ] for every closed 1-form @xmath101 on @xmath61 .
for a closed curve @xmath75 : @xmath106\to m$ ] its rotation vector is defined as @xmath107=\frac{\bar{\gamma}(k)-\bar{\gamma}(0)}k,\ ] ] where @xmath108 stands for a curve in the lift of @xmath75 to the universal covering @xmath109 .
let @xmath110 : @xmath106\to m$ ] be a closed curve such that @xmath111=\omega_k$ ] and @xmath112=\omega_k}\frac 1k\int_0^kl(d\gamma(t),t)dt.\ ] ] the curve determines a periodic orbit @xmath113 of @xmath114 , the lagrange flow determined by the lagrangian @xmath63 , it supports an invariant measure @xmath115 whose rotation vector is @xmath116 .
the measure @xmath115 is not necessarily minimal for @xmath117 . nevertheless , if we choose a sub - sequence of closed curves @xmath118 such that @xmath119=\omega_k}\frac 1k\int_0^kl(d\gamma(t),t)dt,\ ] ] and if
@xmath120\to\omega$ ] , then @xmath121 clearly , there is at least one invariant measure @xmath122 such that @xmath123 , and @xmath122 is a holonomic probability measure with the prescribed rotation vector @xmath124 . according to the definition of holonomic measure , and due to the work in @xcite , we have @xmath125 where @xmath126 is a set of holonomic probability measures with the given rotation vector @xmath127 , not necessarily invariant for @xmath128 . the fenchel - legendre transformation @xmath129 : @xmath130 is defined by the following relation @xmath131 the concept of semi - static curves is introduced by mather and ma ( cf .
@xcite ) . a curve @xmath75 : @xmath76 is called @xmath89-semi - static if in time-1-periodic case we have @xmath132}]=f_c((\gamma(t),t),(\gamma(t'),t'))\ ] ] where @xmath133}]=\int_{t}^{t'}\big ( l(d\gamma(t),t)-\eta_c(d\gamma(t))\big ) dt+\alpha(c)(t'-t),\notag\ ] ] @xmath134 in which @xmath135}].\notag\ ] ] in autonomous case , the period can be considered as any positive number .
consequently , the notation of semi - static curve in this case is somehow simpler @xmath136=f_c(\gamma(t),\gamma(t')),\ ] ] where @xmath137 * convention * : let @xmath138 be an interval @xmath139either bounded or unbounded@xmath140 . a continuous map @xmath75 : @xmath141 is called curve . if it is differentiable , the map @xmath82 : @xmath142 is called orbit .
when the implication is clear without danger of confusion , we use the same symbol to denote the graph , @xmath143 is called curve and @xmath144 is called orbit . in autonomous system , the terminology also applies to the is called curve and @xmath145 is called orbit .
a semi - static curve @xmath146 is called @xmath89-static if , in addition , the relation @xmath136=-f_c((\gamma(t'),\tau'),(\gamma(t),\tau))\ ] ] holds in time-1-periodic case and @xmath136=-f_c(\gamma(t'),\gamma(t))\ ] ] holds in autonomous case .
an orbit @xmath147 is called @xmath89-static ( semi - static ) if @xmath75 is @xmath89-static ( semi - static ) .
we call the ma set @xmath148 the union of @xmath89-semi - static orbits @xmath149 and call the aubry set @xmath150 the union of @xmath89-static orbits
@xmath151 we use @xmath152 , @xmath153 and @xmath154 to denote the standard projection of @xmath93 , @xmath150 and @xmath148 from @xmath79 to @xmath155 respectively .
they satisfy the inclusion relation @xmath156 it is showed in @xcite that the inverse of the projection is lipschitz when it is restricted to @xmath153 as well as to @xmath152 . by adding subscript @xmath157 to @xmath158 ,
i.e. @xmath159 we denote its time-@xmath157-section .
this principle also applies to @xmath160 , @xmath161 , @xmath162 , @xmath163 and @xmath164 to denote their time-@xmath157-section respectively . for autonomous systems ,
these sets are defined without the time component . on the time-1-section of aubry set a pseudo - metric @xmath165 is introduced by mather in @xcite , its definition relies on the quantity @xmath166 .
let @xmath167 @xmath168 the pseudo - metric @xmath165 on aubry set is defined as @xmath169 with the pseudo - metric @xmath165 one defines equivalence class in aubry set .
the equivalence @xmath170 implies @xmath171 , with which one can define quotient aubry set @xmath172 .
its element is called aubry class , denoted by @xmath173 , its lift to @xmath79 is denoted by @xmath174 .
thus , @xmath175 in @xcite mather constructed an example with some quotient aubry set homeomorphic to an interval .
however , it is proved generic in @xcite that , for system with @xmath5 degrees of freedom , each @xmath89-minimal measure contains not more than @xmath176 ergodic components . in this case , each aubry set contains at most @xmath176 classes .
the definition of semi - static curve as well as of ma set depends on which configuration manifold under our consideration .
let @xmath177 be a finite covering , a curve @xmath75 : @xmath76 is said semi - static in @xmath178 if each curve in its lift @xmath179 is semi - static in @xmath178 .
accordingly , we define @xmath180 ( @xmath181 ) as the set containing all @xmath89-semi - static orbits ( curves ) in @xmath178 .
we use the symbol @xmath182 when @xmath61 is defaulted as the configuration manifold .
it is possible that @xmath183 .
for instance , if @xmath184 is a open region such that @xmath185 , @xmath186 and the lift of @xmath187 in @xmath178 has more than one connected component , then this phenomenon takes place .
but we have let @xmath177 be a finite covering space , then @xmath188 pick up any @xmath189 and any small @xmath54 , by definition , there exists sufficiently large @xmath190 as well as a curve @xmath191\to\bar m$ ] such that @xmath192 and @xmath193<\delta$ ] .
let @xmath194 denote the project of @xmath195 down to @xmath61 , clearly , we have @xmath196=[a_c(\bar\xi)]<\delta$ ] .
let @xmath197 , clearly , @xmath198 .
we use variational method to prove the result . since the work of mather @xcite , the variational method has become a powerful tool for the study of dynamical instability in positive definite lagrange systems with multiple degrees of freedom . in the study of arnold diffusion in _ a priori _ stable systems with three degrees of freedom , mainly due to the work of mather @xcite , it has been widely known that the main difficulty takes place near double resonance . to describe what puzzled us and to explain the strategy of our proof , let us recall the example of arnold and previous study on _ a priori _ unstable systems . in the example of arnold , there exists a 2-dimensional cylinder in the phase space , which is invariant and normally hyperbolic for the time-1-map determined by the hamiltonian flow .
this cylinder is foliated into a family of invariant circles , each of them has stable and unstable manifold which intersect each other transversally .
consequently , the unstable manifold of some circle intersects the stable manifold of other circles nearby , it implies the existence of a sequence of successively connected heteroclinic orbits . this structure is called transition chain by arnold .
diffusion orbits are then constructed shadowing these heteroclinic orbits .
such argument heavily depends on the geometric structure and variational method turns out to have wider range of application .
let us interpret the proof by variational language .
each invariant circle is the aubry set for certain cohomology class , the stable as well as the unstable manifold is actually the graph of the differential of the weak kam solution .
expressed as the difference of backward and forward weak kam , the barrier function reaches its minimum at primary intersection points of these two manifolds .
these homoclinic orbits and the aubry set constitute the ma set in certain finite covering space . for positive definite systems , the transversal intersection implies the minimality of the homoclinic orbits as well as local heteroclinic orbits , along which the lagrange action reaches the minimum among all those curves with the same boundary conditions .
the diffusion orbits are obtained by searching for global minimizers which generate orbits shadowing a sequence of local heteroclinic orbits .
the variational arguments still work even if there does not exist such nice geometric structure , provided the following conditions are satisfied for each cohomology class 1 , _ the aubry set is lower dimensional : @xmath199 _ ; 2 , _ the stable manifold " intersects the unstable manifold " transversally_. however , it turns out very difficult to verify whether the stable manifold " intersects the unstable manifold " transversally for each cohomology class along a path of first cohomology class . as uncountably many stable and unstable manifolds " have to be considered , one can not verify the genericity of the transversal intersection by taking the intersection of countably many open - dense sets .
one possible way is to study some regularity of barrier functions with respect to some parameter , with which one obtains the finiteness of hausdorff dimensions of the set of barrier functions . as such regularity
is obtained in the case when a normally hyperbolic cylinder exists we are succeeded in solving the problem in _ a priori _ unstable case @xcite .
in fact , the regularity was obtained in @xcite only for those barrier functions for which the minimal measure is supported on an invariant circle .
it was extended in @xcite to all other barrier functions , which allows us to construct diffusion orbits of which the picture looks like what was constructed by arnold , while in our previous work , the constructed orbits keep close to the cylinder when they pass through strong resonance ( birkhoff instability region ) .
gap " problem was then solved .
intuitively , diffusion orbits in _ a priori _ stable systems may be constructed along some resonant path . in terms of rotation vector ( first homology class ) ,
each point on this curve satisfies at least one resonant condition for the system with @xmath200 degrees of freedom . in integrable systems
, each resonant path corresponds to an invariant cylinders without any hyperbolicity . under generic perturbations ,
it breaks into many pieces of normally hyperbolic cylinder , but may disappear around double resonant points .
it implies a bad consequence : we lost a handhold to get certain regularity of barrier functions in suitable parameter .
it then becomes unclear whether there is a transition chain near double resonance . in terms of first cohomology
, strong double resonance corresponds to a convex disc with size @xmath201 if the perturbation is of order @xmath202 , each piece of normally hyperbolic cylinder corresponds to a channel which extends to a small neighborhood of the disc .
but it is unclear whether these channels are connected to the disc of double resonance .
the method we use to overcome this difficulty bases on following discoveries : first , each double resonant disc is surrounded by an annulus foliated into a family of paths , along each of these paths , there is another invariant ( a coordinate component of the cohomology class ) besides the average action . for each class in this annulus , the intersection of stable manifold " with unstable manifold " is nontrivial although it may not be transversal .
this annulus has width of order @xmath202 .
next , incomplete intersection of the stable and unstable manifold " of an aubry set does not implies that it can not be connected to any other aubry set nearby . it
does if they are @xmath89-equivalent .
finally , the channels of normally hyperbolic cylinder reach to somewhere @xmath203-close to double resonant disc ( @xmath54 ) , i.e. it has overlap with the annulus . for each class in the channel ,
the relative homology of the aubry set is non - trivial .
therefore , we are able to find a path close to prescribed one , for each class on the path , the aubry set is connected to another one nearby if the class is also on the path .
all of these connecting orbits are minimal in local sense .
the diffusion orbits are constructed shadowing these successively connected orbits .
we organize the proof in following way .
section 2 is used to establish the concept of elementary weak kam , with which one has simpler expression of barrier function .
it thus becomes easier to study genericity of transition chain .
section 3 , 4 and 5 are devoted to study the structure of ma set and of aubry set . since these sets are symplectic invariants @xcite , we do it by studying the normal form which is put into the appendix .
the truncated hamiltonian of the normal form is a system with two degrees of freedom . in section 3 ,
we study the dynamics around the double resonance , and the modulus continuity of the period on energy ( average action ) . with these preliminary works ,
normally hyperbolic cylinder is shown to get very close to the double resonance in section 4 , the existence of annulus of @xmath89-equivalence is established in section 5 .
section 6 is devoted to establish two types of local connecting orbits .
the local minimality of these local connecting orbits are naturally given , it enables us to construct global connecting orbit shadowing these local connecting orbits .
it is obtained by searching for the minimizer of certain modified lagrangian , which is done in section 7 .
finally in section 8 , we verify the cusp - residual property of the transition chain in nearly integrable hamiltonian systems with three degrees of freedom .
consequently , the main result of this paper is proved .
the concept of elementary weak kam solution is introduced for @xmath89-minimal measure with finitely many ergodic components , this condition has been shown to be generic in @xcite . each ergodic component
@xmath204 determines a pair of elementary weak kam @xmath205 , with which we introduce a barrier function @xmath206 with this formula , it is easier to show that , generically , the set @xmath207 is totally disconnected .
this property is crucial for the construction of diffusion orbits .
the concept of @xmath89-semi - static curves can be extended to the curves only defined on @xmath208 , which are called forward or backward @xmath89-semi - static curves respectively .
usually one uses @xmath209 : @xmath210\to m$ ] to denote backward @xmath89-semi - static curve such that @xmath211 , and uses @xmath212 : @xmath213 to denote forward @xmath89-semi - static curve such that @xmath214 . in autonomous case ,
one uses the notation @xmath215 such that @xmath216 .
let @xmath217}\ \text{\rm is \it c\rm - semi - static}\},\notag\end{aligned}\ ] ] where @xmath218 , @xmath219 denotes the standard projection along the tangent fiber and @xmath220 denotes the orbit of the lagrangian flow with the initial value @xmath221 at the time @xmath222 .
the corrsponding orbits are called forward ( backward ) semi - static orbit set respectively .
these two sets are upper semi - continuous for the cohomology class .
[ weakpro1 ] if the lagrangian @xmath63 is of tonelli type , for each point @xmath223 , there is at least one @xmath224 which is forward @xmath139backward@xmath140 semi - static curve .
as both the @xmath127-limit set of @xmath225 and the @xmath95-limit set of @xmath226 are in the aubry set one can define @xmath227 and call @xmath228 the stable set , @xmath229 the unstable set of the @xmath89-minimal measure respectively .
if @xmath230 holds for some @xmath223 , passing through the point @xmath231 the orbit is either in the aubry set or homoclinic to this aubry set .
when the aubry set contains only one class , the stable as well as the unstable set has its own generating function @xmath232 such that @xmath233 holds almost everywhere @xcite .
these functions are weak kam solutions , which are the fixed points of so called lax - oleinik operator .
we use @xmath232 to denote the weak kam solution for the lagrangian @xmath234 , where @xmath235 is a closed form with @xmath236=c$ ] .
these functions are lipschitz , thus differentiable almost everywhere . at each differentiable point @xmath237 , @xmath238 uniquely determines backward @xmath89-semi static curve @xmath239 : @xmath240\to m$ ] such that @xmath241 , @xmath242 .
similarly , @xmath243 uniquely determines forward @xmath89-semi static curve @xmath239 : @xmath213 such that @xmath244 , @xmath245 .
given a class @xmath246 , we use @xmath247 to denote the set of aubry classes , use @xmath248 to denote the projected set along the tangent fibers , where @xmath249 is the subscript set : @xmath250 .
we also use the notation @xmath251 where @xmath252 is an ergodic component of the @xmath89-minimal measure @xmath90 .
let @xmath253 .
[ weakpro2 ] ( @xcite ) let @xmath254 and @xmath255 be two weak - kam solutions for @xmath89 .
their difference keeps constant when they are restricted on an aubry class @xmath256 .
recall the definition of @xmath166 in the introduction .
we use the symbol @xmath257 to denote the quantity defined in the same way for @xmath63 with @xmath258 , and drop the subscript @xmath63 when it is clearly defined .
the quantity @xmath259 is a weak - kam solution if we consider it as the function of @xmath260 or of @xmath261 .
let us consider the case that the @xmath89-minimal invariant measure has finitely many ergodic components . in this case , this function has some kind of continuity .
[ weakthm1 ] let @xmath262 be a sequence of lagrangian , converging to @xmath63 in @xmath62-topology as @xmath263 when they are restricted on any bounded regions of @xmath79 .
we assume that the minimal measure for @xmath63 consists of finitely many ergodic components @xmath264 , @xmath265 and the distance from @xmath266 to the aubry set for @xmath267 , @xmath268 as @xmath263 .
then @xmath269 we only need to prove it on the time-1-section , e.g. for @xmath270 .
so we omit the notation for the component @xmath222 . for each @xmath271
, there exists @xmath272 such that @xmath273 .
let @xmath95 and @xmath274 denote the minimal average action of @xmath63 and @xmath267 respectively .
let @xmath275 : @xmath106\to m$ ] be the curve such that @xmath276 , @xmath277 and @xmath278=\int_0^{k}l ( d\gamma^{k}(t),t)dt+k\alpha = h^{k}(x , x').\ ] ] for any @xmath279 , let @xmath280 : @xmath281\to m$ ] be an absolutely continuous curve such that @xmath282 , @xmath283 for @xmath284 $ ] and @xmath285})]= \int_0^{k'-k}l_{\delta } ( d\zeta(t),t)dt+(k'-k)\alpha_{\delta}=h_{l_{\delta}}^{k'-k}(x , x)$ ] .
therefore , we have @xmath286&\le h_{l_\delta}^{k'-k}(x , x)+h^{k}(x , x')+k|\alpha-\alpha_{\delta}|\\ & + \left|\int_0^k(l - l_{\delta})(d\gamma^k(t),t)dt\right|.\end{aligned}\ ] ] since @xmath287 as @xmath263 we see that @xmath288 as @xmath263 . since @xmath95 is continuous in the lagrangian and @xmath289
is arbitrarily small we see that @xmath290 so , in order to complete the proof , we only need to show @xmath291 let @xmath292 : @xmath293\to m$ ] be a curve such that @xmath294 , @xmath295 and @xmath296=h_{l_\delta}^{k_{\ell}}(x , x')\to h_{l_\delta}^{\infty}(x , x'),\ ] ] where @xmath297 is a sequence of integers .
let @xmath298 denote the @xmath289-neighborhood of the set @xmath299 . for small @xmath271 ,
there exist some @xmath300 with @xmath301 , an integer @xmath302 $ ] and @xmath303 such that @xmath304 provided @xmath305 is sufficiently large .
let us consider those ergodic components of minimal measure for @xmath63 of which the support is approached by @xmath306 as @xmath263 : @xmath307 .
we number some @xmath300 as @xmath308 if some @xmath309 exists such that @xmath310 and for each @xmath311 , @xmath312 does not fall into @xmath289-neighborhood of any @xmath313 .
let @xmath314 be the integer such that @xmath315 and @xmath316 for all @xmath317 .
we number some @xmath318 if some @xmath319 and some @xmath320 exists such that @xmath321 , let @xmath322 be the integer such that @xmath323 and @xmath324 for all @xmath325 .
inductively , one obtains @xmath326 ( @xmath327 ) and @xmath328 obviously , there exist small @xmath329 and large integer @xmath330 such that @xmath331 provided @xmath332 .
otherwise , there would exist also an ergodic component @xmath333 of the minimal invariant measure such that @xmath334 for all @xmath335 , but it is absurd
. given small @xmath271 , let @xmath305 be the integer such that @xmath336 .
let @xmath337 , @xmath338 , we choose an absolutely continuous curve @xmath339 : @xmath340\to m$ ] such that @xmath341 , @xmath342 and @xmath343=h^{k_{\ell}^j}(\bar x_j,\tilde x_j)$ ] .
as @xmath344 we can choose sufficiently large @xmath345 such that @xmath346 where @xmath347 is a constant depending on @xmath63 only . as @xmath348 , for any positive integer @xmath349
we have @xmath350 for any large integer @xmath351 with @xmath352 , we construct an absolutely continuous curve @xmath280 : @xmath281\to m$ ] joining @xmath71 with @xmath353 such that @xmath354 where @xmath355 , @xmath356 . the action of @xmath63 along this curve is easily estimated @xmath357-h_{l_\delta}^k(x , x')\\ & \le & 2m'(c\epsilon + k|\alpha-\alpha_{\delta}|)\\ & & + \sum _ { j=1}^{m'}\left|\int_{k'_{j-1}}^{k_j}(l - l_{\delta})(d\gamma_{\delta}^{\ell}(t),t)dt\right|.\end{aligned}\ ] ] as @xmath358 , and @xmath359 is independent of @xmath49 when @xmath49 is sufficiently close to @xmath360 , we see that the inequality ( [ weakeq2 ] ) holds .
this completes the proof .
let @xmath361 be a sequence of cohomology classes such that @xmath362 .
the @xmath89-minimal measure is assumed consisting of finitely many ergodic components @xmath363 , @xmath364 , @xmath365 , @xmath366 and @xmath367 for some @xmath335 , as @xmath362 . then @xmath368 from the proof
one can see that the function @xmath166 is lower semi - continuous in @xmath89 if the @xmath89-minimal measure is assumed to have finitely many ergodic components : @xmath369 let us introduce the concept of _ elementary _ weak kam solution if the @xmath89-minimal measure has finitely many ergodic components .
one can choose finitely many non - negative functions @xmath370 : @xmath371 such that its support has no intersection with a small neighborhood of @xmath372 and the minimal measure for the lagrangian @xmath373 is uniquely supported on @xmath372 . by the theory of weak kam ( @xcite ) , there is exactly one pair of weak kam solutions denoted by @xmath374 and @xmath375 let @xmath376 , in virtue of theorem [ weakthm1 ] , one has @xmath377 as @xmath378 .
since @xmath379 in the neighborhood of @xmath372 , @xmath380 remains unchanged as @xmath378 .
thus , there is a lipschitz function @xmath381 such that @xmath382 as @xmath378 .
clearly this @xmath381 is a weak - kam solution for @xmath383 .
similarly , we can see that @xmath384 as @xmath378 .
( elementary weak - kam solution ) .
assume that the minimal measure for @xmath383 consists of finitely many ergodic components @xmath363 , @xmath385 .
a weak kam solution @xmath386 of @xmath383 is called _ elementary _ for @xmath252 if @xmath387 where @xmath374 is the weak kam solution of @xmath388 , of which the minimal measure @xmath389 is uniquely ergodic and @xmath390 as @xmath378 .
it is not necessary that @xmath386 is a pair of conjugate weak kam .
clearly , if @xmath391 @xmath392 these elementary weak kam solutions generate all weak kam solutions in the following sense .
[ weakpro3 ] assume the minimal measure consists of @xmath393 ergodic components .
for each weak kam solution @xmath394 , there exist @xmath395 @xmath396 constants @xmath397 and @xmath395 open domains @xmath398 such that they do not overlap each other , @xmath399 and @xmath400 it is deduced from the lipschitz property of @xmath401 that it is differentiable almost every where .
let @xmath71 be a point where @xmath401 is differentiable , @xmath402 determines a unique backward semi static orbit @xmath403 : @xmath404\to m$ ] whose @xmath95-limit set is in certain aubry class @xmath405 . by definition
we have @xmath406 let @xmath407 such that @xmath408 , it follows from proposition [ weakpro2 ] that @xmath409 if @xmath410 is another point where @xmath411 determines a backward semi - static orbits whose @xmath95-limit set is also contained in @xmath405 , we then obtain ( [ weakeq5 ] ) for @xmath412 with the same @xmath413 .
all these points constitute a set connected with @xmath414 .
there are not more than @xmath393 connected sets such that ( [ weakeq4 ] ) holds .
let @xmath362 be a sequence of cohomology and assume that the minimal measure consists of finitely many ergodic components for each @xmath415 and @xmath89 .
let @xmath416 , @xmath417 be the support for the ergodic minimal measure @xmath418 and @xmath419 respectively , let @xmath420 and @xmath421 be the corresponding elementary weak kam solution . if @xmath422 as @xmath362 , then @xmath423 in @xmath424-topology .
it follows from the continuity of @xmath425 in @xmath89 shown in theorem [ weakthm1 ] and the definition of the elementary weak kam solution . in terms of conjugate pair of weak kam solution
, one has a definition of ma set in @xcite . for the purpose of this paper
, we would like to use elementary weak kam solution .
recall the definition of the barrier function in @xcite : @xmath426 if the minimal measure consists of finitely many ergodic components , we introduce barrier functions in terms of elementary weak kam solutions : given @xmath427 , we set @xmath428 which measures the minimum of the action along those curves passing through @xmath260 and joining @xmath429 to @xmath430 .
for autonomous systems , this barrier function is independent of time .
obviously , each @xmath89-semi static curve corresponds to a minimum of @xmath431 if its @xmath95-limit set intersects @xmath432 and its @xmath127-limit set intersects @xmath433 . to extend the concept of elementary weak kam solution to universal covering space , let us reveal some properties of minimal homoclinic orbit to aubry set .
given a curve @xmath75 : @xmath76 , we call @xmath82 a homoclinic orbit to some aubry set @xmath434 if it does not stay in the aubry set , but its @xmath127-limit set as well as the @xmath95-limit set is contained in the aubry set : @xmath435 correspondingly , we call @xmath75 homoclinic curve . the existence of homoclinic orbits to aubry sets has been studied in a few papers , see @xcite . the existence of homoclinic orbits is closely related to the issue whether the @xmath436ech homology group @xmath437 is non - trivial ( @xmath438 for time - periodically dependent lagrangian ) .
it is defined as the inverse limit @xmath439 , where @xmath440 is an open neighborhood of @xmath441 .
there exists a small open neighborhood @xmath442 of @xmath443 such that @xmath444 provided @xmath445 .
let @xmath178 be a covering of @xmath61 such that @xmath446 where @xmath447 denotes the hurewicz homomorphism .
the group of deck transformation of this covering space is @xmath448 let @xmath440 be an open neighborhood of @xmath441 such that @xmath449 .
let @xmath450 and @xmath451 , then @xmath452 is a free abel group . to each orbit
@xmath453 : @xmath76 homoclinic to @xmath434 , an element @xmath454\in g$ ] is associated .
if the group @xmath452 is non - trivial , there is a flat @xmath455 of the @xmath95-function containing the cohomology class .
a set @xmath456 is called flat if the function @xmath95 is affine when it is restricted on @xmath455 , not affine for any set properly contains @xmath455 .
the dimension of this flat is not smaller than @xmath457 and the aubry set is the same for all classes in the interior of @xmath455 ( see @xcite ) . in this paper
, we are interested in so - called _ minimal _ homoclinic orbits .
let @xmath458 be a covering manifold of @xmath61 such that @xmath459 .
a curve @xmath75 : @xmath76 is called @xmath458 semi - static if the lift of @xmath75 to @xmath458 , @xmath460 : @xmath461 is semi - static .
a homoclinic orbit @xmath462 is called _ minimal _ if the lift @xmath460 : @xmath463 is semi - static . [ homothm1 ] if there is only one aubry class and @xmath464 , then there are at least @xmath465 minimal homoclinic orbits .
if @xmath466 for @xmath467 and @xmath468 , then there are infinitely many @xmath89-minimal homoclinic orbits .
the existence of at least @xmath465 homoclinic orbits is proved in @xcite , they are actually minimal . the infinity of minimal homoclinic orbits are proved in @xcite .
let us briefly describe how to find these @xmath465 minimal homoclinic orbits .
given a point @xmath469 and @xmath470 , we denote by @xmath471 : @xmath472\to m$ ] the minimizer of @xmath473=g}}\int_{-k}^kl(\xi_k(s),\dot\xi_k(s))ds+2k\alpha\ ] ] obviously , the set @xmath474\}$ ] is uniformly bounded for @xmath475 . because of positive definiteness of @xmath63 , the set @xmath476\}$ ] is also uniformly bounded for each @xmath77 .
let @xmath477 there exists a subsequence of @xmath478 such that @xmath479 .
the quantity @xmath480 keeps constant on each aubry class . by diagonal extraction argument
we can find a subsequence of @xmath481 which @xmath74-uniformly converges , on each compact interval , to a @xmath74-curve @xmath75 : @xmath76 . in this sense , @xmath75 : @xmath76 is called an accumulation point of @xmath482 .
each accumulation point is @xmath483 semi - static and there is at least one accumulation point @xmath484 with non - zero homology @xmath485\neq 0 $ ] .
as the relative homology of the aubry set is non - trivial , some @xmath23 exists such that @xmath486 holds for each class @xmath470 and @xmath487 as @xmath488 .
therefore , for each @xmath470 , there are finitely many accumulation points of @xmath482 with non - zero homology , denoted by @xmath489
. clearly @xmath490=g$ ] . as @xmath452 is @xmath491-dimensional ,
at least @xmath465 geometrically different minimal homoclinic orbits exist .
let us look at these homoclinic orbits from another point of view . for certain finite covering manifold ,
the lift of the aubry set has several connected components ( several aubry classes ) .
these aubry classes are connected by semi - static orbits @xcite .
the projection of these semi - static orbits are nothing else but minimal homoclinic orbits . for a finite covering manifold @xmath492 : @xmath493
, the fiber @xmath494 contains finitely many points . for a closed curve @xmath495 : @xmath496\to m$ ]
such that @xmath497 , there is a lift of @xmath498 such that @xmath499 . by the monodromy theorem ,
@xmath500 is uniquely determined by its class @xmath501\in\pi_1(m)$ ] .
let @xmath502 be the generators of @xmath452 , @xmath503 be closed path so that @xmath504=g_i$ ] , @xmath505 for @xmath506 . if @xmath507 is chosen so that @xmath508 and @xmath509 , there will be at least @xmath510 aubry classes for this covering manifold . among the semi - static orbits connecting different aubry classes for the covering manifold , there are at least @xmath465 orbits whose projection is different from each other
let @xmath511 be defined such that @xmath512 if and only if some minimal homoclinic orbit @xmath462 exists such that @xmath454=g$ ] .
we say that there are @xmath77-types of minimal homoclinic orbits if @xmath513 contains exactly @xmath77 elements .
[ homothm2 ] if @xmath514 , @xmath515 and @xmath441 contains a set homeomorphic to @xmath516 , there exist exactly two types of minimal homolcinic orbits . in this case @xmath517 .
by the condition , we can assume that each standard generator @xmath518 with @xmath519 can be represented by a closed curve in @xmath441 .
let @xmath520 with @xmath521 . if there is a minimal homoclinic orbits @xmath453 such that @xmath454=g$
] , there must be some points @xmath522 but @xmath523 .
as @xmath469 , there is a unique vector @xmath524 such that @xmath525 . given any @xmath271 ,
there is static curve @xmath194 : @xmath76 and @xmath526 such that @xmath527 are in @xmath289-neighborhood of @xmath71 , @xmath528 , @xmath529 and @xmath530}]<\epsilon$ ] .
let @xmath531 , @xmath532 , @xmath533 and @xmath534 be suitably small numbers .
we join @xmath535 to @xmath536 by the curve @xmath537 : @xmath538\to m$ ] which minimizes the action @xmath539}]=\inf_{\stackrel{\zeta(-\tau_1 ^ -)=\gamma(t^- ) } { \scriptscriptstyle \zeta(\tau_1^+)= \xi(s_0^+)}}\int_{-\tau_1 ^ -}^{\tau_1^+}l(\zeta(s),\dot\zeta(s))ds + ( \tau_1^++\tau_1 ^ -)\alpha,\ ] ] and join @xmath540 to @xmath541 by the curve @xmath542 : @xmath543\to m$ ] which minimizes the action @xmath544}]=\inf_{\stackrel{\zeta(-\tau_2
^ -)=\xi(s_1 ^ - ) } { \scriptscriptstyle \zeta(\tau_2^+)= \gamma(t^+)}}\int_{-\tau_2 ^ -}^{\tau_2^+}l(\zeta(s),\dot\zeta(s))ds + ( \tau_2^++\tau_2 ^ -)\alpha.\ ] ] we define a continuous curve @xmath545 : @xmath76 by @xmath546,\\ \zeta_1(t-\delta_1),&t-\delta_1\in [ -\tau_1 ^ -,\tau_1^+],\\
\xi(t-\delta_2),&t-\delta_2\in [ s_0^+,s_1 ^ -],\\ \zeta_2(t-\delta_3),&t-\delta_3\in [ -\tau_2 ^ -,\tau_2^+],\\ \gamma(t-\delta_4 ) , & t-\delta_4\in [ t^+,\infty ) , \end{cases}\ ] ] where @xmath547 , @xmath548 , @xmath549 and @xmath550 . by exploiting the _ curve
shorten _ lemma in riemannian geometry as did in @xcite we find that @xmath551}]+[a(\xi)|_{[s_0,s_0^+]\cup[s_1 ^ -,s_1 ] } ] > [ a(\zeta_1)|_{[-\tau_1 ^ -,\tau_1^+]}]+[a(\zeta_2)|_{[-\tau_2 ^ -,\tau_2^+]}]\ ] ] if @xmath552 and @xmath553 . as @xmath469 , @xmath554 , @xmath555 can be arbitrarily close to @xmath556 by choosing suitable @xmath557 and @xmath558 , this inequality still hold in our case .
note that the quantity @xmath530}]$ ] can be arbitrarily close to zero , we see that @xmath559}]>[a(\gamma')|_{[t_{-1},t_1+\delta_4]}]\ ] ] if @xmath560 and @xmath561 . as @xmath562=[\gamma]$ ] , this property contradicts the fact that @xmath75 is minimal . on the other hand , from theorem [ homothm1 ]
, we obtain the existence of 2 minimal homoclinic orbits .
this completes the proof . for each class @xmath470
, we define @xmath563=g}}\int_{0}^{k } l(\xi(s),\dot\xi(s))ds + k\alpha,\ ] ] @xmath564 it is easy to see that @xmath565 as @xmath566 .
indeed , it follows from the fact @xmath567 that @xmath568 for any @xmath569 .
if @xmath570 remains bounded as @xmath566 , there would be a minimal measure whose support is obviously not contained in @xmath434 , but it is absurd .
if the aubry set contains only one class , as a function of @xmath71 , @xmath570 keeps constant on the aubry set .
so it makes sense let @xmath571 for @xmath469 . obviously , one has @xmath572 and [ homopro1 ] if there is an infinite sequence @xmath573 such that @xmath574 then @xmath513 contains infinitely many elements .
the definition of @xmath575 can be extended @xmath576 for @xmath577 .
let us recall that the covering space @xmath578 : @xmath579 is defined such that @xmath459 , where @xmath440 is a open neighborhood of @xmath580 so that @xmath581 .
let @xmath582 be the fundamental domain for @xmath583 and use the same symbol to denote its projection to @xmath483 as well .
for each closed path @xmath495 : @xmath496\to m$ ] , there is a unique curve @xmath584 in the lift of @xmath495 such that @xmath585 . because of the monodromy theorem
, @xmath586 is uniquely determined by the homological type @xmath501\in h_1(\mathbb{t}^n , u,\mathbb{z})$ ] . for a curve @xmath194 : @xmath106\to m$ ] with @xmath587 , @xmath588 , we denote by @xmath589 the curve in the lift of @xmath194 such that @xmath590 .
we say @xmath591=g$ ] if @xmath501=g$ ] holds for any closed curve @xmath495 such that @xmath592 and @xmath593 .
therefore , the following is well - defined : @xmath594=g}}{\scriptscriptstyle \xi\in c^1}}\int_{0}^{k}l(\xi(s),\dot\xi(s))ds + k\alpha,\ ] ] @xmath595 clearly , @xmath596 as @xmath566 .
indeed , let @xmath597 such that @xmath598 , let @xmath599 : @xmath496\to\bar m$ ] be a straight line such that @xmath600 , @xmath601 and denoted by @xmath280 the projection of @xmath599 down to @xmath61 , we obviously have that @xmath602\ ] ] holds for each class @xmath603 . as @xmath604
$ ] is a finite number , we verify the claim .
[ homopro2 ] there exists positive number @xmath23 such that @xmath605 holds for each @xmath606 and for large @xmath607 .
obviously , there exists a positive number @xmath608 such that @xmath609 holds for each @xmath569 and each @xmath190 .
if the proposition does not hold , for any small @xmath610 there would exists @xmath370 and @xmath611 such that @xmath612 let @xmath613 : @xmath614\to m$ ] be the minimizer of @xmath615 .
let @xmath616 be a sequence so that @xmath617 and @xmath618}]\ne 0 $ ] and there does not exist @xmath619 such that @xmath620 , both @xmath621}]\ne 0 $ ] and @xmath622}]\ne 0 $ ] .
there are two possibilities for this sequence .
either some @xmath623 exists such that @xmath624 as @xmath625 or @xmath626 remains bounded for all @xmath627 . in the first case ,
let @xmath628}^*\nu_i$ ] where @xmath629 is a probability measure evenly distributed on the interval @xmath630 $ ] . by weak@xmath631-compactness a probability measure @xmath122 exists such that @xmath632 .
clearly , @xmath122 is invariant for the lagrange flow , @xmath633 and the support of @xmath122 is not contained in the aubry set .
but it is absurd . in the second case
, one has @xmath634 . by choosing sufficiently small neighborhood @xmath440 of @xmath441
, the distance @xmath635 can be sufficiently small . as there is only one aubry class
, there is a closed curve @xmath280 and a sequence of time @xmath636 ( @xmath637 such that @xmath604<\epsilon$ ] and @xmath638 .
let @xmath639 be the minimizer connecting @xmath640 to @xmath641 , we have @xmath642-[a(\zeta|_{[t'_{i , j-1}-t'_{i , j}]})]|\le c'\epsilon$ ] , where the constant depends only on the lagrangian . by construction ,
the curve @xmath643}\ast\zeta'_j$ ] is a closed curve with @xmath644}\ast\zeta'_j]\ne 0 $ ] and @xmath645}\ast\zeta'_j)]>a'$ ] .
therefore , @xmath646-\epsilon\\ & \ge\sum_{j } [ a(\gamma_i|_{[t_{i , j}-t_{i , j-1}]}\ast\zeta'_j)]-(cc'\|g\|+1)\epsilon\\ & \ge c\|g\|a'-(cc'\|g\|+1)\epsilon.\end{aligned}\ ] ] it contradicts the assumption .
this proves the proposition in the case that @xmath647 . for @xmath577
, we use a straight line connecting @xmath353 to @xmath71 .
the action along this line is bounded .
therefore the proposition is also true for @xmath577 . for the configuration space @xmath583 ,
each weak kam solution is 1-periodic in @xmath648 for @xmath649 , where @xmath650 denotes the configuration coordinate .
if a finite covering of @xmath583 is considered to be configuration space , weak kam solution may not be 1-periodic for each coordinate .
we assume that the minimal measure contains finitely many ergodic components @xmath363 , @xmath385 for the cohomology class @xmath89 . in this case , the elementary weak kam solution for each @xmath252 is well - defined .
the lift of @xmath252 to a finite covering @xmath651 may contain several ergodic components .
for instance , if @xmath652 , then there are two ergodic components in the lift of @xmath653 for @xmath654 .
however , there are cases that the minimal measure is always uniquely ergodic for any finite covering manifold , for instance , if the measure is supported on a kam torus . given @xmath655 with @xmath656 for each @xmath657
, we define an equivalence relation @xmath658 in @xmath109 : we say @xmath659 if @xmath660 for some @xmath661 ( @xmath649 ) . clearly , @xmath662 : @xmath663 is a finite covering of @xmath583 . in the following , we shall also use the symbols : @xmath664 : @xmath665 and @xmath666 : @xmath667 to denote the projection . for a bounded domain @xmath668 ,
if the topology of @xmath669 is trivial , we use the same symbol to denote its projection @xmath670 .
let @xmath671 and @xmath672 be the lift of mather set @xmath673 to the universal covering space as well as to @xmath674 respectively .
the connected components are denoted by @xmath675 and @xmath676 correspondingly .
obviously , the unit cube @xmath677 intersects finitely many connected components of @xmath671 , denoted by @xmath675 with @xmath678 ( @xmath679 ) .
let @xmath680 .
some @xmath681 exists such that for any @xmath682 with @xmath683 , @xmath684 holds for @xmath685 and @xmath686 . in this case
, we use the notation @xmath687 for @xmath688 .
let @xmath689 denote the elementary weak kam for @xmath690 with respect to the configuration manifold @xmath674 .
[ globallem1 ] for each bounded region @xmath668 , there exists @xmath691 such that for any @xmath692 with @xmath693 , @xmath694 holds for each @xmath695 .
we only need to study the case that the minimal measure is uniquely ergodic .
if there are finitely many ergodic components , we obtain this result by perturbing the lagrangian so that it is uniquely ergodic and applying theorem [ weakthm1 ] . each weak kam solution for @xmath583
is a weak kam solution for any @xmath674 .
if the lift of the minimal measure to any finite covering space is still uniquely ergodic , the elementary weak kam solution remains the same .
let us consider the case that there are more than one connected component in @xmath696 with @xmath683 .
remember @xmath697 for @xmath335 where @xmath698 intersects the fundamental domain @xmath699 .
considered as a function defined in @xmath109 , the elementary weak kam solution @xmath700 determined by @xmath701 is @xmath702-periodic in the @xmath349-th coordinate . by the definition of elementary weak kam solution , a sequence of functions @xmath703 exists such that @xmath704 as @xmath378 , where @xmath703 is the weak kam solution for the lagrangian @xmath705 .
this lagrangian satisfies the following conditions : 1 , it is the same as @xmath63 when it is restricted on the tangent bundle of a neighborhood @xmath440 of @xmath690 , i.e. @xmath706 ; 2 , the minimal measure is uniquely ergodic whenever @xmath707 , supported on @xmath690 ; 3 , @xmath708 as @xmath378 . starting from each @xmath709 ,
there exists at least one backward semi - static curve for @xmath710 , @xmath711 : @xmath404 $ ] with @xmath712 . clearly , @xmath713 .
let @xmath714 be the sequence so that @xmath715 , let @xmath95 stand for the average action , then we have @xmath716 again , the lift of @xmath717 to the universal covering space may contain many connected components , denoted by @xmath718 , among which only @xmath719 intersects the fundamental domain @xmath720 . let @xmath721 so that @xmath664 : @xmath722 is an injection and @xmath723 .
let @xmath724 be the points such that @xmath725 .
let @xmath726 be the lift of @xmath727 to the universal covering space so that @xmath728 .
it is possible that @xmath729 .
the curve may approach to another connected component of @xmath718 .
let @xmath730 .
note that @xmath710 is a small perturbation of @xmath63 . in virtue of proposition [ homopro2 ]
we claim that @xmath726 approaches to @xmath719 provided @xmath731 , @xmath289 is suitably small and @xmath732 is sufficiently large .
let us assume the contrary , i.e. @xmath733 approaches to another connected component of @xmath671 .
in this case , @xmath734\|$ ] would be sufficiently large provided @xmath732 is sufficiently large . by proposition [ homopro2 ] the action of @xmath63 along @xmath735 @xmath736\|a\ ] ] with certain @xmath23 . as @xmath710 is a small perturbation of @xmath63
, the action of @xmath710 along @xmath735 would approach infinity as @xmath737 .
the absurdity verifies the claim .
the set @xmath738 is compact as @xmath378 . for each accumulation point
@xmath524 , there is a subsequence of @xmath378 such that @xmath739 .
the initial value @xmath556 uniquely determines an orbit @xmath740 of @xmath63 .
the curve @xmath741 : @xmath404\to m_k$ ] is a backward semi - static curve for @xmath63 which may not approach to @xmath742 .
when @xmath378 , @xmath727 may approach not only one but a family of semi - static curves for @xmath63 including the curves connecting different connected components of @xmath696 .
more precisely , there might be several connected components @xmath743 and semi - static curves @xmath744 of @xmath63 for @xmath674 ( @xmath745 ) such that @xmath746 , @xmath747 , @xmath748 and each @xmath744 is approached by @xmath727 as @xmath378 .
these curves have their natural projection down to @xmath583 , denoted by the same symbol .
we define the quantity @xmath749 : @xmath750 @xmath751=\inf_{\stackrel{\stackrel{\gamma(0)=x_i}{\scriptscriptstyle \gamma(k)=x_j}}{\scriptscriptstyle k\in\mathbb{z}_+}}\int^{k}_{0 } l(d\gamma(s))ds+k\alpha.\ ] ] by definition of weak kam , for almost every point @xmath71 , @xmath752 uniquely determines a backward semi - static curve @xmath727 . since this semi - static curve approaches to several curves : @xmath753 we obtain that @xmath754\notag\end{aligned}\ ] ] where @xmath714 is a sequence such that @xmath755 , @xmath756 . for each @xmath757 ,
the backward semi - static curve @xmath727 approaches to @xmath690 provided @xmath732 is sufficiently large . for different @xmath758 satisfying this condition , @xmath727 and @xmath759 may converge to different curves , @xmath760 and @xmath761 as @xmath378 .
but the action of @xmath63 along @xmath762 is the same as along @xmath763 .
indeed , the action of @xmath710 along @xmath764 is almost the same as the action of @xmath765 along @xmath766 provided the perturbation is sufficiently small .
therefore , we obtain from the formula [ globaleq3 ] that @xmath767 if both @xmath732 and @xmath768 are sufficiently large .
the function @xmath769 is called globally elementary weak kam solution for @xmath770 if for each bounded domain @xmath668 , there exists @xmath691 such that for any @xmath682 with @xmath771 , @xmath772 holds for each elementary weak kam solution @xmath773 : @xmath774 for @xmath690 . from lemma [ globallem1 ]
, we obtain the existence of a globally elementary weak kam solution for each @xmath698 .
to investigate the properties of globally elementary weak kam solution , let us consider a special case first , namely , the mather set contains a connected component homeomorphic to @xmath516 . in this case , each @xmath775 divided @xmath109 into two parts , denoted by @xmath776 and @xmath777 .
[ globalthm1 ] if the mather set contains a connected component contains a @xmath675 homeomorphic to @xmath516 , then the globally elementary weak kam solution @xmath778 : @xmath779 has a decomposition @xmath780 where @xmath781 is periodic and @xmath782 is affine when they are restricted in the half space @xmath777 as well as in another half space @xmath776 . we only need to consider the case that the minimal measure is uniquely ergodic , as we did in the proof of lemma [ globallem1 ] . according to theorem [ homothm2 ] , there are exactly two types of minimal homoclnic orbits to the aubry set , we pick up two representative elements @xmath783 , @xmath784 : @xmath785 .
let @xmath786 where @xmath787 is chosen such that @xmath788 and @xmath789 . as the set @xmath775 is co - dimension one , we are able to number all connected components by @xmath790 ( @xmath791 ) such that any path from @xmath792 to @xmath793 must pass through @xmath790 .
denote by @xmath794 the strip bounded by @xmath790 and @xmath793 .
@xmath795 separates @xmath109 into two parts , denoted by @xmath796 and @xmath797 such that @xmath798 and @xmath799 .
let @xmath800 denote a curve in the lift of @xmath801 to @xmath109 such that @xmath802 .
then , either @xmath803 , @xmath804 , or @xmath805 , @xmath806 .
we only need to study one case , let s say , the first case . given a bounded domain @xmath668 . from the definition of globally elementary weak kam solution
, we see that @xmath807 whenever @xmath732 is suitably large .
clearly , the function @xmath808 is periodic when it is restricted @xmath809 , i.e. @xmath810 if @xmath811 and @xmath812 . for each @xmath813 with @xmath814
, there exists at least one point @xmath815 such that @xmath816 . by definition ,
we find that @xmath817 .
obviously , @xmath818 if @xmath813 with @xmath819
. pick up a point @xmath820 . for each non - zero integer vector @xmath682 ,
the point @xmath821 stays in certain @xmath675 . along the ray @xmath822 with @xmath823
, we define @xmath824 clearly , @xmath825 is affine and @xmath826 is periodic when they are restricted in @xmath796 as well as in @xmath797 . given an ergodic component of a minimal measure with higher co - dimensions , it is unclear what condition guarantees the decomposition of the globally elementary weak kam solutions .
it appears closely related to the problem whether there are infinitely many types of minimal homoclinic orbits to the aubry class .
[ globalpro1 ] let @xmath827 be the globally elementary weak kam solution for @xmath775 .
then , @xmath827 remains bounded on the whole ray @xmath828 for each @xmath829 , where @xmath830 is an abury class ; for @xmath831 , @xmath827 grows up linearly , or asymptotically linearly on the ray @xmath828 . for arbitrarily large @xmath832 , there exists @xmath833 such that @xmath834 and @xmath835 where @xmath720 is the unit cube containing the origin .
let @xmath836 , @xmath195 be a curve connecting @xmath837 to @xmath71 , @xmath838 , then @xmath591\in h_1(\mathcal{a},\mathbb{z})$ ] . by definition ,
@xmath839\in h_1(\mathcal{a},\mathbb{z})}\inf_{\stackrel{\xi(0)=\xi(k)}{\scriptscriptstyle k\in\mathbb{z}_+}}\int_0^k l(d\xi(t),t)dt=0\ ] ] it proves the first conclusion . for the second
, one can see from proposition [ homopro2 ] that it grows up at least linearly .
given @xmath840 finitely many elements @xmath841 exists such that for each @xmath842 with @xmath843 .
thus , @xmath844 for each @xmath845 , @xmath846=g$ ] , we have @xmath847 this completes the proof .
given a tonelli lagrangian @xmath63 : @xmath848 , let @xmath849 .
any minimal measure with zero - rotation vector must be @xmath850-minimal measure . in this section
we study the dynamics around the mather set for the class @xmath850 .
the motivation comes from following argument .
let us consider the normal form of a nearly integrable hamiltonian @xmath851 around a complete resonant point .
let @xmath852 denote the frequency vector of the unperturbed system .
a frequency @xmath127 is called complete resonant of ( minimal ) period @xmath853 if @xmath854 and @xmath855 for each @xmath856 . by
finitely many steps of kam iteration and one step of linear coordinate transformation on torus , one obtains a normal form of nearly integrable hamiltonian ( see appendix a ) @xmath857 where @xmath858 , @xmath859 , @xmath860 , @xmath861 is well - defined in @xmath862 , @xmath863 and @xmath864 is a higher order term .
since @xmath865 , there exists some function @xmath866 solving the equation @xmath867 provided @xmath868 , which defines a time - periodic hamiltonian system with @xmath869-degrees of freedom . here
@xmath870 plays the role of time .
one can write @xmath871 where @xmath872 is a higher order term of @xmath289 and @xmath873 , i.e. the complete resonance reduces to zero frequency . omitting the higher order term , one obtains hamiltonian with @xmath874 degrees of freedom @xmath875 it determines a lagrangian we shall study in this section . by definition , a subset
is called a flat of certain @xmath95-function if , restricted on this set , the @xmath95-function is affine , and no longer affine on any set properly containing the flat . as @xmath95-function is convex with super - linear growth , each flat is a convex and bounded set . given an @xmath5-dimensional flat @xmath455 , a subset in @xmath876 is called an edge if it is contained in a @xmath877-dimensional hyperplane .
since each flat is convex , each edge is also convex .
[ flatthm1 ] given a class @xmath878 , if the minimal measure is uniquely ergodic , supported on a hyperbolic fixed point , then there exists an @xmath5-dimensional flat @xmath879 such that this point supports a @xmath89-minimal measure for all @xmath880 .
* remark * : the condition of this theorem does not exclude topological non - triviality of the aubry set .
an example is the product of @xmath5 pendulums .
the aubry set covers the whole torus @xmath583 if the lagrangian @xmath63 is replaced by @xmath881 with @xmath89 being on the boundary of the flat . by translation
one can assume that the fixed point is at @xmath882 , by adding a closed 1-form and a constant to the lagrangian , one can assume @xmath883 and @xmath884 .
to each closed curve @xmath194 : @xmath885\to\mathbb{t}^n$ ] with @xmath886 a first homology class @xmath591=g\in h(\mathbb{t}^n,\mathbb{z})$ ] is associated .
we consider the quantity @xmath887=g } } \int_{-t}^tl(d\xi(t))dt.\ ] ] by the condition assumed on @xmath63 , one has that @xmath888 for any @xmath569 .
there exist at least @xmath176 irreducible classes @xmath889 and @xmath176 minimal homoclinic orbits @xmath890 such that @xmath891)=a(\gamma_i)$ ] @xcite .
clearly , @xmath892 can be generated by the homology classes of all minimal homoclinic curves over @xmath893 .
we abuse the notation @xmath603 to denote homology class @xmath894 or to denote a point @xmath895 .
for each curve @xmath896 : @xmath885\to\mathbb{r}^n$ ] with @xmath897 and @xmath898 , one has @xmath899 recall the definition of globally elementary weak - kam and note that the point @xmath900 is the support of the minimal measure .
let @xmath901 ( @xmath902 ) denote the backward ( forward ) globally elementary weak - kam for @xmath903 , we have @xmath904 by setting @xmath905 , we claim @xmath906 the quantity @xmath907 is achieved may not by a curve connecting the origin to @xmath895 , but may by the conjunction of several curves @xmath908 . let @xmath909 : @xmath910 denote a curve ( @xmath911 ) , the conjunction implies that @xmath912 .
let @xmath913 , where @xmath914 denotes the standard projection . in this case , @xmath915 are minimal homoclinic curves such that @xmath916 $ ] . by the definition of elementary weak - kam ,
each @xmath909 is a @xmath917-calibrated curve .
let @xmath918 $ ] .
obviously , some large @xmath919 exists such that @xmath920 stays in the local stable manifold of the point @xmath921 whenever @xmath922 .
therefore , some constant @xmath923 exists such that @xmath924 where we use @xmath925 and @xmath926 to denote the globally elementary - kam based on @xmath927 . clearly , @xmath926 and @xmath925 generate the local stable and unstable manifold around the point @xmath921 respectively .
because that @xmath928 is @xmath63-dominate function , for @xmath929 with suitably small @xmath54 we have ( see theorem 5.1.2 in @xcite ) @xmath930 remember that @xmath931 . if @xmath932 substituting @xmath926 in ( [ flateq2.1 ] ) by the expression in ( [ flateq2 ] ) we see that some @xmath933 exists so that @xmath934 here , @xmath933 is chosen so that @xmath935 for @xmath922 . since @xmath902 is an @xmath63-dominate function and
@xmath909 is a @xmath917-calibrated curve for each @xmath936 , @xmath937 hold for any @xmath938 .
this induces that @xmath939 holds for all @xmath46 , and induces in turn the inequality for @xmath940 , and
finally we have @xmath941 since @xmath942 holds in a small neighborhood of @xmath360 @xmath943 on the other hand , as the fixed point @xmath944 is hyperbolic , some @xmath54 exists such that
@xmath945 if we set @xmath946 .
this contradiction proves the formula ( [ flateq1 ] ) .
let @xmath947=g,\ a(\gamma)=0\}.\ ] ] @xmath948 is said to generate a rational direction @xmath895 over @xmath893 if there exist @xmath949 and @xmath950 such that @xmath951 it is an immediate consequence of the formula ( [ flateq1 ] ) that once @xmath948 generates a rational direction @xmath895 over @xmath893 , then it can not generate the direction @xmath952 over @xmath893 .
therefore , the set @xmath953 is a cone properly restricted in half space .
thus , there exists an @xmath5-dimensional cone @xmath954 such that @xmath955 since the minimal measure for zero cohomology class is supported on the fixed point , @xmath956 is composed of those minimal homoclinic orbits along which the action equals zero . according to the upper semi - continuity of ma set in cohomology class , any minimal measure @xmath90
is supported by a set lying in a small neighborhood of these homoclinic orbits if @xmath957 is very small . consequently .
we have @xmath958 , where @xmath959 denotes the rotation vector of @xmath90 .
let us consider a cohomology class @xmath89 such that @xmath960 and @xmath961 .
we claim that the @xmath89-minimal measure is also supported on the fixed point . indeed ,
if it is not true , we would have positive average action of @xmath63 : @xmath962 , since the minimal measure for zero class is assumed unique and supported on the fixed point . by the choice of @xmath89 one
has that @xmath963 .
thus , one obtains @xmath964 it deduces absurdity . for this class @xmath89 ,
the action of the lagrangian @xmath965 along any minimal homoclinic curve @xmath75 is positive , @xmath966\rangle > 0,\ ] ] namely , the aubry set for this class is also a singleton .
consequently , @xmath967 is also supported on this point if @xmath968 is sufficiently close to @xmath89 .
this verifies the existence of @xmath5-dimensional flat .
* eigenvalues of the fixed point * let us consider the eigenvalues of the fixed point by assuming the hyperbolicity , denoted by @xmath969 @xmath970 . under the hyperbolic assumption ,
half of these have positive real part , other half have negative real part . in general , these eigenvalues may have non - zero imaginary part .
but in nearly integrable systems , all eigenvalues are real .
[ flatpro1 ] if the lagrangian is a small perturbation of integrable one @xmath971 where @xmath972 is positive definite in @xmath973 , then for generic @xmath14 and for sufficiently small @xmath289 , all eigenvalues at the fixed point are real and different .
let @xmath974 , @xmath975 , @xmath976 evaluated at the fixed point . as the minimal measure is supported on a fixed point , @xmath977 is positive definite .
we consider the linearized equation and assume the solution with the form of @xmath978 , then @xmath979 let @xmath980 , evaluated at the fixed point . for generic @xmath977 , all solutions of the equation @xmath981
are real and different from each other : @xmath982 , @xmath983 if @xmath686 . since ( [ flateq3 ] ) is a small perturbation of ( [ flateq4 ] ) , all solutions of ( [ flateq3 ] ) are different , and consequently , real . if there was a complex solution @xmath984 , @xmath985 would be solution also , which is guaranteed by the hamiltonian structure . it implies the existence of more than @xmath986 solutions , but it is absurd . * the shape of the flat * here
, we are concerned about the flat @xmath987 .
it is a @xmath5-dimensional flat if the @xmath89-minimal measure is supported on the hyperbolic fixed point for each @xmath988 . by coordinate translation
, we assume it is at the origin : @xmath989 . correspondingly , in canonical coordinates the fixed point is also at the origin @xmath990 .
let @xmath991 denote the eigenvector for @xmath992 , where @xmath993 is for the @xmath71-coordinates , @xmath994 is for the @xmath29-coordinates .
we assume 1 , all eigenvalues are real number and different ; 2 , all minimal homoclinic curves approach to the fixed point in the direction @xmath995 as @xmath996 .
the condition 1 is obviously generic . to see the genericity of the condition 2 ,
let us remind reader that there is , generically , at most one minimal homoclinic curve for each homology class . by further perturbation
, it approaches to the fixed point in the direction of @xmath995 .
since there are countably many homology classes at most , the genericity is obtained . for @xmath997 and @xmath998
, we define a cone @xmath999 and let @xmath1000 [ flatpro2 ] assume that @xmath1001 is a hyperbolic fixed point for @xmath40 , where all eigenvalues are real and different : @xmath1002 let @xmath991 denote the eigenvector for @xmath992 , where @xmath993 is for the @xmath71-coordinates , @xmath994 is for the @xmath29-coordinates .
let @xmath1003 be an orbit such that @xmath1004 passes through a ball @xmath1005 , @xmath1006 , @xmath1007 and @xmath1008 for all @xmath1009 .
then , for suitably small @xmath54 and @xmath1010 , there exist sufficiently large @xmath1011 such that for @xmath1012 one has @xmath1013 for @xmath1014 , one has @xmath1015 for certain @xmath1016 . by certain symplectic coordinate transformation
, the hamiltonian is assumed to have the normal form @xmath1017 where @xmath1018 is a higher order term . by the method of variation of constants , we obtain the solution of the corresponding hamilton equation @xmath1019 where @xmath1020 are constants determined by boundary condition and @xmath1021 substituting @xmath42 with the formula ( [ flateq5 ] ) in the hamiltonian we obtain a constraint for the constants @xmath1020 : @xmath1022 let us estimate the size of the constants @xmath1023 by the boundary conditions @xmath1024 , @xmath1025 and assuming @xmath1026 for @xmath1027 , @xmath1028 implies ( [ flateq7 ] ) holds .
since the curve @xmath1029}$ ] stays inside of the ball @xmath1030 and @xmath853 is sufficiently large , the orbit @xmath1031}$ ] stays near the stable and unstable manifold of the fixed point .
note @xmath1032 , we obtain from the theorem of grobman - hartman that @xmath1033 for sufficiently large @xmath190 , it deduces from the assumption ( [ flateq7 ] ) that @xmath1034 and @xmath1035 since @xmath1036 for each @xmath1037 , @xmath1038 if @xmath853 is sufficiently large . in this case , we obtain from ( [ flateq6 ] ) that @xmath1039 it contradicts the assumption that @xmath1040 .
let @xmath1014 , one easily sees the last conclusion .
this proposition tells us following fact . in the energy level @xmath1041 there
does not exist such an orbit passing through @xmath1030 in the way that it enters into the ball in a direction close to @xmath1042 and leaves in a direction close to @xmath1043 .
[ flatthm2 ] let @xmath987 be an @xmath5-dimensional flat of the @xmath95-function .
each minimal homoclinic curve @xmath75 is assumed approaching to the fixed point in the direction of the eigenvectors corresponding to the smallest eigenvalue @xmath1044 all eigenvalues are assumed real and different .
it is also assumed that , for each @xmath880 @xmath139including the boundary@xmath140 , the minimal measure is uniquely supported on the hyperbolic fixed point .
then , there exists a finite set @xmath1045 such that @xmath454\in h_{\mathbb{f}_0}$ ] if @xmath75 is a minimal homoclinic curve .
consequently , the flat @xmath1046 is a polygon with finitely many edges , denoted by @xmath1047 .
each edge @xmath1048 is associated with a homological class @xmath370 such that @xmath163 is composed by minimal homoclinic curves with homological type @xmath370 if @xmath89 is in the interior of @xmath1048 and @xmath1049 as the minimal measure is uniquely ergodic and supported on a point for each class in @xmath1046 , the ma set consists of homoclinic orbits and the point itself . for each class @xmath1050 , we claim that the ma set contains at least one homoclinic orbit .
otherwise , for each class @xmath1051 very close to @xmath89 , the homology of the ma set is trivial , the same as that for @xmath89 .
it is guaranteed by the upper semi - continuity of ma set in cohomology class .
it follows that @xmath1052 and @xmath1053 however , as @xmath1051 , one has @xmath1054 .
the contradiction verifies our claim
. approached by minimal periodic curves , each minimal homoclinic curve @xmath75 stays in certain aubry set : @xmath1055)\subset \mathbb{f}_0.\ ] ] where the limit is in the sense of hausdorff .
if @xmath1056 contains infinitely many elements , there would be infinitely many minimal homoclinic curves @xmath1057 such that @xmath1058\neq [ \gamma_j]$ ] provided @xmath1059 .
thus we have two possibilities . 1
, a neighborhood @xmath1060 of the fixed point exists such that each minimal homoclinic curve @xmath75 hits the sphere @xmath1061 exactly twice , i.e. @xmath1062 @xmath1063 such that @xmath1064 for all @xmath1065\cup [ t^+,\infty)$ ] and @xmath1066 for all @xmath1067 ; 2 , for any small @xmath1068 , there are infinitely many minimal homoclinic curves @xmath1069 passing through the sphere @xmath1061 in finite time , i.e. @xmath1062 @xmath1070 such that @xmath1064 for all @xmath1065\cup[t^-_0,t_0^+]\cup [ t^+,\infty)$ ] and @xmath1066 holds for some @xmath1071 as well as for some @xmath1072 .
let us study the first possibility .
denote by @xmath1073 the time when the minimal homoclinic curve @xmath613 hits the sphere @xmath1061 . for each @xmath613 , there is a segment @xmath1074 staying outside of @xmath1060 .
each @xmath1075 generates a probability measure @xmath1076 on @xmath1077 such that @xmath1078 holds for each continuous function @xmath1079 : @xmath848 .
as all these curves have different homology class , @xmath1080\|\to\infty$ ] as @xmath625 . as the speed along these curves
are uniformly bounded , we have @xmath1081 let @xmath1082 be the class such that @xmath1083 and let @xmath1084 be an accumulation point of @xmath361 , some invariant probability measure @xmath1085 exists such that @xmath1086 , it is @xmath1087-minimal .
clearly , @xmath1085 is not supported on the fixed point , it contradicts the assumption that the minimal measure is always uniquely ergodic for each @xmath880 .
let us study the second possibility . in this case , for suitably small @xmath54 , there is an infinite sequence of homoclinic curves @xmath613 and correspondingly the sequence of time @xmath1073 such that @xmath1088 for each @xmath1089 $ ] , @xmath1090 and @xmath1091 as @xmath625 .
indeed , if @xmath1092 remains bounded , one can choose @xmath1093 such that these curves hit the sphere @xmath1094 twice only .
it is the first case again . by using proposition [ flatpro2 ] , we find that some @xmath997 exists such that one of the inequalities in the following holds for each @xmath349 @xmath1095 provided @xmath349 is sufficiently large .
it implies that there exists some minimal homoclinic curve @xmath75 as well as some eigenvector @xmath1096 or @xmath1097 with @xmath1098 and @xmath1099 such that at least one of the following holds @xmath1100 this leads to a contradiction to the assumption , then verifies the finiteness of @xmath1056 .
let @xmath1101 be two minimal homoclinic curves contained in the aubry set @xmath163 , @xmath1102 respectively .
let @xmath1103\}$ ] . if @xmath1104 intersects the interior of @xmath1046 , then
@xmath454\neq [ \gamma']$ ] .
indeed , by definition we have @xmath1105\rangle&=0 , \qquad a(\gamma')-\langle c,[\gamma']\rangle \ge 0;\\ a(\gamma')-\langle c',[\gamma']\rangle&=0 ,
\qquad a(\gamma)-\langle c',[\gamma]\rangle \ge 0,\end{aligned}\ ] ] it follows from @xmath454=[\gamma']$ ] that @xmath1106 . consequently , @xmath1107\rangle)+(1-\xi)(a(\gamma')-\langle c',[\gamma']\rangle ) \\ & = a(\gamma)-\langle\xi c+(1-\xi)c',[\gamma]\rangle\\ & = a(\gamma')-\langle\xi c+(1-\xi)c',[\gamma']\rangle.\end{aligned}\ ] ] it implies that both @xmath75 and @xmath545 lie in the aubry set for @xmath1108 . on the other hand ,
the aubry set for each class in the interior of @xmath1046 contains the fixed point only .
the contradiction implies that @xmath454\neq [ \gamma']$ ] .
therefore , @xmath1046 is a polygon with exactly @xmath393 edges , each edge corresponds to one homology type of minimal homoclinic curve .
let @xmath75 be a minimal homoclinic curve lying in the aubry set for @xmath1109 .
then , one has @xmath1110\rangle=0 $ ] .
as the aubry set remains the same for all classes in the interior of the edge , one has @xmath1111\rangle=0 $ ] for each @xmath1109 .
consequently , one has @xmath1112\rangle = 0 $ ] for all @xmath1113 . as it is @xmath877-dimensional , each edge
@xmath1048 determines a unique homology class @xmath370 such that @xmath454=g_i$ ] if @xmath75 is a minimal homoclinic curve lying in the aubry set .
let @xmath1114 be a minimal homoclinic curve approaching to the fixed point in the direction of @xmath1115 corresponding to the smallest eigenvalue @xmath1116 , let @xmath1117 be a edge of @xmath1046 .
in this subsection we assume that 1 , for each @xmath880 , the mather set contains exactly one fixed point @xmath882 ; 2 , for each @xmath1118 , the aubry set consists of the fixed point and one minimal homoclinic curve @xmath1114 : @xmath1119 ; 3 , there exist a sequence of positive numbers @xmath1120 and sequence of ergodic minimal measure @xmath1121 such that @xmath1122 $ ] , and @xmath1123 for each @xmath1124)$ ] .
by definition , there exists an elementary weak kam for each @xmath1121 corresponding to the energy @xmath1125))$ ] .
the main purpose of this section is to study the modulus of continuity of some functions in terms of energy at @xmath1126 .
* dependence of the average speed on energy*. according to birkhoff s ergodic theorem , there is an orbit @xmath1127 : @xmath785 of @xmath128 in the mather set such that @xmath1128})\to a(\mu_i)\ \
\text{\rm and}\ \
\frac 1{2t}(\bar\zeta_i(t)-\bar\zeta_i(-t))\to\rho(\mu_i ) \ \ \
\text{\rm as}\ t\to\infty,\ ] ] where @xmath1129 stands for a lift of @xmath1130 to the universal covering space . by the upper semi - continuity of ma set
, the curve @xmath1130 passes through the ball @xmath1030 infinitely many times if @xmath629 is small .
denoted by @xmath1131 and @xmath1132 the time when @xmath1133 enters and leaves the ball respectively , i.e @xmath1134 for each @xmath1135 $ ] , @xmath1136 .
clearly , @xmath1137 if @xmath1130 is a periodic curve , some @xmath1138 exists so that @xmath1139 holds for all @xmath1140 .
let @xmath1141 such that the minimal homoclinic curve @xmath1114 enters @xmath1030 at @xmath1142 and leaves @xmath1030 at @xmath1143 . by the upper semi - continuity of ma set one
has @xmath1144 as each minimal homoclinic curve approaches to the fixed point in the direction @xmath995 , @xmath1145 holds if @xmath1146 is suitably small and @xmath1147 is suitably large .
each segment @xmath1148}$ ] solves the hamilton equation . in the coordinates of normal form
, it is given by eq .
( [ flateq7.1 ] ) and the integral constants @xmath1020 satisfy the constraint ( [ flateq6 ] ) . the condition ( [ regularenergyeq1 ] ) induces the following @xmath1149 for small @xmath1146 and sufficiently large @xmath1150 , we have @xmath1151 @xmath1152 where the constant @xmath977 depends only on the function @xmath1153 .
so , for suitably small @xmath1146 and sufficiently large @xmath1150 , we obtain from ( [ flateq6 ] ) that @xmath1154 under the same condition , @xmath1155 is obviously upper bounded by @xmath1156 therefore , we find the dependence of speed on the energy @xmath1157 is uniformly bounded for each @xmath1140 : @xmath1158 obviously , @xmath1159 as @xmath625 , some @xmath1160 exists such that @xmath1161 set @xmath1162 , one has @xmath1163 recall the meaning of @xmath629 : @xmath1164 $ ] is the rotation vector of the minimal measure@xmath1076 . in this section
we restrict ourselves to the special case that the system has two degrees of freedom : @xmath1165 .
the task of this section is to study the structure of the mather sets as well as of the ma sets in a neighborhood of the resonant point . under the coordinate transformation ( [ normaleq10 ] )
, it corresponds to a fixed point . since each aubry set is a lipschitz graph over the configuration manifold which is two dimensional here , each orbit in an aubry set has to be _
parallel _ to any other orbit in the same set in the sense that these curves do not intersect each other . on the other hand , for autonomous system , @xmath1166 , regarded as the function of @xmath1167 , is differentiable at each @xmath1168 ( see @xcite ) .
thus , we have [ flatpro3 ] assume that @xmath63 is an autonomous tonelli lagrangian defined on @xmath1169 .
for each non - zero rational vector , the mather set consists of periodic orbits with the same rotation vector .
each minimal measure with zero - rotation corresponds to the minimum of the @xmath95-function .
there are two _ nondegenerate _
cases for the set of minimal point @xmath1170 .
we call a case _ nondegenerate _ if it persists under small perturbation . 1
, @xmath1046 is a two - dimensional flat .
typically , for each class in the interior of @xmath1046 , the minimal measure is supported on a fixed point , or a shrinkable periodic orbit @xmath1171 , i.e. @xmath454=0 $ ] . the fixed point ( periodic orbit )
is of hyperbolic type . 2 ,
@xmath1046 is one - dimensional .
typically , the minimal measure is supported on two periodic orbits @xmath1172 and
@xmath1173 with the property : @xmath1174/\|[\gamma_-]\|=-[\gamma_+]/ \|[\gamma_+]\|.\ ] ] the set @xmath1046 is a singleton only when the @xmath95-function is differentiable at this point . otherwise , the @xmath98 function can not have a two - dimensional flat . in this case , the mather set contains two circles with different rotation direction , but it violates the lipschitz property .
let us study the first case and assume that @xmath1175 , the point @xmath882 supports the minimal measure .
then , @xmath1176 for all @xmath1177 .
the study is similar if it is supported on a shrinkable closed orbit .
given the @xmath95 as well as the @xmath98-function , let us recall the fenchel - legendre transformation @xmath129 : @xmath130 is defined as @xmath1178 let @xmath1179 it may be non - empty . here is an example : @xmath1180 where @xmath1181 , the potential satisfies the following conditions : @xmath900 is the minimal point of @xmath1182 only ; there exist two numbers @xmath1183 such that for any closed curve @xmath75 : @xmath496\to\mathbb{t}^2 $ ] passing through the origin with @xmath454\ne 0 $ ] one has @xmath1184 @xmath1185 when it is restricted a neighborhood of circle @xmath1186 with @xmath1187 mod 1 . in this case , @xmath1188 .
indeed , @xmath1189 the mather set for @xmath1190 consists of the point @xmath900 and the periodic curve @xmath1191 . clearly , the set @xmath1192 is closed with respect to @xmath1046 .
if it is non - empty , the existence of infinitely many @xmath178-minimal homoclinic orbits has been proved in @xcite .
these orbits are associated with different homological classes .
if @xmath1193 , there are at least three minimal homoclinic orbits to the fixed point . the existence of homoclinic orbit to some aubry set is closely related to the existence of the flat of the @xmath95-function . [ flatlem1 ] given @xmath1194 , let @xmath1195 . then @xmath1196 using argument in @xcite , for any curve @xmath75 : @xmath76 , we have @xmath1197=\lambda [ a_c(\gamma|_{i})]+(1-\lambda)[a_{c'}(\gamma|_{i})],\qquad \forall\ i\subset\mathbb{r}.\ ] ] as both @xmath1198 and @xmath1199 , one has that @xmath1200=[a_{c'}(\gamma)]=0 $ ] if @xmath1201=0 $ ] . [ flatlem2 ]
let @xmath1046 be a 2-dimensional flat , the mather set is a singleton for each class in the interior of @xmath1046 , let @xmath1048 be an edge of @xmath1046 , then @xmath1202 holds for @xmath1203 @xmath1204 and @xmath1205 @xmath1206 respectively .
as the mather set is a singleton for each @xmath988 , each orbit in the aubry set is either the fixed point itself , or a homoclinic orbit to the point with zero first homology .
indeed , let @xmath454 $ ] denote its first homology of the homoclinic curve @xmath75 in the aubry set , then @xmath1207\rangle=0\ ] ] holds for each @xmath988 .
it follows that @xmath1112\rangle=0 $ ] for @xmath1208 .
since @xmath1046 shares the same dimension of the configuration space , @xmath454=0 $ ] .
in fact , for classical mechanical system , the aubry set consists of the fixed point only for @xmath988 . if @xmath1209 , as shown in the proof of theorem [ flatthm2 ] , the aubry set @xmath1102 contains at least one minimal homoclinic curve with non - zero first homology .
if @xmath1210 , the certain @xmath968-minimal measure @xmath967 exists with @xmath1211 . in both cases , @xmath1212 if @xmath988 .
let @xmath1048 be an edge . for @xmath1109 ,
the aubry set contains one or more homoclinic curves , all of them share the same homology class , denoted by @xmath1213 which is of course non - zero .
if @xmath164 contains other curves , these curves also share the same rotation vector as @xmath1214 holds for @xmath1113 .
let @xmath1215 and @xmath1109 , one chooses @xmath1216 arbitrarily close to @xmath968 . as the straight line connecting @xmath89 to @xmath1087 passes through the interior of @xmath1217 , we obtain from lemma [ flatlem1 ] that @xmath1218 with @xmath1219 .
for any curve @xmath280 contained in @xmath1220 , it follows from the formulation @xmath1221\rangle\ ] ] that @xmath1222\rangle\neq0 $ ] holds .
we claim @xmath1223\neq g(\mathbb{e}_i)$ ] .
let us assume the contrary and consider the case that @xmath280 is a homoclinic curve and @xmath163 contains a homocilinic curve @xmath75 . in this case ,
by assuming that @xmath1224 for @xmath880 , we have @xmath1225\rangle=0 , \qquad \int_{-\infty}^{\infty } l(d\gamma)dt -\langle c , g(\mathbb{e}_i)\rangle=0.\ ] ] since the class @xmath1087 is not on the straight line containing @xmath1048 , we have @xmath1226 . if @xmath1227 we would have @xmath1228\rangle=\int_{-\infty}^{\infty } l(d\gamma)dt-\langle c,[\gamma]\rangle- \langle c^*-c , g(\mathbb{e}_i)\rangle<0\ ] ] if @xmath1223=g(\mathbb{e}_i)$ ] and @xmath1229 we would have @xmath1230\rangle=\int_{-\infty}^{\infty } l(d\gamma)dt-\langle c^*,[\zeta]\rangle+ \langle c^*-c , g(\mathbb{e}_i)\rangle<0\ ] ] both cases are absurd as @xmath1231 .
because @xmath1223\neq g(\mathbb{e}_i)$ ] , some @xmath1232 remains far away from @xmath163 .
let @xmath1233 , the accumulation point of these points does not fall into @xmath163 , it implies @xmath1212 .
the proof is similar if @xmath194 as well as @xmath75 is a curve lying in the mather set .
recall the definition of @xmath513 in the section 2 : a first homology class @xmath512 if and only if there exists a minimal homoclinic orbit @xmath462 such that @xmath454=g$ ] .
let @xmath1234 be defined such that @xmath1235 if and only if there exists a minimal homoclinic orbit @xmath462 in @xmath161 such that @xmath454=g$ ] .
we say that there are @xmath77-types of minimal homoclinic orbits in @xmath161 if @xmath1236 contains exactly @xmath77 elements . for an edge
we define @xmath1237 for each @xmath1238 , from the proof of lemma [ flatlem2 ] one can see that it makes sense .
[ flatthm3 ] let @xmath1046 be a two dimensional flat , @xmath1239 is a singleton for @xmath1240 .
let @xmath1048 denote an edge of @xmath1046 ( not a point ) , then 1 , either @xmath1241 or @xmath1242 ; 2 , if @xmath1241 , then @xmath1243 contains exactly one element , if @xmath1242 , all curves in @xmath1244 have the same rotation vector ; 3 , if @xmath1245 and @xmath1246 then @xmath1236 contains exactly two elements ; 4 , if @xmath1247 , then either @xmath1048 and @xmath1248 are disjoint , or @xmath1249 ; 5 , if @xmath1242 , @xmath1250 holds for @xmath1251 and @xmath1252 . for the conclusion 1 , as @xmath1253 if @xmath1113 @xcite , we only need to consider @xmath1245 .
if it is not true , there would exist an invariant measure @xmath90 , not supported on the singleton and minimizing the action @xmath1254 but not minimizing the @xmath968-action for @xmath1255 . as the configuration space is @xmath1169 , the lipschitz graph property of aubry set
will be violated if the rotation vector of the measure @xmath959 is not parallel to @xmath1256 .
so , @xmath1257 holds for @xmath1255 , thus @xmath90 also minimizes the action for @xmath1255 .
this leads to a contradiction .
since @xmath1192 is closed , once @xmath1258 , then whole edge is also contained in @xmath1192 .
the conclusion 2 follows from the fact that @xmath1112\rangle=0 $ ] holds for any @xmath1113 and any @xmath1259 , the conclusion 3 follows from that @xmath1260 if @xmath1255 .
if the conclusion 4 was not true , for the cohomology class in @xmath1261 the mather set would contain two closed circles with different homology , but it violates the lipschitz graph property of aubry set . with the same reason we have the conclusion 5 . by this theorem , each edge @xmath1242 aslo uniquely determines a class @xmath1213 so that for each @xmath1238 , the rotation vector of each @xmath89-minimal measure has the form @xmath1262 ( @xmath1263 ) . for brevity
, we also use the notation @xmath1264 for @xmath1265 .
, @xmath1266 .
the blue curve is in @xmath163 for @xmath89 at one end point of @xmath1048 , the red curve is in @xmath1102 for @xmath968 at another end point of @xmath1048.,width=249,height=120 ] given two homology classes @xmath1267 , we call them adjacent if @xmath1268 , @xmath1269 , @xmath1270 , @xmath1271 , @xmath1272 and @xmath1273 are adjacent
. the special topology of two - dimensional torus induces some restrictions on adjacent homologies .
[ flatlem3 ] let @xmath1274 be two adjacent edges and assume @xmath1275 .
if @xmath1276 and @xmath1277 , then one has that @xmath1278 .
the aubry set @xmath161 contains homoclinic orbits with two two classes @xmath1279 and @xmath1280 , both are irreducible .
guaranteed by the lipschitz graph property , these curves intersect each other only at the fixed point . in the universal covering space @xmath1281 , each curve in the lift of the homoclinic curves are determined by the equation @xmath1282 the solution of the equations corresponds to the intersection point which are lattice points in @xmath1283 for any @xmath1284 . to guarantee this property , the necessary and sufficient condition is @xmath1285 . for each indivisible homological class @xmath1286 , either @xmath1287 for any @xmath1263 , or some @xmath1288 exists such that @xmath1289 . in the first case , @xmath1290 as @xmath1291 , at least one periodic curve @xmath1292 exists for @xmath1293 with @xmath1198 .
it is impossible that @xmath1294 holds for @xmath1295 , as in that case certain @xmath89-minimal measure @xmath90 would exist so that @xmath959 is not parallel to @xmath454 $ ]
. it will violate the lipschitz property .
generically , @xmath1296 is hyperbolic and @xmath1297 is an interval if @xmath1198 .
if @xmath1256 , then @xmath1298 approaches to certain edge @xmath1048 .
if @xmath1299 with indivisible @xmath1300 , @xmath1301 , @xmath1302 , @xmath1048 and @xmath1303 are two adjacent edges .
as @xmath1291 , the interval will shrink to a vertex where @xmath1048 is joined to @xmath1303 , and we have a sequence of closed orbits @xmath1304 .
its kuratowski upper limit set is obviously in the aubry set for certain @xmath1305 , thus , consists of minimal homoclinic orbits to the fixed point . as @xmath89 approaches to the vertex ,
the mather set approaches to a set of figure eight : @xmath1306 where @xmath1307 is a minimal homoclinic orbit such that @xmath1308=g_{\ell}$ ] for @xmath1309 .
on the red line in the channel , the aubry set is a closed orbit in the cylinder , it approaches to a curve of figure eight.,width=264,height=113 ] to be more precise , let us study this phenomenon in the finite covering space @xmath1310 where @xmath1311 for @xmath1312 if we write @xmath1313 for @xmath1314 .
let @xmath1315 : @xmath1316 be a permutation such that the cardinality @xmath1317 and @xmath1318 .
the lift of homoclinic curve @xmath613 as well as @xmath1319 to @xmath178 contains several curves , not closed .
pick up one curve @xmath1320 in the lift of @xmath1321 , it determines a unique curve @xmath1322 such that the end point of @xmath1320 is the starting point of @xmath1322 , and so on .
see figure [ fig3 ] .
= ( 1,0)$ ] , @xmath1323=(1,1)$ ] , @xmath1324 , @xmath1325 . for each class @xmath89 on the red line , @xmath1326 + 2[\gamma_2])$ ] .
the solid blue line represent a periodic curve , the solid purple line represent the conjunction of the minimal homoclinic curves .
the dashed lines represent the image of the deck - transformation.,width=264,height=158 ] we are going to show below that there exists a unique permutation @xmath1315 such that one aubry class in @xmath1327 @xmath1328 as @xmath89 approaches to the vertex along the path in the channel . as the minimal curve @xmath1329 is periodic with the homological class @xmath1330=k_ig_i+k_{i+1}g_{i+1}$ ] , the permutation @xmath1315 : @xmath1331 is @xmath1332-periodic . since @xmath702 is prime to @xmath1333 , we have @xmath1334 if @xmath1335 .
[ flatlem4 ] the permutation is uniquely determined by @xmath702 and @xmath1333 .
if @xmath1336 , the following holds for @xmath1337 @xmath1338 by the assumption , there exists only one minimal homoclinic curve @xmath1339 such that @xmath1340=g_j$ ] for @xmath1314 .
because of the lemma [ flatlem3 ] , we can assume that @xmath1341 and @xmath1342 by choosing suitable coordinates on @xmath1169 .
we choose two sections @xmath1343 and @xmath1344 in a small neighborhood of the origin such that , emanating from the origin , these homoclinic curves pass through @xmath1343 and @xmath1344 successively before they return back to the origin as @xmath1345 . in the section @xmath1346 we choose disjoint subsections @xmath1347 and @xmath1348 such that the curve @xmath1339 passes through @xmath1349 for @xmath1314 .
let @xmath1329 be the minimal periodic curve with rotation vector @xmath1350 . for small @xmath1198
, @xmath1329 falls into a small neighborhood of these two homoclinic curves .
so it has to pass either through @xmath1347 or through @xmath1348 .
let @xmath1351 be the time for @xmath1329 passing through @xmath1346 with @xmath1352 , and it does not tough these sections whenever @xmath1353 . by definition ,
the period of the curve equals @xmath1354 .
if the curve intersects @xmath1355 at @xmath1356 and intersects @xmath1357 at @xmath1358 , then the segment @xmath1359}$ ] keeps close to @xmath613 and @xmath1360}$ ] keeps close to @xmath1319 , so one has @xmath1361 and @xmath1362 .
[ fig4 ] as the curve @xmath1329 is minimal , it does not have self - intersection .
thus , once there exists @xmath1363 such that @xmath1364 and @xmath1365 , then there does not exist @xmath1366 such that @xmath1367 and @xmath1368 . therefore
, there is a set @xmath1369 with cardinality @xmath1370 such that for @xmath1371 one has @xmath1372 , for @xmath1373 one either has @xmath1374 and @xmath1375 or has @xmath1376 and @xmath1377 . by introducing coordinate transformation on @xmath853 : @xmath1378 such that @xmath1379 @xmath1380 , let us think the curve @xmath1381 as a straight line projected down to the unit square , a fundamental domain of @xmath1169 . starting from a point @xmath1382
, the line successively reaches to the points @xmath1383 where @xmath1384 with small @xmath1385 . to connect the point @xmath1386 to the point @xmath1387
, the curve @xmath1381 does not touch the vertical boundary lines if @xmath1388=\big[m\frac{k_{i+1}}{k_i}\big],\ ] ] where @xmath1389 $ ] denote the largest integer not bigger than the number @xmath1390 , and it has to pass through the vertical lines at some point @xmath1391 if @xmath1388 + 1=\big[m\frac{k_{i+1}}{k_i}\big].\ ] ] we define an order @xmath1392 for these @xmath1393 points such that @xmath1394 iff @xmath1395 and @xmath1396 iff @xmath1397 + 1=[(j+1)k_i / k_{j+1}]$ ] .
returning back to the original coordinates , the curve @xmath1329 falls into a neighborhood of the curves @xmath1398 and @xmath1319 , intersects the horizontal line @xmath1399 at @xmath1400 and intersects the vertical line @xmath1401 at @xmath1402 , @xmath1403=g_{i+1}$ ] and @xmath1404=g_i$ ] .
naturally , the map @xmath853 induces the order among these points : @xmath1405 if and only if @xmath1406 .
if the curve passes the point @xmath1407 at @xmath1408 , the segment @xmath1409}$ ] falls into a neighborhood of @xmath613 , otherwise , it falls into a neighborhood of @xmath1319 . in this way
, we obtained a unique permutation @xmath1315 up to a translation .
[ fig5 ] in the second case , some @xmath1288 exists such that @xmath1410 .
it is typical that certain edge @xmath1048 exists such that @xmath1411 , the @xmath89-minimal measure is supported on a hyperbolic periodic orbit for each @xmath1412 .
thus , there is a channel @xmath1413 endding at @xmath1048 such that the family of periodic orbits @xmath1414 constitutes a normally hyperbolic cylinder .
, the aubry set is a closed orbit in the cylinder ( in purple ) , the closed circle in blue is in the aubry set for @xmath1415 , the closed orbit in green is not global minimum.,width=264,height=132 ] in the following , we shall study the dynamics on certain energy level of truncated normal form @xmath1416 .
since one obtains the condition @xmath1417 from the normal form , some function @xmath1418 solves the equation @xmath1419 . treat @xmath1420 as the new hamiltonian and let @xmath1421 be the time , one obtains a system with two degrees of freedom , which is equivalent to the dynamics on the energy level @xmath1422 .
[ flatthm5 ] for the hamiltonian @xmath1423 we assume that @xmath1424 on @xmath1425\}$ ] .
let @xmath1426 be the solution of @xmath1419 @xmath1427 .
let @xmath1428 and @xmath1429 be the @xmath95-function for @xmath1430 and @xmath1431 respectively , where @xmath1432 @xmath1433 then for @xmath1434 $ ] we have @xmath1435 .
let @xmath1436 , @xmath1437 , @xmath1438 and @xmath1439 .
let @xmath75 be @xmath89-minimal curve for the lagrange flow @xmath1440 , @xmath1441 is then @xmath1442-minimal curve for the lagrange flow @xmath1443 if @xmath1444 and @xmath1441 is re - parameterized @xmath1445 . if @xmath1446 is a solution of @xmath1440 , one obtains @xmath1447 from the hamiltonian equations . since @xmath1448 , we find @xmath1449&=\int\big(\big\langle\frac{dx}{d\tau},y - c\big\rangle - y_n+\alpha_y(c)\big)d\tau\\ & = \int(\langle\dot{\tilde x},\tilde y-\tilde c\rangle -h+e)dt\\ & = [ a_h(\tilde\gamma)].\end{aligned}\ ] ] this completes the proof .
let @xmath1450 be the projection @xmath1451 . by this theorem
, @xmath1452 is a homeomorphism for @xmath1453 , the @xmath1454-neighborhood of the flat @xmath1046 .
thus , what we obtained in this subsection have their counterpart in the energy level set @xmath1422 where the class @xmath1455 .
in this section , normally hyperbolic invariant cylinder is proved to exist in certain neighborhood of double resonant point .
it uses the normal form which is obtained in the appendix where several steps of kam iteration and one step of linear coordinate transformation were carried out .
all these coordinate transformations are symplectic .
since aubry set and ma set are symplectic invariants @xcite , it is good enough to study these objects by considering the normal form .
the normal form of the hamiltonian takes the form @xmath1456 where @xmath1457 and @xmath1458 .
since @xmath11 is positive definite , a unique curve exists along which @xmath1459 .
this curve passes through the energy level @xmath1460 transversally at a unique point @xmath1461 . as @xmath1462 ,
one has @xmath1463 . by coordinate translation
, we assume @xmath1464 . as the dynamics we are going to study
is restricted on an energy level @xmath1465 , it can be reduced to a system with two and half degrees of freedom . since @xmath1466 , the equation
@xmath1467 uniquely determines a smooth function @xmath1468 in certain neighborhood of @xmath1469 . treating @xmath1470 as a new hamiltonian and @xmath1471 as a new time variable ,
we obtain a time - periodic system with two degrees of freedom .
correspondingly , we have the normal form @xmath1472 where @xmath1473 , @xmath1474 and @xmath872 is as small as @xmath864 .
each @xmath1475 determines a resonant curve @xmath1476 .
normally hyperbolic cylinder is searched when @xmath29 varies along this curve .
recall that the normal form remains valid in the domain @xmath1477 ( @xmath1478 ) .
the @xmath1479-neighborhood of the curve is covered by as many as @xmath1480 small balls with radius @xmath201 .
given a ball centered at @xmath1481 , by rescaling variables @xmath1482 , @xmath1483 , one obtains an equivalent hamiltonian equation @xmath1484 which corresponds to the hamiltonian @xmath1485 where @xmath1486 , @xmath1487 , @xmath1488 and @xmath1489 according to appendix a , one has @xmath1490 and @xmath1491 , where the @xmath62-norm is with respect to @xmath1492 only . to guarantee the covering property shown in the appendix , one choose @xmath1493 , so we have @xmath1494 . obviously , one has [ homogenizationpro ] each orbit @xmath1495 of the hamiltonian flow @xmath1496 uniquely determines an orbit @xmath1497 of @xmath1498 .
if @xmath1499 and @xmath1500 , then @xmath1501 . the hamiltonian @xmath1502 is a small perturbation of the homogenized hamiltonian @xmath1503 which determines the lagrangian @xmath1504 let us first consider the case when @xmath1505 , it follows that @xmath1506 .
the variable @xmath29 is restricted in the domain @xmath1507 , where the higher order term is bounded by @xmath1508 ( see ( [ normaleq1 ] ) ) .
correspondingly , let @xmath1509 and @xmath1510 for @xmath1505 : @xmath1511 for this hamiltonian system , the maximal point of @xmath1182 determines a stationary solution which corresponds to a minimal measure of @xmath1512 , where the matrix @xmath1513 has 4 real eigenvalue @xmath1514 .
by translation of coordinates , it is generic that ( * h1 * ) : _ @xmath1182 attains its maximum at @xmath900 only , the hessian matrix of @xmath1182 at @xmath900 is negative definite .
all eigenvalues are different : @xmath1515_. such a hypothesis leads to certain hyperbolicity of minimal homoclinic orbits . let us consider the case : for @xmath1050 the aubry set @xmath1516 , where @xmath280 : @xmath76 is a minimal homoclinic curve . by the assumption * h1 * , the fixed point @xmath1517 has its locally stable manifold @xmath1518 as well as the locally unstable manifold @xmath1519 .
they intersect each other transversally at the origin .
as each homoclinic orbit entirely stays in the stable as well as in the unstable manifolds , along such orbit their intersection can not be transversal in the standard definition , but transversal module the curve : @xmath1520 holds for @xmath71 is on minimal homoclinic curves . without danger of confusion , we also call the intersection transversal .
if we denote by @xmath1521 the eigenvector corresponding to the eigenvalue @xmath969 , where @xmath1522 and @xmath1523 are for the @xmath71- and @xmath29-coordinate respectively , then the eigenvector for @xmath1524 will be @xmath1525 .
it is also a generic condition that ( * h2 * ) : _ with each @xmath1526 , there is at most one minimal orbit associated , the stable manifold intersects the unstable manifold transversally along each minimal homoclinic orbit .
each minimal homoclinic orbit approaches to the fixed point along the direction @xmath1527 : @xmath1528 as @xmath1529 .
_ recall the set @xmath1192 . if @xmath1530 , the @xmath89-minimal measure consists of two or more ergodic components . because of theorem [ flatthm3 ] and the countability of homology classes of all homoclinic curves , we have another generic condition ( * h3 * ) : _ for each @xmath1295 , the aubry set does not contain minimal curve homoclinic to the origin @xmath139fixed point@xmath140 .
_ let us start with a hamiltonian with two and half degrees of freedom : @xmath1531 where @xmath1532 , @xmath1490 , @xmath1508 ( see theorem [ normalthm1 ] ) where the @xmath62-norm is with respect to @xmath1492 . recall the homogenized hamiltonian as well as the homogenized lagrangian @xmath1533 where @xmath1534 . by the assumption ( * h1 * ) , the fixed point @xmath1535 each @xmath880 which is a 2-dimensional flat .
as classified before , for each @xmath1536 the aubry set consists of minimal homoclinic orbits plus the fixed point .
given an irreducible class @xmath1526 , it is not necessary that some minimal homoclinic curve @xmath75 exists such that @xmath454=g$ ] .
let us consider the @xmath89-minimal measure for @xmath1537 with @xmath1263 . as we shall see later , it is generic that the minimal measure is supported on at most two periodic orbits , denoted by @xmath1538 , both are hyperbolic , namely , it has the stable and unstable manifold .
these periodic orbits constitute a two - dimensional cylinder .
however , it appears not reasonable to assume the normal hyperbolicity for @xmath1539 in usual sense as the speed along the orbit may undergo large variation , especially , when it is very close to some homoclinic orbit .
therefore , one can not see the separation of the spectrum of @xmath1540 in normal and in tangent direction . as the first step ,
let us study the case when @xmath1541 .
there are two cases alternatively as @xmath333 decreases : @xmath1542 as @xmath333 decrease to zero , or @xmath1062 @xmath1543 such that @xmath1544 . in the first case
, the cohomology class approaches to some edge @xmath1545 or to some vertex where two adjacent edges @xmath1546 joint together . under the hypothesis ( * h2 * ) , for class @xmath89 in the interior of the edge , the aubry set @xmath163 contains exactly one minimal homoclinic curve and the point of the origin .
let @xmath1339 be the minimal homoclinic curve related to @xmath1248 for @xmath1314 , it determines the minimal homoclinic orbit @xmath1547 .
denote by @xmath1548 $ ] the homology class . if @xmath1549 , @xmath1550 as @xmath333 decreases to zero .
if there exist two positive integers @xmath1551 such that @xmath1299 , the curve @xmath1552 approaches to the set @xmath1553 ( figure eight ) as @xmath1554 , folding @xmath702 and @xmath1333-times along @xmath648 and @xmath1555 respectively .
the periodic curve @xmath1556 determines a periodic orbit @xmath1557 in the phase space .
it stays in certain energy level set @xmath2 . for @xmath1299 and suitably small @xmath17 , by the study in section 3.2 ( eq.([regularenergyeq2 ] ) ) , the period @xmath853 is related to the energy by the formula @xmath1558 where @xmath1559 as @xmath1560 , both @xmath1561 and @xmath1562 is bounded . to study the dynamics around the minimal homoclinic orbits @xmath1563
( @xmath1564 ) , we use a new canonical coordinates @xmath1492 such that , restricted in a small neighborhood of @xmath1565 , one has the form @xmath1566 where @xmath1567 . without losing generality , we assume @xmath1568 as @xmath1569 , @xmath1570 as @xmath1571 and @xmath1572 as @xmath1573 . here
the notation is taken as granted : @xmath1574 .
we choose 2-dimensional disk lying in @xmath1575 @xmath1576 because of the special form of @xmath1577 , one has @xmath1578 let @xmath1519 ( @xmath1518 ) denote the unstable ( stable ) manifold of the fixed point which entirely stays in the energy level set @xmath1579 . if @xmath1580 , the tangent vector of @xmath1581 has the form @xmath1582 .
so , the tangent vector of @xmath1581 takes the form @xmath1583 where both @xmath1584 and @xmath1585 are small .
denote by @xmath1586 the time when the homoclinic orbit @xmath1587 passes through @xmath1588 .
as @xmath1589 holds at the point @xmath1590 , both homoclinic orbits @xmath1591 and @xmath1592 approach in the same direction to the fixed point , the section @xmath1593 as well as @xmath1594 intersects these two homoclinic orbits transversally .
let @xmath1595 denote the intersection point of @xmath1587 with @xmath1588 . in a small neighborhood of that point @xmath1596 ,
one obtains a map @xmath1597 : @xmath1598 in following way , starting from a point @xmath260 in this neighborhood , there is a unique orbit which moves along @xmath1599 and comes to a point @xmath1600 after a time approximately equal to @xmath1601 .
let us fix small @xmath1602 .
there exists @xmath1603 ( depending on @xmath720 ) such that @xmath1604 holds for both @xmath1605 and @xmath1606 .
clearly , one has @xmath1607 as @xmath1608 .
as the homoclinic curves approach to the origin in the direction of @xmath1609 in @xmath71-space , for small @xmath1610 , there exists a constant @xmath1611 such that @xmath1612 as @xmath1608 and @xmath1613 the hamiltonian flow @xmath1614 defines a map @xmath1615 : @xmath1616 and a map @xmath1617 : @xmath1618 : emanating from a point in @xmath1594 ( @xmath1619 ) there exists a unique orbit which arrives @xmath1620 ( @xmath1593 ) after a time bounded by the last formula . restricted in the ball @xmath1621 ,
let us consider the variational equation of the flow @xmath1622 along the homoclinic orbit @xmath1623 .
it follows from the normal form of the homogenized hamiltonian @xmath1577 that the tangent vector @xmath1624 satisfies the variational equation @xmath1625 where @xmath1626 .
clearly , @xmath1627 with @xmath1628 if @xmath1629 is small . since the homoclinic orbit approaches to the fixed point in the direction of @xmath1630 , one has @xmath1631 for the initial value @xmath1632 satisfying the condition @xmath1633 @xmath1634 one obtains from the hyperbolicity that @xmath1635 holds for some constant @xmath1636 depending on @xmath969 as well as on @xmath14 .
therefore , for each vector @xmath1637 which is nearly parallel to @xmath1638 : @xmath1639 holds for @xmath1640 we obtain from the last two formulae that @xmath1641 similarly , one has @xmath1642 where @xmath1643 also depends on @xmath969 as well as on @xmath14 , @xmath1644 and @xmath1645 as @xmath1608 .
by the construction , the 2-dimensional disk @xmath1646 intersects the unstable manifold @xmath1519 along a curve .
let @xmath1647 be a very short segment of the curve , passing through the point @xmath1648 .
pick up a point @xmath1649 on the homoclinic orbit @xmath1650 far away from the fixed point and take a 2-dimensional disk @xmath1651 containing the point @xmath1649 and transversal to the flow @xmath1652 in the sense that @xmath1653 . the hamiltonian flow
@xmath1652 sends each point of @xmath1654 to this disk provided it is close to @xmath1655 . in this way ,
one obtains a map @xmath1656 : @xmath1657 .
let @xmath1658 . according to the assumption ( * h2 * )
, one has @xmath1659 .
thus , one also has @xmath1660 .
it follows from the @xmath101-lemma that @xmath1661 keeps @xmath74-close to @xmath1662 at the point @xmath1663 and @xmath1664 keeps @xmath74-close to @xmath1665 at the point @xmath1666 provided @xmath54 is sufficiently small . as @xmath1667 , one obtains @xmath1668 and @xmath1669 where @xmath1670 .
see the figure below .
recall the definition , @xmath1671 is a two - dimensional disk lying in the energy level set @xmath1575 . for @xmath17 sufficiently small , @xmath1671 is @xmath1672-close to @xmath1588 respectively .
let @xmath1673 be the minimal periodic orbit staying in the energy level set @xmath1575 , it approaches to the homoclinic orbit as @xmath3 decreases to zero .
thus , for sufficiently small @xmath17 , it passes through the section @xmath1674 as well as @xmath1675 @xmath1676 times for one period .
we number these points as @xmath1677 ( @xmath1678 ) by the role that emanating from a point @xmath1679 , the orbit reaches to the point @xmath1680 after time @xmath1681 , then to the point @xmath1682 and so on .
note that @xmath1681 remains bounded uniformly for any @xmath17 .
restricted on small neighborhoods of these points , denoted by @xmath1683 , the flow @xmath1684 defines a local diffeomorphism @xmath1685 : @xmath1686 . because of the smooth dependence of ode solutions on initial data , a small @xmath1687 exists such that , for the vector @xmath1688 @xmath1689-parallel to @xmath1690 in the sense that @xmath1691 holds for some @xmath1692 , we obtain from the hyperbolicity of @xmath1597 ( see the formulae above figure [ fig7 ] ) that @xmath1693 and @xmath1694 where @xmath1695 , @xmath1696 as @xmath1608 .
if the vector @xmath1697 is chosen @xmath1689-parallel to @xmath1698 then the vector @xmath1699 is @xmath1689-parallel to @xmath1700 . for @xmath17 ,
the hamiltonian flow @xmath1684 defines local diffeomorphism @xmath1701 : @xmath1702 . to make @xmath1703 one
has @xmath1704 as @xmath1560 . according to the study in section 3.2 ( cf .
formula ( [ regularenergyeq2 ] ) ) , starting from @xmath1675 , the periodic orbit comes to @xmath1674 after a time approximately equal to @xmath1705 in which @xmath1706 is uniformly bounded as @xmath263 .
given a vector @xmath524 , we use @xmath1707 denote the @xmath1708-component . for a vector @xmath1709 @xmath1689-parallel to @xmath1710 ,
there is @xmath1711 such that @xmath1712 . from eq.([cylindereq3 ] ) one obtains @xmath1713 where @xmath1714 as @xmath1715 .
it follows that the vector @xmath1716 is @xmath1689-parallel to @xmath1717 and @xmath1718 where @xmath1719 and @xmath1720 as @xmath1721 .
similarly , for a vector @xmath1697 @xmath1689-parallel to @xmath1722 , one sees that the vector @xmath1723 is @xmath1689-parallel to @xmath1724 and @xmath1725 the composition of the two maps constitutes a poinare map @xmath1726 , it maps a small neighborhood of the point @xmath1679 in @xmath1674 to a small neighborhood of the point @xmath1682 in @xmath1674 . for a vector @xmath1697 @xmath1689-parallel to @xmath1724
the vector @xmath1727 is still @xmath1689-parallel to @xmath1724 @xmath1728 and for a vector @xmath1709 @xmath1689-parallel to @xmath1722 the vector @xmath1729 is still @xmath1689-parallel to @xmath1722 @xmath1730 holds for each @xmath77 , where @xmath1731 , @xmath1732 as @xmath1608 .
therefore , each point @xmath1679 is a hyperbolic fixed point for the map @xmath1733 , @xmath1734 is a hyperbolic orbit of @xmath1735 . by lemma [ flatlem4 ] ,
these points are uniquely ordered , @xmath1393 is the minimal period . as these points approach to the fixed point as @xmath1736 , the hyperbolicity guarantees the uniqueness
. it also guarantees the smooth continuation of periodic orbits .
therefore , we have [ cylinderlem1 ] assume the conditions and and let @xmath1526 be a class .
if @xmath1542 as @xmath1737 , then there exists @xmath1738 such that for each @xmath1537 with @xmath1739 $ ] the mather set @xmath161 consists of exactly one periodic orbit .
let @xmath1740 be a two - dimensional disk transversally intersecting the orbit at @xmath1741 such that @xmath1742 for @xmath1743 and let @xmath1744 : @xmath1745 be the return map naturally determined by the flow @xmath1684 , there exists some @xmath1746 independent of @xmath1747 such that @xmath1748 @xmath1749 where @xmath1750 is the point where the periodic orbit intersects @xmath1751 , @xmath1752 denotes the stable @xmath139unstable@xmath140 manifold of the periodic orbit .
therefore , the periodic orbits for @xmath1753 $ ] constitute a smooth cylinder . in the second case ,
some @xmath1543 exists such that @xmath1754 .
it is typical that @xmath1755 is an edge of @xmath1242 , where each mather set consists of a hyperbolic periodic orbit @xmath1756 and the fixed point .
the uniqueness of minimal periodic orbit for @xmath333 close to @xmath1757 follows from the implicit function theorem .
given @xmath1526 , the set @xmath1758 is a channel in @xmath1759 .
there is a path @xmath1760 reaching to the boundary @xmath1761 , along which the @xmath95-function monotonely decreases to the minimum as the cohomology class approaches to @xmath1761 . with the argument as above and theorem [ appenhyperth1 ] shown in the appendix , we find the following condition also generic ( * h4 * ) : _ given a class @xmath1526 and an energy @xmath1762 , there are at most finitely many @xmath1763 $ ] such that for @xmath1764 with @xmath1765 the mather set consists of two periodic orbits , for all other @xmath1764 with @xmath1766 , the mather set consists of exactly one periodic orbit .
all these periodic orbits are hyperbolic .
_ we call these @xmath1767 bifurcation points .
let @xmath1768 be the smallest one .
each energy @xmath1769 uniquely determines a hyperbolic periodic orbit @xmath1770 .
thus , we introduce a notation @xmath1771=g , e\in [ e_0,e_1],s\in\mathbb{r}\},\ ] ] with small @xmath1160 .
it is a two - dimensional cylinder composed of a family of periodic orbits , may approach to a curve of figure - of - eight as @xmath1772 decreases to zero . obviously , the cylinder is invariant for the hamiltonian flow @xmath1684 of @xmath1577 .
let @xmath1773 denote the period of the periodic orbit in @xmath1575 , one has @xmath1774 the cylinder may be slant and crumpled , not standard . to see how the symplectic area is related to the usual area of the cylinder ,
let us study the dependence of the fixed point @xmath1775 of the poincar ' e return map @xmath1735 on @xmath3 . by definition ,
the fixed point is a solution of the equation @xmath1776 emanating from a point @xmath1777 the orbit reach a point @xmath1778 after a time @xmath1779 which remains bounded as @xmath1780 .
let @xmath1781 be the point corresponding to @xmath1782 , obtained in the same way .
the difference of the @xmath1783-coordinate of these two points is bounded by @xmath1784 where @xmath1785 depends on @xmath49 .
let @xmath1786 be the solution of the variational equation ( [ cylindereq3 ] ) along the periodic solution @xmath1787 passing through a neighborhood of the origin , let @xmath526 be the time such that the first coordinate @xmath1788 and @xmath1789 , the quantity @xmath1790 is bounded by ( [ regularenergyeq2 ] ) . in virtue of the formula ( [ cylindereq5 ] ) , it yields @xmath1791 where @xmath1792 as @xmath263 .
therefore , we find that @xmath1793 as the quantity @xmath1794 is bounded by ( [ cylindereq6 ] ) , the quantity for the inverse of @xmath1795 is bounded by ( [ cylindereq7 ] ) , we obtain from the equation ( [ cylindereq8 ] ) that @xmath1796 it yields a relation between the symplectic area @xmath127 and the usual area @xmath299 of the cylinder @xmath1797 we assume the conditions . for each @xmath1798 $ ] , the cylinder @xmath1799 is normally hyperbolic for the map @xmath1800 , where @xmath1801 . for any @xmath1802 ,
the cylinder @xmath1799 is a 2-dimensional symplectic sub - manifold , invariant for the hamiltonian flow @xmath1652 .
however , it is not clear whether this cylinder is normally hyperbolic for the time-1-map @xmath1803 , as it is possible that @xmath1804 and we do not have the estimate on the norm of @xmath1805 acting on the normal bundle . as the first step of the proof , let us search for the normal hyperbolicity of @xmath1652 with large @xmath157 for the cylinder @xmath1806 , where @xmath1807 denotes the largest value such that the formulae ( [ cylindereq6 ] ) and ( [ cylindereq7 ] ) hold for each @xmath1747
from these formulae , one sees that the smaller the energy reaches , the stronger hyperbolicity the map @xmath1735 obtains .
the strong hyperbolicity is obtained by passing through small neighborhood of the fixed point .
however , on the other hand , the smaller the energy decreases , the longer the return time becomes .
let @xmath1808 denote the time interval such that , starting from @xmath1679 , the periodic orbit comes to @xmath1682 after time @xmath1808 . in virtue of the study in section 3 ,
( [ regularenergyeq2 ] ) , @xmath1809 where @xmath1810 is uniformly bounded and we take @xmath1810 if the segment of the periodic orbit keeps close to the homoclinic orbit @xmath1811 . note @xmath1812 is much larger than @xmath1813 .
thus , starting from any point @xmath260 on the minimal periodic orbit @xmath1814 , @xmath1815 passes through the neighborhood of the fixed point after time @xmath1816 .
it implies that the map @xmath1817 obtains strong hyperbolicity on normal bundle . to measure how the map @xmath1818 acts on the tangent bundle ,
let us study how the map @xmath1652 elongates or shortens small arc of the periodic orbit . as the orbit passes through the neighborhood of the origin @xmath1819 in a time approximately equal to @xmath1820 the variation of the length of short arc is between @xmath1821 and @xmath1822 , where @xmath1823 is small .
because of the relation between the symplectic area @xmath127 and the usual area @xmath299 of the cylinder , given by the formula ( [ cylindereq10 ] ) , the variation of @xmath1824 , restricted on the tangent bundle of the cylinder , is between @xmath1825 and @xmath1826 . because of periodicity , it is independent of @xmath157 .
thus , the normally hyperbolic property becomes clear : the tangent bundle of @xmath61 over @xmath1799 admits @xmath1827-invariant splitting @xmath1828 and some @xmath1829 , @xmath1830 and small @xmath1263 exist such that @xmath1831 @xmath1832 @xmath1833 hold for @xmath1834 ( cf .
( [ cylindereq6 ] ) and ( [ cylindereq7 ] ) ) .
note that @xmath1835 provided @xmath1263 is suitably small .
the formula ( [ cylindereq11 ] ) satisfies the definition of normal hyperbolicity . for each @xmath1836 $ ] , let @xmath1814 be the minimal periodic orbit ,
@xmath1837 be a 2-dimensional disk intersecting @xmath1814 transversally at the point @xmath1750 , @xmath1744 : @xmath1838 be the poincar return map . by the generic property
, @xmath1750 is the hyperbolic fixed point of @xmath1751 and @xmath1839 exists such that @xmath1840 @xmath1841 as the cylinder is foliated into periodic orbits and @xmath1842 preserves the symplectic form , some @xmath1843 exists such that @xmath1844 holds for any @xmath1845 . choosing @xmath1846 such that @xmath1847 , one obtains the normal hyperbolicity for @xmath1848 with @xmath1849 , where @xmath1850 is the period of the periodic solution in @xmath1851 , @xmath1852 with @xmath1853 $ ] . by choosing suitably small @xmath1160
, we have @xmath1854 . to apply the theorem of normally hyperbolic manifold @xcite to the hamiltonian @xmath1502 of ( [ cylindereq3 ] )
, we note that the hamiltonian @xmath1855 is autonomous with two degrees of freedom . as @xmath1490 , and because of the non - degeneracy assumption for @xmath1182 ( ( * h1*,*h2 * ) )
, we see that for each suitably small @xmath271 , the map @xmath1856 admits invariant cylinder also , denoted by @xmath1799 still , with the normally hyperbolic properties ( see formulae ( [ cylindereq11 ] ) ) , independent of the size of @xmath289 .
let us consider the persistence of @xmath1857 with @xmath1858 with @xmath1068 .
as the perturbation depends on time @xmath157 , we use @xmath1859 to denote the map from the time @xmath557-section to the time @xmath157-section , omit the symbol @xmath557 if @xmath1860 . since these hyperbolic properties are posed for the map @xmath1861 with large @xmath1862 , one has to measure how large the quantity @xmath1863 will be . as @xmath1864 is autonomous , @xmath1865 .
[ cylinderlem2 ] let the equation @xmath1866 be a small perturbation of @xmath1867 , let @xmath1868 and @xmath1869 denote the flow determined by these two equations respectively
. then @xmath1870 where @xmath1871 and @xmath1872 .
let @xmath1873 denote the solution of the equations @xmath1874 for @xmath1875 respectively , and @xmath1876 .
let @xmath1877 , then @xmath1878 and @xmath1879 where @xmath1880 $ ] .
therefore , one has @xmath1881 let @xmath1882 , we have @xmath1883 .
it follows from gronwell s inequality that @xmath1884 along the orbit @xmath1873 , the differential of the flow @xmath1885 obviously satisfies the equation @xmath1886 therefore , for each tangent vector @xmath524 attached to @xmath1887 one has @xmath1888 to study the differential of @xmath1889 , let us consider the equation of secondary variation .
let @xmath1890 be the solution of the variational equation @xmath1891 for @xmath1892 respectively , where @xmath1873 solves the equation @xmath1893 and @xmath1876 .
to measure the size @xmath1894 with the condition @xmath1876 , we make use of the relations such as @xmath1895 , @xmath1896 and find that @xmath1897 let @xmath1898 , we have @xmath1883 . using gronwell s inequality again
, one obtains an upper bound of the variation of the differential @xmath1899 note that @xmath524 represents initial tangent vector , it completes the proof .
let us applying this lemma to the hamiltonian @xmath1502 .
treating @xmath1900 as the function of @xmath1492 we find that there exist some constants @xmath1901 independent of @xmath289 such that @xmath1902 as @xmath1903 ( see theorem [ normalthm1 ] ) . since the function @xmath1864 comes from @xmath1904 which is @xmath15-smooth ( @xmath16 ) , one has @xmath1905 . for @xmath1906
one obtains from lemma [ cylinderlem2 ] that @xmath1907 if the condition @xmath1908 is satisfied , then @xmath1909 as @xmath378 .
it allows one to apply the theorem of normally hyperbolic manifold to obtain the existence of invariant cylinder for the flow @xmath1859 in the extended phase space @xmath1910 , which is a small deformation of @xmath1911 .
be aware of the fact that @xmath1857 is a cylinder with boundary , normally hyperbolic and invariant for @xmath1912 , where @xmath1913 , we do not expect that the whole cylinder survives small perturbation , it may lose some part close to the boundary . to measure to what range the cylinder survives ,
we see that the variation of the energy along each orbit of @xmath1914 is bounded by @xmath1915 here , the estimate @xmath1916 is used ( see theorem [ normalthm1 ] ) .
assume that the number @xmath1454 satisfies the condition @xmath1917 one sees that , starting from the energy level @xmath1918 , after a time of @xmath1919 , the orbit of @xmath1859 can not reach the energy level @xmath1920 if @xmath289 is suitably small so that @xmath1921 .
indeed , under such condition one has @xmath1922 to use the theorem of normally hyperbolic invariant manifold , let us introduce a modified hamiltonian .
let @xmath1923 : @xmath1924 be a smooth function so that @xmath1925 for @xmath1926 and @xmath1927 for @xmath1928 , @xmath1929 it coincides with @xmath1502 for @xmath1930 with @xmath1931 and coincides with @xmath1864 for @xmath1930 with @xmath1932 . for small @xmath1454 satisfying the condition ( [ cylindereq13 ] ) and small @xmath289 , the cylinder @xmath1857 survives the perturbation @xmath1933 and the bottom remains invariant for @xmath1934 . by the definition
, one has @xmath1935 for @xmath1930 with @xmath1936 and it follows from ( [ cylindereq14 ] ) that @xmath1937 provided @xmath1938 and @xmath1939 $ ] .
so , for @xmath1940 the invariant cylinder @xmath1941 persists under the perturbation @xmath1942 , denoted by @xmath1943 .
a point @xmath1944 implies @xmath1945 $ ] .
the invariance is in the sense that , emanating from any point in @xmath1943 , the orbit has to pass through the bottom of the cylinder if it is going to leave the cylinder .
* location of aubry set in the cylinder * as the working space here is phase space , we say that an aubry set @xmath161 is located in the cylinder @xmath1943 if for each @xmath89-static curve @xmath75 , the orbit in the phase space @xmath1946 . recall the hamiltonian @xmath1502 defined in ( [ cylindereq3 ] ) and note the hamiltonian @xmath1947 is autonomous .
let @xmath1948 and @xmath1949 denote the @xmath95-function for the lagrangians determined by @xmath1502 and @xmath1864 respectively . since @xmath1950 is a hyperbolic cylinder , invariant for the hamiltonian flow @xmath1951 , the channel @xmath1952}\mathscr{l}_{\beta}(\nu g)$ ] has a foliation into a family of segments of line ( one - dimensional flat ) , where @xmath1953 is chosen so that @xmath1954 for each @xmath1537 if @xmath1955 .
we claim that the @xmath95-function @xmath1949 is smooth in @xmath1956 . indeed , restricted on each of these flats
the function @xmath1949 keeps constant , while restricted on a line @xmath1957 orthogonal to these flats , the function is smooth because @xmath1864 can be treated as a hamiltonian with one degree of freedom when it is restricted on the cylinder .
if @xmath1299 we consider the hamiltonian in the finite covering space @xmath1310 where @xmath1311 for @xmath1312 if we write @xmath1313 for @xmath1314 . in that space there are @xmath1393 fixed points . because of hartman s theorem for two - dimensional system , it is @xmath74-conjugate to a linear equation @xmath1958 , @xmath1959 around each fixed point .
therefore , for small @xmath17 some @xmath74-function @xmath1960 exists such that the period of the frequency @xmath1961 is @xmath1962 ( see ( [ regularenergyeq2 ] ) ) . since @xmath1963 @xmath1964 as @xmath1965 remains constant when it is restricted on each of its flats which are orthogonal to the line @xmath1957
, we find that for @xmath1966 @xmath1967 where @xmath524 is the direction of @xmath1957 and @xmath1968 .
it follows that @xmath1969 holds for @xmath1970 , @xmath1971 @xmath1764 , @xmath1972 .
let @xmath1087 be the class so that @xmath1973 . to measure the difference of @xmath1974
, we find @xmath1975 . as the @xmath95-function undergoes small variation : @xmath1976 for small perturbation @xmath1977 with @xmath1978 @xcite , we find @xmath1979 when @xmath89 is restricted on the path @xmath1957 .
therefore , we obtain that @xmath1980 with @xmath1981 , any orbit does not hit the energy level set @xmath1982 with @xmath1983 . if the lemma does not hold , there would exist an orbit @xmath1495 in the aubry set for @xmath1984 , which hits the energy level @xmath1985 at the time @xmath1986 , i.e. @xmath1987 . since the orbit entirely stays in the invariant cylinder and the perturbation is of order @xmath1988 , it returns back to the neighborhood of @xmath1989 after a time @xmath1990 where @xmath1991 remains bounded as @xmath378 . to see how close it could be , we obtain from ( [ cylindereq12 ] ) that @xmath1992 as @xmath1993 is an invariant circle for @xmath1951 , the perturbed cylinder is @xmath1994-close to the original one @xcite and the cylinder may be crumpled but at most up to the order @xmath1995 ( cf . ( [ cylindereq9 ] ) ) , some large @xmath1140 exists such that @xmath1996 and @xmath1997 since the curve @xmath1998 is assumed @xmath89-static , it follows that @xmath1999 as the cylinder @xmath1941 is @xmath1988-close to @xmath1943 , there is a @xmath968-minimal orbit @xmath2000 of the hamiltonian flow @xmath1912 on @xmath1799 such that @xmath2001 and @xmath2002 .
let @xmath2003 and @xmath2004 where @xmath2005 is the period of @xmath2006 , we have an estimate on the hausdorff distance @xmath2007 .
consequently , we have @xmath2008 as @xmath2009 we have @xmath2010 let @xmath2011 be the lift of @xmath1998 to the universal covering space , it follows that @xmath2012 since it follows from ( [ cylindereq19 ] ) that @xmath2013 holds for all @xmath2014 $ ] , we find @xmath2015 to verify the second inequality , let @xmath1087 be the class such that @xmath2016 , then we obtain from the formula ( [ cylindereq18 ] ) that @xmath2017 for small @xmath289 such that @xmath2018 we find from ( [ cylindereq15 ] ) that @xmath2019 holds for @xmath2020 and @xmath2021 .
therefore , one obtains from ( [ cylindereq16 ] ) and ( [ cylindereq17 ] ) that @xmath2022 from which one obtains the second inequality of ( [ cylindereq21 ] ) from the first one .
as @xmath2023 is very small , the formula ( [ cylindereq21 ] ) contradicts ( [ cylindereq20 ] ) .
it completes the proof . for @xmath1068 satisfying the condition ( [ cylindereq13 ] ) ,
going back to the original coordinates ( @xmath2024 , @xmath2025 and @xmath1483 ) , we obtain [ cylinderthm2 ] for an irreducible class @xmath1526 , there exists a 3-dimensional cylinder @xmath2026 associated with a channel @xmath2027 such that 1 , the cylinder @xmath1943 is a small deformation of the cylinder @xmath2028=g,(\tau,\tau^*)\in\mathbb{t}^2,e\in[e_0,e_1]\},\ ] ] where @xmath2029 , @xmath2030 is the minimal periodic curve for @xmath2031 , the function @xmath2032 solves the equation @xmath2033 , @xmath2034 and @xmath2035 ; 2 , the cylinder is invariant for the hamiltonian flow @xmath2036 : @xmath2037 @xmath2038 , if @xmath2039 holds for @xmath2040 then @xmath1062 @xmath2041 such that @xmath2042 is on the boundary of @xmath1943 , where @xmath2043 solves the equation @xmath2044 ; 3 , @xmath1943 is normally hyperbolic for @xmath2036 with @xmath2045 ; 4 , this channel reaches to a small neighborhood of the flat @xmath1046 in the sense @xmath2046 for each @xmath2034 with @xmath2047 , the aubry set entirely stays in the cylinder .
let us consider the autonomous hamiltonian @xmath41 with the form of ( [ homogenizedeq1 ] ) .
let @xmath2048 , @xmath866 be the function solves the equation @xmath2049
. then @xmath2043 has the form of ( [ homogenizedeq2 ] ) .
the energy @xmath3 of @xmath452 corresponds to the coordinate @xmath2050 for the autonomous hamiltonian @xmath41 .
applying theorem [ flatthm5 ] one obtains [ cylinderthm3 ] assume @xmath2048 .
given an irreducible @xmath1526 , there is a @xmath200-dimensional cylinder @xmath2051 associated with a channel @xmath2052 such that 1 , the cylinder @xmath1943 is a small deformation of the cylinder @xmath2053=g,(x_3,x^*_3)\in\mathbb{t}^2,e\in[e_0,e_1]\};\ ] ] 2 , the cylinder @xmath1943 is invariant for the hamiltonian flow @xmath2054 : for each @xmath2055 , if @xmath2056 holds for certain @xmath823 then @xmath1062 @xmath2057 such that @xmath2058 is on the boundary of @xmath1943 ; 3 , @xmath1943 is normally hyperbolic for @xmath2059 with @xmath2060 ; 4 , for each @xmath2061 with @xmath2062 , the aubry set is contained in that cylinder @xmath2063 . by * h4 * , it is generic that there are finitely many bifurcation points , denoted by @xmath2064 corresponding to the energy @xmath2065 . for each point @xmath2066 , the @xmath89-minimal measure for the hamiltonian @xmath1864
is uniquely supported on a hyperbolic periodic orbit . for the class @xmath415 ,
the mather set is composed of two hyperbolic periodic orbits .
the periodic orbits for @xmath2067 constitute a piece of cylinder @xmath794 .
as these periodic orbits are hyperbolic , the cylinder @xmath794 can be extended a little bit consisting of periodic orbits which are hyperbolic also , but do not support minimal measure .
they are local minimal .
these cylinders are clearly invariant and normally hyperbolic for the hamiltonian flow @xmath1684 with suitably large @xmath832 . applying the theorem of normally hyperbolic manifold
, one can see that there exists some @xmath2068 which is invariant for the hamiltonian flow determined by @xmath452 , and keeps close to @xmath2069 . for each @xmath1764 , the mather set for @xmath452 stays in the cylinder . except for finitely many @xmath2070 very close to @xmath2071 ,
the time-@xmath1479-section of the mather set for other @xmath2072 is an invariant circle , or periodic points or aubry - mather set in the cylinder , for bifurcation point @xmath2073 , the mather set consists of two parts , one stays in @xmath2068 another one stays in @xmath2074 .
some normally hyperbolic invariant cylinder has been shown to reach @xmath2075-neighborhood of the double resonant point .
this cylinder extends to the place a bit far away from the double resonant point . to see how to transit from double resonance to single resonance ,
let us homogenize the hamiltonian in a region @xmath2076 and choose different @xmath2077 .
recall that the normal form remains valid in the domain @xmath1477 ( @xmath1478 ) .
the @xmath1479-neighborhood of the curve is covered by as many as @xmath2078 small balls with radius @xmath201 .
such approach is based on the following : for nearly integrable lagrangian @xmath2079 , each orbit in mather set @xmath453 @xmath2080 the result is proved in @xcite for time-1-map .
it is also true for the hamiltonian flow i.e. @xmath2081 .
it makes sense for us to homogenize the hamiltonian in the range @xmath2082 with suitably large @xmath2083 . using new variables @xmath1482 and @xmath1483 , the homogenized hamiltonian equation turns out to be the following form @xmath2084 where @xmath2085 , the frequency @xmath2086 satisfies a resonant condition .
the corresponding lagrangian reads @xmath2087 with the potential @xmath2088 on the torus one associates its time average @xmath2089 $ ] along the orbits of the linear flow defined by @xmath127 : @xmath2090 @xmath2091(x)=\frac 1t\int_0^tv(x+\omega t)dt,\ ] ] where @xmath853 is the period of the frequency @xmath127 .
the function @xmath2089 $ ] is then defined on a circle .
let @xmath2092 be the maximal point of @xmath2089 $ ] , it corresponds to a circle on @xmath1169 .
the averaged hamiltonian is also associated with a lagrangian @xmath2093(\dot x , x)=\frac 12\big\langle a^{-1}\big(\dot x-\frac{\omega}{\sqrt{\epsilon}}\big),\big(\dot x-\frac{\omega } { \sqrt{\epsilon}}\big)\big\rangle-[v](x).\ ] ] let @xmath2094 be the period of the frequency @xmath2095 , @xmath2096 : @xmath2097\to\mathbb{t}^2 $ ] be the minimizer of the action @xmath839=g_{\omega}}\int_0^{t_{\omega,\epsilon}}[l](d\xi(s))ds,\ ] ] then it is a curve of maximal points of @xmath2089 $ ] with constant speed @xmath2098 .
consider @xmath2089 $ ] as a function defined on @xmath2099 and denote by @xmath2089''$ ] the second derivative for the variable of @xmath2099 , where we use @xmath2096 to denote the circle @xmath2100 } \xi_{\omega,\epsilon}(t)$ ] .
let @xmath2101 denote a translation of @xmath2096 such that @xmath2102 and let @xmath2103 : @xmath2097\to\mathbb{t}^2 $ ] be the minimizer of the action @xmath839=g_{\omega}}\int_0^{t_{\omega,\epsilon}}l(d\xi(s))ds,\ ] ] then we have assume @xmath2104 $ ] is non - degenerate at its minimal point @xmath2104''>\lambda$ ] , and assume some @xmath1198 exists so that @xmath2105 .
then there exist some constants @xmath2106 such that the minimizer @xmath2103 entirely stays in @xmath2107-neighborhood of the circle @xmath2101 , i.e. @xmath2108 holds for each @xmath2109 $ ] and @xmath2110 . since the minimizer solves the lagrange equation , its second derivative remains bounded .
thus , as the first step , we claim that the minimizer stays entirely in @xmath2107-neighborhood of @xmath2101 , a translation of the circle @xmath2096 .
if not , the oscillation of @xmath2103 in the direction perpendicular to @xmath127 would not be smaller than @xmath2111 . as the average speed is @xmath2112 , there would be a point on the minimizer where @xmath2113 .
since the potential is bounded @xmath2114 , one obtains that @xmath2115 if we choose @xmath2116 . on the other hand
, the action along the curve @xmath2117 would be not bigger then @xmath2118 .
the contradiction implies our claim .
let us compare the action of @xmath63 along the curve @xmath2103 with that along the curve @xmath2096 .
if @xmath2119 , by what we have proved , some @xmath2120 exists such that @xmath2121 it follows that @xmath2122(x_1)+[v](x_0))\epsilon^{\lambda}\\ > & \big(\frac 12{c_{22}}d'^2-|\max\partial v|\big)\epsilon^{2\lambda}\end{aligned}\ ] ] it contradicts the minimality of the curve @xmath2103 if @xmath2123 is chosen suitably large .
recall the picture of minimal periodic orbit close to double resonance , one can see from this proposition how the shape of the periodic orbit changes when it moves away from double resonance to single resonance .
let us also start with the hamiltonian @xmath1502 defined by formula ( [ cylindereq3 ] ) , it has two and half degrees of freedom . given any two homology class @xmath1267 , the theorem [ cylinderthm2 ] confirms the existence of two wedge - shaped regions @xmath2124 and @xmath2125 which reach to the boundary of the annulus @xmath2126 for each class in @xmath2124 and @xmath2125 , the aubry set lies in the normally hyperbolic cylinder and , by the result for _ a priori _ unstable systems , can be connected to other aubry set lying in the cylinder under certain generic conditions .
however , it seems unclear whether these two wedges can reach to the flat @xmath1046 .
thus , a notable difficulty rises as these cylinders are separated by an annulus @xmath2127 around the flat @xmath1046 , it is the problem of crossing double resonance .
it is the goal of this section to find an annulus @xmath2128 where those two wedge - shaped regions are plugged into and for each class in that annulus , the stable set of the aubry set
intersects " the unstable set non - trivially , possibly incomplete . in other words , for each class in this region , the ma set does not cover the whole configuration space . as the first step ,
let us consider the hamiltonian @xmath1577 and study all cases when the ma set covers the whole configuration space . for each @xmath2129 , except for the minimal measure @xmath122 supported on the fixed point @xmath2130 , some minimal measure exists with non - zero rotation vector . in the covering space @xmath2131
: @xmath2132 , a disk @xmath1030 is contained in a strip bounded by two @xmath89-static curves @xmath2133 and @xmath2134 , both curves are in the mather set : @xmath2135 no other @xmath89-static curve in the mather set touches the interior of this strip .
let @xmath2136 and @xmath2137 denote the elementary weak kam solution determined by @xmath2133 and @xmath2134 respectively , we investigate what happens when @xmath2138 holds in this strip . as the configuration space is two dimensional , for each @xmath71 in this region
, @xmath2139 uniquely determines a @xmath89-semi static curve which lies entirely in this strip .
the @xmath89-semi static curves considered here are all determined by @xmath2140 .
it is possible that some curve approaches to the origin as @xmath2141 , in this case , because of * h3 * , it approaches to the curve @xmath2133 ( @xmath2142 ) as @xmath2143(@xmath2144 ) .
supported on the fixed point , the measure @xmath122 is minimal for all @xmath880 .
thus , there always exists some semi - static curve @xmath2145 connecting the fixed point to the support of other @xmath89-minimal measure @xmath90 @xmath2146 as all eigenvalues are assumed different , generically , all minimal homoclinic curves approach to the fixed point in the direction @xmath2147 , associated to the smallest eigenvalue : @xmath2148 but this does not exclude the possibility that some @xmath89-semi static curves approach to the point in the direction of @xmath2149 .
it provides us a criterion to classify the cases when the ma set covers the whole configuration manifold .
* case 1 * : no @xmath89-semi static curve approaches the origin in the direction of @xmath2147 . in this case , as @xmath2150 , there exist exactly two semi - static curves @xmath2151 such that @xmath2152 as @xmath1529 .
they approach the origin in the direction @xmath2149 and @xmath2153 other cases are classified under the condition that there exist some @xmath89-static curves approaching the origin in the direction of @xmath2147 . since the curves @xmath2133 as well as @xmath2154 is disjoint with the origin , some number @xmath54 exists such that these two curves do not hit the ball @xmath1030 .
the number @xmath49 seems depending on @xmath89 .
let @xmath2155 be a semi - static curve approaching the origin as @xmath1345 , it intersects the circle @xmath2156 at some point .
let @xmath2157 be such a set that passing through each point @xmath2158 a @xmath89-semi static curve approaches to the origin , as @xmath1529 , in the direction of @xmath2147 . by assumption ,
the set @xmath1344 is not empty .
obviously , @xmath1344 does not occupy the whole circle and can be made closed by adding at most two points , through which some semi - static curves approach the origin in the direction of @xmath2149 .
passing through a point @xmath2159 very close to @xmath1344 , there is a unique @xmath89-semi static curve , determined by @xmath2160 . because of proposition [ flatpro2 ]
, the curve @xmath2161 will get very close to the origin and leave in a direction far away from @xmath2147 .
let @xmath1355 be a connected component of @xmath1344 , it may be a point or an interval .
if it is a point , let @xmath2162 be two sequences of points such that they approach @xmath1355 from different sides .
let @xmath613 ( @xmath2163 ) be the semi static curve passing through @xmath2164 ( @xmath2165 ) , it shall intersect the circle @xmath2156 at a point @xmath2166 ( @xmath2167 ) respectively . some @xmath2168 exist so that @xmath2169 , @xmath2170 as @xmath625 .
if @xmath2171 , it determines a @xmath89-semi static curve approaches the origin as @xmath2172 . because of proposition [ flatpro2 ] , it approaches in the direction of @xmath2149 .
this leads to * case 2 * : there exists exactly one @xmath89-semi static curve approaching origin in the direction of @xmath2149 . if @xmath2173 , let @xmath2174 denote the arc bounded by these two points , not containing @xmath1355 .
one can see from the proof of proposition [ flatpro2 ] that the angle of this arc is not smaller than @xmath2175 . passing from each point in the interior of the arc , the @xmath89-semi static curve approaches to the origin as @xmath2143 and these curves constitute a sector .
since the fixed point is hyperbolic , it has its stable and unstable manifolds @xmath2176 .
thus , some some generating function @xmath2177 and @xmath2178 exist such that @xmath2179 .
as the orbits determined by the curves entirely lie in the unstable manifold of the fixed point , one has @xmath2180 in the sector .
therefore , the size of the sector - shaped region is independent of the size of @xmath49 .
this leads to * case 3 * : in the disk @xmath2181 there is a sector - shaped region with the field angle not smaller than @xmath2175 . in this sector
, one has @xmath2180 .
let @xmath2182 ( @xmath2183 ) be @xmath89-semi static curve passing through a point in @xmath2184 ( @xmath2174 ) respectively , then they approach the origin in opposite direction as @xmath1529 respectively , i.e. @xmath2185 .
to verify this claim , let us assume the contrary .
thus , these two curves cut the ball @xmath1030 into two parts , one is sharp wedge - shaped , denoted by @xmath2186 .
[ fig8 ] we choose a @xmath89-semi static curve lying in @xmath2186 and keeping very close to the curves @xmath2187 and @xmath2155 . in canonical coordinates
such that @xmath2188 , the set @xmath2186 has a vertex at the origin .
as both @xmath2189 and @xmath2190 approach the origin in the direction of @xmath2147 , there exists @xmath2191 such that @xmath2192 if @xmath2193 and @xmath2194 . since the fixed point is hyperbolic , it has local stable and unstable manifold , determined by the generating functions @xmath2195 and @xmath2196 respectively .
restricted in @xmath2186 , these functions satisfy the condition @xmath2197 pick up two points @xmath71 and @xmath353 very close to @xmath2145 respectively , through which some @xmath89-semi static curve @xmath2161 passes , namely , some @xmath2198 exist such that @xmath2199 and @xmath2200 .
note the orbit determined by @xmath2155 ( @xmath2155 ) lies in the stable ( unstable ) manifold , by definition ones has @xmath2201}]\ge\frac 34\big(u_c^-(x')-u_c^+(x)\big)\ge\frac { \lambda_1 ^ 2}4(\|x'\|^2+\|x\|^2).\ ] ] if we choose @xmath71 sharing the same first coordinate with @xmath353 and connect them with a straight line @xmath280 : @xmath2202\to\mathbb{t}^2 $ ] , then @xmath2203 and the action along this curve one has @xmath2204\le o(\|x\|^3)$ ] .
it contradicts the minimality of @xmath2161 , thus the claim is proved .
we claim that @xmath1344 has only one connected component .
otherwise , there would be two connected component @xmath2205 and @xmath1355 . by the definition , passing through a point @xmath2206 ( @xmath2207 ) there is a @xmath89-semi static curve @xmath2208 ( @xmath2209 ) which approaches the curve @xmath2133 as @xmath1345 and approaches the origin as @xmath1345 .
these two curves divided the strip into two parts @xmath2210 , where @xmath2211 is such a strip that passing through any point @xmath2212 , the @xmath89-semi static curve will approach the origin as @xmath1345 .
however , there exists a point @xmath2213 , it implies that passing thorough @xmath837 , the @xmath89-semi - static curve will approach the curve @xmath2134 , namely , it would intersect either @xmath2208 or @xmath2209 .
it is absurd .
thus , we obtain the left picture in figure [ fig9 ] . by similar argument applying to the set @xmath1343 , we have either the case 2 again or * case 4 * : in the disk @xmath2181 there is a sector - shaped region with the field angle not smaller than @xmath2175 , where @xmath2214 , see the right picture in figure [ fig9 ] .
we claim that all of these cases do not occur for generic potential @xmath1182 .
the first two cases takes place at most for four invariant measures , as there are only four curves which approaches the origin in the direction of @xmath2149 .
each of these curves approaches at most one mather set .
these mather sets correspond to at most four edges of @xmath1046 .
let @xmath2215 be a non - negative function such that its support does not touch these four curves as well as the support of the minimal measure . by perturbing the potential @xmath2216
, one can see that the mather set remains unchanged , but the ma set does not cover @xmath1169 for each of these four cohomology classes . both case 3 and 4
take place also for at most four mather sets , as each sector - shaped region has the field angle not smaller than @xmath2175 , and the orbit determined by @xmath2217 approaches one mather set only .
let us destruct it one by one .
if some sector - shaped region @xmath2218 exists where @xmath2219 , @xmath2220 .
we divide it into three sub - sectors @xmath2221 , each of which is composed of @xmath89-semi static curves approaching the origin as @xmath1345 and @xmath2222 is disjoint with @xmath2223 .
we introduce another potential @xmath2224 such that the function @xmath2215 is non - negative , @xmath2225 ( @xmath2226 ) . for
the perturbed lagrangian determined by @xmath2227 , the minimal measure for the class @xmath89 is the same as that for unperturbed hamiltonian .
let @xmath2228 , @xmath2229 be the elementary weak kam solutions of the perturbed hamiltonian , associated to the minimal measure @xmath90 and @xmath2230 respectively , one has @xmath2231 under such perturbation , there might be another cohomology class @xmath968 such that @xmath2232 holds on the whole torus and a sector @xmath2233 exists where @xmath2234 .
note @xmath2235 , we split it into three sub - sectors @xmath2236 , each of which is composed by @xmath968-semi static curves approaching to the origin as @xmath2143 and @xmath2237 is disjoint with @xmath2238 .
we introduce again a perturbed potential @xmath2224 such that the function @xmath2239 is non - negative , @xmath2240 ( @xmath2226 ) . for the new perturbed lagrangian ,
the minimal measure for the class @xmath968 is the same as that for unperturbed one .
let @xmath2241 , @xmath2242 be the elementary weak kam solutions of the perturbed hamiltonian , determined by @xmath2243 and @xmath2244 respectively , one also has @xmath2245 for suitably small @xmath2178 ,
the hamiltonian flow determined by @xmath1577 is well approximated by its linearized flow when they are restricted in the ball @xmath2246 .
for the linearized flow , if @xmath44 is an orbit in the unstable manifold , @xmath2247 is an orbit in the stable manifold .
therefore , some sectors @xmath2248 ( @xmath2249 ) exist so that @xmath2250 , @xmath2251 and @xmath2252 hold for @xmath2249 , each @xmath2253 consists of @xmath89-semi static curves which approach the origin as @xmath1345 with @xmath2254 each @xmath2255 consists of @xmath968-semi static curves which approach the origin as @xmath2143 and @xmath2256 since there are at most two sectors corresponding to unstable manifold and two sectors corresponding to stable manifold , there are at most four pairs of static curves @xmath2257 ( @xmath2258 ) for which the case 3 and 4 takes place .
these four pairs of static curves @xmath2257 ( @xmath2258 ) corresponds to four edges ( points ) contained in @xmath1192 . by construction
, the ma set does not cover the whole torus @xmath1169 for each cohomology class one these four edges .
for any other class @xmath1295 , the ma set can not cover the whole torus also .
otherwise , there would be a sector @xmath2259 of @xmath2246 where @xmath2260 , but it is absurd since some @xmath2261 where @xmath2262 holds for some @xmath2263 .
for each @xmath2264 , @xmath2265 determines an orbit of the lagrangian flow which approaches both to the support of @xmath2266 and to the support of @xmath2267 as @xmath1529 , it is impossible . therefore , we have [ beltlem1 ] it is an open and dense condition for the potential @xmath1182 that for all class @xmath1295 , the ma set does not cover the torus : @xmath2268 .
this set contains at most countably many vertexes . indeed ,
if both @xmath2270 and @xmath1192 are non - empty , there do exist countably many vertexes ( cf .
theorem [ flatthm3 ] ) .
let @xmath2271 be an edge joined to other two edges at the vertex @xmath415 , @xmath2272 respectively . by theorem [ flatthm3 ] , the aubry set for @xmath2273 consists of two minimal homoclinic curves @xmath2274 and @xmath1339 .
denote by @xmath2275 the homology class of @xmath1339 , then the matrix @xmath2276 is uni - module . by introducing suitable coordinates on @xmath1169
, we can assume @xmath1341 .
in this coordinate system , @xmath2277 and @xmath2278 . in this figure ,
each unit square represents a fundamental domain of @xmath1169 in the universal covering space , the horizontal line represents the lift of the homoclinic curve @xmath613 , which stays in the aubry set for each @xmath2279 .
the blue dashed lines represent the lift of the @xmath2280 which stays in the abury set for the class at one end - point of @xmath1155 .
the purple dashed lines represents the lift of the @xmath1319 which stays in the abury set for the class at another end - point of @xmath1155 .
let us consider weak kam solution @xmath2281 in the strip bounded by the lines @xmath2282 and @xmath2283 . according to lemma [ cylinderlem1 ] , for @xmath2284 very small
, there exists an interval @xmath2285 such that for each @xmath2286 , the mather set consists of only one closed curve @xmath2287 such that @xmath2288=g_i$ ] . in the universal covering space ,
let @xmath2289 be a component of the lift of @xmath2287 which approaches @xmath2282 as @xmath2290 .
let @xmath2291 be the elementary weak kam solution determined by @xmath2289 . as @xmath2290 , @xmath2292 in @xmath424-topology .
the function @xmath2293 is obtained in the same way .
the function @xmath2294 determines backward semi - static curves approaching to the line @xmath2295 as the time approaches to minus infinity , @xmath2296 determines forward semi - static curves approaching to the line @xmath2295 as the time approaches to positive infinity respectively . if we remove the coercive condition on these weak kam solutions that @xmath2297 determines a backward ( forward ) semi - static curve approaching @xmath2282 ( @xmath2283 ) , then the weak kam depends on @xmath2279 ( no longer elementary ) .
as @xmath2298 for each @xmath2299 , starting from a point close to the line @xmath2282 ( @xmath2283 ) , the backward ( forward ) semi - static curve will approach to @xmath2282 ( @xmath2283 ) .
let @xmath2300 .
for each @xmath2301 , by proposition [ weakpro3 ] , the strip is divided into two connected parts @xmath2302 and @xmath2303 such that @xmath2304 and @xmath2305 .
let @xmath2306 be the forward semi - static curve determined by @xmath2307 , starting from the point @xmath71 . if @xmath2308 , the curve @xmath2309 is calibrated for @xmath2310 which induces @xmath2311 where both @xmath2312 and @xmath2313 exist .
note @xmath2314 where @xmath2315 is the standard projection .
one can see from figure [ fig10 ] that @xmath2316 it follows that the inequality ( [ belteq-1 ] ) also holds for @xmath2317 , which implies that @xmath2318 also .
clearly , @xmath2302 expands and @xmath2303 shrinks as @xmath101 increases . as the limit , we see that @xmath2319 and @xmath2320 occupies the whole strip
. therefore , we have : [ beltpro1 ] assume @xmath2271 be an edge joined to other two edges at the vertex @xmath415 , @xmath2272 respectively .
let @xmath2321 be the globally elementary weak kam for @xmath2322 with @xmath1314 .
then , for each @xmath2323 , the weak kam solution is completely determined by @xmath2324 and @xmath2325 in the following sense : there is a partition @xmath2326 such that @xmath2327 is connected , @xmath2328 where @xmath2329 and @xmath2330 if @xmath2331 , and @xmath2332 . in virtue of this proposition
, we see that , for all @xmath1305 , the set of barrier functions @xmath2333 are determined by countably many weak kam solutions . for each edge @xmath1155
, there is an open - dense set in @xmath2334 space such that @xmath2335 holds for each @xmath2279 if the potential @xmath1182 takes value in this set .
therefore , in virtue of the lemma [ beltlem1 ] , we have [ beltthm1 ] let @xmath2336 .
a residual set @xmath2337 exists such that for each @xmath2338 and for each @xmath1050 , the ma set does not cover the whole configuration space : @xmath2339 .
because of this theorem , we have generic hypothesis ( * h5 * ) : the potential is chosen so that the ma set does not cover the whole torus @xmath2340 for each @xmath1050 .
the hamiltonian under consideration is given by the formula ( [ homogenizedeq1 ] ) which is autonomous .
to start with , let us assume assume @xmath2341 and study elementary weak kam solutions of the hamilton - jacobi equation @xmath2342 where @xmath452 solves the equation @xmath2343 and @xmath1421 .
expanding the hamiltonian into taylor series of @xmath289 , replacing @xmath1923 by @xmath2344 and rescaling @xmath222 by @xmath1483 we obtain the equation @xmath2345 where @xmath2346 is the hessian matrix of @xmath11 in @xmath29 at @xmath2347 , @xmath2348 ( cf . formula ( [ cylindereq3 ] ) ) . in this equation ,
only higher order term depends on the time @xmath157 . as the first step to study the weak kam of this hamilton - jacobi equation , we omit the higher order term and let @xmath2349 . since the potential @xmath1182 is independent of the time @xmath157 , all elementary weak kam solutions discussed in the last two subsections solve the equation @xmath2350 and for each @xmath2351 and each @xmath1050 , the weak kam solutions of this equation define a ma set which does not cover the whole torus : @xmath2352 . by the upper semi - continuity of the set of semi - static curves , some small @xmath2353 exists such that for each positive @xmath2354 and each @xmath2355 ( we use @xmath2356 to denote the @xmath95-function determined by the hamiltonian @xmath1577 ) the ma set does not cover the torus .
it implies that @xmath2357 holds if @xmath2358 , both @xmath2359 and @xmath2360 are the weak kam solutions of the equation [ belteq2 ] with @xmath2354 . for each average action @xmath2361 ,
the dynamics on the energy level @xmath2362 is similar to twist and area - preserving maps .
first of all , the rotation vector of each minimal measure is not zero .
thus , any minimal measure is not supported on fixed points .
secondly , for each class @xmath2358 , all @xmath89-minimal measures share the same rotation rotation direction , otherwise , the lipschitz graph property will be violated .
if the rotation direction is rational , each ergoidc minimal measure is supported on a periodic orbit . from these properties
one derives the following : for each @xmath2355 , there exists a circle @xmath2363 such that each semi - static curve passes through it transversally and @xmath2364 .
since the set @xmath2365 is closed , there exist finitely many intervals @xmath2366 disjoint to each other such that @xmath2367 .
as these functions are independent of @xmath1483 , all of these functions can be thought as the weak kam solutions of the homogenized hamilton - jacobi equation @xmath2368 if they are thought as the function of the variable @xmath2369 .
here , the cohomology class takes value on the circle : @xmath2355 .
it follows that , for each class @xmath2355 , there exists non - degenerate embedded two - torus @xmath2370 and finitely many intervals @xmath2366 disjoint to each other such that each @xmath89-semi - static curve passes through the two - torus transversally and @xmath2371 . here
the circle @xmath2372 is for the time @xmath1483 .
let us return back to the hamilton - jacobi equation ( [ belteq1 ] ) .
recall the normal form of the hamiltonian , we see that in the remainder @xmath201 , one contribution is from @xmath2373 ( see the formula ( [ cylindereq3 ] ) ) , other contributions are independent of @xmath157 . again , by the upper semi - continuity of the set of semi - static curves , we have [ beltthm2 ] under the hypotheses @xmath139*h1@xmath2374h3 , h5*@xmath140 , some positive numbers @xmath2353 and @xmath2375 exist , depending on the potential @xmath1182 , such that for each @xmath2376 , each @xmath2377 and each @xmath2378 , all semi - static curves pass through transversally the two - torus @xmath2379 , @xmath2380 where @xmath2381 is a circle located in a 2-torus @xmath2382 , @xmath2366 are closed intervals , disjoint to each other and independent of the time @xmath157 . here , the semi - static curves are in the sense of extended configuration space , i.e. if @xmath2383 is a curve , we also call its graph a curve @xmath2384
. let us go back to the original scale . by theorem [ beltthm2 ] ,
there exist a annulus - shaped region @xmath2385 such that the condition ( [ belteq3 ] ) holds for each @xmath2386 .
recall this time - periodic system is deduced from the autonomous system restricted on certain energy level @xmath2387 . in virtue of theorem [ flatthm5 ] , the counterpart of @xmath2388 in @xmath2389 is @xmath2390 where we notice that the sphere @xmath2391 is located in the upper half space of @xmath20 and touches the plane @xmath2392 where @xmath880 . in
the original coordinates @xmath2393 , theorem [ beltthm2 ] states such a fact : for each @xmath2394 , all @xmath1442-semi static curves pass through the 2-torus @xmath2395 transversally and all intersection points are restricted in the strips @xmath2396 .
however , the condition ( [ belteq3 ] ) dost not guarantee complete intersection " of the stable set with unstable set , in the sense that the set @xmath2397 contains some disconnected points . therefore , we call @xmath2398 the annulus of incomplete intersection . in this case , we do not expect to construct orbits connecting each aubry set to any other aubry set nearby , possible incompleteness may block some direction
. however , once non - trivial intersection exists , it opens way to connect some aubry set nearby .
we shall show it in the subsection 7.2 .
as @xmath2399 provided @xmath271 is sufficiently small and @xmath1068 , we obtain the following : * overlap property * : _ given any two irreducible @xmath1267 , there exists a positive number @xmath2400 such that the wedge - shaped regions intersects the annulus - shaped region : @xmath2401 and @xmath2402 _ provided @xmath2403 .
to construct orbits connecting some aubry set to another one nearby , we introduce two types of modified tonelli lagrangian , namely , the time - step and the space - step lagrangian . they satisfy the conditions of positive definiteness , super - linear growth and completeness . the time - step lagrangian @xmath2404 is not periodic in @xmath832 on the whole @xmath2405 , instead , it is periodic when it is restricted either on @xmath2406 or on @xmath2407 , i.e. @xmath2408 if @xmath2409 or @xmath2410 .
the second type of lagrangian is defined on some covering space .
let @xmath2411 .
the space - step lagrangian @xmath2412 is not periodic in one component of spaces coordinates @xmath2413 .
it is periodic in @xmath2413 when it is restricted either on @xmath2406 or on @xmath2407 , i.e. @xmath2414 if @xmath2415 or @xmath2416 .
the existence of local connecting orbits is established based on some upper semi - continuity of minimal curves for the modified lagrangian . * time - step lagrangian * : let us consider time - step lagrangian first
. a curve @xmath2383 is called minimal if @xmath2417 holds for any @xmath2418 and for any absolutely continuous curve @xmath2419\to m$ ] with @xmath2420 and @xmath2421 .
let @xmath2422 denote the set of minimal curves for @xmath63 .
let @xmath2423 , @xmath2424 where @xmath2425 is the standard projection .
[ semicontinuitythm1 ] the set - valued map @xmath2426 is upper semi - continuous .
consequently , the map @xmath2427 is also upper semi - continuous .
let @xmath359 be the diameter of the closed manifold @xmath61 , namely , @xmath2428 where @xmath2429 denotes the length of the shortest geodesic connecting @xmath71 with @xmath353 .
let @xmath2430 as @xmath63 is assumed periodic for @xmath2431 as well as for @xmath2432 , @xmath2433 is finite .
let @xmath75 be a shortest geodesic connecting the point @xmath71 to the point @xmath353 .
given time interval @xmath2434 $ ] with @xmath2435 , we re - parameterize the geodesic @xmath2436 by @xmath2437 , then @xmath545 : @xmath2434\to m$ ] is @xmath74-curve such that @xmath2438 , @xmath2439 .
clearly , the action along this curve is not bigger than @xmath2440 . obviously , there is an upper bound uniformly for all minimizing action of @xmath2441 if it is close to @xmath63 on @xmath2442 , still denoted by @xmath2443 if the lagrangian has super - linear growth , some positive numbers @xmath2444 exist such that @xmath2445 for all @xmath2446 and for all @xmath2441 close to @xmath63 . therefore
, if @xmath75 is a minimizer , one obtains @xmath2447 as ( [ semicontinuityeq1 ] ) holds for any @xmath2435 , it implies that there must be some @xmath2448 $ ] for each @xmath2449 such that @xmath2450 .
as it holds for any @xmath2451 , therefore , some positive number @xmath2452 exists such that @xmath2453 holds for all @xmath2454 $ ] and for all relevant @xmath349 .
it implies that @xmath2455 holds for all @xmath2456 $ ] .
let @xmath2457 be a sequence converging to @xmath63 in the following sense : there exists some @xmath2458 , as well as a sequence of @xmath2459 as @xmath625 such that @xmath2460 .
let @xmath2461 : @xmath2434\to m$ ] be the minimizer of @xmath2462 with @xmath2463 , we then have @xmath2464 for all @xmath2465 $ ] .
the set @xmath2466 is compact in the @xmath2467,m)$]-topology . indeed , since @xmath2468 is positive definite one can write the lagrange equations in the form of @xmath2469 , which implies @xmath2461 is bounded in @xmath62-topology .
let @xmath2470 : @xmath2434\to m$ ] be one of the accumulation points of this set . clearly , @xmath75 : @xmath2434\to m$ ] is the minimizer of @xmath63 .
let @xmath2471 $ ] and let @xmath2472 and @xmath2473 , we obtain a sequence of minimizers of @xmath2462 , @xmath613 : @xmath2474 . by diagonal extraction argument some subsequence of @xmath613 which converges @xmath74-uniformly on each compact set to a @xmath74-curve @xmath75 : @xmath76 .
obviously , it is a minimal curve of @xmath63 .
this proves the upper semi - continuity . in application
, the set @xmath2422 seems too big for the construction of connecting orbits . for time - periodic lagrangian
, ma set can be a proper subset of @xmath2475 , @xmath2476 .
it is closely related to the problem whether the lax - oleinik semi - group converges or not ( cf .
@xcite ) . for time - step lagrangian , pseudo connecting curve
is introduced to play roles similar to what semi - static curve does .
each time - step lagrangian @xmath63 uniquely determines two time - periodic lagrangian @xmath2477 and @xmath2478 such that @xmath2479 and @xmath2480 .
let @xmath2481 denote the minimal average action of @xmath2482 .
for @xmath2483 and @xmath2484 , we define @xmath2485 clearly the limit infimum is bounded @xmath2486 let @xmath2487 and @xmath2488 be the sequence of positive integers such that @xmath2489 ( @xmath2490 ) as @xmath625 and the following limit exists @xmath2491 let @xmath2492 : @xmath2493\to m$ ] be a minimizer connecting @xmath2494 and @xmath2495 @xmath2496 from the proof of theorem [ semicontinuitythm1 ] one can see that for any compact interval @xmath2497 $ ] there is some @xmath2498 such that the set @xmath2499 is pre - compact in @xmath2500,m)$ ] .
[ semicontinuitylem1 ] let @xmath75 : @xmath76 be an accumulation point of @xmath2466 .
then for @xmath2501 @xmath2502)=&\inf_{\stackrel{s_1-s\in\mathbb{z } , \tau_1-\tau\in\mathbb{z } } { \stackrel{s_1,\tau_1\ge\delta}{\stackrel{\gamma^*(-s_1)=\gamma(-s)}{\scriptscriptstyle \gamma^*(\tau_1 ) = \gamma(\tau)}}}}\int_{-s_1}^{\tau_1}l(d\gamma^*(t),t)dt \\ & + ( s_1-s)\alpha^-+(\tau_1-\tau)\alpha^+.\notag\end{aligned}\ ] ] : to prove the lemma let us suppose the contrary .
thus there would exist @xmath2503 , @xmath2504 , @xmath2505 , @xmath2506 and a curve @xmath2507 : @xmath2508\to m$ ] with @xmath2509 , @xmath2510 such that @xmath2511)\ge\int_{-s_1}^{\tau_1}l(d\gamma^*(t),t)dt+(s_1-s)\alpha^-+(\tau_1-\tau)\alpha^+ + \delta.\ ] ] let @xmath2512 . by the definition of limit
infimum there exist @xmath2513 and @xmath2514 such that @xmath2515 and there exist subsequences @xmath2516
( @xmath2490 ; @xmath2517 ) such that @xmath2518 , @xmath2519 and @xmath2520 holds for each @xmath272 .
let @xmath2521 be the minimizer of @xmath2522 . by taking a subsequence further
one can assume @xmath2523 .
in this case , for sufficiently large @xmath77 , we are able to construct a curve @xmath2524 : @xmath2508\to m$ ] which has the same endpoints as @xmath2521 : @xmath2525 , @xmath2526 and satisfies the following @xmath2527)\ge\int_{-s_1}^{\tau_1}l(d\gamma_{i_k}^*(t),t)dt + ( s_1-s)\alpha^-+(\tau_1-\tau)\alpha^++\frac 23\delta.\ ] ] extending @xmath2524 from @xmath2508 $ ] to the @xmath2528 $ ] by @xmath2529 and defining @xmath2530 , @xmath2531 we find that @xmath2532)+t'_0\alpha^-+t'_1\alpha^+\\ \le & a_{l}(\gamma_{i_k}|[-t_0^{i_k},t_1^{i_k}])+t^{i_k}_0\alpha^-+t_1^{i_k}\alpha^+-\frac 23\delta\\ \le & h_{l}^{\infty}(m_0,m_1)-\epsilon.\end{aligned}\ ] ] but this contradicts the definition of the limit infimum as @xmath2533 and @xmath2534 .
we define so - called pseudo connecting curve set @xmath2535 in application , we usually choose time - step lagrangian so that the aubry set of @xmath2478 is different from that of @xmath2477 .
clearly , for @xmath2536 , the orbit @xmath2537 approaches the aubry set @xmath2538 as @xmath2143 and approaches @xmath2539 as @xmath2540 .
that is why we call it pseudo connecting curve .
let @xmath2541 clearly , if @xmath63 is periodic in @xmath832 , then @xmath2542 and @xmath2543 .
[ semicontinuitythm2 ] the map @xmath2544 is upper semi - continuous . as the special case , the map @xmath2545 as well as the map @xmath2546 is upper semi - continuous . :
let @xmath2547 be a sequence of time - step lagrangian , let @xmath2548 and let @xmath75 be an accumulation point of the set @xmath2549 .
we claim that @xmath2536 .
if @xmath2550 , there would be two point @xmath2436,@xmath2551 connected by another curve @xmath2507 : @xmath2552\to m$ ] and @xmath2503 such that @xmath2553)-n_1\alpha^--n_2\alpha^++\delta\ ] ] where @xmath2554 , @xmath2555 .
since @xmath75 is an accumulation point of @xmath613 , for any small @xmath271 , there would be sufficiently large @xmath349 such that @xmath2556}<\epsilon$ ] , and above inequality also holds for @xmath2557})$ ] .
it follows that @xmath2558 , contradicting to the assumption . * space - step lagrangian * : let @xmath514 and @xmath2411 , where @xmath2405 is for the coordinate @xmath2413 . the space - step lagrangian @xmath63 is introduced to handle the problem of incomplete intersection .
a space - step lagrangian also uniquely determines two lagrangian @xmath2478 and @xmath2477 : @xmath2559 such that @xmath2560 and @xmath2561 if we treat @xmath2482 as its natural lift to @xmath2562 .
let @xmath2563 denote minimal measure of @xmath2482 with @xmath360-cohomology class , @xmath2564 denote the rotation vector .
we assume some conditions on the lagrangian : 1 , @xmath2565 for each ergodic minimal measure @xmath2563 ; 2 , @xmath2566 , without losing of generality , it equals zero ; 3 , @xmath2567 .
it is shown in @xcite that some coordinates exists such that the first condition holds provided @xmath2568 .
as the minimal average action of @xmath2482 is achieved on @xmath2569 with @xmath2570 , one can see that @xmath2571 , so the third condition makes sense . to introduce minimal curve for space - lagrangian ,
we define @xmath2572 [ semicontinuitylem2 ] if the rotation vector of each ergodic minimal measure has positive first component @xmath2565 , @xmath2573 , then @xmath2574 let @xmath2575 : @xmath885\to\bar m$ ] be the minimizer of @xmath2576 . let @xmath2577 , @xmath2578 , @xmath280 : @xmath496\to m$ ] be a smooth curve connecting @xmath2495 to @xmath2494 , @xmath2579 and @xmath2580 .
the action of @xmath2477 along @xmath280 is clearly bounded , thus for any @xmath271 , one has @xmath2581 provided @xmath853 is sufficiently large .
the curve @xmath2582 determines a holonomic probability measure @xmath2583 such that @xmath2584 since @xmath2585 is bounded for any @xmath190 , one has @xmath2586 as @xmath1014 . by using the third condition
, we obtain @xmath2587 it implies that @xmath2588 . the case for @xmath2589
is a consequence of the super - linear growth of @xmath63 in @xmath973 . as an intermediate step in introducing pseudo - connecting curve ,
we define a set of minimal curve @xmath2422 .
a curve @xmath2590 is in @xmath2422 if @xmath2591})=\inf_{t'\in\mathbb{r}_+ } h_{l}^{t'}(\bar\gamma(-t),\bar\gamma(t)).\ ] ] we claim that @xmath2592 .
indeed , denote by @xmath2593\to m$ ] the minimizer such that @xmath2594 , @xmath2595 and @xmath2596 because of lemma [ semicontinuitylem2 ] , this infimum is attained for finite @xmath190 if @xmath2597 and @xmath2598 are two different points in @xmath178 .
the super - linear growth of @xmath63 in @xmath973 guarantees that @xmath1014 as @xmath2599 , where @xmath2600 denotes the first coordinate of @xmath2601 .
given an interval @xmath885 $ ] , for sufficiently large @xmath2602 , the set @xmath2603}\}$ ] is pre - compact in @xmath2604,\bar m)$ ] .
let @xmath1014 . by diagonal extraction argument , there is a subsequence of @xmath2605 which converges @xmath74-uniformly on any compact set to a @xmath74-curve @xmath179 : @xmath2606 . obviously ,
@xmath2607 , and [ semicontinuitypro1 ] some number @xmath2083 exists so that @xmath2608 holds for any curve @xmath2609 and any @xmath190 . by the assumption
, one has @xmath2610 .
so , some @xmath2611 exists such that @xmath2612 holds for any interval @xmath2613 provided it is a forward ( backward ) semi - static curves for @xmath2614 ( @xmath2478 ) .
also , some @xmath2615 exists such that @xmath2616 we claim that @xmath2617 .
if there exists some @xmath2607 and some @xmath190 such that @xmath2618 , we join @xmath2619 to @xmath2620 by another curve @xmath2621 where @xmath2622 is a lift of backward semi - static curve @xmath783 for @xmath2283 such that @xmath2623 , denote by @xmath2624 the intersection point of this curve with the section @xmath2625 , @xmath2626 is a lift of forward semi - static curve @xmath784 for @xmath2282 such that @xmath2627 , denote by @xmath2628 the intersection point of this curve with the section @xmath2629 , @xmath280 is a minimal curve of @xmath63 that connects the point @xmath2624 to @xmath2628 .
obviously , one has @xmath2630 , but it contradicts the definition of @xmath2422 .
each @xmath1140 defines a deck transformation @xmath2631 : @xmath2632 .
let @xmath2633 , @xmath2634 .
[ semicontinuitydef2 ] a curve @xmath2607 is called pseudo connecting curve if the following holds @xmath2591})=\inf_{\stackrel{\stackrel{t'\in\mathbb{r}_+ } { \scriptscriptstyle { \bf k}^-\bar\gamma(-t)\in \bar m^-_{\delta}}}{\scriptscriptstyle { \bf k}^+ \bar\gamma(t)\in \bar m^+_{\delta } } } h_{l}^{t'}({\bf k}^- \bar\gamma(-t),{\bf k}^+\bar\gamma ( t))\ ] ] for each @xmath2635 and @xmath2636 .
denote by @xmath2637 the set of pseudo connecting curves .
[ semicontinuitylem3 ] the set @xmath2637 is non - empty .
let us start with a curve @xmath2607 .
given @xmath2503 , if some interval @xmath2638 $ ] exists such that @xmath2639 can be connected to @xmath2640 by another curve @xmath1130 with smaller action @xmath2641})-a_l(\zeta_i)\ge\delta>0,\ ] ] then one obtain a curve @xmath2642}\ast\zeta\ast{\bf k}^-_i \bar\gamma|_{[t^+_i,\infty)}$ ] by one step of such surgery .
given any @xmath2503 , we claim that there are finitely many intervals @xmath2643 $ ] with @xmath2644 such that @xmath2639 can be connected to @xmath2640 by another curve @xmath1130 with the action @xmath2645 smaller than the original one .
let us assume the contrary .
then , for any positive integer @xmath393 , some large @xmath190 exists such that @xmath885\supset \cup_{i=1}^m [ t_i^-,t_i^+]$ ] .
we can choose arbitrarily many of such intervals such that either @xmath2646 or @xmath2647 .
in the first case , let @xmath2648 and @xmath2649 . by assumption ,
these two points can be connected by a curve @xmath280 along which the action @xmath2650 as it follows from proposition [ semicontinuitypro1 ] that @xmath2651})\le k$ ] . since @xmath393 can be arbitrarily large , it implies the existence of a curve along which the action of @xmath63 approaches to minus infinity , it also contradicts proposition [ semicontinuitypro1 ] .
given a curve @xmath2607 and any small @xmath610 , by finitely many steps of such surgery , we obtain a curve @xmath2652 with following properties : 1 , for each small @xmath610 , some large @xmath611 exists such that @xmath2653 , @xmath2636 and @xmath2654})\le\inf_{\stackrel{\stackrel{t'\in\mathbb{r}_+ } { \scriptscriptstyle { \bf k}^-\bar\gamma(-t)\in \bar m^-_{\delta}}}{\scriptscriptstyle { \bf k}^+ \bar\gamma(t)\in \bar m^+_{\delta}}}h_{l}^{t'}({\bf k}^-\bar\gamma_i(-t),{\bf k}^+\bar\gamma_i(t ) ) + \epsilon_i.\ ] ] 2 , @xmath909 is smooth everywhere except for two points which fall beyond the region @xmath2655 , and @xmath2656 as @xmath2657 .
let @xmath2658 such that @xmath2659 .
because of lemma [ semicontinuitylem2 ] , we see that @xmath2660 as @xmath2656 . in virtue of the argument before , for any large @xmath853 @xmath1062 @xmath2661 such that the set @xmath2662}:i\ge i_0\}$ ] is pre - compact in @xmath2604,\bar m)$ ] .
let @xmath1014 , by diagonal extraction argument , there is a subsequence of @xmath2663 which converges @xmath74-uniformly on each compact set to a @xmath74-curve @xmath179 : @xmath2606 .
obviously , @xmath2664 .
[ semicontinuitythm3 ] the map @xmath2544 is upper semi - continuous .
let @xmath2665 , @xmath2547 .
if @xmath2663 converges @xmath74-uniformly on each compact set to a @xmath74-curve @xmath179 , it is obvious that @xmath2664 .
it is an immediate consequence of definition [ semicontinuitydef2 ] that [ semicontinuitypro ] if the space - step lagrangian @xmath63 is periodic in @xmath2413 , then a curve @xmath2664 if and only if its projection @xmath2666 : @xmath76 is semi - static .
similar to the definition for time - step lagrangian , we define @xmath2667 if @xmath63 is periodic in @xmath2413 , then @xmath2668 and @xmath2669 .
an orbit @xmath462 ( a curve @xmath75 ) is said connecting one aubry set @xmath161 to another one @xmath2670 if the @xmath95-limit set of the orbit @xmath462 is contained in @xmath161 and the @xmath127-limit set is contained in @xmath2670 .
it is called local connecting orbit if these two classes are close to each other .
it is called global when the two classes are far away from each other . in this subsection , we show how to construct local connecting orbits of type-@xmath89 by using so - called @xmath89-equivalence .
this type of connecting orbits are found in the annulus of incomplete intersection and plays key role in establishing transition chain crossing strong double resonance .
for this purpose , we use the new version of @xmath89-equivalence introduced in @xcite .
the concept of @xmath89-equivalence was introduced in @xcite for the first time , but it does not apply in interesting problems of autonomous system . the new version is defined not on the whole @xmath61 , but on a non - degenerate embedded @xmath877-dimensional torus .
we call @xmath2671 non - degenerately embedded ( @xmath2672)-dimensional torus by assuming a smooth injection @xmath2673 : @xmath2674 such that @xmath2671 is the image of @xmath2673 , and the induced map @xmath2675 : @xmath2676 is an injection .
let @xmath2677 be a connected set where we are going to define @xmath89-equivalence .
for each class @xmath2678 , we assume that there exists a non - degenerate embedded @xmath877-dimensional torus @xmath2679 such that each @xmath89-semi static curve @xmath75 transversally intersects @xmath2671 .
let @xmath2680 here @xmath2681 : @xmath2682 denotes inclusion map .
@xmath2683 is defined to be the annihilator of @xmath2684 , i.e. if @xmath2685 , then @xmath2686 if and only if @xmath2687 for all @xmath2688 . clearly ,
@xmath2689 note that there exists a neighborhood @xmath440 of @xmath2690 such that @xmath2691 and @xmath2692 ( see @xcite ) .
we say that @xmath2693 are @xmath89-equivalent if there exists a continuous curve @xmath1104 : @xmath496\to \mathfrak{c}$ ] such that @xmath2694 , @xmath2695 , @xmath2696 keeps constant for all @xmath2697 $ ] , and for each @xmath2698 $ ] there exists @xmath54 such that @xmath2699 whenever @xmath2697 $ ] and @xmath2700 .
let @xmath2701 be the standard basis of @xmath2702 , one obtains @xmath5-dimensional vectors @xmath2703 . because @xmath2673 is injection
, there is a vector @xmath2704 such that the @xmath2705 matrix @xmath2706 is uni - module , i.e. @xmath2707 . in new coordinates system
@xmath2708 , the lagrangian @xmath2709 is also @xmath2710-periodic in @xmath71 . in new coordinates ,
let @xmath2711 be a covering space of @xmath583 , @xmath2712 .
the lift of @xmath2671 , @xmath2713 has infinitely many compact components @xmath2714 . if @xmath2673 is linear , @xmath2715 . for the section @xmath2716 we have @xmath2717 while for @xmath2718 , the lift @xmath2719 consists of only one connected component .
[ typecthm1 ] assume the cohomology class @xmath1087 is @xmath89-equivalent to the class @xmath968 through the path @xmath1104 : @xmath496\to h^1(\mathbb{t}^n,\mathbb{r})$ ] . for each @xmath2697 $ ] , the following are assumed : 1 , there exists a coordinate systems @xmath2720 where the first component of rotation vector is positive , @xmath2721 for each ergodic @xmath2722-minimal measure @xmath2723 ; 2 , for the covering space @xmath2724 in the coordinate system the lift of non - degenerately embedded codimension - one torus @xmath2725 has infinitely many connected and compact components , each of which is also a codimension - one torus .
then there exist some classes @xmath2726 on this path , closed 1-forms @xmath2727 and @xmath2728 on @xmath61 with @xmath2729=c_i$ ] and @xmath2730=c_{i+1}-c_i$ ] , and smooth functions @xmath2731 on @xmath178 for @xmath2732 , such that the pseudo connecting curve set @xmath2733 for the space - step lagrangian @xmath2734 possesses the properties : , each curve @xmath2735 determines an orbit @xmath2736 of @xmath128 ; , the orbit @xmath2736 connects @xmath2737 to @xmath2738 , i.e. , the @xmath95-limit set @xmath2739 and @xmath127-limit set @xmath2740 . by the definition of @xmath89-equivalence , there exists a path @xmath1104 : @xmath496\to h^1(m,\mathbb{r})$ ] with @xmath2741 , @xmath2695 such that for each @xmath2742 ( @xmath2697 $ ] ) on the path , there exists @xmath271 such that @xmath2743 whenever @xmath2744 $ ] and @xmath2745 .
thus , there exist a non - degenerately embedded ( @xmath2672)-dimensional torus @xmath2671 , a closed form @xmath2746 and a neighborhood @xmath440 of @xmath2747 such that @xmath2748=\gamma(s')-c$ ] and @xmath2749 .
in the new coordinates @xmath2750 on the torus as above , the codimension one hypersurface @xmath2751 separates @xmath178 into two parts , the upper part @xmath2752 and the lower part @xmath2753 .
@xmath2754 extends to where the first coordinate @xmath2755 .
let @xmath2756 denotes the @xmath49-neighborhood of @xmath2751 in @xmath178 , we introduce a smooth function @xmath2757)$ ] such that @xmath2758 if @xmath2759 , @xmath2760 if @xmath2761 .
let @xmath2762 and @xmath2763 are closed 1-forms on @xmath61 such that @xmath2764=c$ ] and @xmath2765=c'$ ] .
these forms have natural lift on @xmath178 , with the same notation .
a sufficiently small @xmath54 can be chosen so that @xmath2766 it follows from the upper semi - continuity of @xmath2767 w.r.t .
@xmath63 , we find @xmath2768 if @xmath2769 is @xmath424-sufficiently small . as @xmath2763
is carefully chosen so that its support is disjoint from @xmath440 , each curve @xmath2770 is clearly a solution of the euler - lagrange equation determined by @xmath63 , the term @xmath2769 has no contribution to the equation along @xmath179 . in other words ,
each curve in @xmath2771 generates an orbit @xmath462 of @xmath128 : @xmath2772 .
the definition of @xmath2773 tells us that for each curve @xmath2774 , @xmath2775}$ ] is backward @xmath2722-semi static once @xmath2776}$ ] falls entirely into @xmath2777 , @xmath2778 is forward @xmath2779-semi static once @xmath2780 falls entirely into @xmath2781 .
therefore , @xmath2782 as @xmath2143 and @xmath2783 as @xmath1345 .
because of the compactness of @xmath496 $ ] , there are finitely many numbers @xmath2784 $ ] such that above argument applies if @xmath157 and @xmath2785 are replaced respectively by @xmath2786 and @xmath2787 .
set @xmath2788 .
[ typeccor1 ] let @xmath415 , @xmath2727 , @xmath2728 and @xmath2731 be evaluated as in theorem [ typecthm1 ] .
let @xmath2789 be an open neighborhood of @xmath2790 such that @xmath2791 .
then , there exist large @xmath2792 , @xmath2793 and small @xmath54 such that for each @xmath2794 , with @xmath2795 , @xmath2796 , the quantity @xmath2797 reaches its minimum at some @xmath2798 and the corresponding minimizer @xmath2799 satisfies the condition @xmath2800 there is some flexibility to choose the coordinate system and the non - degenerately embedded codimension one torus .
let @xmath2801 : @xmath2802 be a covering space such that @xmath2803 in the coordinate system @xmath2720 .
[ typecdef1 ] for @xmath2697 $ ] , the non - degenerately embedded codimension one torus @xmath2804 is called admissible for the coordinate system @xmath2720 if the lift of @xmath2804 to the covering space @xmath2805 consists of infinitely many connected and compact components , the first component of the rotation vector is positive @xmath2806 for each ergodic @xmath2722-minimal measure .
let us describe how the equivalence relation is established between two classes near strong double resonance .
let @xmath2807 be a curve skirting around the flat @xmath1046 , along which the @xmath95-function keeps constant and the third coordinate @xmath2808 keeps constant as well . for each @xmath2809
, there exists certain coordinate system and finitely many intervals @xmath2810 for @xmath2811-coordinates such that each @xmath89-semi static curve passes through the section @xmath2812 and @xmath2813 clearly , some open set @xmath2814 such that @xmath2815 , from which one obtains that @xmath2816 . for each class @xmath2817 very close to @xmath89 ,
one has @xmath2818 , thus , there exists a closed 1-form @xmath2763 such that @xmath2819=c'-c$ ] and @xmath2820 therefore , all classes along the curve @xmath1104 are equivalent in this case .
another type of local connecting orbits look like heteroclinic orbits .
therefore , we call them local connecting orbits of type-@xmath11 .
it is used to handle a typical case when an aubry set falls in a neighborhood @xmath187 of some lower dimensional torus such that @xmath2821 .
equivalence relation seems not exist among those classes if the aubry sets is located in @xmath187 .
however , each of these aubry sets has homoclinic orbit , it may lead to the existence of heteroclinic orbits . towards this goal ,
let us work in suitable finite covering manifold @xmath578 : @xmath2822 .
in this covering space , these homoclinic orbits turn out to be semi - static orbits .
we assume that the aubry set @xmath2823 consists of finitely many classes @xmath2824 @xmath2825 , @xmath458 is chosen so that the lift of @xmath187 , @xmath2826 with @xmath521 , @xmath2827 and @xmath2828 provided @xmath686 . in the following ,
we denote by @xmath2829 the open neighborhood such that each @xmath2829 contains one aubry class @xmath2830 . if an aubry set contains finitely many static classes only , denoted by @xmath2831 ( @xmath2832 ) , then these classes are transitive in the following sense : by rearranging the subscripts , there exist @xmath77 semi - static curves @xmath2833 @xmath2834 such that @xmath2835 and @xmath2836 @xcite
. it does not exclude the case that some semi - static curve @xmath2837 exists such that @xmath2838 @xmath2834 , @xmath2839 and @xmath2840 .
we say that @xmath2831 is connected to @xmath2841 through @xmath2842 with @xmath2843 if there exist semi - static curves @xmath2844 such that @xmath2845 and @xmath2846 .
the aubry set @xmath2847 is said to be directly connected to the aubry set @xmath2848 if a semi - static curve @xmath75 : @xmath76 exists such that @xmath2849 and @xmath2850 . that @xmath2847 is directly connected to @xmath2848
does not imply that @xmath2848 is directly connected to @xmath2847 .
pick up two points @xmath2851 , @xmath2852 , we consider the quantity @xmath2853 by standard notation , @xmath2854 let @xmath2855 : @xmath885\to m$ ] be the minimal curve realizing the quantity @xmath2856 .
let @xmath2857 $ ] be the sub - interval of @xmath885 $ ] such that @xmath2858 for @xmath2859 but @xmath2860 and @xmath2861 . in the case that @xmath2847 is directly connected only to @xmath2848 , @xmath2862 is upper bounded uniformly for @xmath190 . some sequence of time @xmath2863 and a positive number @xmath2503 such that @xmath2864\subset ( t_{i , t},t_{j , t})$ ] for sufficiently large @xmath853 .
the set of curves @xmath2865}\}$ ] is compact in @xmath74-topology .
let @xmath2866}$ ] be the accumulation point which can be uniquely extended to whole line @xmath75 : @xmath76 .
clearly , @xmath2867 and @xmath2868 .
if @xmath2847 is directly connected also to other @xmath2869 , one can also obtain such a sequence of curves by introducing small perturbation so that @xmath2847 is directly connected only to @xmath2848 and the support of the perturbation does not touch the semi - static curves connecting @xmath2847 to @xmath2848 . given a semi - static curve
one can choose an @xmath877-dimensional disk @xmath2870 intersecting the curve transversally .
this disk also intersects semi - static curves nearby .
a semi - static curve is said _ disconnected _ to other semi - static curves if the intersection point is disconnected to the intersection points of all other semi - static curves .
[ typehthm1 ] ( connecting lemma ) assume that the aubry set contains finitely many classes @xmath2871 , there exist open domains @xmath2872 such that @xmath2873 for each @xmath2874 and @xmath2828 provided @xmath686 .
if each semi - static curves connecting different aubry sets is disconnected to all other semi - static curve , then there exists some orbit @xmath2875 of @xmath128 connecting @xmath161 to @xmath2876 provided @xmath2877 , the class @xmath968 is close to the class @xmath89 , @xmath2878 and two sets @xmath2829 , @xmath2879 exist such that @xmath2880 and @xmath2881 . in autonomous case , @xmath161 can be connected to @xmath2670 only if @xmath2877 .
if both @xmath89 and @xmath968 are the minimal points of the @xmath95-function , then @xmath2882 ( see @xcite ) , it is trivial to connect an aubry set to itself . thus we only need to work on the energy level set @xmath2 with @xmath4 , the minimum of the @xmath95-function . in this case
, we obtain from @xcite that [ typehpro1 ] let @xmath2883 be an autonomous lagrangian of tonelli type , the class @xmath89 not be the minimal point of the @xmath95-function , and @xmath2884 be the flat of the @xmath98-function such that @xmath2885 then , there exists a coordinate system such that each rotation vector in this flat has positive first component @xmath2886 .
the existence of such connecting orbits is derived from the upper - semi continuity of pseudo - connecting orbit set ( see definition [ semicontinuitydef2 ] ) . for the definition of this set in autonomous case
, we need to work in certain covering space @xmath2887 where @xmath2888 holds for each ergodic minimal measure @xmath90 . by proposition [ typehpro1 ] ,
it is possible if we choose suitable coordinate system .
let @xmath179 denote the lift of the curve @xmath2383 , @xmath2889 denote the first coordinate .
let @xmath2890 be a codimension one hyperplane separating @xmath178 into two parts , the upper part @xmath2752 connected to @xmath2891 and the lower part @xmath2892 connected to @xmath2893 .
let @xmath2894 denote the @xmath49-neighborhood of @xmath2895 in @xmath178 , we introduce a smooth function @xmath2896)$ ] such that @xmath2897 if @xmath2898 , @xmath2899 if @xmath2900 .
let @xmath2762 and @xmath2763 be closed 1-forms on @xmath61 such that @xmath2764=c$ ] and @xmath2765=c'$ ] .
they have natural lift on @xmath178 .
let @xmath2901 .
we carefully choose smooth function @xmath2902 such that @xmath2903 as @xmath2904 ( the construction will be demonstrated later ) and let @xmath2905 let @xmath2906 be two points in @xmath178 , we define @xmath2907 for small @xmath122 and @xmath2908 , the lagrangian @xmath2909 satisfies the conditions required for space - step lagrangian . in the following
we shall use the notation @xmath2910 to denote the relevant set of the pseudo - connecting curves .
let us recall a graph property .
given two aubry classes @xmath2831 and @xmath2841 , let @xmath2911 be the set of all semi - static orbits whose @xmath95-limit set is in @xmath2831 and the @xmath127-limit set is in @xmath2841 .
let @xmath2912 , where @xmath2913 denotes the standard projection @xmath2914 .
then , the inverse of @xmath2913 , restricted on @xmath2915 , is lipschitz .
the proof is the same as that for the graph property of aubry set .
if @xmath2831 is directly connected to @xmath2841 , there exists a semi - static orbit @xmath2916 connecting @xmath2831 to @xmath2917
. pick up a curve @xmath2918 in the lift of @xmath2919 to @xmath178 such that its intersection point @xmath2092 with the section @xmath2920 is not close to @xmath2921 , the lift of @xmath2922 to @xmath178 .
denote by @xmath2923 the velocity of @xmath2919 at @xmath2924 , obviously , @xmath2925 .
let @xmath2926 be a smooth function in @xmath157 such that @xmath2927 for @xmath2928 , @xmath2929 for @xmath2930 and @xmath2931 for @xmath2932 , where @xmath54 is suitably small .
let @xmath2933 , then @xmath2934 where @xmath2935 .
we choose an @xmath877-dimensional plane @xmath2936 .
since the set of semi - static curves is totally disconnected , we can choose , for each @xmath2937 $ ] , two suitably small @xmath877-dimensional topological disks @xmath2938 located in @xmath2939 and small @xmath2940 such that @xmath2941 , @xmath2942 , certain semi - static curve @xmath2919 passes through the disk @xmath2943 and no semi - static curve in @xmath2915 passes through @xmath2944 .
these disks can be chosen so that the hausdorff distance @xmath2945 and @xmath2946 as @xmath2947 . let @xmath2948}d'_{ij , s}$ ] , @xmath2949}d_{ij , s}$ ] .
we choose a smooth non - negative function @xmath2950 : @xmath2951 such that @xmath2952 and @xmath2953 if @xmath2954 . for different @xmath2955 ,
it is possible that @xmath2956 .
but it does not make trouble , as @xmath2957 .
let @xmath2958 be the graph of a lipschitz map @xmath2959 containing @xmath2911 .
therefore , we can choose a smooth function @xmath2960 : @xmath2961 $ ] such that @xmath2962 when @xmath2963 and @xmath2964 when @xmath2965 , where @xmath2966 are small numbers .
as there are finitely many aubry classes , we have @xmath2967 if @xmath2955 .
let us consider what curves contained in the set @xmath2968 by assuming @xmath2969 since the term @xmath2970 for @xmath2971 , we do not care about how @xmath2950 is defined on those @xmath877-dimensional plane @xmath2939 with @xmath2972 $ ] .
the set @xmath2973 may extend to infinity , but it does not make trouble since the support of @xmath2950 is contained in @xmath2943 when @xmath2974 .
let @xmath2975 be the function defined on @xmath2559 such that @xmath2976\ } } = \psi_0 $ ] . by the construction of @xmath2908
we see that @xmath2975 is well - defined and smooth .
the aubry set for the lagrangian @xmath2977 is the same as for @xmath2978 .
as there is no semi - static curve of @xmath2978 touches the tube @xmath2979 , each semi - static curve of @xmath2980 also solves the euler - lagrange equation determined by @xmath2977 . because of the upper semi - continuity of @xmath2981
, each semi - static curve for @xmath2982 stays in a small neighborhood of @xmath2983 . since @xmath2984 if @xmath2985 and @xmath2903 if @xmath2986 , a curve is still semi - static for @xmath2977 if it is semi - static for @xmath2978 and passes through the solid cylinder @xmath2987 .
it is based on following argument .
since the 1-form @xmath2988 is closed in @xmath2987 , @xmath2989 on each @xmath2939 with @xmath2972 $ ] , this term has no contribution to the euler - lagrange equation along this semi - static curve , i.e. this curve solves the euler - lagrange equation determined by @xmath2982 also .
any other semi - static curve for @xmath2980 is no longer minimal for @xmath2982 if it connects @xmath2847 to @xmath2990 but does not pass through @xmath2987 , for there exists some semi - static curve @xmath2919 of @xmath2980 passing through @xmath2987 , along which the action is smaller than the action along @xmath2991 .
let us go back to the covering space @xmath178 . for small @xmath2908 , realized by choosing small @xmath2950 ,
each curve @xmath2992 stays in a small neighborhood of certain curve belong to the lift of the semi - static curve .
it is due to the upper semi - continuity of @xmath2544 . by the discussion above
, a curve does not belong to @xmath2968 if its projection does not belong to the aubry set for @xmath2980 , or dose not pass through @xmath2987 although it is semi - static and connects @xmath2993 to @xmath2994 .
let @xmath2918 be a curve in @xmath2968 passing through @xmath2987 .
its projection @xmath2995 connects @xmath2993 to @xmath2994 .
let @xmath2996 with @xmath1140 denote its shift .
each of these curves solves the euler - lagrange equation determined by @xmath2997 . however , except for @xmath2918 , any other curve @xmath2998 with @xmath2999 does not belong to @xmath2968 because they do not pass through @xmath2987 , the action along these curves is bigger than the action along @xmath2918 .
it can be easily seen from the definition [ semicontinuitydef2 ] : minimal property persists under translation . in the cylinder @xmath178
we choose two sections @xmath3000 and @xmath3001 such that : 1 , both are the deformation of the section @xmath3002 , they divide @xmath178 into three parts , @xmath2752 , @xmath2892 and @xmath3003 .
@xmath2752 is homeomorphic @xmath3004 , @xmath2892 is homeomorphic @xmath3005 and @xmath3003 is homeomorphic to @xmath3006 .
let @xmath3007 denote the boundary of @xmath2754 shared with @xmath3003 ; 2 , there exists @xmath3008 such that @xmath3009 ; 3 , for each @xmath3010 , both @xmath3011 and @xmath3012 are connected , i.e. if one moves into @xmath3013 along the curve as @xmath1529 then it stays in @xmath3013 forever .
let @xmath3014 be the tube connecting @xmath2987 to @xmath2752 , @xmath3015 , each @xmath3010 passes through @xmath3014 , does not touch the boundary of @xmath3014 before it moves forward into @xmath2752 .
similarly , we define the tube @xmath3016 connecting @xmath2987 to @xmath2892 such that @xmath3017 , each of those curve passes through @xmath3016 , does not touch the boundary of @xmath3016 before it is going to retreat back into @xmath2892 .
since there are finitely many aubry classes only , by choosing suitably small @xmath3018 we can assume @xmath3019 if @xmath3020 .
a closed 1-form @xmath2763 clearly exists such that @xmath2819=c'-c$ ] and @xmath3021 .
let @xmath3022 $ ] be a smooth function such that @xmath3023 for @xmath2928 , @xmath3024 for @xmath2930 and let @xmath3025 be an open set containing the closure of @xmath3026 and @xmath3027 if @xmath3020 .
we define a smooth function @xmath3028 : @xmath3029 $ ] such that @xmath3030 if @xmath2954 where @xmath3031 and @xmath3032 , @xmath2899 if @xmath3033 and @xmath3034 if @xmath3035 . by the construction of @xmath2754 , @xmath3036 and @xmath3037
, we see the existence of such function .
let us now study the lagrangian @xmath2909 with @xmath2901 . by condition , @xmath2880 , @xmath2881 and @xmath686 .
thus , there exist @xmath3038 and @xmath3039 .
let @xmath3040 and @xmath3041 be two points in @xmath178 such that @xmath3042 and @xmath3043 and let @xmath3044 and @xmath3045 where @xmath3046 .
let @xmath3047 : @xmath885 \to\bar m$ ] be the minimizer of @xmath3048 and let @xmath3049 , we obtain a sequence of @xmath3050 .
let @xmath179 : @xmath3051 be the accumulation point of the sequence .
due to the upper semi - continuity of @xmath3052 with respect to @xmath3053 , the curve @xmath179 must pass through @xmath3054 if @xmath3055 is suitably small .
thus , along the curve @xmath179 the term @xmath3056 does not contribute the lagrange equation , namely , the curve determines an orbit of @xmath128 .
since this curve is in the set @xmath3057 , therefore , it connects @xmath161 to @xmath2670 .
this completes the proof .
the orbit @xmath462 obtained in theorem [ typehthm1 ] is locally minimal in the sense we define in the following .
it is crucial for the variational construction of global connecting orbits .
the set of local minimal curve will not be empty if the aubry set @xmath163 has some totally disconnected minimal homoclinic orbit , the 1-form @xmath122 as well as the function @xmath2908 is carefully chosen for the modified lagrangian . here is the definition for autonomous systems : [ localdef2 ] let @xmath3058 be open domains such that @xmath2828 @xmath2825 .
we assume that @xmath3059 , @xmath2764=c$ ] , @xmath3060=c'$ ] , @xmath2877 and the first component of both @xmath89- and @xmath968-minimal measures is positive @xmath2888 , @xmath3061 .
let @xmath3062 be the covering space , denote by @xmath179 the lift of a curve @xmath2383 .
then , @xmath462 : @xmath3063 is called local minimal orbit of type-@xmath11 that connects @xmath161 to @xmath2670 if 1 , @xmath462 is an orbit of @xmath128 , @xmath3064 and @xmath3065 .
there exist @xmath3066 such that @xmath3067 and @xmath3068 ; 2 , there exist two @xmath877dimensional disks @xmath3069 , @xmath3070 and positive numbers @xmath3071 such that @xmath3072 , @xmath3073 , @xmath75 transversally passes @xmath3074 and @xmath3075 at the time @xmath3076 and @xmath853 respectively , and @xmath3077 holds for each @xmath3078)$ ] , @xmath3079 and @xmath3080 .
where @xmath3081 and @xmath3082 are the sequences such that @xmath3083 and @xmath3084 . in this definition ,
the term @xmath3085 measures the smallest action of @xmath2909 along those curves which join @xmath2494 to @xmath2495 with time @xmath3086 such that @xmath3087 is an accumulation point of these curves as @xmath2172 and @xmath3088 is an accumulation point of the curves as @xmath1345 .
* remark*. in the space of curves , a neighborhood of the curve @xmath75 consists of those curves that start from @xmath3089 and reach @xmath3090 within a time between @xmath3091 and @xmath3092 . different time scale determine orbits in different energy levels , that is why we consider the time scale @xmath3093 $ ] as variable while we search for the local minimum .
* remark*. this definition applies also to the case that there exists only one aubry class staying in the small neighborhood of lower - dimensional torus . in that case
, we can consider a suitable finite covering of the configuration manifold . in the finite covering configuration space
, there are more than one aubry class .
the following is the version for time - periodic systems [ localdef3 ] let @xmath3058 @xmath2825 be open domains with@xmath2828 .
we assume that @xmath3094 , @xmath2764=c$ ] , @xmath3060=c'$ ] .
then , @xmath462 : @xmath3063 is called local minimal orbit of type-@xmath11 that connects @xmath161 to @xmath2670 if 1 , @xmath462 is an orbit of @xmath128 , the @xmath95-limit and the @xmath127-limit sets of @xmath462 are contained in @xmath161 and @xmath2670 respectively , @xmath3095 and @xmath3096 with @xmath686 ; 2 , there exist two open balls @xmath3069 , @xmath3097 and two positive integers @xmath3098 such that @xmath3099 , @xmath3100 , @xmath3101 , @xmath3102 and @xmath3103 holds @xmath2037 @xmath3104 , @xmath3105 , @xmath3106 , where @xmath3107 are the sequences such that @xmath3108 and @xmath3109 .
the set of curves starting from @xmath3089 and reaching @xmath3090 with time @xmath3110 constitutes a neighborhood of the curve @xmath75 in the space of curves .
once a curve @xmath1441 touches the boundary of this neighborhood , the action of @xmath2909 along @xmath1441 will be larger than the action along @xmath75 . as @xmath3089 , @xmath3090 and therefore @xmath1068 can be chosen arbitrarily small , it is reasonable to call it locally minimal .
in this section we show how to construct global connecting orbits by variational method , provided a generalized transition chain exists . in the next section ,
the main result ( theorem [ mainthm ] ) is proved by showing the genericity of such transition chain .
the concept of transition chain was proposed by arnold in his celebrated paper @xcite where it is formulated in geometric language .
the generalized transition chain formulated in our previous work @xcite is in variational version which need less information about the geometric structure .
[ chaindef1 ] ( autonomous case ) let @xmath89 , @xmath968 be two cohomolgy classes in @xmath3111 .
we say that @xmath89 is joined with @xmath968 by a generalized transition chain if a continuous curve @xmath1104 : @xmath496\to h^1(m,\mathbb{r})$ ] exists such that @xmath2694 , @xmath2695 , @xmath3112 and for each @xmath2697 $ ] at least one of the following cases takes place : @xmath3113 , the aubry set is composed of finitely many classes only .
there exist certain finite covering : @xmath3114 , two open domains @xmath3115 with @xmath3116 , an @xmath877 dimensional disk @xmath3117 and small numbers @xmath3118 such that i , the aubry set @xmath3119 , @xmath3120 and @xmath3121 for each @xmath3122 , ii , @xmath3123 is non - empty and totally disconnected ; @xmath3124 , for each @xmath3125 , @xmath2779 is equivalent to @xmath2722 .
some section @xmath2804 and some neighborhood @xmath440 of @xmath3126 exist such that @xmath3127 .
each class @xmath2779 is associated with an admissible section @xmath3128 and an admissible coordinate system @xmath3129 .
* remark*. because of upper semi - continuity of ma set , it is possible that there exist some classes for which both cases take place . in the case
( h1 ) , if the aubry set contains only one aubry class , one can take some finite covering @xmath3130 non trivial if @xmath515 .
a typical case is that @xmath3131 is contained in a small neighborhood of lower dimensional torus .
one takes suitable finite covering space so that @xmath3132 contains exactly two connected components .
if @xmath3131 contains more than one class , we choose @xmath3133 .
the existence of generalized transition chain implies that there exists a sequence of local connecting orbits .
more precisely , there exists a sequence of locally minimal curve @xmath613 , a sequence of numbers @xmath2786 @xmath3134 such that @xmath3135 and @xmath3136 .
global connecting orbits are constructed shadowing these local connecting orbits .
one can also define generalized transition chain for time - periodic systems .
( time - periodic case ) let @xmath89 , @xmath968 be two classes in @xmath3111 .
we say that @xmath89 is joined with @xmath968 by a generalized transition chain if a continuous curve @xmath1104 : @xmath496\to h^1(m,\mathbb{r})$ ] exists such that @xmath2694 , @xmath2695 and for each @xmath2697 $ ] at least one of the following cases takes place : @xmath3113 , the aubry set is composed of finitely many classes only .
there exist certain finite covering : @xmath3114 , two open domains @xmath3137 with @xmath3116 and small number @xmath3118 such that i , the aubry set @xmath3138 , @xmath3139 and @xmath3140 for each @xmath3122 , ii , @xmath3141 is non - empty and totally disconnected ; @xmath3124 , for each @xmath3125 , @xmath2779 is equivalent to @xmath2722 , namely , there exists a neighborhood of @xmath3142 , denoted by @xmath440 , such that @xmath3127 .
given @xmath3143 and @xmath246 , there exists at least a forward ( backward ) @xmath89-semi static curve @xmath2182 : @xmath3144 ( @xmath2183 : @xmath404\to m$ ] ) such that @xmath3145 .
it determines certain velocity @xmath3146 , for almost all points , the velocity is uniquely determined . before proving the main theorem of this subsection ,
let us formulate and prove a proposition .
[ constructionpro1 ] given an aubry set , the aubry distance from any aubry class @xmath2993 to all other aubry classes is assumed have positive lower bound , namely , some @xmath1068 exists such that @xmath3147 for all @xmath3148 .
let @xmath3149 then for all @xmath3150 and for any @xmath3151 one has @xmath3152 for any @xmath3150 and any @xmath3153 one has @xmath3154 : for each pair of points @xmath3155 , we claim that there exists some aubry class @xmath2994 such that @xmath3156 holds for each @xmath3157 .
indeed , let @xmath3158 be a subsequence of integers such that @xmath3159 let @xmath3160 : @xmath3161\to m$ ] be the minimizer for @xmath3162 .
there exists at least one point @xmath469 which is the accumulation point of @xmath3163 .
otherwise , the quantity @xmath3164 as @xmath3158 .
given @xmath3165 , we claim that ( [ constructioneq1 ] ) and ( [ constructioneq2 ] ) hold if @xmath3166 .
let @xmath297 be a sequence such that @xmath3167 and let @xmath3168 : @xmath3169\to m$ ] be the minimizer of @xmath3170 .
there is a positive number @xmath3171 such that the ordinary distance @xmath3172 for any @xmath3173 $ ] and @xmath3148 . otherwise along the curve @xmath3174
there exists a point getting closer and closer to a point @xmath3175 .
consequently , one would obtain from the property that @xmath3147 for each @xmath3148 that @xmath3176 where @xmath3177 . on the other hand
, we have @xmath3178 it is a contradiction . therefore
, some @xmath3177 and @xmath3179 $ ] exist such that @xmath3180 as @xmath297 and @xmath3181 .
this proves ( [ constructioneq1 ] ) in case @xmath3182 . for different points
@xmath3150 and @xmath3183 with @xmath3148 , let @xmath3184 : @xmath472\to m$ ] be the curve which minimizes the quantity @xmath3185 , let @xmath3186 be the subsequence of @xmath77 such that @xmath3187 . in autonomous case , it converges as @xmath3049 .
similarly , we let @xmath3188 : @xmath472\to m$ ] be the curve which minimizes the quantity @xmath3189 , let @xmath3190 be the sequence of @xmath77 such that @xmath3191 .
let @xmath3192 , @xmath3193 : @xmath472\to m$ ] be the minimizer of @xmath3194 and let @xmath305 be the subsequence of @xmath77 such that @xmath3195 . by the proof we just finished , there exists @xmath3196 and integer @xmath3197 $ ] such that @xmath3198 and @xmath3199 as @xmath297 .
let @xmath3200 : @xmath472\to m$ ] be the minimizer of @xmath3201 , @xmath3202 be the subsequence of @xmath77 such that @xmath3203 , let @xmath3204 : @xmath106\to
m$ ] be the minimizer of @xmath3205 , @xmath3206 be the subsequence of @xmath77 such that @xmath3207 . given arbitrarily small @xmath54 , we have sufficiently large @xmath478 , @xmath3190 , @xmath3208 , @xmath3209 , @xmath3210 and @xmath3211 such that @xmath3212 since @xmath3213 , we have @xmath3214 .
consequently , @xmath3215 since @xmath71 is in aubry class @xmath2994 , while @xmath3216 , one has @xmath3217 because @xmath49 can be arbitrarily small , by subtracting ( [ constructioneq3 ] ) from ( [ constructioneq4 ] ) we obtain @xmath3218 it verifies ( [ constructioneq2 ] ) . since ( [ constructioneq2 ] ) holds for each @xmath3183 with @xmath3148 and for any @xmath3150 , ( [ constructioneq1 ] ) hold for each @xmath3151 and for any @xmath3150 .
this completes the proof of the proposition .
[ constructionthm1 ] if @xmath89 is connected to @xmath968 by a generalized transition chain , then 1 , there exists an orbit of the lagrange flow @xmath128 , @xmath462 : @xmath2772 which connects the aubry set @xmath161 to @xmath2670 , namely , @xmath3219 and @xmath3220 ; 2 , given @xmath2451 and arbitrarily small @xmath54 , there exists an orbit @xmath453 of @xmath128 passing through @xmath49-neighborhood of the points @xmath3221 and @xmath3222 successively , namely , @xmath3223 such that @xmath3224 and @xmath3225 .
we only need to study the autonomous case .
time periodic case can be treated in the same way . therefore , one has that @xmath3226 . by adding suitable constant on the lagrangian , we assume @xmath1126 to simplify notation .
first of all , as a generalized transition chain @xmath2722 is assumed , the aubry set @xmath3227 can be connected to some @xmath3228 by locally minimal orbits of either type-@xmath11 , or type-@xmath89 if @xmath2785 close to @xmath157 .
so , there is a sequence @xmath3229 such that for each @xmath3230 , @xmath3231 is connected to @xmath3232 by some local minimal orbits .
the global connecting orbits are constructed shadowing such a sequence of orbits .
let @xmath3233 .
we divide the set @xmath3234 into @xmath393 parts @xmath3235 the rule to make such a partition is that for all @xmath3236 , @xmath3237 is connected to @xmath2738 by a local minimal orbit of the same type .
more precisely , let @xmath3238 and @xmath3239 be the subset of @xmath3240 , @xmath3241 , @xmath3242 . if @xmath3243 , then for all @xmath3244 , @xmath3237 is connected to @xmath2738 by a local minimal orbit of type-@xmath3245 ( @xmath3245 = @xmath89 , or @xmath11 ) .
since the map @xmath3246 is upper semi - continuous , once the ma set @xmath3247 is in the case ( h1 ) ( or h2 ) , then for @xmath2785 sufficiently close to @xmath157 , the set @xmath3247 is also in the case ( h1 ) ( or h2 ) .
thus , for each @xmath3248 , by choosing @xmath3249 and @xmath3250 sufficiently close to @xmath3251 and @xmath3252 respectively , we can assume that both @xmath3249 and @xmath3250 satisfy the condition ( h1 ) also . with the class @xmath415 we associate an admissible coordinate system @xmath3253 and let @xmath3254^t$ ] denote the inverse of @xmath3255 .
because we consider the problem on @xmath2 with @xmath4 , we can choose @xmath3255 for each @xmath3256 ( see proposition [ typehpro1 ] ) such that , in the new coordinate system , the first component of @xmath3257 is positive for each ergodic @xmath415-minimal measure .
in virtue of the upper semi - continuity of ma set on cohomology class , one can assume that @xmath3258 holds for each ergodic component @xmath2266 as @xmath3259 is chosen suitably close to @xmath415 .
it means that the @xmath3260 holds in the coordinates not only determined by @xmath3255 , but also determined by @xmath3261 as well as by @xmath3262 . therefore ,
@xmath1062 @xmath3263 such that @xmath3264 holds for each @xmath415-semi - static curve @xmath613 , where @xmath3265 denotes a curve in the lift of @xmath613 to universal covering space and @xmath3266 with @xmath2198 . as the second step ,
let us describe the minimal properties of local connecting orbits of type-@xmath11 as well as of type-@xmath89 . *
the case of type-@xmath11*. for each integer @xmath3267 , the condition ( h1 ) holds for generalized transition chain .
namely , in certain finite covering space @xmath3114 , the aubry set for @xmath349 and @xmath3268 consists of more than one but finitely many classes @xmath3269 for @xmath1564 . by the assumption of ( h1 ) ,
some open domains @xmath3270 exist such that @xmath3271 , @xmath3272 , @xmath3273 , @xmath3274 and @xmath3275 is totally disconnected , with small @xmath3276 .
the new coordinate system @xmath3253 of @xmath583 is introduced such that ( [ constructioneq5 ] ) holds .
if writing @xmath3277 , we introduce the covering space @xmath3278 . we shall work with this covering space @xmath3279 .
for closed 1-forms @xmath2727 and @xmath2728 on @xmath458 we use the same symbol to denote their natural lift to @xmath178 . recall the proof of theorem [ typehthm1 ] .
some decomposition of @xmath178 exists such that @xmath3280 such that @xmath3281 is diffeomorphic to @xmath3282 , @xmath3283 is diffeomorphic to @xmath404\times\mathbb{t}^{n-1}$ ] and @xmath3284 is diffeomorphic to @xmath3285 .
some open and connected disks @xmath3286 and @xmath3287 exist such that @xmath3288 , @xmath3289 and @xmath3290 , the intersection of any two of these sets is empty and the closure of @xmath3291 , denoted by @xmath3292 , is connected , see figure [ fig11 ] .
as it was studied in the subsection of 6.3 ( local connecting orbit of type-@xmath11 ) , some function @xmath3293 , @xmath3294 : @xmath3029 $ ] , some closed 1-form @xmath2727 , @xmath2728 , @xmath2731 and some small constant @xmath3287 exist such that @xmath2729=c_i$ ] , @xmath2730=c_{i+1}-c_i$ ] , @xmath3295 , @xmath3296 , @xmath3297 , @xmath3298 if @xmath3299 and @xmath3300 if @xmath3301 .
let @xmath3302 , @xmath3303 we introduce a space - step lagrangian for the coordinate system @xmath3304 @xmath3305 in virtue of theorem [ typehthm1 ] , some curve @xmath3306 ( pseudo connecting orbit set ) such that its projection down to @xmath458 , @xmath3307 , determines an orbit @xmath3308 connecting certain aubry class @xmath3309 to another aubry class @xmath3310 .
here , the subscript @xmath349 indicates the aubry set is for the cohomology class @xmath415 , the superscript @xmath300 indicates which aubry class it belongs to .
such curve stays entirely in the interior of @xmath3311 .
therefore , along such curve both @xmath1076 and @xmath3312 do not contribute to the euler - lagrange equation , i.e. @xmath1127 is an orbit of @xmath128 .
as pointed out in definition [ localdef2 ] , such local connecting orbit of type-@xmath11 is endowed with certain kind of local minimality . there
exist two @xmath877-dimensional disks @xmath3069 and @xmath3313 with @xmath3314 , @xmath3315 , large number @xmath3316 , suitably small @xmath3317 and quite small @xmath3318 such that @xmath3319 , @xmath3320 and @xmath3321)\big\}\notag\\ \ge\min&\big\{h_{c_i}^{\infty}(x^-,m_0)+h_{\eta_i,\mu_i,\psi_i}^{t}(\bar m_0,\bar m_1 ) + h_{c_{i+1}}^{\infty}(m_1,x^+):\notag\\ & ( \bar m_0,\bar m_1,t)\in v_i^-\times v_{i+1}^+\times[t_i^+-d_i , t_i^++d_i ] \big\ } + 5\epsilon_i^*,\notag\end{aligned}\ ] ] where @xmath3322 , @xmath3323 .
the disks @xmath3324 and @xmath3325 are chosen so that @xmath1129 intersects them transversally , for each @xmath3326 @xmath3327 $ ] , the minimizer of @xmath3328 , @xmath3329 @xmath3330 has the property @xmath3331.\ ] ] let @xmath3332 be a locally minimal curve such that the orbit @xmath3333 connecting @xmath3334 to @xmath3335 .
denote by @xmath3336 the aubry class which contains the @xmath127-limit set of @xmath3333 .
it is possible that @xmath3337 is different from @xmath3338 which contains the @xmath95-limit set of @xmath1127 .
remember that we have assumed that each aubry set consists of finitely many classes , let s say , @xmath702 classes . by the result in @xcite ,
the subscript of aubry classes can be rearranged such that some @xmath415-semi - static curve @xmath2837 exists such that @xmath3339 and @xmath3340 for @xmath3341 @xmath3342 .
so some positive integer @xmath3343 exists such that @xmath3344 @xmath3342
. let @xmath3345 .
we choose some @xmath877-dimensional small disks @xmath3346 with @xmath3347 such that @xmath3348 , @xmath3349 , @xmath3350 is located within @xmath3351 , a small neighborhood of @xmath3338 such that @xmath3352 holds for each @xmath3353 and each @xmath3354 ( the definition of @xmath3351 is the same as @xmath2829 in proposition [ constructionpro1 ] ) , @xmath2837 intersects @xmath3355 as well as @xmath3356 transversally .
these curves also have locally minimal property similar to the form of ( [ constructioneq6 ] ) : @xmath3357 ) \big\}\\ \ge\min&\big\{h_{c_i}^{\infty}(x^-,m_0)+h_{c_i}^{t}(m_0,m_1)+h_{c_{i}}^{\infty}(m_1,x^+):\notag\\ & ( \bar m_0,\bar m_1,t)\in v_{i , j}^-\times v_{i , j+1}^+\times[t^+_{i , j}-\tau_i , t^+_{i , j}+\tau_i ] \big\ } + 5\epsilon_i^*,\notag\end{aligned}\ ] ] where @xmath3358 , @xmath3359 , @xmath3360 is the time such that @xmath3361 if @xmath3362 . as semi - static curves are totally disconnected , @xmath3355 and @xmath3356
are chosen so that any curve in @xmath3363 does not touch the boundary of @xmath3355 and of @xmath3356 .
note that @xmath3364 in the autonomous case @xcite , we find from proposition [ constructionpro1 ] that , for any @xmath3365 , there exists @xmath3366 such that @xmath3367 holds for each @xmath3368 , each @xmath3369 and @xmath3354 .
let @xmath3370 , @xmath3371 , @xmath3372 . a curve @xmath75 : @xmath3373\to\check{m}$ ]
is called _ admissible _ for @xmath3346 if @xmath3374 where @xmath3375 and @xmath3376 are chosen to satisfy the condition @xmath3377 where @xmath3378 is chosen in the inequality ( [ constructioneq8 ] ) , the condition @xmath3379 which is set so that the minimal curve does not touch the boundary of @xmath3355 provided it passes through @xmath3380 at @xmath3381 and through @xmath3356 at @xmath3382 and the condition @xmath3383 which is set so that the minimal curve does not touch the boundary of @xmath3380 provided it passes through @xmath3384 at @xmath3385 and through @xmath3355 at @xmath3386 .
these conditions define non - empty set for @xmath3387 if we choose suitably large @xmath3388 .
we consider the minimum of the following action among all admissible curves : @xmath3389 let @xmath3390 : @xmath3373\to\check{m}$ ] be the minimizer of the action . if @xmath3391 is sufficiently large , the minimizer is smooth at each @xmath3392 .
first of all , we claim that @xmath3393)$ ] holds for each @xmath3394 .
if it does not hold for certain @xmath3394 , one obtains from ( [ constructioneq8 ] ) that @xmath3395 where @xmath3396 , @xmath3397 , @xmath3398 as well as @xmath3399 is the intersection point of a semi - static curve @xmath3400 with @xmath3355 and with @xmath3356 respectively .
the orbit @xmath3401 connects @xmath3338 to @xmath3402 .
this contradicts the minimality of @xmath75 .
the smoothness follows from the property that @xmath3403 is in the interior of the domain .
if the minimizer @xmath3390 is not smooth at @xmath3404 , we join the points @xmath3405 and @xmath3406 by the minimizer of @xmath3407 as the minimizer approaches @xmath3404 from both sides of @xmath3355 as @xmath832 approaches @xmath3408 from opposite direction @xcite , this minimizer also passes through @xmath3355 .
thus , one obtains a curve @xmath545 by replacing the segment of @xmath3409 with this minimizer .
let @xmath3410 be the time of this curve passing through @xmath3355 , clearly , @xmath3411 , the action along this curve is clearly smaller than the original one . but
this is absurd . *
the case of type-@xmath89*. for each integer @xmath3412 , there exist an admissible section @xmath3413 , a neighborhood @xmath2789 of @xmath3414 , two closed 1-forms @xmath2727 and @xmath2728 on @xmath61 with @xmath2729=c_i$ ] , @xmath2730=c_{i+1}-c_i$ ] and @xmath3415 .
correspondingly , an admissible coordinate system @xmath3416 on @xmath61 is chosen such that in the new coordinates , one has a covering space @xmath3417 : @xmath3418 , the set @xmath3419 consists of infinitely many compact components @xmath3420 @xmath3421 .
@xmath3422 is separated by @xmath3423 into upper part @xmath3424 and lower part @xmath3425 .
a smooth function @xmath3294 : @xmath3029 $ ] is constructed such that @xmath3426 for @xmath3427 and @xmath3428 for @xmath3429 . the number @xmath3287 is chosen so small such that @xmath3430)+\delta_i ) \subset u_i$ ] , ( cf . formula ( [ typeceq1 ] ) ) .
let @xmath3302 . to make notation simpler , for each integer @xmath3431 , we let @xmath3432 , @xmath3433 and @xmath3434 in the coordinate system @xmath3416 ( see the corollary [ typeccor1 ] for the definition of @xmath3435 ) . again , let @xmath3305 since the class @xmath415 is equivalent to the class @xmath2272 and they are close to each other , one sees from theorem [ typecthm1 ] that each curve @xmath3436 determines a locally minimal orbit type-@xmath89 @xmath462 which is an orbit of @xmath128 and connects @xmath3437 to @xmath3438 .
let @xmath3439 , @xmath3440 and let @xmath3441\to\bar m$ ] be the minimizer of @xmath3442 according to lemma [ semicontinuitylem2 ] and corollary [ typeccor1 ] , @xmath1062 @xmath2792 , @xmath3443 , there exists @xmath3444 such that @xmath3445 provided the first coordinate of @xmath3446 as well as of @xmath3447 satisfies the condition that @xmath3448 and @xmath3449 .
the minimizer @xmath3441\to\bar m$ ] satisfies @xmath3450 by the definition , the disks @xmath3451 and @xmath3324 are codimension one torus in different coordinate systems , their relative position needs to be fixed in the universal covering space . for this purpose
, we define the following covering spaces : @xmath3452 where @xmath3453 and @xmath3454 in the coordinate system @xmath3416 . for simplicity of notation and without danger of confusion , we use the same symbol for a fundamental domain of @xmath3455 in @xmath109 , i.e. restricted on @xmath3455 the projection is a homeomorphism , @xmath3456 and @xmath3457 . both @xmath3324 and @xmath3325
are some translation of unit @xmath877-dimensional disk @xmath3458 . clearly , @xmath3459 is parallel to @xmath3460 and there exist @xmath877 irreducible integer vectors @xmath3461 tangent to @xmath3462 ( @xmath3325 ) such that @xmath3463 given a fundamental domain @xmath3464 and a @xmath415-semi static curve @xmath613 , there exists a curve in the lift @xmath613 to the universal covering space , denoted by @xmath3265 , which intersects @xmath3465 .
clearly , one can choose a fundamental domain @xmath3466 ( up to a translation ) so that it intersects the curve @xmath3265 and @xmath3467 is above " the @xmath3468 in the following sense : some suitably large @xmath3469 exists such that @xmath3470 .
note that @xmath3324 is usually not parallel to @xmath3464 .
we say that the two fundamental domains @xmath3464 and @xmath3324 are @xmath3471-related if they satisfy this condition .
let @xmath3464 and @xmath3324 be @xmath3471-related fundamental domains .
given positive integers @xmath3472 , we define @xmath3473 then @xmath3474 .
let @xmath3475 , @xmath3476 one defines the minimal action of @xmath3477 connecting these two points @xmath3478 let @xmath3479 clearly , for fixed @xmath3480 , some positive number @xmath610 and suitably large integer @xmath3481 exist such that @xmath3482 does not touch @xmath3483 , @xmath3484 to understand this property let us consider those curves in the lift of @xmath415-semi static curves which pass through @xmath3485 .
there exists @xmath3486 such that the intersection points of these curves with @xmath3487 locate in the disk @xmath3488 .
the rotation vector " of this segment of the orbit can not be too far away from @xmath3489 .
one can also define related fundamental domains @xmath3324 and @xmath3325 . for a fundamental domain @xmath3324 and a curve @xmath3490
, we pick up a curve in the lift of this curve to the universal covering space , denoted by @xmath1441 , which intersects the section @xmath3324 . some fundamental
domain @xmath3325 exists where this curve intersects .
recall the projection of the two fundamental domains takes the form @xmath3491 and @xmath3492 in the configuration space @xmath3422 .
we say that the two fundamental domains @xmath3324 and @xmath3325 are @xmath3493-related .
as the lagrangian @xmath3494 is well - defined in the universal covering space , let us consider its action in the universal covering space : @xmath3495 where @xmath3496 stands for a translation of @xmath3497 and @xmath3498 . obviously , one has @xmath3499 where the term @xmath3500 was defined before by considering the action in the configuration space @xmath3422 .
as above , one defines @xmath3501 again , for fixed @xmath3481 , some positive number @xmath610 and suitably large @xmath3502 exist such that @xmath3483 does not touch @xmath3503 and @xmath3504 let @xmath3505 . by induction
, these sections @xmath3506 are well defined such that @xmath3464 and @xmath3324 are @xmath3471-related , @xmath3324 and @xmath3325 are @xmath3493-related , the formulae ( [ constructioneq14 ] ) and ( [ constructioneq15 ] ) are satisfied .
as the third step of the construction , let us clarify what conditions the candidates of minimal curve are required to satisfy .
let @xmath75 : @xmath3507\to m$ ] be an absolutely continuous curve joining @xmath393 to @xmath395 , i.e. @xmath3508 and @xmath3509 .
we split the interval @xmath3507 $ ] into @xmath3510 subintervals @xmath3511=[t_0^+,t^-_0]\cup [ t^-_0,t^+_1]\cup\cdots\cup [ t^+_{i_m},t_{i_m}^-],\ ] ] where @xmath3512 , @xmath3513 .
correspondingly , we divide the curve into @xmath3510 segments @xmath3514}$ ] , @xmath3515}$ ] for @xmath3516 , and @xmath3517}$ ] .
we fix a curve @xmath1441 in the lift of @xmath75 to the universal covering space @xmath109 by choosing @xmath3518 .
correspondingly , each @xmath3519 has its lift @xmath3520 to @xmath109 .
the curve @xmath75 is required to satisfy the conditions : 1 , for each @xmath3521 , there is some @xmath3522 such that @xmath3523 2 , for @xmath3524 , @xmath3525 .
let @xmath3526 and @xmath3527 . to formulate the conditions for @xmath3528 ,
let us consider the quantity @xmath3529 one obtains from the proof of lemma [ semicontinuitylem2 ] that @xmath3530 as @xmath3531 or @xmath3532 .
thus , if @xmath3533 is defined as the quantity such that @xmath3534 , then we find @xmath3535 .
since both @xmath3536 and @xmath3537 are compact , there exist @xmath3538 such that @xmath3539 holds for each @xmath3540 and @xmath3541 .
let @xmath3542.\ ] ] the range of @xmath3528 is somehow implicitly defined .
let @xmath3543,\qquad \forall i\in\bigcup_{i_j\in\lambda_h}\{i_j , i_j+1 , \cdots , i_{j+1}-1\}\ ] ] @xmath3544 , \qquad \delta t_i^-=[\hat t_i^-,\breve{t}_i^-],\qquad \text{\rm for other}\ i\le i_m.\ ] ] see ( [ constructioneq6 ] ) , ( [ constructioneq13 ] ) for the definition of @xmath3545 and ( [ constructioneq17 ] ) for the definition of @xmath3546 respectively .
the conditions for @xmath3528 are the following : 1 , @xmath3547 for all @xmath3548 ; 2 , @xmath3549 for @xmath3550 ; 3 , for @xmath3551 , as it is assumed that the aubry set @xmath3552 contains finitely many classes , an orbit connects @xmath3553 to @xmath3552 by approaching the aubry class @xmath3337 as @xmath1345 , another orbit connects @xmath3552 to @xmath3554 by approaching the aubry class @xmath3338 as the time retreat back to @xmath3555 . for the time interval
@xmath3556 $ ] , one has the partition @xmath3557= [ t^+_i , t^-_{i , j'}]\cup[t^-_{i , j'},t^+_{i , j'+1}]\cup\cdots\cup[t^+_{i , j } , t^-_i],\ ] ] and has restrictions for these quantities , formulae ( [ constructioneq10 ] ) , ( [ constructioneq11 ] ) , ( [ constructioneq12 ] ) and @xmath3558 @xmath3559 with suitably large @xmath3560 . 4 , for @xmath3561 with @xmath3248 , by definition , @xmath415 is equivalent to @xmath3259 , one has @xmath3562 5 , for @xmath3561 with @xmath3563 , by choosing @xmath3259 suitable close to @xmath415 one can also assume that @xmath415 is equivalent to @xmath3259 .
thus , one has @xmath3564 as the system is autonomous , by choosing sufficiently large @xmath3565 , these conditions defines non - empty set for @xmath3566 .
finally , let us introduce a modified lagrangian and verify the smoothness of the minimizer of the action .
recall @xmath1076 and @xmath3312 are defined on @xmath3567 in the coordinate system @xmath3416 , @xmath3568 is well defined .
we introduce a modified lagrangian @xmath3569 where @xmath3570 is a translation of @xmath3571 : @xmath3572 on @xmath3573 and the integer @xmath702 is chosen so that ( [ constructioneq16 ] ) holds .
let @xmath492 : @xmath3574 be the universal covering space . for a curve @xmath1441 : @xmath3507\to\mathbb{r}^n$ ] , let @xmath3575 : @xmath3507\to m$ ] .
let @xmath3576 , @xmath3577 , we consider the minimal action @xmath3578 where the infimum is taken over all absolutely continuous curves @xmath1441 : @xmath3507\to\mathbb{r}^n$ ] with the boundary conditions @xmath3579 , @xmath3580 for @xmath3581 , @xmath3508 , @xmath3509 and satisfying the condition ( [ constructioneq16 ] ) .
moreover , restricted on @xmath3373 $ ] , @xmath75 is admissible for the condition ( [ constructioneq9 ] ) . as the system is autonomous
, the quantity @xmath3582 remains constant if @xmath3583 is subject to a translation .
thus , it is a function of @xmath3584 , @xmath3585 and @xmath3586 .
denote by @xmath3587 the domain where @xmath3588 takes its value .
let @xmath3589 , where all entries have been well defined in the previous proof . denote by @xmath3590 the curve along which the quantity of ( [ construction 24 ] ) is realized , it obviously depends on the value @xmath3591 and it may not be smooth at @xmath3592 .
let @xmath3593 and @xmath3588 range over the set @xmath3594 and @xmath3587 respectively , one obtains a minimizer .
the purpose of the following steps is to show that the minimizer is a solution of the euler - lagrange equation determined by @xmath63 .
let @xmath3595 be the minimum of @xmath3582 over @xmath3594 in @xmath3593 and over @xmath3587 in @xmath3588 respectively : @xmath3596 denote the minimal curve by @xmath3597 , we claim that @xmath3598 is a solution of the euler - lagrange equation of @xmath63 if @xmath359 and @xmath3599 are sufficiently large .
to verify this claim , we need to show that 1 , @xmath3600 solves the euler - lagrange equation determined by @xmath63 .
restricted on @xmath3601 , it obviously solves the euler - lagrange equation . 2
, @xmath3597 has no corner at @xmath3602 and @xmath3603 for each @xmath3604 , i.e. it is smooth for the whole @xmath3605 $ ] . for each @xmath3606 , as @xmath3607 is the minimizer for the curves admissible for the condition ( [ constructioneq9 ] ) , it is smooth at each @xmath3608 . indeed ,
if @xmath3267 , we obtained from ( [ constructioneq7 ] ) that @xmath3609 where @xmath3610 .
since the support of @xmath2728 has no intersection with @xmath2789 and @xmath3312 is closed in @xmath2789 , while @xmath1076 is closed and @xmath3432 in the region @xmath3611\}$ ] , the term @xmath1076 and @xmath3312 have no contribution to the euler - lagrange equation along @xmath3612 . for other @xmath349 , the conclusion is obtained from ( [ constructioneq13 ] ) by similar argument .
this proves the first conclusion .
recall the disks @xmath3613 and @xmath3614 are defined in the covering space @xmath109 .
we claim that @xmath1441 does not touch the boundary of @xmath3615 $ ] for @xmath3551 .
let us assume the contrary , i.e. @xmath3616)$ ] holds for some @xmath3551 .
let @xmath3617 and @xmath3618 . by the condition ( [ constructioneq16 ] ) we see that @xmath3619 and @xmath3620 .
then , in @xmath3253-coordinates , we obtain from ( [ constructioneq6 ] ) and ( [ constructioneq18 ] ) that @xmath3621 where @xmath3622 and @xmath3623 are the intersection points of a curve in @xmath3624 with @xmath3625 and with @xmath3313 respectively , @xmath3626 and @xmath3627 .
this contradicts the minimality of @xmath75 , thus it verifies our claim . to see that the curve @xmath1441 is smooth at @xmath3628 ,
let us assume the contrary again .
let @xmath3629 and @xmath3630 , here @xmath49 is chosen so small that @xmath3631 $ ] .
this is possible since @xmath3632)$ ] implies that @xmath3633 .
we join these two points by a minimizer @xmath3634\to m$ ] with @xmath3635 and @xmath3636 @xmath3637})]=\inf_{\stackrel{\zeta(-\delta)=x'}{\scriptscriptstyle \zeta(\delta)=x^ * } } \int_{-\delta}^{\delta}(l-\eta_{i})(d\xi(s))ds.\ ] ] if @xmath194 passes through @xmath3638 , we obtain a curve @xmath545 by replacing the segment of the minimizer @xmath3639}$ ] with @xmath3634\to m$ ] .
let @xmath3640 be the time for @xmath545 passing through @xmath3324 , then @xmath3641 and @xmath3642 .
thus , we obtain an absolutely continuous curve which is admissible for each required condition ( see ( [ constructioneq18 ] ) and ( [ constructioneq20 ] ) ) . along this curve
we obtain smaller action @xmath3595 , but this is absurd .
so , we only need to show that @xmath194 passes through @xmath3638 .
indeed , as @xmath3069 is chosen small and transversal to the local connecting curve in @xmath3624 , @xmath3643 approaches @xmath3069 from different sides as @xmath3644 and @xmath3645 respectively .
otherwise , the minimality of @xmath75 would be violated .
one refers to @xcite for the details .
the smoothness at @xmath3646 can be proved similarly .
the smoothness of @xmath75 at @xmath3647 for @xmath3648 is obvious . because of the formulae ( [ constructioneq14 ] ) and ( [ constructioneq15 ] )
, @xmath1441 does not touch the boundary of @xmath3506 at the time of @xmath3649 respectively .
indeed , @xmath1441 approaches @xmath3650 from different sides as @xmath3651 and @xmath3652 respectively .
if @xmath1441 has a corner at @xmath3653 , let @xmath280 : @xmath3654\to\mathbb{r}^n$ ] be the minimizer of the action @xmath3655})= \inf_{\stackrel{\scriptscriptstyle\xi(t^{\pm}_i-\delta ) = \tilde\gamma(t^{\pm}_i-\delta)}{\scriptscriptstyle \xi(t^{\pm}_i+\delta ) = \tilde\gamma(t^{\pm}_i+\delta ) } } \int_{t^{\pm}_i-\delta}^{t^{\pm}_i+\delta}l_{c_i}(d\xi(s))ds,\ ] ] then the curve @xmath280 passes through the disk @xmath3506 . replacing @xmath3656}$ ] by this minimizer
@xmath280 one obtains a curve with smaller action .
the contradiction verifies the smoothness .
if @xmath3259 is connected to @xmath415 by a type-@xmath11 orbit and @xmath3648 , then @xmath3464 is a small disk . by the same argument as above
, one obtains the smoothness of @xmath1441 at @xmath3657 and the smoothness at @xmath3658 from the arguments for type-@xmath11 .
as the system is autonomous , the following limit exists @xmath3659 we pick out a sequence of @xmath3597 for large @xmath359 and @xmath3599 . obviously , the set @xmath3660 has at least one accumulation point @xmath3661 : @xmath76 with the property @xmath3662 and @xmath3663 .
as we have shown , it is an orbit of @xmath3664 .
this proves the first conclusion of the theorem .
for any two points @xmath2451 , the sequence @xmath3665}\}$ ] approaches to a forward @xmath89-semi static curve as @xmath3666 , which starts from the point @xmath71 , and @xmath3667}\}$ ] approaches to a backward @xmath968-semi static curve as @xmath3668 , which approach to the point @xmath353 .
therefore , for sufficiently large @xmath3669 , the initial value @xmath3670 falls into any prescribed @xmath49-neighborhood of the points @xmath3221 and the orbit reaches the @xmath49-neighborhood of @xmath3222 at the time @xmath3671 .
this completes the proof .
the proof for time - periodic system is similar , and a bit easier from technical point of view , since one can treat the time variable @xmath832 as the first angle variable and take @xmath3672 the section for all classes .
one does not need to introduce various coordinate systems @xmath3673 for different cohomology class .
we omit the details here .
once one obtains the existence of a generalized transition chain in the system ( [ introeq1 ] ) , theorem [ mainthm ] is proved by applying theorem [ constructionthm1 ] .
therefore , the main purpose of this section is to show the genericity of such transition chains .
let us consider the hamiltonian ( [ introeq1 ] ) .
any integer vector @xmath3674 determines a plane @xmath3675 , which passes through the origin .
let @xmath3676 , it is diffeomorphic to a @xmath3677-sphere with the origin inside if @xmath3678 , because the hamiltonian @xmath11 is assumed convex . thus , the set @xmath3679 is a closed curve , denoted by @xmath3680 .
given any positive number @xmath54 , some positive integer @xmath3681 exists such that @xmath3682 constitutes a @xmath49-grid on @xmath3683 in the sense that the @xmath3684-neighborhood of @xmath3682 cover the whole sphere @xmath3683 , where @xmath3685 .
therefore , there exists a resonant path @xmath3686 such that each rotation vector falls into its @xmath3684-neighborhood , where @xmath3687 represents a resonant path determined by one resonant relation ( one integer vector ) .
it is possible that @xmath3687 and @xmath3688 are determined by the same resonant relation @xmath3689 .
.,width=151,height=151 ] obviously , the map @xmath3690 is a global diffeomorphism , which maps each closed curve @xmath3687 onto the @xmath3677-sphere @xmath3691 , @xmath3692 .
the @xmath49-grid on @xmath3683 induces a @xmath49-grid on @xmath3691 : the @xmath49-neighborhood of @xmath3693 covers the whole sphere . under the inverse of the frequency map
@xmath3694 we obtain a path @xmath3695 in action variable space , where @xmath3696 .
let @xmath3697 be the lagrangian determined by the hamiltonian @xmath11 , @xmath3698 be the lagrange flow . as the system is integrable , the action variable @xmath3699 keeps constant along each orbit of @xmath3698 which obviously lies in the support of certain @xmath89-minimal measure with @xmath3700 .
in this sense , one obtains a path @xmath3701 and @xmath3702 if we identify @xmath3703 . by the study of normal form ,
finitely many points @xmath3704 exist such that each @xmath3705 is rational frequency vector with period @xmath3706 and @xmath3707 where @xmath3708 , @xmath3709 , see ( [ normaleq9 ] ) .
obviously , @xmath187 depends on @xmath289 , the size of perturbation . under five steps of kam iteration
, we obtain the normal form @xmath3710 which is valid in the domain @xmath3711 .
in which @xmath3712 and @xmath3713 is resonant with respect to @xmath3714 @xmath3715 where the summation is made over all those @xmath3716 spanned by @xmath3717 : @xmath3718 in which @xmath3719 , @xmath3720 is irreducible and the integer vector @xmath3721 is used to determine the resonant path @xmath3722 , i.e. @xmath3723 holds for each @xmath3724 . for these two integer vectors @xmath3720
there is another @xmath3725 such that the matrix @xmath3726 is uni - module .
there are infinitely many @xmath3727 satisfying the condition , we choose one so that its norm is as small as it can be .
the coordinate transformation : @xmath3728 is obviously symplectic and @xmath3729 is also a function of @xmath3730 defined in @xmath583 . let @xmath3699 be the point where @xmath3731 , then the gradient of @xmath3732 satisfies @xmath3733 and @xmath3734 is independent of @xmath3735 , namely , @xmath3736 if we write @xmath3737 and @xmath3738 .
let us still use @xmath3739 to denote the new coordinate system .
therefore , around a strong resonance point @xmath3740 the normal form takes the form @xmath3741 in the new coordinate system ( [ chaineq1 ] ) .
this form remains valid in @xmath3742 and at @xmath3743 one has @xmath3744 with @xmath3745 . for our purpose ,
it is not necessary consider the hamiltonian @xmath3746 on the whole disk @xmath3747 .
instead , we choose finitely many @xmath3748 with @xmath3749 such that @xmath3750 and @xmath3751 where @xmath2083 is a suitably large number .
the results obtained in section 4 and 5 can be applied to the hamiltonian when it is restricted on each domain @xmath3752 , especially on the domain @xmath3753 .
let @xmath3754 be the solution of the equation @xmath3755 , where @xmath3746 is given by ( [ chaineq2 ] ) .
it can be written in the form of @xmath3756 the truncated form of @xmath3757 @xmath3758 is determined by @xmath3759 , the truncated form of @xmath3746 .
denote @xmath3760 and let @xmath3761 , @xmath1483 , we obtain from @xmath3762 the hamiltonian @xmath3763 for @xmath3764 we have @xmath3765 where @xmath3766 , @xmath3767 , @xmath3768 , @xmath3769 , @xmath3770 , @xmath3771 and @xmath3772 where the @xmath62-norm is with respect to @xmath1492 only . by the choice of @xmath3773 and the convexity of
@xmath11 we can see that @xmath3774 where @xmath3775 is the lower bound of the eigenvalues of @xmath3776 .
therefore , the @xmath95-function @xmath3777 for the lagrangian determined by @xmath3778 does not reach its minimum when the action variable is restricted on the disk @xmath3779 .
as the frequency @xmath3780 satisfies certain resonant condition , the existence of normally hyperbolic cylinder is guaranteed by theorem [ appenhyperth1 ] ( see appendix b ) for generic @xmath3781 .
therefore , all functions @xmath3778 ( @xmath3782 ) are treated as _ a priori _ unstable hamiltonian and the hamiltonian @xmath3783 is considered as the problem of double resonance .
recall the fenchel - legendre transformation @xmath129 : @xmath130 , determined by the @xmath98-function .
let @xmath3784 , @xmath3785 and @xmath3786 be the @xmath98-function for @xmath11 , @xmath3787 and @xmath3788 respectively .
obviously , @xmath3789 is still a curve .
as it was studied in subsection 4.3 , @xmath3790 is composed of a flat @xmath1046 joined with two channels .
see figure [ fig13 ] below .
these channels are joined to the flat either at a point or along an edge .
the former case was thought difficult to handle . , represented by the thick solid red curve . along the segment from @xmath3791 to @xmath977
, @xmath2808 keeps constant .
the purple dashed curve represents the curve @xmath3792.,width=359,height=151 ] let @xmath3793 be @xmath95-function determined by the hamiltonian @xmath3794 respectively .
as it has been studied in the subsection 5.3 , the double resonance corresponds to a flat @xmath3795 , around which there exists a annulus of incomplete intersection @xmath3796 the following has been proved generic in theorem [ beltthm2 ] . for each @xmath3797
, the ma set does not cover the whole @xmath200-torus .
thus , some @xmath3317 exists such that for each @xmath3797 the set @xmath3798 does not cover the whole @xmath200-torus , where @xmath3799 and @xmath3800 are the elementary weak kam solutions .
such results are obtained under the hypothesis ( * h1@xmath23744 * ) proposed in the section 5 . as the truncated system is independent of @xmath3801 ,
the ma set for @xmath3787 is independent of @xmath3801 @xmath3802 where @xmath2804 is a co - dimension one section on which @xmath3803 .
let @xmath3804 : @xmath3805 be the standard projection : @xmath3806 , let @xmath3807 and @xmath3808 be the lagrangian determined by @xmath3809 and @xmath3762 respectively , @xmath3810 and @xmath3811 denote the ma set for the lagrangian determined by @xmath3809 and @xmath3762 respectively . as @xmath3812 solves the equation @xmath3813 , one has @xmath3814 .
if @xmath3815 does not cover the @xmath200-torus , @xmath3816 does not cover the @xmath3677-torus .
because of @xmath3817 , each @xmath89-minimal measure possesses non - zero rotation vector .
therefore , there exists some circle @xmath3818 non - degenerately embedded into the @xmath3677-torus such that each @xmath89-minimal curve passes through @xmath3818 transversally and @xmath3819 is topologically trivial , i.e. some open intervals @xmath3820 exist such that @xmath3821 one can suitably choose @xmath3822 so that @xmath3823 as @xmath3824 , some @xmath610 exists such that the ma set for the hamiltonian @xmath3746 @xmath3825 let @xmath3826 where @xmath3827 it is a closed curve . by the preliminary works as above , some @xmath89-equivalence along the curve is established .
indeed , for each @xmath3828 , let @xmath3829 by the construction , each @xmath1442-semi static curve passes through the section @xmath3830 transversally .
recall @xmath3831 one sees that @xmath3832 provided @xmath3833 is close to @xmath1442 , @xmath3834 and @xmath3835 , i.e. @xmath3836 . in this case , some open set @xmath3837 such that @xmath3838 , from which one obtains that @xmath3839 . for each class @xmath3840 close to @xmath1442 ,
one has @xmath3841 , thus , there exists a closed 1-form @xmath2763 such that @xmath2819=c'-c$ ] and @xmath3842 thus , any two classes along the curve @xmath3843 is equivalent .
therefore , a transition chain for incomplete intersection is established , see figure [ fig13 ] , the thick solid red curve from the point @xmath3791 to the point @xmath977 . by the study in the subsection 4.2 ( see theorem [ cylinderthm2 ] ) , there are two wedge - shaped channels @xmath3844 and @xmath3845 which extend into the annulus @xmath2398 .
corresponding to these two channels there exist two normally hyperbolic cylinder @xmath1943 and @xmath3846 respectively , which are three - dimensional and invariant for the hamiltonian flow : for each @xmath3847 , the ma set @xmath3848 if @xmath2062 .
we are now in the situation that there is a normally hyperbolic cylinder @xmath3849 homeomorphic to @xmath3850 , the aubry set is located on this cylinder for each cohomology class under consideration .
if the aubry set is a two - dimensional torus , it has its own stable and unstable manifold .
it implies that the forward ( backward ) weak kam solution is differentiable when it is restricted in a neighborhood of this 2-torus . because weak
kam is a viscosity solution , any @xmath74 viscosity solution for tonelli hamiltonian must be @xmath3851 @xcite .
therefore , in a small neighborhood of the aubry set , the stable and unstable manifold are lipschitz graphs . as the cylinder is smooth , the aubry set is also a lipschitz graph over two - torus .
let @xmath3852 be a four - dimensional section intersecting each orbit in the aubry sets transversally .
the set @xmath3853 is a two - dimensional cylinder . in a neighborhood of @xmath3854
the hamiltonian flow defines a return map on the section @xmath2870 .
restricted on the cylinder @xmath3854 , each aubry set is either periodic orbit , or aubry - mather set or invariant circle .
each circle is a lipschitz curve .
a piece of the cylinder @xmath3854 , bounded by two invariant circles , is invariant for the return map which preserves some area " element .
let @xmath2908 : @xmath3855\times\mathbb{t}\to\pi$ ] be the map , it pulls back the standard closed 2-form @xmath3856 to a 2-form on @xmath3854 .
since the second de rham cohomology of a cylinder is trivial , by moser s theorem on the isotopy of symplectic forms , there exists a diffeomorphism @xmath3857 which transforms this form to the standard 2-form , namely @xmath3858 since the return map @xmath3859 preserves the form @xmath127 , one has @xmath3860 let us consider those aubry sets which are invariant two - torus , denoted by @xmath3861 .
we use the same notation for their intersection with @xmath3854 , which are circles . fix one circle @xmath3862
, other circles are parameterized by the area " @xmath1315 .
given any other circle @xmath3861 , we obtain the algebraic area @xmath1315 of the region bounded by these two circles .
if each circle is regarded as the graph of a function , then there is a regularity result @xcite @xmath3863 because the cylinder is normally hyperbolic , there is an segment of a line @xmath3864 such that all cohomology classes located in this segment share the same aubry set , an invariant 2-torus , so we have a map @xmath3865 . for a small segment of cylinder ,
some neighborhood @xmath3866 of a two - torus exists so that all aubry sets on this cylinder fall into this neighborhood : @xmath3867 . in a suitably coordinate system
we take a finite covering space @xmath458 so that the lift of @xmath187 consists of two connected components @xmath3868 and @xmath3869 .
the ma set satisfies the condition @xmath3870 to construct transition chain in this situation , one need to show it consists of totally disconnected semi - static curves when the aubry set is a two - torus .
let us consider the covering space @xmath3871 such that the lift of @xmath187 contains infinitely many connected components , each of which is still a neighborhood of two - torus .
we consider two adjacent components @xmath3868 and @xmath3869 in the lift of @xmath187 , i.e. @xmath3872 and no other component in the lift is located between them .
the subscript @xmath491 means right " and @xmath3873 means
correspondingly , denote by @xmath3874 and @xmath3875 the connected component in the lift of @xmath3876 respectively , @xmath3877 and @xmath3878 .
the barrier function takes the form @xmath3879 where @xmath3880 and @xmath3881 are the elementary weak kam solution determined by @xmath3874 and @xmath3875 respectively .
the elementary weak - kam solution @xmath3880 is uniquely determined by @xmath3882 , all classes in @xmath3882 share the same elementary weak - kam solution .
it is why we use the subscript @xmath1315 .
a point @xmath3883 if and only if @xmath3884 let @xmath3885 be a segment of @xmath3886 bounded by @xmath3887 and @xmath3888 .
the problem turns out to be the version : whether does the set @xmath3889 consist of totally disconnected semi - static curves ?
we only need to follow the argument in @xcite if we are satisfied with the generic property in the category of lagrangian , where the perturbations are functions also defined on @xmath2559 : @xmath3890 . in this paper
, we are also going to prove the generic property in the sense of ma , i.e. the perturbations are imposed on the potential @xmath3891 .
let us construct the potential perturbations .
choose a 2-dimensional disk @xmath720 which transversally intersects the backward semi - static curves @xmath3892\to\bar m$ ] with @xmath3893 .
these curves approach @xmath3894 as @xmath2143 . in suitable coordinate system
we can assume that @xmath720 is located in the section @xmath3895 where @xmath3896 . let @xmath3897 .
we write the curve @xmath3898 in the coordinate form @xmath3899 where @xmath3900 is monotonely increases for @xmath3901 $ ] .
since continuous function can be approximated by smooth function , for any small @xmath54 , a tubular neighborhood of the semi - static curve @xmath3902}$ ] admits smooth foliation of curves @xmath3903 : @xmath3904\to\mathbb{t}^3 $ ] such that each semi - static curve @xmath3905}$ ] remains @xmath49-close to @xmath3903 in the sense that @xmath3906 for all @xmath3907 $ ] .
the tubular neighborhood is defined by the form @xmath3908 let @xmath3028 : @xmath3909 be a smooth function such that @xmath3910 , @xmath3911 if @xmath3912 $ ] with small @xmath919 and @xmath3913 if @xmath3914 . as @xmath3915 is a smooth foliation of the tubular domain , it can be thought as a differeomorphism @xmath3916 : @xmath3917\to\text{\uj c}$ ] , namely , for @xmath3918 there exists unique @xmath3919 $ ] such that @xmath3920 . with a smooth function @xmath1182 : @xmath3921
we obtain a smooth function @xmath3922 defined on @xmath3923 @xmath3924 since unique @xmath3919 $ ] is determined by certain @xmath3918 , some constant @xmath3925 exists such that @xmath3926 we construct the potential perturbation in the form of ( [ completeeq5 ] ) where @xmath1182 ranges over the function space spanned by @xmath3927 where each parameter of @xmath3928 ranges over an unit interval @xmath3929 $ ] . if we construct a grid for the parameters
@xmath3928 by splitting the domain equally into a family of cubes and setting the size length by @xmath3930 the grid consists of as many as @xmath3931 $ ] cubes .
let us choose a neighborhood @xmath3932 of the point @xmath3933 which satisfies the conditions : 1 , for each @xmath3934 with @xmath3935 and @xmath3936 , there is a unique backward semi - static curve @xmath3937 such that @xmath3938 and @xmath3939 as @xmath2143 .
it is guaranteed by the existence of unstable manifold and if @xmath720 is chosen close to @xmath3894 . by the definition , @xmath3940 for @xmath3901 $ ] and @xmath3935
, so each @xmath3936 defines a linear operator @xmath3941 2 , as each curve @xmath3942 stays in @xmath49-neighborhood of the fiber @xmath3903 for @xmath3901 $ ] with small @xmath54 , by choosing suitably small neighborhood @xmath3932 ( depending on the size of @xmath720 ) some constant @xmath3943 exists such that @xmath3944 with @xmath3945 .
indeed , as @xmath1182 is a linear combination of the functions @xmath3946 , there exists some number @xmath3947 depending on the size of @xmath720 only such that the hausdorff distance @xmath3948 where @xmath3949 and @xmath3950 . by requiring @xmath1315 suitably close to @xmath3933 and using the notation @xmath3951 , we have @xmath3952 and @xmath3953 therefore , one obtains ( [ completeeq8 ] ) from ( [ completeeq5 ] ) , ( [ completeeq6 ] ) and ( [ completeeq7 ] ) ; 3 , for each @xmath3954 and each @xmath3935 , the forward semi - static curve @xmath3955 , determined by @xmath3956 with @xmath3957 , does not touch the support of @xmath3958 and approaches @xmath3875 as @xmath832 increases to infinity . for the perturbed system @xmath3959 , we use @xmath3960 and @xmath3961 to denote the weak kam solution . by the construction of perturbation
, the invariant cylinder remains unchanged . restricted on the disk @xmath720 ,
the forward weak - kam solution @xmath3960 is also unchanged @xmath3962 , but the backward weak kam solution undergoes small perturbation @xmath3963 . to see how it is related to the potential ,
let us recall the following relations @xmath3964 if @xmath3965 is a semi - static curve determined by @xmath3966 with @xmath3967 and @xmath3968 .
we also have @xmath3969 clearly , for suitably large @xmath832 the backward weak - kam solution @xmath3970 shall retreat into a small neighborhood of @xmath3874 where the weak kam solution @xmath3966 also remains unchanged
. therefore we deduce from the last two formulae that @xmath3971 in a similar way , we find @xmath3972 where @xmath3973 stands for the backward semi - static curve determined by the elementary weak - kam solution @xmath3961 with @xmath3974 .
as @xmath71 is located in the region where the weak kam solution is differentiable , we have @xmath3975 as @xmath3976 , guaranteed by the upper - semi continuity of semi - static curves .
therefore , it follows that for @xmath3935 @xmath3977 where the linear operator @xmath3978 is defined in ( [ completeeq7 ] ) and @xmath3979 .
next , let us consider all backward weak - kam solutions for @xmath3980 .
each parameter @xmath3980 determines an interval @xmath3981 for cohomology class .
we restricted ourselves on a curve of first cohomology classes contained in the set @xmath3982 and intersecting each @xmath3981 transversally . in this sense
, we think the class defined on the interval @xmath3983 and the map @xmath3984 is continuous . as @xmath3985 if @xmath3986 and @xmath3987
if @xmath3988 , we obtain from lemma 6.4 in @xcite @xmath3989 we split the interval @xmath3990 equally into @xmath3991 $ ] parts and split the interval @xmath3992 equally into @xmath3993 $ ] , where @xmath3994,\qquad k_c=\big[l_c\frac{12c_4}{c_3}\big],\ ] ] @xmath3995 and @xmath383 are the length of @xmath3990 and of @xmath3992 respectively .
the grid over @xmath3996 consists of as many as @xmath3997 $ ] cuboids in which @xmath3998 are independent of @xmath1689 .
we pick up all cuboids which contain the points @xmath3999 and denote them by @xmath4000 with @xmath4001 , then the cardinality of the set @xmath4002 is not bigger than @xmath3997 $ ] . according to the definition
, a point @xmath4003 corresponds to a barrier function @xmath4004 .
let us assume that some parameters @xmath4005 exist such that @xmath4006 where @xmath4007 is defined as in ( [ completeeq5 ] ) with @xmath4008 determined by the parameters .
we consider another perturbation determined by the parameters @xmath4009 @xmath4010 and set @xmath4011 . by using the formula ( [ completeeq9 ] )
we write the identity @xmath4012 for any point @xmath4013 , in virtue of the formulae in ( [ completeeq10 ] ) the first term on the right - hand - side of the identity is not bigger than @xmath4014 . for small @xmath4015 we have @xmath4016 . as both @xmath4017 and @xmath4018 are independent of @xmath3801 ,
if the parameters @xmath4009 satisfy @xmath4019 we find from above identities and the estimate ( [ completeeq8 ] ) that @xmath4020 it implies that , for each small rectangle @xmath4000 we only need to cancel out at most @xmath4021 @xmath1689-cubes from the grid for @xmath4022 so that the formula ( [ completeeq11 ] ) holds for the all other cubes .
let @xmath300 ranges over the set @xmath4002 , we obtain a set @xmath4023,b_{\ell}\in[1,2]:\ell=1,2\}$ ] with lebesgue measure @xmath4024 such that the formula ( [ completeeq11 ] ) holds for each @xmath4025 and for each @xmath3936 .
by taking @xmath4026 , in the same way we can see that some set @xmath4027,d_{\ell}\in[1,2]:\ell=1,2\}$ ] with lebesgue measure @xmath4028 such that the formula @xmath4029 for each @xmath4030 and each @xmath3936 .
therefore , for each @xmath4031 , the formulae ( [ completeeq11 ] ) and ( [ completeeq12 ] ) implies that for all @xmath3936 the diameter of each connected component of the set @xmath4032 is smaller than @xmath720 . as @xmath1687 can be arbitrarily small , for each disk @xmath720 , an open - dense set @xmath4033 exists such that this disconnect property holds for the system @xmath4034 with @xmath4035 .
since @xmath1315 is restricted on a closed set in the line which can be covered by finitely many @xmath4036 , this property is also open - sense for all @xmath1315 under our consideration .
each section @xmath720 admits a hierachy of partition of small disks @xmath4037 such that the size @xmath4038 approaches zero as @xmath3049 , the intersection @xmath4039 is a residual set .
therefore , we have proved [ chainthm1 ] it is an open - dense condition for @xmath41 such that the set @xmath4040 consists of totally disconnected semi - static curves .
given a perturbation @xmath4041 , it is natural to ask , along the resonant path @xmath1104 , how many many double resonant points need to be treated as strong resonance .
along a segment of resonant path @xmath3687 the resonance condition @xmath4042 is always satisfied and at each double resonant point some other @xmath4043 exists such that @xmath4044 is linearly independent of @xmath3716 and @xmath4045 recall the process of kam iteration , the main part of the resonant term is obtained by averaging the perturbation over a circle determined by these two resonant relations .
it takes the form @xmath4046 where @xmath4047 since @xmath14 is @xmath15-function , the coefficient @xmath4048 is bounded by @xmath4049 which deduces the estimation @xmath4050 where @xmath4051 depends on @xmath3716 . the function @xmath4052 is periodic in @xmath4053 . in virtue of the theorem [ appenbthm1 ] ( see appendix b ) , the following hypotheses is obviously open and dense : ( * h1.1 * ) : for each @xmath4054 , @xmath4052 is non - degenerate at its maximal point , i.e. @xmath4055 holds provided @xmath3730 is a maximal point . given some @xmath4052 satisfying the hypothesis ( * h1.1 * ) , certain @xmath1198 exists such that for each @xmath4054 , @xmath4056 holds at the maximal point .
assume at @xmath4057 , the second resonant condition @xmath4058 is also satisfied .
one thus obtains the normal form ( [ chaineq2 ] ) , by performing the coordinate transformation ( [ chaineq1 ] ) .
the homogenized form of the truncated hamiltonian takes the form @xmath4059 the hamiltonian flow determined by @xmath4060 admits a normally hyperbolic invariant cylinder @xmath4061 if @xmath4062 is a non - degenerate maximal point of @xmath4063 . applying the theorem of normally hyperbolic manifold , one obtains from the estimate ( [ criterioneq1 ] ) that some positive number @xmath4064 exists such that @xmath1684 also admits a normally hyperbolic and invariant cylinder @xmath4065 close to @xmath4066 provided @xmath4067 it is a criterion to see whether the double resonance is thought as weak resonance and can be treated in the way for _ a priori _ unstable system .
there are only finitely many @xmath4043 not satisfying this condition , thus are treated as strong double resonance .
therefore , once a perturbation @xmath14 is chosen so that ( * h1.1 * ) is satisfied , there are finitely many double resonant frequencies which need to be treated as strong double resonance .
the number is independent of the size of @xmath289 . at strong double resonance ,
it is also open and dense condition that ( * h1.2 * ) : at each strong double resonance point , the maximal point of @xmath4068 is non - degenerate , two eigenvalues of the hessian matrix are positive and different @xmath4069 for @xmath4070 . indeed
, there exists @xmath1263 such that @xmath4071 and @xmath4072 .
given @xmath4073 we have chosen a resonant path @xmath4074 so that @xmath4075 passes through each @xmath49-neighborhood of these points .
let @xmath1 satisfy all hypothesis above .
in order to make things convenient for readers , we formulate them here again : ( * h1 * ) _ for each strong double resonance , the potential @xmath4076 attains its maximum at one point only , the hessian matrix of @xmath4076 at that point is negative definite .
all eigenvalues are different : @xmath1515_. ( see * h1 * in subsection 5.1 , * h1.1 * and * h1.2 * in subsection 8.4 ) ; ( * h2 * ) _ for the hamiltonian flow @xmath4077 , the stable and unstable manifold of the fixed point intersect transversally along each minimal homoclinic orbit .
each minimal homoclinic orbit approaches to the fixed point along the direction @xmath1527 : @xmath1528 as @xmath1529 . _
( see * h2 * in subsection 4.1 .
the function @xmath3809 solves the equation @xmath4078 , the transversality is in the sense that , at the intersection points , the tangent space of the stable and unstable manifold span the tangent space of the energy level set . )
( * h3 * ) : _ for each @xmath4079 , the aubry set does not contain minimal curve homoclinic to the origin @xmath139fixed point@xmath140 . _
( see * h2 * in subsection 4.1 , each strong double resonance is related to a flat @xmath4080 corresponding to the hamiltonian @xmath3809 . )
( * h4 * ) : _ for each @xmath1526 , there are finitely many @xmath4081 such that , for each rotation vector @xmath4082 , the mather set consists of two periodic orbits , for other rotation vector @xmath4083 , the mather set consists of one periodic orbit only .
all these periodic orbits are hyperbolic . _
( see * h4 * in subsection 4.2 , also formulated for the hamiltonian @xmath3809 . )
( * h5 * ) : _ for each @xmath4079 there is a disk disjoint either with the support of @xmath90 or of @xmath122 , restricted on which , the set @xmath4084 is non - empty .
the size of the disk is independent of @xmath89 . _
( see * h5 * in subsection 4.2 , also formulated for the hamiltonian @xmath3809 . ) along the resonant path @xmath4074 , the strong double resonance points are denoted by @xmath4085 , where the number @xmath393 depends on @xmath14 .
each flat @xmath4086 is surrounded by a annulus @xmath4087 . for each segment of @xmath4074 connecting @xmath3714 to @xmath4088 , denoted by @xmath4089 , @xmath4090 constitutes a channel connecting @xmath4091 to @xmath4092 . split the unit interval into @xmath4093 segments @xmath4094=[0=s_{0,i},s_{0,c}]\cup[s_{0,c},s_{1,i}]\cup\cdots\cup [ s_{m-1,i},s_{m , c } ] \cup[s_{m , c},s_{m , i}=1]\ ] ] and
let @xmath4095 : @xmath4096\to\alpha^{-1}_h(e)$ ] , @xmath4097 : @xmath4098\to\alpha^{-1}_h(e)$ ] denote the paths such that @xmath4099 , @xmath4097 falls into the annulus @xmath4100 along which the component @xmath2808 keeps constant in the local coordinate system and @xmath4095 falls into the channel @xmath4101 connecting @xmath4091 to @xmath4092 .
let @xmath4102 and @xmath4103 . the subscript
@xmath89 " is used to indicate complete intersection and the subscript
@xmath349 " denotes the incomplete intersection .
we choose the conjunction of these curves as candidate of generalized transition chain @xmath4104 indeed , restricted on the segment @xmath4097 ( @xmath4105 ) it has been proved satisfying the condition ( h2 ) in the definition [ chaindef1 ] ( @xmath89-equivalence ) by using the hypothesis ( * h1@xmath23745 * ) . to guarantee the condition ( h1 ) in the definition [ chaindef1 ] when it is restricted on each segment @xmath4095 ( @xmath4106 ) , one need to impose some condition which has been proved to be generic in subsection 8.3 , theorem [ chainthm1 ] : ( * h6 * ) : _ if the aubry set covers certain @xmath3677-torus in @xmath4107 for @xmath4108 , then certain finite covering manifold @xmath483 and certain two - dimensional section @xmath2671 exist such that @xmath4109 is totally disconnected . _ under these hypothesis , namely ( * h1@xmath23746 * ) , the path @xmath1104 defined in ( [ proofeq1 ] ) is a transition chain .
choose suitably many @xmath4110 ( @xmath4111 such that 1 , each @xmath3437 is connected to @xmath2738 by local minimal orbit either of type-@xmath89 or of type-@xmath11 ; 2 , among these classes , some classes @xmath3251 @xmath4112 exist such that @xmath3251 is very close to @xmath4113 ( the prescribed action variables in theorem [ mainthm ] ) if one thinks both @xmath3251 and @xmath2077 as points in @xmath20 .
recall the proof of theorem [ constructionthm1 ] .
let @xmath75 : @xmath3507\to\mathbb{t}^3 $ ] be the minimizer of the action ( see ( [ constructioneq23 ] ) ) satisfying the boundary conditions @xmath4114 and @xmath4115 . dividing the time interval @xmath3507 $ ] into @xmath3510 parts @xmath3511=[t_0^+,t^-_0]\cup [ t^-_0,t^+_1]\cup\cdots\cup [ t^+_{i_m},t_{i_m}^-],\ ] ] imposing some constraints on @xmath75 at @xmath3647 and conditions on sufficiently large @xmath4116 and @xmath4117 , one then proves that @xmath75 is a solution of the lagrange equation determined by @xmath41 .
the curve @xmath75 determines an orbit of the hamiltonian flow @xmath40 : @xmath4118 for each @xmath3143 , the set @xmath4119 is defined as follows : a vector @xmath4120 if and only if a backward @xmath89-semi static curve @xmath4121 for the lagrangian @xmath63 exists such that @xmath4122 . the set
@xmath4123 is defined for forward semi - static curve similarly .
clearly , one has the set - valued map @xmath4124 is upper - semi continuous . to see that this orbit visits the ball @xmath4125 and the balls @xmath51 @xmath4126
, we use this proposition . as @xmath11 is integrable , any backward ( forward ) @xmath89-semi static curve is @xmath89-static for all @xmath46 . along any @xmath89-minimal curve
it holds that the action variable @xmath4127 . since the perturbation @xmath4128 is small , @xmath4129 is close to @xmath4130 if @xmath4131 is sufficiently large .
it follows that @xmath4132 is very small provided @xmath289 is sufficiently small . in the same way
, one can see that @xmath4133 is also very small .
since ma set is upper semi continuous with respect to lagrangian , at the time @xmath4134 , @xmath4135 is very close to @xmath3437 , @xmath4136 is very small .
this proves that the hamiltonian flow @xmath40 admits an orbit that visits these balls in turn . to complete the proof of theorem [ mainthm ]
, we only need to show the generic property . towards this goal ,
let us observe a fact : under the rescaling @xmath4137 , @xmath4138 , the hamiltonian equation determined by @xmath4139 is the same as it for the function @xmath4140 .
therefore , some open - dense set @xmath4141 exists , some @xmath4142 is associated to each @xmath4143 such that the hamiltonian flow @xmath40 satisfies the conditions * h1@xmath23745 * provided @xmath4144 , because the number of strong double resonant points is independent of the size of @xmath289 . from the proof of theorem [ chainthm1 ] in the subsection 8.3 , the condition * h6 * is required for the intersection of countably many open - dense set contained in @xmath4145 : @xmath4146 .
clearly , there exists a residual set @xmath4147 , for each @xmath4148 there exists a set @xmath34 residual in @xmath4149 $ ] such that @xmath4150 .
take the intersection of these sets , we obtain the cusp - residual property . @xmath4151
in this appendix , we study the normal form of nearly integrable hamiltonian , from which one obtains some information about the relevant mather sets , aubry sets as well as ma sets . here , the system is assumed to have arbitrary @xmath5-degrees of freedom @xmath4152 the perturbation can be autonomous as well as time-1-periodic . as @xmath832 can be treated
as the @xmath4153-th angle coordinate , we replace @xmath5 by @xmath176 when we consider time-1-periodic perturbation @xmath4154 .
thus , we consider autonomous hamiltonian only .
let @xmath4155 denote the frequency vector of the unperturbed system . for autonomous case
, a frequency @xmath127 is called rational of ( minimal ) period @xmath853 if @xmath4156 and @xmath4157 for each @xmath856 .
let the frequency @xmath127 be rational of period @xmath853
. with a function @xmath4158 on the torus one associates its time average @xmath4159 $ ] along the orbits of the linear flow defined by @xmath127 : @xmath2090 @xmath4160(x ,
y)=\frac 1t\int_0^tg(x+\omega t , y)dt.\ ] ] we say that @xmath603 is _ resonant _ ( with respect to @xmath127 ) if @xmath4161 $ ] , it implies that @xmath603 is constant along the orbits of the linear flow @xmath4162 .
let @xmath4163 be the ball of radius @xmath4164 around the origin , then there are positive numbers @xmath4165 such that @xmath4166 let @xmath1315 and @xmath4167 denote positive number such that @xmath4168 the value of @xmath1315 will be specified later to satisfy certain covering property .
denoted by @xmath4169 the set of frequencies which are rational of period @xmath853 with @xmath4170 . clearly , @xmath4171 is a finite index set
. let @xmath4172 .
let @xmath4173 , namely , @xmath4174 is non - negative @xmath2037 @xmath4175 .
let @xmath4176 , @xmath4177 .
let @xmath4178 denote the @xmath4179-norm on the domain @xmath720 , we omit the notation @xmath720 when it is clearly implied .
[ normalthm1 ] for a nearly integrable hamiltonian @xmath4180 we assume that both @xmath11 and @xmath4181 are @xmath15-smooth with @xmath16 , and @xmath4182 .
some small @xmath4183 exists such that for each @xmath4184 and each rational frequency @xmath4185 with a period @xmath4186 , a canonical transformation @xmath4187 is well defined on @xmath4188 which reduce the hamiltonian into the normal form @xmath4189 where @xmath4190 is resonant with respect to @xmath4191 with @xmath4192 , @xmath4193 is a higher order term when it is restricted in @xmath4194 @xmath4195 @xmath4196 where the constant @xmath1602 is independent of @xmath289 . restricted in the region @xmath4197 , we have a sharper estimate @xmath4198 the canonical transformation @xmath4187 is the composition of five steps of coordinate transformations @xmath4199 . each step of transformation @xmath4200 is defined as the time-1-map of @xmath4201 , the hamiltonian flow determined by the generating function @xmath4202 . for the first step of coordinate transformation @xmath4203
, we set @xmath4204 which solves the equation @xmath4205.\ ] ] let @xmath4206 $ ] , it follows that @xmath4207 where @xmath4208 obviously , @xmath4209=0 $ ] and one has the form @xmath4210 .
if we write @xmath4211 where @xmath4212 and for the coordinate transformation @xmath4200 ( @xmath4213 ) , we set by induction @xmath4214 then the hamiltonian @xmath4215 takes the form @xmath4216 where @xmath4217 also , @xmath4218=0 $ ] and we can write @xmath4219 . by the construction , we see that @xmath4220 is @xmath4221-smooth and @xmath4222 is @xmath4223-smooth .
some constants @xmath4224 , independent of @xmath289 , exists such that for @xmath4225 @xmath4226 where we have used the relations that @xmath4227 , @xmath4228 and @xmath4229 .
if we write the canonical transformation @xmath4200 in the form @xmath4230 then @xmath4231 it maps @xmath4232 for @xmath4184 if we set @xmath4233 the estimate on @xmath4234 and @xmath4235 follows from the formulae ( [ estimate ] ) .
if the system is real analytical , the higher order term can be reduced to the order @xmath4236 ( see @xcite ) , with which one obtains the nekhoroshev s estimate .
recall that the set of frequencies @xmath4169 each of which is rational of period @xmath853 with @xmath4170 , and the domains @xmath4237 where the iteration of kam is carried ( see theorem [ normalthm1 ] for definition ) .
the following covering property holds @xmath4238 to show the covering property , we use dirichlet s approximation theorem . for real
@xmath71 one has @xmath4239+\{x\},\ ] ] where @xmath4240\in\mathbb{z}$ ] the integer part , and @xmath4241 .
we use notation @xmath4242 if @xmath4243 one sets @xmath4244 [ normalpro1 ] ( dirichlet , see for examples , @xcite ) let @xmath4245 and @xmath359 a real number with @xmath4246 .
there exists an integer @xmath77 , @xmath4247 , such that @xmath4248 for any @xmath4245 , by applying dirichlet s theorem , we find some rational vector @xmath4249 existing such that @xmath4250 with @xmath4251 and @xmath4252 here , @xmath4249 is a rational vector of period @xmath853 .
since @xmath11 is assumed strictly convex , there exist two points @xmath4253 such that @xmath4254 , @xmath4255 and @xmath4256 the condition @xmath4257 is guaranteed if we choose @xmath4258 as @xmath4170 is required , the following should be satisfied : @xmath4259 that is , referring to ( [ normaleq8 ] ) , @xmath4260 which determines a threshold for @xmath289 provided : @xmath4261 as we choose @xmath1315 satisfying this condition , the covering property ( [ normaleq2 ] ) is proved . for the purpose of this paper , the covering property ( [ normaleq2 ] ) for the whole space is not necessary , instead , it is good enough to cover a neighborhood of a resonant path .
denote by @xmath4262 a @xmath4263 matrix , where @xmath4264 are integer vectors .
we consider the @xmath2672 resonance line @xmath4265 if the covering property ( [ normaleq2 ] ) in theorem [ normalthm1 ] is replaced by covering a neighborhood of the line @xmath4266 then it works if @xmath4267 indeed , as all frequencies are on a @xmath877-resonance line , by using dirichlet approximation theorem ( proposition [ normalpro1 ] ) for @xmath4268 we obtain a threshold @xmath4269 .
recall that the term @xmath4190 in ( [ normaleq1 ] ) is resonant with respect to @xmath127 , some rational frequency of period @xmath4170 , namely , it has the form @xmath4270 note that @xmath4271 is an indivisible integer vector , i.e. @xmath4272 for any @xmath4273 .
there are @xmath2672 integer vectors @xmath4274 such that the matrix @xmath4275 is indivisible , @xmath4276 and @xmath4277 holds for each @xmath4278 .
clearly , there is another integer vector @xmath4279 such that the matrix @xmath4280 is uni - module .
each integer vector @xmath682 with @xmath4281 uniquely determines an integer vector @xmath4282 such that @xmath4283 .
we introduce a coordinate transformation : @xmath4284 such that @xmath4285 this coordinate transformation is symplectic , @xmath4286 is also a function defined on @xmath583 with respect to @xmath3730 .
let @xmath29 be the point where @xmath4254 , then the gradient of @xmath4287 satisfies @xmath4288 and @xmath4289 is independent of @xmath4290 , thus we can write @xmath4291 if we use the natation @xmath4292 and @xmath4293 .
next , let us consider the time-1-periodical non - autonomous case .
assume @xmath4294 is an indivisible integer vector .
as @xmath4190 is resonant with respect to @xmath127 , we have @xmath4295 thus , there are @xmath5 integer vectors @xmath4296 such that @xmath4297 for each @xmath4298 . for each @xmath4299 with @xmath4300 , there is uniquely determined @xmath4301 such that @xmath4302 by choosing suitable @xmath4303 , we can make @xmath4280 be uni - module .
introduce the coordinate transformation ( [ normaleq10 ] ) , let @xmath29 be the point where @xmath4254 , then the gradient of @xmath4304 satisfies @xmath4305 note that each @xmath4306 with @xmath4307 uniquely determines @xmath4308 such that @xmath4309 .
as we have @xmath4310 in the new coordinates the resonant term @xmath4311 .
let @xmath4312 and let @xmath4313 , we find the hamiltonian equation of @xmath4314 is the same as the hamiltonian equation of @xmath4315 .
_ by _ chong - qing cheng and min zhou in the section 4 , we made the hypothesis ( * h4 * ) on the hyperbolicity of minimal periodic orbits in hamiltonian systems with two degrees of freedom . in the subsetion 8.4
, we need the hypothesis ( * h1.1 * ) . in @xmath15-topology with @xmath4316 these hypotheses are shown generic in @xcite . for the sake of completeness and convenience of reader
, we present the proof in this appendix .
[ appenbthm1 ] let @xmath4317 : @xmath4318 be a family of @xmath2334-functions depending on the parameter @xmath4319 $ ] @xmath4320 . if @xmath4317 is lipschitz continuous in @xmath101 , then there exists an open - dense set @xmath4321 such that for each @xmath4322 and each @xmath4323 $ ] , each global minimum of @xmath4324 is non - degenerate , namely , the second derivative is positive at each global minimizer .
since the openness is obvious , we only need to show the density . for this goal
, we introduce a set of perturbations with four parameters : @xmath4325 where @xmath4326\times[1,2]\times[1,2]\times[1,2]$ ] .
let @xmath4327 , we are going to show that , for any small numbers @xmath4328 there exists @xmath4329 such that @xmath4330 holds for each @xmath4323 $ ] whenever the point @xmath837 is a global minimizer of @xmath4331 .
it implies the second derivative is positive .
indeed , if it equals zero , the third derivative will be zero also .
consequently , the above formula does not hold . by choosing sufficiently large integer @xmath77 , @xmath4332 can be arbitrarily small .
let @xmath4333 , @xmath4334 $ ] and @xmath4335 , then @xmath4336 and @xmath4337 restricted on each interval @xmath4303 , each function @xmath2338 is approximated by taylor series ( module constant ) @xmath4338 given two points @xmath4339 and @xmath4340 , we have two functions @xmath4076 and @xmath4341 in the form of taylor series .
let @xmath4342 , @xmath4343 , @xmath4344 and @xmath4345 , we have @xmath4346 and @xmath4347 using the notation @xmath4348
it follows from the identities above that @xmath4349 we construct a grid for the parameters @xmath4339 by splitting the domain for @xmath4339 equally into a family of cuboids and setting the size length by @xmath4350 these cuboids are denoted by @xmath4351 with @xmath4352 , the cardinality of the set of the subscripts is bounded by @xmath4353,\ ] ] where the integer @xmath359 is independent of @xmath1454 .
let @xmath4354 denote the center of each cuboid and let @xmath4355 define @xmath4356 we find from the formula ( [ appenbeq2 ] ) that following holds for suitably small @xmath1068 @xmath4357 let us define a subset @xmath4358 in the following way .
a subscript @xmath4359 if and only the set @xmath4360 : \text{\rm osc}_{i_i}(f_{\lambda}-\epsilon v_{ij } ) < 2md^4\}\neq\varnothing.\ ] ] is non - empty . by definition , we have @xmath4361\ \text{\rm and}\ j\in\mathbb{j}_i\backslash\mathbb{j}'_i.\ ] ] using the lipschitz property @xmath4362 , we claim an estimate on the cardinality of this subset @xmath4363 , \ \ \
\text{\rm where}\ \ \
\frac{\log k_d}{|\log d|}\to 0 \ \
\text{\rm as}\ \
d\to 0.\ ] ] in order to prove it , let us replace @xmath4364 by @xmath4365 , which is still lipschitz in @xmath101 , and denote the set of functions by @xmath4366\}.\ ] ] it follows from the lipschitz property that the box dimension of the set @xmath4367 equals one in @xmath424-topology .
let @xmath4368 denote a cube in @xmath424-function space , centered at the origin with the size equal to @xmath4369 , where @xmath1602 depends on the upper bound of @xmath4370 .
the set @xmath4371 is non - empty only if @xmath4372 holds for @xmath4373 .
since the box dimension of the set @xmath4367 equals one , we see that the set @xmath4374 can be covered by as many as @xmath4375 $ ] cubes with the size of @xmath4376 , where the number @xmath4377 satisfies the condition that @xmath4378 as @xmath1704 .
let us keep in mind that , by the definition , each @xmath4359 corresponds to a non - empty set @xmath4379 $ ] .
if the cardinality @xmath4380 $ ] , by pigeonhole principle , there would be at least 28 different subscripts @xmath4381 such that certain @xmath4382 , and the 28 functions @xmath4383 fall into one small cube with the size of @xmath4376 .
on the other hand , since the parameter space is three dimensional , in these 28 different subscripts , there must be @xmath4384 such that @xmath4385 it follows from ( [ appenbeq4 ] ) that @xmath4386 on the other hand , as both @xmath4387 and @xmath4388 fall into the same cube where @xmath4389 we find that @xmath4390 this contradiction proves the claim that @xmath4391 $ ] .
the gird for @xmath4339 induces a grid for the parameters @xmath4397 determined by the equation @xmath4398= \left[\begin{matrix } -\sin x_i & \cos x_i & -2\sin 2x_i & 2\cos 2x_i \\ -\cos x_i & -\sin x_i & -4\cos 2x_i & -4\sin 2x_i \\ \sin x_i & -\cos x_i & 8\sin 2x_i & -8\cos 2x_i \end{matrix}\right ] \left[\begin{matrix } a_1\\ b_1\\ a_2\\ b_2 \end{matrix}\right]\ ] ]
the coefficient matrix is non - singular for each @xmath4399 .
indeed , let @xmath4400 be the @xmath4401 matrix formed by first three columns and let @xmath4402 be the matrix formed by the first , second and the fourth column , we have @xmath4403 note that @xmath4404 , the grid for @xmath4339 induces a grid for @xmath4397 .
it contains as many as @xmath4405 $ ] 4-dimensional strips , denoted by @xmath4406 with @xmath4407 .
each @xmath4406 is mapped onto @xmath4351 by the equation ( [ appenbeq6 ] ) . for each @xmath4408 with @xmath4409 , the inequality ( [ appenbeq5 ] ) holds for any @xmath4319 $ ] .
let us consider all intervals @xmath4303 with @xmath4410 .
different @xmath4303 induces different gird for the parameters @xmath4397 .
in general , @xmath4406 is not parallel to @xmath4411 if @xmath4412 . for each @xmath4303
, one can define two set of subscripts @xmath4413 in the way as above .
thus , one obtains the cardinality of the disjoint union set @xmath4414=27k_d\pi[d^{-7/2}]\ll k[d^{-9/2}].\ ] ] since @xmath4378 as @xmath1704 , we obtain a lebesgue measure estimate @xmath4415 let @xmath4416 , we obtain the lebesgue measure estimate @xmath4417 by the definition @xmath4418 and @xmath4419 , one can see that for any @xmath4420 and any @xmath4319 $ ] the variation of @xmath4331 is bounded from below @xmath4421 it implies that the inequality ( [ appenbeq1 ] ) holds .
this completes the proof .
let @xmath4422 be a tonelli lagrangian with two degrees of freedom ( @xmath4316 ) , let @xmath41 be the hamiltonian determined by @xmath63 . because of topological property of two torus , each ergodic minimal measure is supported on closed orbits if the rotation vector satisfies certain resonant condition .
it is a natural question whether these periodic orbits are hyperbolic .
once it is true , one then obtains normally hyperbolic cylinder composed by these periodic orbits .
given a lagrangian @xmath63 and a rotation direction @xmath4423 , by fenchel - legendre transformation , we obtain a channel in @xmath9 @xmath4424 typically , it is foliated into segments of line ( flat of the @xmath95-function ) , along which the @xmath95-function keeps constant , all cohomology classes share the same mather set .
thus , it makes sense to write @xmath4425 with @xmath4426 and @xmath4427 .
[ appenhyperth1 ] given a class @xmath4423 and a closed interval @xmath4428\subset \mathbb{r}_+$ ] with @xmath4429 , there exists an open - dense set @xmath4430 with @xmath4316 such that for each @xmath4143 , each @xmath4431 $ ] , the mather set @xmath4432 for @xmath4433 consists of hyperbolic periodic orbits . indeed , except for finitely many @xmath4434 $ ] where the mather set consists of two hyperbolic periodic orbits , for all other @xmath4431 $ ] it consists of exactly one hyperbolic periodic orbit .
this theorem will be proved by showing the non - degeneracy of the minimal point of certain action function . toward this goal ,
let us split the interval into suitably many subintervals @xmath4428=\cup_{i=0}^{k}[e_i-\delta_{e_i},e_i+\delta_{e_i}]$ ] with suitably small @xmath4435 .
once the open - dense property holds for each small subinterval , then it hold for the whole interval .
let us explain how the interval @xmath4428 $ ] is split .
in the channel , one can choose a path along which the @xmath95-function monotonely increases .
restricted on this path , we obtain a family of lagrangians with one parameter . by using the method of @xcite , we can see that it is typical that the minimal measure is supported at most on two periodic orbits for each class on this path .
thus , the mather set @xmath4432 consists of at most two periodic orbits for each @xmath4431 $ ] . without of losing generality ,
we assume @xmath4436 , all of these minimal curves are associated with the homological class . restricted on the neighborhood @xmath4437 of a minimal curve @xmath4438 for certain energy @xmath1155 ,
we introduce a configuration coordinate transformation @xmath4439 such that along the curve @xmath4440 one has @xmath4441 . in the new coordinates
, the lagrangian reads @xmath4442 which is obviously positive definite in @xmath4443 .
as @xmath4444 is a solution of the euler - lagrange equation determined by @xmath63 , the curve @xmath4445 solves the equation determined by @xmath2441 and is minimal for the action of @xmath2441 .
as there are at most two minimal curves for each energy , the neighborhood of these two curves can be chosen not to overlap each other .
therefore , one can extend the coordinate transformation to the whole torus .
let @xmath4446 be the hamiltonian determined by @xmath2441 through the legendre transformation .
the minimal curve determines a periodic solution for the hamiltonian equation . by construction
, @xmath4447 holds along the periodic solution which entirely stays in the energy level set @xmath4448 .
we choose suitably small @xmath4435 such that for @xmath4449 $ ] each minimal periodic curve in @xmath4450 falls into the strip @xmath4451 and @xmath4447 holds along each minimal periodic orbit . for brevity of notation , we still use @xmath71 to denote the configuration coordinates , use @xmath63 and @xmath41 to denote the lagrangian and hamiltonian , for which the condition @xmath4452 holds along minimal periodic orbits for @xmath4449 $ ] . under such conditions ,
the lagrangian as well as the hamiltonian can be reduced to a time - periodic system with one degree of freedom when it is restricted on energy level set .
the new hamiltonian @xmath4453 solves the equation @xmath4454 with @xmath4455 , from which one obtains a new lagrangian @xmath4456 where @xmath4457 solves the equation @xmath4458 . in the following
we omit the subscript 1 " , i.e. let @xmath4459 if no danger of confusion occurs .
we introduce a function of lagrange action @xmath4460 : @xmath4318 : @xmath4461 a periodic curve @xmath75 is called the minimizer of @xmath4462 if the lagrange action along this curve reaches the quantity @xmath4463 .
as there might be two or more minimizers if @xmath71 is not a minimal point , the function @xmath4462 may not be smooth in global .
however , we claim that it is smooth in certain neighborhood of minimal point .
to verify our claim , we let @xmath4464 and define the function of action @xmath4465 @xmath4466 there will be two or more minimizers of @xmath4465 if the point @xmath71 is in the cut locus " of the point @xmath353 .
however , the minimizer is unique if @xmath71 is suitably close to @xmath353 , denoted by @xmath4467 . in this case
, it uniquely determines a speed @xmath4468 such that @xmath4469 .
let @xmath4470 denote a periodic configuration ( @xmath4471 ) , we introduce a function of action
@xmath4472 as @xmath4473 is suitably small and the lagrangian is positive definite in the speed , the boundary condition @xmath4474 for @xmath1314 uniquely determines the speed @xmath4475 for @xmath4476 .
indeed , the function @xmath4477 generates an area - preserving twist map from the time-@xmath611-section to the time-@xmath4478-section @xmath4479 : @xmath4480 @xmath4481 where @xmath4482 .
as the lagrangian is positive definite in @xmath973 , it implies that the initial condition @xmath4483 smoothly depends on the boundary condition @xmath4484 in this case . because of the smooth dependance of solution of ordinary differential equation on initial condition , the function is smooth . obviously , each minimal point of @xmath4485 uniquely determines a @xmath89-minimal measure for @xmath4486 , supported on a periodic orbit @xmath4487 with @xmath2288=g$ ] .
let @xmath4488 , it satisfies the discrete euler - lagrange equation @xmath4489 we shall show later that the periodic orbit is hyperbolic if and only if the minimal configuration is non - degenerate , namely , the jacobi matrix is positive definite : @xmath4490\ ] ] where @xmath4491 and @xmath4492 .
let @xmath4493 be a minimal configuration of the function @xmath4494 , where the jacobi matrix is non - negative and the smallest eigenvalue is simple .
indeed , as the lagrangian is positive definite , the generating function @xmath4465 determines an area - preserving and twist map @xmath4479 , we have @xmath4495 .
consequently , by using a theorem in @xcite , we find that the smallest eigenvalue is simple .
let @xmath969 denote the @xmath349-th eigenvalue of the matrix , at the minimal configuration one has @xmath4496 let @xmath4497 be the eigenvector for @xmath969 . by choosing @xmath4498 we have @xmath4499 for @xmath4500 ( see lemma 3.4 in @xcite ) .
at the minimal configuration , we find the following matrix is positive definite : @xmath4501.\ ] ] if not , there will be a vector @xmath4502 such that @xmath4503 .
it follows that @xmath4504 if we set @xmath4505 .
as the matrix @xmath4506 is non - negative , it implies that @xmath4507 , but it contradicts the fact that all entries of @xmath4508 have the same sign , either positive or negative . in a suitably small neighborhood @xmath4509 of the minimal configuration ,
let us consider the equations @xmath4510 since the matrix @xmath4511 is positive definite at the minimal point , by the implicit function theorem , this equation has a unique smooth solution @xmath4512 when @xmath4513 .
let @xmath75 : @xmath4514\to\mathbb{r}$ ] be a minimizer of @xmath4515 with @xmath4516 , we obtain a configuration @xmath4517 .
clearly , @xmath4518 holds at this configuration for each @xmath4519 .
it implies the uniqueness of the minimizer of @xmath4520 for @xmath4513 .
the minimal point of @xmath4462 uniquely determines a minimal configuration of @xmath4521 , therefore , the function @xmath4462 is smooth in certain neighborhood of its minimal point . in a neighborhood of the minimal point , let us study what change the function of action undergoes when the lagrangian is under a perturbation of potential @xmath4522 , where @xmath14 : @xmath4523 is a potential .
let @xmath4524 denote the reduced lagrangian of @xmath4433 and let @xmath4525 , one has @xmath4526 we denote the minimal curve of @xmath4527 by @xmath4528 such that @xmath4529 .
let @xmath4530 and @xmath4531 be the quantities defined for @xmath4524 as the quantities @xmath4528 and @xmath4527 defined for @xmath1512 . by the definition of minimizer
, we have @xmath4532 and @xmath4533 since the distance between these two curves @xmath75 and @xmath545 is due to the perturbation @xmath14 , we finally obtain @xmath4534 where @xmath4535 and @xmath4536 the field of smooth curves @xmath4537 defines an operator @xmath4538 , which maps functions defined on @xmath1169 into the function space defined on @xmath2372 . obviously , both @xmath4539 and @xmath4540 are smooth in @xmath4541 and in @xmath3 .
let us consider the derivative of @xmath4460 . as the lagrangian is positive definite ,
some positive constants @xmath4548 exist such that @xmath4549 since @xmath4550 , one has @xmath4551 and @xmath4552 where the second equality follows from that @xmath75 solves the euler - lagrange equation . if @xmath4553 and if the lemma does not hold , by choosing @xmath4554 ( @xmath4555 ) we obtain from the taylor series up to second order that @xmath4556 if @xmath4557 , where @xmath4558 . but
it is absurd .
the case @xmath4559 can be proved by choosing @xmath4560 .
this completes the proof .
let @xmath4561 , where @xmath837 is the minimal point of @xmath4562 .
as it was shown above , @xmath4563 smoothly depends on @xmath71 , we have a smooth foliation of curves in a neighborhood of the curve @xmath4564 .
the corner at @xmath4565 , i.e. @xmath4566 approaches to zero as @xmath4567 . for each @xmath71 , if there is a corner at @xmath4568 , we construct a curve @xmath4569 that smoothly connects the point @xmath4570 to the point @xmath4571 with @xmath4572 , where @xmath54 is suitably small . replacing the segment @xmath4573\cup[2\pi-\delta,2\pi]}$ ] by this curve
, we obtain a smooth curve @xmath4569 such that @xmath4574}$ ] and @xmath4572 . indeed ,
as the curve @xmath4575 is @xmath4576-smooth in @xmath71 and @xmath4577 is also @xmath4576-smooth in @xmath832 , for small number @xmath1689 some @xmath4578 exists such that the quantities @xmath4579 , \
k=0,1,2,3.\ ] ] let the curve @xmath4580 : @xmath4581\to\mathbb{r}$ ] be an interpolation polynomial of degree eight such that @xmath4582 and @xmath4583 , then the coefficients of the polynomial are smooth in @xmath71 .
let @xmath4584 , we see that the foliation of the curves @xmath4569 is smooth in @xmath71 and as a function of @xmath832 , @xmath4585 is small in @xmath4576-topology . for each point @xmath4586
, there is a curve @xmath4587 such that @xmath4588 .
it uniquely determines a speed @xmath4589 . by the construction ,
@xmath4590 is @xmath4576-smooth in @xmath237 . as @xmath4591 when it is restricted to a neighborhood of the minimal curves , both @xmath524 and @xmath452 can be approximated by @xmath15-function @xmath4592 and @xmath4593 in @xmath4576-topology respectively i.e. @xmath4594 and @xmath4595 hold for small @xmath1687 .
given a @xmath15-function @xmath4596 : @xmath4318 we obtain a @xmath15-function @xmath4597 : @xmath4523 defined by @xmath4598 if @xmath4588 . by the definition
, we have @xmath4599 where @xmath4600 is defined as follows : passing through the point @xmath4601 there is a unique @xmath353 such that @xmath4602 .
we set @xmath4603 .
we introduce a set of perturbations with four parameters : @xmath4604 where @xmath4326\times[1,2]\times[1,2]\times[1,2]$ ] . by applying the formula ( [ nondegenerate3 ] ) to the function @xmath4605 and @xmath4606
we find that @xmath4607 where @xmath4608 let us study the dependence of the terms @xmath4609 and @xmath4610 on the point @xmath71 .
we claim that there exists constant @xmath4611 as well as small numbers @xmath4612 and @xmath4613 such that for each @xmath4614 , each @xmath4561 , @xmath4615 and @xmath4616 , we have @xmath4617 @xmath4618 by the construction of the curves @xmath2208 , for @xmath4619 and for @xmath4561 we have @xmath4620 and @xmath4621 is small for @xmath4622 and for @xmath4615 .
integrating the them over the set with lebesgue measure @xmath4623 , we find that some small @xmath4611 exists such that the formulae in ( [ nondegenerate5 ] ) hold for @xmath4624 with @xmath4625 being replaced by @xmath4626 if @xmath4593 and @xmath4592 in the formula ( [ nondegenerate3 ] ) are replaced by @xmath452 and @xmath524 respectively .
as both @xmath524 and @xmath452 are approximated by @xmath4592 and @xmath4593 in @xmath4576-topology , by choosing @xmath1687 suitably small , all formulae in ( [ nondegenerate5 ] ) hold for @xmath4624 with @xmath4625 being replaced by @xmath4627 . for other energy @xmath3 ,
let us recall the solution @xmath4512 of eq .
( [ variationeq1 ] ) is smooth .
as the map @xmath4479 : @xmath4480 is area - preserving and twist , it uniquely determines the initial speed @xmath4628 , namely , the initial speed smoothly depends on the initial position and such dependence is also smooth in the parameter @xmath3 . as solution of ode
smoothly depends on its initial conditions , the minimal curve @xmath4542 of @xmath4527 smoothly depends on the parameters @xmath71 and @xmath3 .
thus , the formulae in ( [ nondegenerate5 ] ) hold if the numbers @xmath4612 and @xmath4613 are suitably small . to show the non - degeneracy of the global minimum of @xmath4460 located at the point @xmath71
, we only need to verify that @xmath4632 holds for small @xmath4633 , where @xmath4634 .
assume @xmath4635 is an interval , we define @xmath4636 . to show the non - degeneracy , it is sufficient to verify that @xmath4637 if the minimal point @xmath4638 , where @xmath4639 denotes the length of the interval . the openness is obvious of @xmath4640 . to show the density ,
we are concerned only about the configurations where @xmath4462 takes the value close to the minimum and consider small perturbations from the following set where the parameters @xmath4397 range over the cube @xmath4326\times[1,2]\times[1,2]\times[1,2]$ ] @xmath4641 where each element is a function of @xmath4642 , see the formulae ( [ nondegenerate4 ] ) . recall that both operators @xmath4643 and @xmath4644 are linear and @xmath4645 , see the formula ( [ nondegenerate1 ] ) and the formula ( [ nondegenerate2 ] ) .
we choose sufficiently large integer @xmath359 so that @xmath4646{\pi / k}$ ] can be arbitrarily small .
let @xmath4647 , @xmath4648 $ ] and @xmath4649 , then @xmath4650 .
restricted on each interval @xmath4651 , each @xmath4652-function @xmath4653 is approximated by the taylor series ( module constant ) @xmath4654 given two points @xmath4655 and @xmath4656 , we have two functions @xmath4657 and @xmath4658 in the form of taylor series .
let @xmath4659 , @xmath4660 , @xmath4661 and @xmath4662 , we have @xmath4663 and @xmath4664 it follows that @xmath4665 we construct a grid for the parameters @xmath4655 by splitting the domain for them equally into a family of cuboids and setting the size length by @xmath4666 these cuboids are denoted by @xmath4667 with @xmath4668 , the cardinality of the set of the subscripts is up to the order @xmath4669,\ ] ] where the integer @xmath4670 is independent of @xmath1454 .
if @xmath4671 , we obtain from the formula ( [ nondegenerate7 ] ) that @xmath4672 if @xmath4673 with @xmath4674 the coefficients @xmath4655 depend on the parameters @xmath4397 , the energy @xmath3 and the position @xmath4675 .
the gird for @xmath4655 induces a grid for the parameters @xmath4397 , determined by the equation @xmath4676=({\bf
c_1}{\bf u}+{\bf c_2 } ) \left[\begin{matrix } a_1\\
b_1\\ a_2\\ b_2 \end{matrix}\right]\big(1+t_{\epsilon , e , x_{k}}(a_1,b_1,a_2,b_2)\big)\ ] ] where the map @xmath4677 : @xmath4678 is as small as of order @xmath202 , @xmath4679,\ ] ] @xmath4680 , \left[\begin{matrix } u_2(x_{k } ) & v_2(x_{k})\\ -v_2(x_{k } ) & u_2(x_{k } ) \end{matrix}\right]\right\},\ ] ] each entry of @xmath4681 is a linear function of @xmath4682 , @xmath4683 , @xmath4684 and @xmath4685 with @xmath4686 , @xmath4616 .
both matrices @xmath4687 and @xmath4688 depend on the energy @xmath3 , @xmath4687 is close to the identity matrix .
let @xmath4689 be the matrix composed by the first three columns of @xmath4690 , @xmath4691 be the matrix composed by the first , the second and the fourth column of @xmath4690 .
as we are only concerned about those positions where @xmath4462 takes value close to the minimum and about the energy @xmath3 close to @xmath1772 , in virtue of ( [ nondegenerate5 ] ) we obtain @xmath4692 since @xmath4693 , the grid for @xmath4655 induces a grid for @xmath4397 which contains as many as @xmath4694 $ ] 4-dimensional strips ( @xmath4695 is independent of @xmath1454 ) .
note that the induced partition for the parameters @xmath4397 depends on the energy @xmath3 .
given an energy @xmath4696 $ ] , if there exist taylor coefficients @xmath4655 which determines a perturbation @xmath1182 such that @xmath4697 then for @xmath4656 which determines a perturbation @xmath4698 and satisfies the condition @xmath4699 one obtains from the formula ( [ nondegenerate7 ] ) that @xmath4700 under the map defined by the formula ( [ nondegenerate8 ] ) , the inverse image of a cuboid @xmath4701 with the size @xmath4702 is a strip in the parameter space of @xmath4397 , denoted by @xmath4703 , with the lebesgue measure as small as @xmath4704 .
if the cuboid @xmath4701 is centered at @xmath4655 , then for @xmath4705 the inequality ( [ nondegenerate9 ] ) holds . splitting the interval @xmath4706 $ ] equally into small sub - intervals @xmath4707 with the size @xmath4708
, we obtain as many as @xmath4709 $ ] small intervals .
since the function @xmath4462 is lipschitz in @xmath3 , suitably large positive number @xmath4710 can be chosen so that @xmath4711 therefore , for @xmath4653 with @xmath4712 , one has @xmath4713 pick up one energy @xmath4714 in each small interval @xmath4707 , there are @xmath4709 $ ] strips @xmath4715 .
finally , by considering all small intervals @xmath4651 with @xmath4716 , we find @xmath4717{d}.\ ] ] let @xmath4718 , we obtain the lebesgue measure estimate @xmath4719{d}\to 1,\qquad \text{\rm as}\ \ d\to 0.\ ] ] obviously , for each @xmath4420 , each @xmath4631 $ ] and each @xmath4720 the formula ( [ nondegenerate10 ] ) holds .
this proves that it is open - dense that all minimal points of @xmath4460 are non - degenerate when the energy ranges over the interval @xmath4706 $ ] .
let @xmath837 be a minimal point of the function @xmath4460 and let the curve @xmath4721 : @xmath4318 be the minimal curve of @xmath4722 which is smooth and determines a periodic orbit @xmath4723 of the lagrange flow @xmath4724 . back to the autonomous system
, it determines a periodic orbit @xmath4725 of the lagrange flow @xmath128 , where @xmath4726 , @xmath4727 .
if a periodic orbit is hyperbolic , it has its stable and unstable manifold in the phase space .
consequently , any orbit staying on the stable ( unstable ) manifold approaches to the periodic orbit exponentially fast as the time approaches to positive ( negative ) infinity . in a neighborhood of the minimal periodic curve @xmath75 , each point @xmath71 on the section
@xmath4728 determines at least one forward ( backward ) semi - static curve @xmath4729 : @xmath4730 ( @xmath239 : @xmath4731 ) such that @xmath4732 .
these curves determine forward ( backward ) semi - static orbits @xmath4733 of which the @xmath127-set ( @xmath95-set ) is the periodic orbit @xmath462 . in the configuration space @xmath4734 , these two curves intersect with the section @xmath4728 infinitely many times at the points @xmath4735 and @xmath4736 .
these points are denoted by @xmath648 , they are well ordered @xmath4737 .
it is possible that @xmath4738 . in this case
, we count the point twice . for each point
@xmath648 , there is a curve joining @xmath4739 to @xmath4740 which is composed by some segments of @xmath4729 as well as of @xmath239 .
for instance , in the following figure , by starting from the point @xmath4741 and following a segment of @xmath239 to the point @xmath2346 , then following a segment of @xmath4729 to the point @xmath3791 and finally following a segment of @xmath239 to the point @xmath4742 , we obtain a circle .
clearly , the lagrange action along this circle is not smaller than the quantity @xmath4743 .
let us consider the whole sequence @xmath4744 , we obtain infinitely many circles in that way .
therefore , the sum of the quantities @xmath4745 is obviously not bigger than the total action along all of these circles @xmath4746 the right hand side is nothing else but the barrier function valued at @xmath2092 .
as the periodic orbit supports the minimal measure , both @xmath4735 and @xmath4736 approach the point @xmath837 where the periodic curve intersects the section @xmath4728 as @xmath3049 .
if the periodic orbit is not hyperbolic , the sequence of @xmath4744 approach @xmath71 slower than exponentially , i.e. , for any small @xmath1198 there exists @xmath54 such that @xmath4747 .
it follows that @xmath4748 .
as the periodic curve is assumed non - degenerate minimizer , some @xmath1288 exists such that @xmath4749 by subtracting @xmath4750 from the lagrangian @xmath63 we obtain that @xmath4751 and @xmath4752 where @xmath394 represents the backward ( forward ) weak - kam solution . since @xmath401 is semi - concave and @xmath4753 is semi - convex
, @xmath4754 is semi - concave . since @xmath4755 is a minimal point where @xmath4756 , there exists some number @xmath4757 such that ( cf .
@xcite ) @xmath4758 comparing this with the inequality ( [ hypereq2 ] ) , we obtain from ( [ hypereq1 ] ) a contradiction @xmath4759 if @xmath1198 is suitably small .
this proves the hyperbolicity of the periodic orbit .
_ proof of theorem [ appenhyperth1]_. according to theorem [ theo2 ] and [ theo3 ] , for each @xmath4760 $ ] , a neighborhood @xmath4761 $ ] of @xmath1155 and an open - dense set @xmath4762 exist such that for each @xmath4763 and each @xmath4449 $ ] each minimal orbit of @xmath4764 with homological class @xmath603 is hyperbolic .
as each @xmath4765 is positive , there exists finitely many @xmath1155 such that @xmath4428\subset\cup_i[e_i-\delta_{e_i},e_i+\delta_{e_i}]$ ] .
we take @xmath4766 , the hyperbolicity for @xmath4433 holds for all @xmath4431 $ ] .
once a minimal point is non - degenerate for certain @xmath3 , by the theorem of implicit function it has natural continuation to a neighborhood of @xmath3 .
namely , there exists a curve of minimal points passing through this point , it either reaches to the boundary of @xmath4428 $ ] , or extends to some point @xmath1807 where the critical point is degenerate .
since each global minimal point is non - degenerate , the critical point becomes local minimum when it enters into certain neighborhood of @xmath1807 .
as each non - degenerate minimal point is isolated to other minimal points for the same energy @xmath3 , there are finitely many such curves , denoted by @xmath3722 . for a curve @xmath3722 : @xmath4767 , the definition domain @xmath4303 contains finitely many closed sub - intervals @xmath4768 such that @xmath4769 for all @xmath4770 . by definition , @xmath4771 for @xmath4772 .
let @xmath4773 , we have finitely many curves @xmath4774 such that @xmath4775 reaches global minimum at the point @xmath71 if and only if @xmath4776 for certain subscript @xmath4777 . for each @xmath4778 , by the definition of @xmath4768 , some other subscript @xmath4779 exists such that @xmath4460 reaches the global minimum at the points @xmath4780 and @xmath4781 .
it is obviously an open - dense property that @xmath4782 thus , it is also open - dense that @xmath4428=\cup i_{i , j}$ ] and @xmath4428\backslash\cup\text{\rm int}i_{i , j}$ ] contains finitely points .
this completes the whole proof . @xmath4151
* acknowledgement * this work is supported by nnsf of china ( grant 11171146 , grant 10531050 ) , national basic research program of china ( 973 , 2007cb814800 ) , basic research program of jiangsu province ( bk2008013 ) and a program papd of jiangsu province , china .
i would like to thank my colleagues j. cheng , w. cheng , x. cui , j. yan and m. zhou for helpful discussions .
i also thank marc chaperon , alain chenciner and hakan eliasson for inviting me to give talks on this result at their seminars while i was visiting university of paris 7 for its hospitality .
the main ideas of the proof were presented on the conferences at edinburgh and at ias , princeton in october of 2011 .
arnold v.i .
kozlov v.v . and neishtadt a.i . ,
_ mathematical aspects of classical and celestial mechanics _ , dynamical systems iii , encyclopaedia of mathematical sciences , * 3 * springer - verlag berlin heidelberg , ( 1988 )
. bolotin s. _ homoclinic orbits in invariant tori of hamiltonian systems , _ dynamical systems in classical mechanics , 2190 , amer .
ser . 2 , * 168 * , amer .
soc . , providence , ri , 1995 .
delshams a. , de la llave r. and seara t. m. , _ geometric mechanism for diffusion in hamiltonian systems overcoming the large gap problem : heuristic and rigorous verification of a model _ , memoirs amer . math . soc . * 179*(844 ) , ( 2006 ) .
|
in this paper , arnold diffusion is proved to be generic phenomenon in nearly integrable convex hamiltonian systems with three degrees of freedom : @xmath0 under typical perturbation @xmath1 , the system admits connecting " orbit that passes through any two prescribed small balls in the same energy level @xmath2 provided @xmath3 is bigger than the minimum of the average action , namely , @xmath4 .
|
large transverse single spin asymmetries ( ssas ) have been observed in different hadronic reactions and these observations generated a lot of interest .
such experiments are conceptually rather simple , but their theoretical description proved to be challenging as the leading - twist contributions to such asymmetries vanish , see @xcite for a review . over the past few years
there was a splash of theoretical activity in this field , which mainly followed two lines : the @xmath0 factorization in terms of the transverse - momentum dependent ( tmd ) distributions ( e.g. @xcite ) , or , alternatively , collinear factorization including twist - three contributions in terms of multiparton correlation functions @xcite .
these two techniques have their own domain of validity and were shown to be consistent with each other in the kinematic regime where they both apply @xcite .
however , practically all existing calculations have been so far at the leading order ( lo ) which corresponds , roughly speaking , to the ( generalized ) parton model . in order to test the qcd dynamics and
eventually also reduce the dependence of theory predictions on both the factorization scale and the renormalization scale of the strong coupling , it is necessary to calculate the scale dependence ( evolution ) of the relevant nonperturbative functions .
the corresponding calculation was done in refs .
@xcite where the evolution equation was derived for the gluonic pole contributions to twist - three correlation functions that are relevant to ssa .
unfortunately , it turns out that the equations derived in @xcite are in contradiction to earlier results obtained in the framework of the operator product expansion ( ope ) .
a resolution of this discrepancy presents the main motivation for this study . although we do reproduce the terms that appear in ref .
@xcite , we obtain two additional contributions to the evolution equation both for flavor - nonsinget and flavor - singlet operators .
we have found that part of the disagreement is due to omission of cut diagrams corresponding to the mixing of conventional twist - three correlation functions that are discussed in the context of ssa with the contributions involving a gluon and the quark - antiquark pair on the opposite sides of the cut .
these are the same contributions that have been discussed for some time for semi - inclusive dis @xcite , and the new observation in this work is that they necessarily contribute to the leading logarithmic accuracy through the ( factorization ) scale dependence . for another part we do not have an intuitive explanation .
our ope - based calculation contains additional contributions ( compared to @xcite ) for which sending the gluon momentum to zero has to be done with caution , as will be explained in the text .
another question that we want to address is to clarify the relation of these results to earlier calculations of the scale dependence of twist - three correlation functions that contribute to inclusive reactions , pioneered by the study of the structure function @xmath1 in ref .
we would like to emphasize that the scale dependence of an _ arbitrary _ twist - three operator in qcd , hence _
arbitrary _ twist - three light - cone correlation function can be determined in terms of the two - particle evolution kernels introduced by bukhvostov , frolov , lipatov and kuraev ( bflk ) @xcite .
the bflk approach has become standard in calculations of the spectrum of anomalous dimensions ( dilatation operator ) in supersymmetric theories that are relevant to the ads / cft correspondence ( see e.g. @xcite ) , but , unfortunately , remains to be largely unknown to the broad qcd community .
we will show that evolution equations for the particular parton distributions relevant for ssa @xcite do not require an independent calculation but can be obtained from the bflk kernels by simple algebra . using an updated version of this technique
@xcite we derive the complete evolution equations for all relevant three - particle correlation functions for arbitrary gluon momenta .
this generalization is necessary since the gluon - pole contributions considered in @xcite do not have autonomous scale dependence .
the presentation is organized as follows .
section 2 is introductory .
we review basic properties of three - particle correlation functions and introduce a very convenient decomposition of momenta ( coordinates ) and the field operators in the spinor representation .
this rewriting makes the symmetries explicit and drastically simplifies the forthcoming algebra .
section 3 contains a detailed derivation of the flavor - nonsinglet evolution equation in the bflk approach and the comparison with @xcite .
our result coincides identically with the corresponding equation for the quark - antiquark - gluon correlation function relevant for the structure function @xmath1 .
the flavor - singlet evolution is considered in section 4 . in this case
there are two independent evolution equations for positive and negative c - parity that involve three - gluon correlation functions involving @xmath2 structure constants @xmath3 and @xmath4 , respectively . the equation for positive c - parity coincides , again , with the corresponding equation for the structure function @xmath1 , whereas the equation for negative c - parity is , to our knowledge , a new result ( in this form ) .
finally , in section 5 we summarize .
the full list of the bflk kernels in the momentum representation is collected in the appendix .
following ref . @xcite we will consider the correlation functions @xmath5 , @xmath6 , corresponding to the nucleon matrix elements of the quark - antiquark - gluon light - ray operators @xmath7 and also the correlation functions @xmath8 and @xmath9 corresponding to the three - gluon operators @xmath10 here @xmath11 is the light - like vector , @xmath12 , the `` plus '' projection is defined as @xmath13 .
note that there are two gluon operators with a different color structure ; the factors @xmath14 are written in terms of the @xmath2 structure constants @xmath15 in all cases the path - ordered wilson lines are implied that ensure gauge invariance .
they are not shown for brevity .
the definitions of parton distributions used by kang and qiu are ( cf . eqs .
( 12),(16),(14),(23 ) in @xcite ) : @xmath16 here @xmath17 is the nucleon spin vector normalized by the condition @xmath18 , @xmath19 with @xmath20 being the second light - like vector , @xmath21 , @xmath22 and @xmath23 , @xmath24 .
the spin vector @xmath17 is assumed to be transverse , @xmath25 . in the last two equations @xmath26 and @xmath27 correspond to matrix elements of @xmath28 and @xmath29 , respectively , cf .
( [ cpm ] ) . for our purposes
it is convenient to use a more symmetric notation with quark , antiquark and gluon momentum fractions treated equally @xmath30 where the integration measure is defined as @xmath31 we assume here the standard ( relativistic ) normalization of states ; the factor @xmath32 is necessary to ensure the reparametrization invariance to the choice of the light - cone vector @xmath33 .
obviously @xmath34 and similarly for the other distributions .
writing the definition in such a form ( fourier - transformed ) makes explicit the support properties of the correlation functions : they are defined on the surface @xmath35 and effectively are functions of two variables only .
the support of parton correlation functions in the notation by kang and qiu , i.e. in the @xmath36 plane , is shown in fig .
[ figure1 ] . ) in the @xmath36 plane .
[ figure1 ] , width=188 ] it can be separated in six different regions where @xmath37 , @xmath38 and @xmath39 are positive or negative , respectively .
the parton - model interpretation of each region is different . as discussed in detail in ref .
@xcite , interpretation of light - cone correlation functions as describing sequential emission or absorption of quark- , antiquark- and gluon - partons by the target arises by choice of a particular representation in terms of the sum over intermediate states ( cut diagrams ) that does not involve semidisconnected contributions .
in particular , the upper - right region in fig . [ figure1 ] corresponds to emission of a pair of a quark - parton and a gluon with momentum fractions @xmath40 , @xmath41 , respectively , and subsequent absorption of the quark - parton with @xmath39 .
the picture can be made more symmetric going over to the correlation functions ( [ tdt ] ) and using analogue to barycentric coordinates @xcite as shown in fig .
[ figure2 ] : @xmath42 the six different regions , labeled @xmath43 , @xmath44 , etc . , correspond to different subprocesses at the parton level @xcite ; for each parton @xmath45 `` plus '' stands for emission ( @xmath46 ) and `` minus '' for absorption ( @xmath47 ) .
alternatively , one may think of `` plus '' and `` minus '' labels as indicating whether the corresponding parton appears in the direct or the final amplitude in the cut diagram @xcite . ) in barycentric coordinates .
for the explanation of different regions and lines see text .
[ figure2],width=302 ] it is important that different regions do not have autonomous scale dependence ; they `` talk '' to each other and get mixed under the evolution .
the particular mixing pattern can be understood as follows .
it is easy to see that to one loop accuracy any contribution to the evolution equation only involves two of the three partons .
for example , parton1 and parton2 can exchange a gluon whereas parton3 stays a spectator , etc .
the full evolution kernel can , therefore , be split in three two - particle kernels that involve parton pairs ( 12 ) , ( 23 ) and ( 31 ) , respectively .
schematically @xmath48\otimes t_{\bar q f q}. \label{111}\ ] ] each two - particle kernel is , obviously , a function of the contributing parton momentum fractions , e.g. @xmath49 and due to energy conservation @xmath50 . in other words , the rate for the scale variation of a three - particle parton correlation function with given values of momentum fractions @xmath51 can only involve this function on the line of constant @xmath52 ( for the contribution of @xmath53 ) , as shown in fig . [ figure2 ] ( thick blue line _ * ab * _ ) . since @xmath54 ( in general ) ,
the total momentum fraction carried by the two participating partons is non - zero , @xmath55 .
this situation is familiar from studies of the scale dependence of leading - twist generalized parton distributions ( gpds ) ( see e.g. @xcite ) and the three regions @xmath44 , @xmath43 , @xmath56 traversed by the thick blue line * ab * in fig .
[ figure2 ] are in one - to - one correspondence to the two dglap regions ( @xmath57 and @xmath58 ) and the erbl ( central ) region ( @xmath59 ) in the corresponding evolution equations . as well known @xcite , the scale dependence of gpds in the dglap regions is autonomous , whereas in the erbl mode there are also terms describing `` leakage '' from the dglap regions . in the present context
, this result implies that evolution equation for a generic three - particle light - cone correlation function for the momentum fractions @xmath51 in the @xmath43 region , as in fig .
[ figure2 ] , will receive nontrivial contributions from the @xmath44 , @xmath56 , @xmath60 and @xmath61 regions as well , which have a different partonic interpretation .
so far we have considered the contribution of @xmath53 only .
the contributions of @xmath62 and @xmath63 are , in turn , kinematically constrained to the lines of constant @xmath64 and @xmath65 , respectively , and , for the particular choice of @xmath51 in the @xmath43 region , correspond to the dglap modes of the corresponding gpd - like evolution equations . recall that the dglap evolution is ordered in the momentum fraction ; hence only parts of the kinematically allowed regions contribute , as shown in fig .
[ figure2 ] by the thick red * cd * ( @xmath63 ) and thick green * ef * ( @xmath62 ) lines .
note that evolution of the parton correlation function for given values of @xmath51 only involves the regions of momentum fractions outwards from the center . if one draws a small(er ) hexagon ( see fig .
[ figure2 ] ) inside the big one , then evolution of the correlation function outside of the small hexagon does not depend on the latter inside the small hexagon .
this `` radial '' ordering is a generalization of the usual momentum fraction ordering in dglap equations .
in contrast , there is no ordering / restrictions in the `` azimuthal '' direction and , in principle , all regions `` talk '' to each other .
this property may result in the increase of the evolution rate in the small @xmath37 region .
analysis of symmetry properties and the scale dependence of the light - cone correlation functions can be simplified significantly by going over to the spinor representation .
each covariant four - vector @xmath66 is mapped to a hermitian @xmath67 matrix @xmath37 : @xmath68 where @xmath69 , @xmath70 and @xmath71 are the usual pauli matrices . in components
@xmath72 so that instead of the usual components @xmath73 each four - vector is described by its light - cone coordinates @xmath74 and two complex coordinates in the transverse plane @xmath75 .
the dirac ( quark ) spinor @xmath76 is written as @xmath77 where @xmath78 , @xmath79 are two - component weyl spinors , @xmath80 , @xmath81 .
finally , the gluon strength tensor @xmath82 and its dual @xmath83 can be decomposed as @xmath84 here @xmath85 and @xmath86 are chiral and antichiral symmetric tensors , @xmath87 , which belong to @xmath88 and @xmath89 representations of the lorenz group , respectively .
the antisymmetric tensors , @xmath90 and @xmath91 are normalized by @xmath92 and used for rising and lowering indices @xmath93 the contractions @xmath94 and @xmath95 are defined as @xmath96 the scalar product of two four - vectors @xmath97 and @xmath98 takes the form @xmath99 . for convenience , we present the expressions for dirac matrices in the spinor basis : @xmath100_{\alpha\dot\beta}\\ [ \bar\sigma^{\mu}]^{\dot\alpha\beta}&0 \end{pmatrix } , & & \gamma_5=\begin{pmatrix}-\delta_{\alpha}^\beta&0\\ 0&\delta^{\dot\alpha}_{\dot\beta } \end{pmatrix}\,,\end{aligned}\ ] ] where @xmath101 .
more expressions and various useful identities can be found in @xcite .
a covariant generalization of the decomposition in terms of light - cone coordinates and a transverse plane ( [ xd ] ) can be found by observing that any light - like vector can be represented as a product of two spinors . in particular one can parameterize the light - like vectors @xmath102 and @xmath20 as follows @xmath103 where @xmath104 , @xmath105 .
the standard convention @xmath106 corresponds to the normalization of auxiliary spinors @xmath107 the basis vectors in the plane transverse to @xmath108 can be chosen as @xmath109 and @xmath110 and an arbitrary four - vector @xmath37 represented as @xmath111 where @xmath112 and @xmath113 are real and @xmath114 , @xmath115 complex coordinates in the two light - like directions and the transverse plane , respectively , cf .
( [ xd ] ) . in particular the spin vectors @xmath17 and @xmath116 take the form @xmath117 where @xmath118 .
the `` @xmath119 '' fields ( good components ) are defined as the projections onto @xmath120 : @xmath121 the scale dependence of the correlation functions ( [ qiu1 ] ) , ( [ tdt ] ) is determined by the renormalization properties of the corresponding light - ray operators ( [ tdt ] ) , ( [ fff ] ) . in order to make a connection to the existing results ,
it is convenient to go over , for quark - antiquark - gluon operators , to another basis @xcite @xmath122\gamma_+ q(z_3)\,.\end{aligned}\ ] ] here and below we use a shorthand notation @xmath123 etc . , and
also @xmath124 .
the operators @xmath125 contribute to the structure function @xmath1 for polarized deep - inelastic scattering and their renormalization properties are studied in much detail @xcite .
it is easy to check that @xmath126 the second identity is a consequence of the known representation for the product of two antisymmetric @xmath127tensors as the determinant of the @xmath128 matrix in terms of the metric tensors @xmath129 . in the spinor basis one
finds for the operators @xmath130 the following expressions : @xmath131 , \notag\\ s_{\alpha\dot\alpha}^{-}({z})&=&2g\big[\bar\lambda_{\dot\alpha}\chi_+(z_1 ) f_{+\alpha}(z_2)\bar\chi_+(z_3 ) \notag\\ & & + \lambda_{\alpha}\bar\psi_+(z_1)\bar f_{+\dot\alpha}(z_2)\psi_+(z_3 ) \big].\end{aligned}\ ] ] taking into account transformation properties of the quark and gluon fields under charge conjugation , @xmath132 where @xmath133 , one derives @xmath134 since operators of different @xmath135-parity do not mix under renormalization , it is convenient to introduce the @xmath135-even and @xmath135-odd combinations @xmath136 where @xmath137 . note that these are _ not _ the same combinations as the ones appearing in eq .
( [ svq ] ) since the quark and antiquark field coordinates are interchanged .
the expressions for the operators @xmath138 in terms of chiral fields read @xmath139 where @xmath140 it is easy to see that @xmath141^\dagger$ ] , so that the @xmath135-even ( `` plus '' ) and @xmath135-odd ( `` minus '' ) @xmath142-operators are hermitian and antihermitian , respectively , i.e. @xmath143 .
the operators @xmath138 have autonomous evolution for the flavor - nonsinglet sector , whereas in the singlet sector they mix with three - gluon operators @xmath144 of the same @xmath135-parity : @xmath145 the color factors @xmath146 are given in eq .
( [ cpm ] ) . in the vector notation
this definition corresponds to @xmath147 where @xmath148 and @xmath149 are the permutation operators acting on the field coordinates , e.g. @xmath150 etc .
taking the nucleon matrix elements of the operators @xmath138 and @xmath144 one obtains the corresponding parton correlation functions in momentum space @xmath151 here @xmath152 , the integration measure @xmath153 is defined in eq .
( [ measure ] ) .
we use the same notation for the correlation functions as for the corresponding operators which , hopefully , will not result in a confusion .
note that although definitions of the operators @xmath138 and @xmath144 , eqs .
( [ sdef ] ) and ( [ fdef ] ) , involve the spin - vector @xmath154 , the dependence on @xmath154 actually drops out in the matrix elements ( [ sfm ] ) .
hermiticity and @xmath135-invariance imply that ( cf .
@xcite ) @xmath155 this means , in particular , that the correlation functions @xmath156 and @xmath157 are real functions .
notice also that the function @xmath158 is symmetric and @xmath159 antisymmetric under the interchange of the last two arguments : @xmath160 the six correlation functions defined in eq .
( [ tdt ] ) ( or , equivalently , eq .
( [ qiu1 ] ) ) can be written in terms of two independent quark - antiquark - gluon functions @xmath156 and two three - gluon functions @xmath157 as follows : @xmath161 , \notag\\ \delta t_{\bar q f q}({x})&=-\frac14\big[(1-p_{13})\mathfrak{s}^+({x})+ ( 1+p_{13})\mathfrak{s}^-({x})\big ] , \notag\\ t_{3f } ^\pm({x})&=\frac12(1\mp p_{13})\mathcal{f}^{\pm}({x})\ , , \notag\\ \delta t_{3f } ^\pm({x})&=-\frac12(1\pm p_{13})\mathcal{f}^{\pm}({x})\,.\end{aligned}\ ] ] here @xmath162 , etc . , and @xmath163 are the permutation operators for the corresponding momentum fractions , e.g. @xmath164 . as follows from ( [ relation11 ] ) the functions @xmath165 and @xmath166 are not independent , @xmath167 \ ] ] it follows from ( [ c - inv ] ) and ( [ relation11 ] ) that the correlation functions satisfy the following symmetry relation @xmath168 authors of @xcite also introduce symmetrized quark - antiquark - gluon parton distributions @xmath169 as follows from eq .
( [ symmetry5 ] ) such a symmetrization is not necessary since @xmath170 and @xmath171 . in the expressions given below we
drop the `` tilde '' notation for these functions .
the flavor - nonsinglet light - ray operators @xmath172 and @xmath173 satisfy the same evolution equation , @xmath174 where @xmath175 is an integral operator . the explicit expression for @xmath175 ( at one - loop )
can be restored from the corresponding result for @xmath176 in @xcite ( see also @xcite ) . expanding this equation at short distances @xmath177
one reproduces the mixing matrix for the twist - three local operators @xcite .
alternatively , the answer for @xmath175 can be obtained by simple algebra from the known expressions for the one - loop two - particle kernels @xcite .
this technique is general and applicable to arbitrary twist - three ( and twist - four @xcite ) evolution equations . in the next section
we will use this approach for a more complicated case of flavor - singlet operators so it makes sense to explain the details on the present ( simpler ) example .
the starting observation is that to one - loop accuracy any contribution to the evolution can only involve two partons ; hence @xmath175 can be represented as a sum of two - particle kernels .
schematically @xmath178 where the contractions correspond to the sum of relevant feynman diagrams ( in light - cone gauge ) .
the corresponding expressions were derived originally by bukhvostov , frolov , lipatov and kuraev ( bflk ) @xcite .
the complete list of the bflk kernels for arbitrary chiral fields is given in ref .
@xcite so that it only remains to contract the color indices . for the reader
s convenience we collect all the kernels in table [ table1 ] in the appendix .
after a simple algebra one gets @xmath179 where , in notation of ref .
@xcite , @xmath180 here @xmath181 are two - particle integral operators that act on the light - cone coordinates of the @xmath182-th and @xmath183-th partons : @xmath184(z_i , z_k)=\int dz'_i dz'_k\ , \mathcal{h}(z_i , z_k|z'_i , z'_k)\,\phi(z'_i , z'_k)\,.\ ] ] these kernels are @xmath185-invariant and depend on the conformal spins of partons that they are acting on .
the corresponding values are @xmath186 for quarks and @xmath187 for gluons ( for the `` plus '' components ) so one has to use @xmath188 , @xmath189 , @xmath190 .
one obtains , for example , @xmath191(z_1,z_2,z_3)=\int_0 ^ 1d\alpha\int_0^{\bar\alpha}d\beta\ , \mathfrak{s}^+(z_{13}^\alpha , z_2,z_{31}^\beta)\,,\end{aligned}\ ] ] where @xmath192 taking the nucleon matrix element of eq .
( [ rg1 ] ) one obtains the evolution equation for the corresponding parton distribution function . technically , this corresponds to going over from coordinate to the momentum fraction space @xmath193 .
thanks to energy conservation , two - particle kernels in momentum fraction space can be written in the following generic form @xmath194(x_i , x_k)&=\int_{-\infty}^\infty \mathcal{d}x'\ , \mathcal{h}(x_i , x_k|x'_i , x'_k)\,\varphi(x'_i , x'_k)\,,\end{aligned}\ ] ] where @xmath195 .
it is assumed that restrictions on integration regions over @xmath196 come from support properties of the kernels and the parton distributions ; the corresponding expressions are collected in the appendix . the last step , using the first two relations in eq .
( [ relation11 ] ) we obtain @xmath197 which is our final result .
the `` hamiltonian '' @xmath175 is defined in eqs .
( [ hns ] ) , ( [ hpw ] ) .
explicit expressions for the two - particle kernels @xmath198 , @xmath199 , @xmath200 and @xmath201 are given in the appendix .
note that when applying two - particle kernels to a three - particle parton distribution one has to treat the latter as a function of three independent variables ; the condition @xmath202 is applied afterwards . ) .
[ figure3],width=226 ] to compare our result with the calculation in ref .
@xcite we have to change notation to @xmath203 which are ( anti)symmetric functions under permutation @xmath204 , @xmath205 and @xmath206 . using explicit expressions for the two - particle kernels and taking the gluon - pole limit @xmath207 ( @xmath208 ) we obtain ( it is assumed that @xmath40 ) @xmath209\notag\\ & -n_c \mathcal{t}_{q , f}(x , x ) + \frac1{2n_c}\int_x^{1}\frac{d\xi}{\xi } \biggl[(1 - 2z)\mathcal{t}_{q , f}(x , x-\xi)-\mathcal{t}_{\delta q , f}(x , x-\xi)\biggr]\biggr\}\,,\end{aligned}\ ] ] where @xmath210 , @xmath211\end{aligned}\ ] ] and @xmath212 the partonic interpretation of different contributions in eq .
( [ kqe ] ) is illustrated in fig .
[ figure3 ] .
note that the condition of zero gluon momentum @xmath208 corresponds to the choice of @xmath213 on the ( positive ) horizontal axis .
the first term @xmath214 on the r.h.s . of eq .
( [ kqe ] ) corresponds to the integration over the horizontal line segment * xd * shown in red in fig .
[ figure3 ] .
the terms in @xmath215 are due to the integration along the blue , * ab * , and green , * ef * , lines in the regions @xmath56 and @xmath43 , and the ones in @xmath216 correspond to the contributions along the same lines in the regions @xmath44 and @xmath217 , respectively . by construction ,
the function @xmath218 is symmetric and @xmath219 antisymmetric under the interchange of arguments , @xmath204 . in fig .
[ figure3 ] this corresponds to a reflection symmetry around the horizontal axis .
the evolution equation ( [ kqe ] ) differs from the corresponding result by kang and qiu ( see eq . ( 99 ) in @xcite ) by the two extra terms in the second line .
the last term , proportional to @xmath220 , originates from the kernels @xmath221 and @xmath222 in eq .
( [ hpw ] ) . on the diagrammatic level this contribution corresponds to the `` exchange '' diagrams of the type shown in fig .
[ figure4 ] which correspond to mixing of the regions that have different partonic interpretation : @xmath223 , or @xmath224 .
similar contributions have been discussed in a somewhat different context in ref .
@xcite .
the other difference is the extra term @xmath225 in the second line in eq .
( [ kqe ] ) for which we do not see any obvious explanation .
the origin of this term in our calculation can be traced to the kernels @xmath226 and @xmath227 .
e.g. the second term in the expression for @xmath226 , eq .
( [ a : hhat ] ) , gives @xmath228 .
\label{subtle}\end{aligned}\ ] ] because of an overall @xmath38 factor , it is tempting to put this contribution to zero in the @xmath229 limit .
however , it is easy to see that in the same limit the integral becomes linearly divergent so that at the end a finite contribution arises
. this term can easily be missed if the gluon momentum is put to zero at the beginning of the calculation .
it is instructive to analyze eq .
( [ kqe ] ) in the large@xmath230 limit .
neglecting @xmath220 terms in ( [ kqe ] ) one is left with @xmath231\biggl\}\,,\end{aligned}\ ] ] where @xmath232 is obtained from @xmath233 , eq .
( [ pqq ] ) , omitting the @xmath234 factor .
one sees that the scale dependence of @xmath235 is determined by @xmath215 in the region @xmath236 , which corresponds to the contribution of the blue and green line segments * xb * and * xf * in fig . [ figure3 ] .
it receives no contribution from the `` diagonal '' region @xmath237 ( the red line segment * xd * in fig .
[ figure3 ] ) , the corresponding contributions cancel out between the first and the second terms in eq .
( [ kqe ] ) ( to the @xmath238 accuracy ) . by this reason
the conclusion in ref .
@xcite that the evolution of the function @xmath235 mainly follows a pattern determined by the quark splitting function can be misleading .
the evolution does simplify , however , in the large @xmath37 limit in which case the integration regions shrink to a point .
one obtains @xmath239 where , retaining singular terms at @xmath240 only @xmath241 - n_c\delta(1-z)\,.\ ] ] this result can be compared with the evolution of the usual @xmath242 structure function which involves , to the same accuracy @xmath243,\ ] ] and the twist - three contribution to the structure function @xmath1 @xcite @xmath244 -\frac{n_c}{2}\delta(1-z)\,.\ ] ] the last term in @xmath245 is written in the large-@xmath230 limit .
the contributions @xmath246 are the same in all three cases , which indicates that all three functions @xmath247 , @xmath242 and @xmath248 may have the same functional dependence on the bjorken variable @xmath37 in the @xmath249 limit .
different terms @xmath250 suggest , on the other hand , that twist - three functions are suppressed at large scales @xmath251 compared to the twist - two distribution as @xmath252 the suppression of @xmath248 compared to @xmath242 exactly corresponds to the gap between the lowest anomalous dimension in the spectrum of twist - three operators and the usual twist - two anomalous dimension . for the function
@xmath247 we predict a stronger suppression which translates to scaling violation in ssa .
this result can be phenomenologically relevant .
in the flavor - singlet sector one has to take into account mixing between the quark - antiquark - gluon and three - gluon operators with the same @xmath135-parity .
namely , @xmath172 gets mixed with @xmath159 and @xmath173 with @xmath158 . for each case , the evolution equation takes the matrix form @xmath253 where @xmath254 in what follows we assume that the quark - antiquark - gluon flavor - singlet operator is defined including the sum over @xmath255 light flavors , @xmath256 . for the diagonal entries we obtain @xmath257 where @xmath175 is given by eq .
( [ hns ] ) and @xmath258 with @xmath259 .
the off - diagonal entries in coordinate space take the form @xmath260\pi_0\ , , \notag\\ \mathbb{h}_{fq}^{-}&=-i\frac{n_c^2 - 4}{n_c}(1+p_{23})\frac{1}{z_{13 } } \big[2\mathcal{h}_{13}^+p_{13}-1\big ] .\end{aligned}\ ] ] the corresponding expressions in momentum space are given in the appendix , eqs .
( [ a : qf])([a : last ] ) .
the evolution equations for conventional @xmath261 , @xmath262 , @xmath165 @xmath263 can readily be obtained from eq .
( [ matrix ] ) by symmetrization in the arguments , as specified in eq .
( [ relation11 ] ) . in the limit of zero gluon momentum
our result for the evolution of @xmath264 differs from the corresponding expression in eq .
( 107 ) of @xcite by the same two terms as in the nonsinglet case ; the terms in @xmath265 ( in our calculation due to @xmath266 ) coincide . to compare our results for three - gluon distributions we introduce the functions @xmath267 which coincide with @xmath268 , @xmath269 , @xmath270 , @xmath271 defined in @xcite , respectively .
we remind that the @xmath272 distributions and hence @xmath270 , @xmath271 can be expressed in terms of @xmath273 alias @xmath269 , @xmath274 , so that they do not need to be considered separately . after some algebra one obtains the following equations for @xmath275 : @xmath276\notag\\ & \phantom{= } -(1-z)\left(z+\frac1z\right)t_f^\pm(\xi,\xi ) + \frac{1+z}{2}\big[t_f^\pm(x,\xi)-\delta t_f^\pm(x,\xi)\big ] \notag\\ & \phantom{= } \mp\frac12 ( 1-z)\big[t_f^\pm(x , x-\xi )
- \delta t_f^\pm(x , x-\xi)\big ] + \frac12a^{\pm}\bar p_{gq}(z)\big[{t}_{q ,
f}(\xi,\xi)\pm { t}_{q , f}(-\xi,-\xi)\big ] \biggr\}\biggr),\end{aligned}\ ] ] where @xmath277 and @xmath278 our result does not agree with that of kang and qiu @xcite , eqs .
( 109 ) , ( 110 ) in an overall sign in front of the @xmath279 distribution ( which may , however , be an artifact of a different sign convention for @xmath280 ) , and two extra terms ( up to the @xmath281 factor ) @xmath282 which seem to have the same origin as the extra contributions that we also have for the flavor - nonsinglet distributions : the first term in ( [ miss2 ] ) originates from contributions of the type in eq .
( [ subtle ] ) that involve a subtlety in taking the @xmath283 limit , and the second one corresponds to the contribution of `` exchange '' type diagrams , the kernel @xmath284 in ( [ hff ] ) , that give rise to mixing of regions with different partonic interpretation . closing this section ,
we want to stress that the scale dependence has to be studied using complete evolution equations for the three - particle parton distributions , eq .
( [ matrix ] ) . using the gluon - pole projected equations ( [ tf ] ) with a certain ansatz for the `` off - diagonal '' correlation functions @xmath285
may be misleading as they are modified by the evolution themselves .
we note in passing that the particular ansatz proposed in @xcite @xmath286e^{-{(x_1-x_2})^2/2\sigma^2}\end{aligned}\ ] ] and @xmath287 , is inconsistent with the constraint ( [ tdeltat ] ) .
we have given a complete reanalysis of the scale dependence of twist - three three - particle correlation functions that are relevant for calculations of single transverse spin asymmetries in the framework of collinear factorization .
the calculation is done using the two - particle kernels for the renormalization of light - ray operators in the spinor basis , which are available from ref .
evolution equations are derived for arbitrary parton momentum fractions , for the flavor - nonsiglet quark - antiquark - gluon distribution , eqs .
( [ rg1 ] ) , ( [ hns ] ) , ( [ hpw ] ) , and for the mixing matrix of the flavor - singlet quark - antiquark - gluon and three - gluon distributions with both positive and negative @xmath135-parity , eqs .
( [ matrix])([hff ] ) and ( [ a : qf])([a : last ] ) . specializing to the case of zero gluon momentum
we have compared our results with the recent calculation in ref .
there are two terms where we disagree , and their origin could be identified . as a byproduct of our calculation
we predict logarithmic scaling violation in the ssa at large values of bjorken @xmath37 which may be phenomenologically relevant .
numerical studies of the evolution effects on realistic models of parton distributions will be considered elsewhere .
we are grateful to yu .
koike , k. tanaka and w. vogelsang for correspondence which initiated this study and useful comments .
this work was supported by the german research foundation ( dfg ) , grant 9209282 , rnp grant 2.1.1/1575 and the rffi grants 07 - 02 - 92166 , 09 - 01 - 93108 .
gauge - invariant @xmath288-particle quasipartonic light - ray operators can be defined as a product of `` plus '' fields @xmath289x(z_1)\big)^{i_1 } \ldots \big([0,z_n]x(z_n)\big)^{i_n}\!\!\ ! , \label{3:o}\end{aligned}\ ] ] where @xmath290 , @xmath291 $ ] are wilson lines in the appropriate representation of the gauge group , @xmath292 are color indices and @xmath293 is an invariant color tensor such that @xmath294s_{i_1,\ldots , i_n}=0\,.\end{aligned}\ ] ] here and below it is implied that the generators @xmath295 are taken in the appropriate representation , @xmath296 where @xmath297 are the generators in fundamental representation . the condition in eq .
( [ 3:s - inv ] ) ensures that @xmath298 is a color singlet . for each @xmath288 , the set of quasipartonic operators with the same quantum numbers
is closed under renormalization @xcite .
a renormalized quasipartonic operator is written as @xmath299_r=\mathbb{z}_{ik}\mathcal{o}_{k}(x_0)\ , , \ ] ] where @xmath300 is the bare field .
renormalized operators satisfy the rg equation @xmath301_r=0\,.\end{aligned}\ ] ] here @xmath302 is the ( qcd ) beta function and @xmath303 is the matrix of anomalous dimensions . to the one - loop accuracy one obtains in dimensional regularization ( @xmath304 ) @xmath305 the operator @xmath175 ( hamiltonian ) is given by the sum of two - particles kernels , @xmath306 @xmath307 the general structure of the kernels is @xmath308 \,=\ , } \nonumber\\&= & \sum_{q}\sum_{i'_1 i'_2 } [ c_q]^{i_1i_2}_{i'_1i'_2 } [ \mathcal{h}^{(q)}_{12 } x^{i'_1}\otimes x^{i'_2}](z_1,z_2)\,.\end{aligned}\ ] ] here @xmath309^{i_1i_2}_{j_1j_2}$ ] is a color tensor , @xmath310 is an @xmath311 invariant operator which acts on coordinates of the fields , and @xmath76 enumerates different structures .
explicit expressions for the two - particle ( bflk ) kernels in the light - cone gauge in coordinate space are given in table [ table1 ] @xcite .
we tacitly assume existence of @xmath255 quark flavors ; the kernels * b*@xmath312 and * b*@xmath313 correspond to the flavor - nonsinglet and flavor - singlet quark pairs , respectively . in the kernel * b*@xmath313 the generators @xmath295 should be taken @xmath314 for the @xmath315 pair , and @xmath316 for the @xmath317 pair .
the kernels in table [ table1 ] are written in terms of several `` standard '' @xmath185-invariant operators defined as @xmath318(z_1,z_2)&= \int_0 ^ 1\frac{d\alpha}{\alpha}\big[2\varphi(z_1,z_2 ) \nonumber\\&-\bar\alpha^{2j_1 - 1}\varphi(z_{12}^\alpha , z_2)-\bar\alpha^{2j_2 - 1}\varphi(z_1,z_{21}^\alpha)\big]\ , , \nonumber\\ { } [ { \mathcal{h}}^d\varphi](z_1,z_2)&=\int_0 ^ 1d\alpha\ , \bar\alpha^{2j_1 - 1}\alpha^{2j_2 - 1}\,\varphi(z_{12}^\alpha , z_{12}^\alpha)\ , , \nonumber\\ { } [ \mathcal{h}^+\varphi](z_1,z_2)&=\int_0 ^ 1\!\!d\alpha\!\!\int_0^{\bar\alpha}\!\!d\beta\ , \bar\alpha^{2j_1 - 2}\bar\beta^{2j_2 - 2}\,\varphi(z_{12}^\alpha , z_{21}^\beta)\ , , \nonumber\\ { } [ \widetilde{\mathcal{h}}^+\varphi](z_1,z_2)&=\int_0 ^ 1\!d\alpha\int_0^{\bar\alpha}\!\!d\beta\ , \bar\alpha^{2j_1 - 2}\bar\beta^{2j_2 - 2}\,\left(\frac{\alpha\beta}{\bar\alpha\bar\beta}\right ) \nonumber\\ & \phantom{=}{}\times \varphi(z_{12}^\alpha , z_{21}^\beta)\ , , \nonumber\\ { } [ \mathcal{h}^-\varphi](z_1,z_2)&=\int_0 ^ 1\!\!d\alpha\!\!\int_{\bar\alpha}^1\!\!\!d\beta\ , \bar\alpha^{2j_1 - 2}\bar\beta^{2j_2 - 2}\,\varphi(z_{12}^\alpha , z_{21}^\beta)\ , , \nonumber\\ { } [ \mathcal{h}^{e,(k)}_{12}\!\varphi](z_1,z_2)&=\int_0 ^ 1d\alpha\,\bar\alpha^{2j_1-k-1}\ , \alpha^{k-1 } \varphi(z_{12}^\alpha , z_2 ) , \nonumber\\ \label{4:hh}\end{aligned}\ ] ] where in the last line it is assumed that @xmath319 . here and below we use a shorthand notation @xmath320 .
the kernels depend on the conformal spins of the partons they are acting on , @xmath186 for quarks and antiquarks and @xmath187 for gluons .
another shorthand notation is @xmath321 , \nonumber\\\end{aligned}\ ] ] where @xmath322 is the flavor index ; the sum runs over all possible flavors .
the constants @xmath323 where @xmath324 , correspond to the `` plus '' quark field and transverse gluon field renormalization in the axial gauge @xmath325 finally , @xmath149 and @xmath326 stand for permutation operators in position and color space , respectively .
for example @xmath327 going over from the light - ray operator renormalization to evolution equations for parton distributions corresponds to the fourier transformation of the functions @xmath328 from coordinate to the momentum fraction representation .
a generic two - particle kernel in momentum space has the form @xmath329(u_1,u_2)&=\int_{-\infty}^\infty \mathcal{d}v\ , \mathcal{h}(u_1,u_2|v_1,v_2)\,\varphi(v_1,v_2)\,,\end{aligned}\ ] ] where @xmath330 . in this appendix
we use the letters @xmath331 for momentum fractions .
note that the integration regions over @xmath332 and @xmath333 in eq .
( [ mrep ] ) are ( formally ) infinite ; in practice they are constrained by the support properties of parton distributions and also ( heaviside ) step - functions that are present in the kernels . for later convenience
we introduce a notation @xmath334 we remind that the kernels depend on the conformal spins of fields they acts on , cf .
( [ 4:hh ] ) , so that , in principle , each one has to carry a pair of indices @xmath335 which we omitted for brevity .
we obtain @xmath336 @xmath337 @xmath338(\boldsymbol{u})= } \notag\\ & = & \int \mathcal{d}{v}\biggl\ { \frac{u_1}{v_1}\frac{\theta(u_1,v_1-u_1)}{v_1-u_1 } \left(\varphi(\boldsymbol{u})- \left(\frac{u_1}{v_1}\right)^{2j_1 - 2}\hspace*{-5 mm } \varphi(\boldsymbol{v})\right ) \notag\\ & + & \frac{u_2}{v_2}\frac{\theta(u_2,v_2-u_2)}{v_2-u_2 } \left(\varphi(\boldsymbol{u})- \left(\frac{u_2}{v_2}\right)^{2j_2 - 2}\hspace*{-5 mm } \varphi(\boldsymbol{v})\right)\biggr\}\,.\end{aligned}\ ] ] note that @xmath339 ( [ a : hd ] ) vanishes in the dglap region .
the operator @xmath340 enters the bflk kernels in table [ table1 ] with the values of conformal spins @xmath341 . in momentum fraction
space one obtains @xmath342 the first term contributes to the dglap region only , the other two to the erbl region .
one derives the following expressions : @xmath343 obviously @xmath344 the `` modified '' @xmath345 kernel is needed for the conformal spins @xmath346 only .
it is given by a similar expression @xmath347 with @xmath348 note that the factors @xmath349 ( and , similar , @xmath350 ) satisfy the relation @xmath351 the kernels @xmath352 and @xmath353 are continuous functions in the whole domain although they are given by different expressions in the dglap and erbl regions .
the kernel @xmath354 has a different structure of regions , namely @xmath355 for the spins of interest @xmath356 one gets @xmath357 we also give here the explicit expressions for the off - diagonal kernels ( [ off ] ) in momentum representation .
the @xmath358kernel can be written as @xmath359 where @xmath360 and @xmath361 have the decomposition over different regions of the form ( [ a : h+ ] ) and ( [ a : h- ] ) , respectively .
for the corresponding functions @xmath362 and @xmath363 one obtains @xmath364 , \nonumber\\ { f}_{\mathcal{v}}(\boldsymbol{u}|\boldsymbol{v})&=\phantom{- } \frac{u_3 ^ 2}{v_1 ^ 2(u_1+u_2)^2}\left[\frac{2u_1v_1}{u_3(u_1+u_3)}-1\right].\end{aligned}\ ] ] the @xmath365kernel can be written as @xmath366 , \nonumber\\ \mathbb{h}_{fq}^-&=-\frac{n_c^2 - 4}{n_c}(1+p_{23})\left[\mathcal{w}^++\mathcal{w}^-\right].\end{aligned}\ ] ] here @xmath367 the kernel @xmath368 has the structure ( [ a : h+ ] ) with @xmath369 and the kernel @xmath370 takes the form ( [ a : h- ] ) with @xmath371
the kernels @xmath372 are antisymmetric under @xmath373 .
note that the kernels in ( [ hfq ] ) have discontinuities on the dglap erbl boundaries .
99 m. anselmino , a. efremov and e. leader , phys .
* 261 * , 1 ( 1995 ) [ erratum - ibid . *
281 * , 399 ( 1997 ) ] .
z. t. liang and c. boros , int .
j. mod .
a * 15 * , 927 ( 2000 ) .
v. barone , a. drago and p. g. ratcliffe , phys .
rept . * 359 * , 1 ( 2002 ) .
d. w. sivers , phys .
d * 43 * , 261 ( 1991 ) .
p. j. mulders and r. d. tangerman , nucl .
b * 461 * , 197 ( 1996 ) [ erratum - ibid .
b * 484 * , 538 ( 1997 ) ] .
d. boer , p. j. mulders and o. v. teryaev , phys .
d * 57 * , 3057 ( 1998 )
. s. j. brodsky , d. s. hwang and i. schmidt , phys .
b * 530 * , 99 ( 2002 ) .
s. j. brodsky , d. s. hwang and i. schmidt , nucl .
b * 642 * , 344 ( 2002 ) . j. c. collins , phys . lett .
b * 536 * , 43 ( 2002 ) .
d. boer , p. j. mulders and f. pijlman , nucl .
b * 667 * , 201 ( 2003 ) .
a. bacchetta , m. diehl , k. goeke , a. metz , p. j. mulders and m. schlegel , jhep * 0702 * ( 2007 ) 093 a. v. efremov and o. v. teryaev , sov .
j. nucl .
phys . * 36 * , 140 ( 1982 ) .
a. v. efremov and o. v. teryaev , phys .
b * 150 * , 383 ( 1985 ) .
j. w. qiu and g. sterman , phys .
lett . * 67 * , 2264 ( 1991 ) .
j. w. qiu and g. sterman , nucl .
b * 378 * , 52 ( 1992 ) .
a. efremov , v. korotkiian and o. teryaev , phys .
b * 348 * ( 1995 ) 577 .
j. w. qiu and g. sterman , phys .
d * 59 * , 014004 ( 1998 ) .
y. kanazawa and y. koike , phys .
b * 478 * , 121 ( 2000 ) .
y. kanazawa and y. koike , phys .
d * 64 * , 034019 ( 2001 ) .
h. eguchi , y. koike and k. tanaka , nucl .
b * 763 * , 198 ( 2007 ) .
y. koike and k. tanaka , phys .
b * 646 * , 232 ( 2007 ) [ erratum - ibid .
b * 668 * , 458 ( 2008 ) ] ; phys .
d * 76 * , 011502 ( 2007 ) .
z. b. kang , j. w. qiu , w. vogelsang and f. yuan , phys .
d * 78 * ( 2008 ) 114013 .
x. ji , j. w. qiu , w. vogelsang and f. yuan , phys
. rev .
lett . * 97 * , 082002 ( 2006 ) .
x. ji , j. w. qiu , w. vogelsang and f. yuan , phys .
d * 73 * , 094017 ( 2006 ) .
x. ji , j. w. qiu , w. vogelsang and f. yuan , phys .
b * 638 * , 178 ( 2006 ) .
y. koike , w. vogelsang and f. yuan , phys .
b * 659 * , 878 ( 2008 ) .
z. b. kang and j. w. qiu , phys .
d * 79 * ( 2009 ) 016003 .
w. vogelsang and f. yuan , arxiv:0904.0410 [ hep - ph ] . y. koike and k. tanaka , arxiv:0907.2797 [ hep - ph ] .
a. p. bukhvostov , e. a. kuraev and l. n. lipatov , jetp lett . * 37 * , 482 ( 1983 ) ; sov .
jetp * 60 * , 22 ( 1984 ) .
a. p. bukhvostov , g. v. frolov , l. n. lipatov and e. a. kuraev , nucl .
b * 258 * ( 1985 ) 601 .
a. v. belitsky , v. m. braun , a. s. gorsky and g. p. korchemsky , int .
j. mod .
a * 19 * , 4715 ( 2004 ) .
n. beisert , phys .
* 405 * , 1 ( 2005 ) .
v. m. braun , a. n. manashov and j. rohrwild , nucl .
b * 807 * , 89 ( 2009 ) .
v. m. braun , a. n. manashov and j. rohrwild , arxiv:0908.1684 [ hep - ph ] .
r. l. jaffe , nucl .
b * 229 * ( 1983 ) 205 .
a. f. mbius , _ der barycentrische calcl _ ,
verlag von johann ambrosius barth , leipzig , germany , 1827 .
i. i. balitsky and v. m. braun , nucl .
b * 311 * ( 1989 ) 541 .
a. ali , v. m. braun and g. hiller , phys .
b * 266 * , 117 ( 1991 ) .
d. mueller , phys .
b * 407 * , 314 ( 1997 ) .
y. koike , n. nishiyama and k. tanaka , phys .
b * 437 * , 153 ( 1998 ) .
s. e. derkachov , g. p. korchemsky and a. n. manashov , nucl .
b * 566 * ( 2000 ) 203 .
v. m. braun , g. p. korchemsky and a. n. manashov , phys .
b * 476 * , 455 ( 2000 ) . v. m. braun , g. p. korchemsky and a. n. manashov , nucl .
b * 603 * ( 2001 ) 69 .
x. d. ji and c. h. chou , phys .
d * 42 * , 3637 ( 1990 ) .
|
we reexamine the scale dependence of twist - three correlation functions relevant for the single transverse spin asymmetry in the framework of collinear factorization .
evolution equations are derived for both the flavor nonsinglet and flavor singlet distributions and arbitrary parton momenta .
our results do not agree with the recent calculations of the evolution in the limit of vanishing gluon momentum .
possible sources for this discrepancy are identified .
|
barakat has recently proposed a shifted large
dimension expansion ( slnt ) implemented by means of the asymptotic iteration method ( aim)@xcite . according to the author , this alternative approach corrects `` serious difficulties '' of previous applications of that successful perturbation theory .
he applies this improved slnt to an anharmonic oscillator with a rational potential
energy function .
the results are not impressive because the perturbation series are restricted to order six and contain only four terms becuse the odd coefficients vanish .
however , the article title promises a large
order calculation@xcite .
the purpose of this paper is to calculate some eigenvalues of the anharmonic oscillator with the rational potential accurately by means of the riccati
pad method ( rpm ) that has proved to be successful for other problems@xcite . in sec .
2 we show the model and outline the application of the rpm . in sec .
3 we discuss the results and compare barakat s approach@xcite with other applications of the slnt .
for concreteness we consider the eigenvalue equation @xmath0 \chi ( x)=e\chi ( x ) \label{eq : model}\ ] ] that applies to one dimensional and central field problems . in the former case we have @xmath1 and @xmath2 for even and odd states , respectively , and
in the latter case @xmath3 is the angular momentum quantum number .
one easily generalizes this equation to take into account a central field model for any other space dimension .
it is sufficiently general for our present goals . for comparison purposes we choose @xmath4 to be the radial quantum number@xcite .
the simple function @xmath5 is an exact solution to equation ( [ eq : model ] ) if @xmath6 \nonumber \\ e & = & ( 2l+3)(1 - 2c ) \label{eq : exact_e}\end{aligned}\ ] ] the modified logarithmic derivative of the solution @xmath7 can be expanded in a taylor series about the origin @xmath8 where the coefficients @xmath9 are polynomial functions of the energy @xmath10 .
if we require that the rational approximation @xmath11(x^{2})=\frac{\sum_{j=0}^{n+d}a_{j}x^{2j}}{% \sum_{j=0}^{n}b_{j}x^{2j } } \label{eq:[m / n]}\ ] ] satisfies @xmath11(x^{2})-\sum_{j=0}^{2n+d+1}f_{j}x^{2j}=\mathcal{o}(x^{2(2n+d+2)})\ ] ] then the hankel determinant @xmath12 with elements @xmath13 , @xmath14 vanishes .
@xmath15 is the determinant dimension and @xmath16 .
the main assumption of the rpm is that there are sequences of roots @xmath17 , @xmath18 of the hankel determinant that converge towards the eigenvalues of the equation ( [ eq : model ] ) for a given @xmath19 .
previous applications of this approach have shown that the hankel series of roots exhibit remarkable rate of convergence and yield accurate results for a wide variety of problems@xcite . if @xmath20 ( @xmath21 ) then the hankel series provide lower ( upper ) bounds .
this property has been proved for some simple models@xcite but it is more practical to proceed by inspection of the series .
the rpm yields the exact result in the partially solvable cases ; for example , when @xmath22 and @xmath23 , then @xmath24 , where the roots of the polynomial @xmath25 give the approximations to the other eigenvalues .
notice that the rpm yields the exact result @xmath26 for the ground state of the model ( @xmath27 ) , while , on the other hand , the aim with the large
dimension expansion gives this eigenvalue only approximately@xcite . in those cases that the eigenvalue equation ( [ eq : model ] ) is not exactly solvable
the rpm provides upper an lower bounds . for concreteness
we choose a model with intermediate values of the parameters : @xmath28 .
1 shows @xmath29 , which is is a clear indication of the convergence of the upper @xmath30 and lower @xmath31 bounds towards each other , for the states with @xmath32 , and @xmath33 .
it is well known that the convergence rate of the hankel sequences decreases as the number of radial zeros ( given by @xmath34 ) increases because we need larger values of @xmath35 to take into account the increasing number of poles of @xmath36@xcite .
we clearly appreciate this fact in fig . [
fig : fig1 ] , where the lower curve corresponds to the state with @xmath37 . for this state
we obtain the bounds @xmath38 .
notice that both the variational @xmath39@xcite and aim snlt @xmath40@xcite results lie outside the bounds and display a misleading number of wrong digits .
for the state @xmath33 we conclude that the sixth order aim snlt estimate @xmath41 lies outside our bounds @xmath42 .
we realize that the rpm enables one to estimate the accuracy of other approaches .
we can also monitor the convergence rate of a sequence @xmath43 by means of a sort of logarithmic error @xmath44 .
[ fig : fig2 ] shows this logarithmic error for the sequences @xmath45 and @xmath31 .
we can improve our results by means of an interpolation of the upper and lower bounds : @xmath46 .
if @xmath47 is our largest determinant dimension we obtain the optimal value of @xmath48 from the condition @xmath49 .
[ fig : fig2 ] shows also the logarithmic error for the sequence @xmath50 and we appreciate what appears to be a considerably gain of accuracy . from the interpolated sequence we estimate @xmath51 that is clearly between the bounds given above
we have shown that the rpm yields much more accurate results than the perturbation version of the aim .
the partial sums of the perturbative aim appear to converge towards the rpm results but they may by asymptotic divergent .
it is impossible to state that a series converges by simple inspection of the first four terms . in the particular case of the anharmonic oscillator ( [ eq : model ] ) the rpm provides tightly accurate upper and lower bounds to the eigenvalues that are most useful to estimate the accuracy of the results .
finally , we want to discuss one of barakat s comments regarding an early application of the hpm@xcite .
he states `` however , the previous authors in their work did not give explicit expressions of their algorithm , each order getting progressively much more complicated than the previous one , and the derivations were tediously long .
thus , the need arises here to have a relatively simple , fast and effective method that will provide large - order shifted @xmath52 expansions . ''
this sentence is surprising ( to say the least ) because maluendes et al@xcite gave a recurrence relation from which one obtains perturbation corrections of any order , either numerically ( in those days ) or analytically by the aid of today computer algebra systems . despite the fact that each order `` gets progressively more complicated than the previous one '' maluendes et al@xcite carried out perturbation calculations of order twenty five , while barakat@xcite has not managed to obtain more that six . besides
, that very same hypervirial perturbation method has enabled us to derive analytical expressions of the perturbation coefficients for any order and arbitrary potential energy functions@xcite . on the other hand ,
barakat presents numerical perturbation coefficients for some particular states and particular values of the potential parameters@xcite .
i doubt that the aim may be able to provide analytical expressions like those just mentioned@xcite .
of course , the general analytical expressions@xcite are bound to look more cumbersome than the numbers in table 1 in barakat s paper@xcite that apply to particular cases .
it is also important to remember that the aim _ does not give _ results in terms of arbitrary quantum numbers .
any such result has been obtained by induction from results for @xmath4@xcite .
this is possible for extremely simple problems but it is much more difficult in general . besides , in those cases the authors simply verified that their particular results agreed with the analytical expression derived by other more efficient strategies .
notice that barakat presents only the correction of order zero in terms of @xmath34@xcite . on the other hand , the unperturbed energy @xmath53 appears explicitly in the main equations of the hypervirial perturbative method and , consequently , all the corrections are straightforwardly given in terms of it@xcite . finally , we mention that a variant of the rpm @xcite has been successfully applied to several nonlinear problems of physical interest@xcite .
it is not clear that the aim can be suitable for the treatment of such problems .
|
we obtain tight upper and lower bounds to the eigenvalues of an anharmonic oscillator with a rational potential .
we compare our bounds with results given by other approaches .
|
helical dirac fermions , massless relativistic charged particles with spin locked to their linear momentum , are proposed to exist on the surface of three dimensional topological insulators ( tis)@xcite and later confirmed in experiments@xcite on samples like bi@xmath1se@xmath2 and bi@xmath3sb@xmath4 . for compounds such as hgte and bi@xmath1se@xmath2 , surfaces with a single dirac cone have been found@xcite .
while the low energy spectrum is very similar to that of graphene , the difference in the microscopic origin of the phenomena points to subtle differences .
for example , the presence of a single dirac cone in tis , as opposed to two doubly degenerate ones in graphene , leads to significant reduction in phase space at low energies .
furthermore , the coupling of physical spin , rather than pseudospin ( related to sublattice symmetry ) , in tis suggests that the response in magnetic fields will be markedly different .
the key property of the surface states of tis is the magnetoelectric effect .
unfortunately , the presence of bulk carriers has made it hard to observe it in transport measurements@xcite .
one way to overcome this is to identify unique signatures , in transport , of the surface states .
in particular the anomalous dependence of transport coefficients on the gyromagnetic ratio ( g - factor ) can be exploited to this end .
the advantage of this approach is that the anomalous contribution can be enhanced by applying an in plane electric field@xcite . in this article
we study the thermopower and magnetotransport properties of these surface dirac fermions in the presence of randomly distributed impurities . in the conventional two dimensional electron gas under a quantizing magnetic field
, the diagonal thermopower , @xmath5 , of the clean sample at low temperatures shows a series of peaks near the landau levels @xmath6 .
the peak values are @xmath7 independent of magnetic field strength@xcite .
one of our goals is to show how these universal features are modified in graphene and tis . for electrons exhibiting a dirac spectrum , such as graphene
, this height of @xmath5 is expected to vary as @xmath8 , as required by the berry phase effects leading to a @xmath9 integer shift in the landau level index@xcite .
we verify this result for @xmath10 and discuss the singular behavior near @xmath11 .
we use self consistent born approximation ( scba ) to consider the effect of short ranged , randomly distributed impurities on transport and magnetotransport@xcite of the dirac fermions and compute diagonal and off diagonal thermopower numerically .
similar to the results in conventional 2d metals , the height of @xmath5 at low temperature does not have universal value after the inclusion of impurities@xcite . we also show that for intermediate impurity strength , not considered in previous studies which have focused on the unitarity limit@xcite , the particle hole symmetry of the original dirac spectrum is broken under scba while for weak and strong impurity strength the particle hole symmetry is restored .
another source of scattering in graphene is charged coulomb impurities and its effect on conductivity and thermopower , both within perturbation@xcite and scba@xcite , have been extensively studied .
comparison to available data@xcite suggests that a single scattering mechanism can not account for all observed features@xcite .
our formulation for tis , with the g - factor set to zero , is similar to the case of graphene .
thus , in addition to exploring the difference in the transport properties induced by the spin orbit coupling , we can also compare our prediction with the available data .
qualitative agreement with thermopower data is obtained for all but the zeroth landau level .
the failure is related to the underestimation of the longitudinal conductivity .
the important new feature in topological insulators , as compared to graphene , is the spin - orbit coupling .
in addition to orbital quantization , an external magnetic field couples to the momentum via its zeeman coupling with the spin .
such an interaction reveals itself in novel signatures in transport unique to topological insulators .
for example , the universal amplitudes in thermopower depend on the gyromagnetic ratio ( @xmath0 ) , but no splitting of peaks occurs as the surface states are derived from a single dirac cone . while the dimensional coupling constant @xmath12 is quite small for @xmath13 ( of order @xmath14 ) to yield a measurable signature , we derive the expected behavior in a model system where @xmath15 is enhanced by an order of magnitude .
the enhancement is achievable in principle by applying an inplane electric field@xcite .
the paper is organized as follows : in sec . [ s2 ] we describe the general hamiltonian and t - matrix formulation used to perturbatively compute the effects of impurities . in sec .
[ s3 ] we present our results , within linear response , for conductivity and the diagonal thermopower @xmath5 in the absence of magnetic fields . in sec .
[ s4 ] we consider the linear response conductivity and the thermopower @xmath5 and @xmath16 in the presence of the magnetic field .
we compare our numerical results with those obtained in the graphene experiments and other theoretical results .
conclusion are summarized in sec .
the surface state of the electrons on a topological insulator under electric and magnetic fields is@xcite @xmath18 \psi(\vec{r})\end{aligned}\ ] ] where @xmath19 is the fermi velocity ( @xmath20 is restored in computing current current correlation ) , @xmath21 denotes the pauli matrices in spin space , @xmath22 is the @xmath23 identity matrix , @xmath24 is the annihilation operator for dirac spinor ( @xmath17 denotes the transpose of a row vector)and @xmath25 is the canonical momenta . electric field
is assumed to be pointed in x - direction and magnetic field is perpendicular to the plane ( z - direction ) .
the coupling to local impurities of strength @xmath26 is modeled as @xmath27 where @xmath28 denotes spin degree of freedom of electron operator @xmath29 and sum over @xmath30 indicates summing all impurity positions .
the full hamiltonian of the system is given by @xmath31 . in the dilute impurity limit
we compute transport coefficients to linear order in impurity density @xmath32 .
the finite temperature green s functions of the electrons are @xmath33 where @xmath34 is the two dimensional momentum of the surface band , @xmath28 and @xmath35 are the spin indices , and @xmath36 with @xmath37 being the time ordering operator .
to linear order in @xmath38 , the greens function satisfy @xmath39 where @xmath40 is the fermionic matsubara frequency and @xmath41 the @xmath17 matrix @xmath42 is @xmath43^{-1}\end{aligned}\ ] ] in the following two sections we discuss the nature of transport without and with the magnetic field .
we use the above t - matrix formulation to compute the greens function .
transport coefficients are obtained within linear response using the appropriate current - current correlation function .
in zero magnetic field and @xmath44 the hamiltonian for the helical metal without impurities reduces to @xmath45 the eigenfunction of the schordinger equation @xmath46 in the energy eigenstate is given by @xmath47 . here
@xmath48 is the vacuum , @xmath49 denotes spinless fermion operator , @xmath50 , and the spin part is described by the two component spinor function @xmath51 is @xmath52 with @xmath53 denoting conduction / valence bands , @xmath54 , and @xmath55 is the area of the system . following
the full self consistent born approximation@xcite we have the self energy of electron given by @xmath56^{-1}}\end{aligned}\ ] ] the derivation of eq.([sp2 ] ) is in the appendix.[a0 ] .
this self energy is solved numerically as shown in fig.[figf1 ] .
the main feature of this self energy is that for moderate impurity strength ( @xmath57 ) the particle hole symmetry is not preserved , while for strong ( @xmath58 ) or weak ( @xmath59 ) impurity interaction the particle hole symmetry is restored .
the scale is set by the bandwidth which in our calculations is @xmath60 . , @xmath61 , and @xmath57 for blue lines and @xmath58 for dashed purple lines .
the inset of right figure shows imaginary part of self energy for weak impurity potential @xmath59.,title="fig : " ] , @xmath61 , and @xmath57 for blue lines and @xmath58 for dashed purple lines .
the inset of right figure shows imaginary part of self energy for weak impurity potential @xmath59.,title="fig : " ] the physical quantity we are interested in is the transport properties of surface states . in the kubo formulation the linear response conductivity is related to current current correlation . written in matsubara formulation the particle current current correlator is@xcite @xmath62 here
@xmath63 is the area of the system , with @xmath6 being the number of unit cells and @xmath64 denoting unit cell area .
@xmath65 is related to the linear response charge conductivity @xmath66 via @xmath67 the particle current operator in @xmath68 direction in the heisenberg representation is @xmath69=-v_f\sigma_k\end{aligned}\ ] ] note that the electron current is related to the spin of the electron and hence sensitive to the zeeman coupling . the corresponding particle current operator in the energy eigenbasis , eq.([sp1 ] ) ,
is @xmath70 within the limit of dilute random impurities we use the scba which is valid as long as @xmath71 ( with @xmath72 as electron s mean free path)@xcite .
the kubo formula for the real part of particle current current correlation is then given as @xmath73\\\times\sum_{\vec{k}}\im g(\vec{k},\epsilon+i0^+)\im g(\vec{k},\epsilon+\omega+i0^+)\end{aligned}\ ] ] with @xmath74 ( fermi dirac distribution function ) and @xmath75 .
quantum effects such as weak localizations are not considered in this article as the major contributions to the conductivity comes from non crossing diagrams .
the momentum integral @xmath68 is cut off by linear spectrum energy boundary @xmath76 with @xmath77 .
this upper cutoff is also related@xcite to unit cell size @xmath64 by @xmath78 .
the zero frequency thermal response function @xmath79 is related to @xmath65 by@xcite @xmath80 and the thermopower @xmath81 is @xmath82 in zero magnetic field only diagonal component of @xmath81 is nonzero . for a given self energy @xmath83
the dc conductivity is @xmath84 the explicit form of @xmath85 is given in the appendix [ a1 ] .
notice that for @xmath86 the low temperature conductivity is proportional to @xmath87 .
thus the low temperature conductivity at dirac node has an universal value @xmath88 .
close to zero temperature eq.([cond ] ) to eq.([tp ] ) lead to the generalized mott formula for the thermopower@xcite @xmath89_{im}[\partial\sigma/\partial\mu]_{mj}\end{aligned}\ ] ] for a clean surface state we take @xmath90 and @xmath91 and thus away from the half filling we have @xmath92 . near zero temperature in the clean limit
the thermopower takes the form @xmath93 since @xmath94 in general is nonzero , the thermopower in the clean limit is susceptible to impurity interaction and does not show universal behavior for a given chemical potential , different from what occurs in the case of finite magnetic field which we show in the next section . the dc conductivity and diagonal thermopower for general impurity strength
are computed numerically and shown in fig.([figf2 ] ) .
similar to the results of self energy in fig.([figf1 ] ) the dc conductivity shows asymmetry in @xmath95 for moderate impurity strength .
the dip near @xmath96 for @xmath57 case is the remnant signature of the clean sample , which shows up as a singularity in thermopower of the form @xmath97 for @xmath86 .
( left ) and thermopower @xmath98 ( right ) at @xmath99 vs chemical potential @xmath100 for @xmath101 ( blue lines ) and @xmath58 ( dashed purple lines ) .
the conductivity at @xmath96 for dashed purple line is around @xmath88 .
the dotted brown line on the right figure is the first term in eq.([cls ] ) which is the dominant thermopower in large @xmath95.,title="fig : " ] ( left ) and thermopower @xmath98 ( right ) at @xmath99 vs chemical potential @xmath100 for @xmath101 ( blue lines ) and @xmath58 ( dashed purple lines ) .
the conductivity at @xmath96 for dashed purple line is around @xmath88 .
the dotted brown line on the right figure is the first term in eq.([cls ] ) which is the dominant thermopower in large @xmath95.,title="fig : " ]
we assume the magnetic field @xmath102 is perpendicular to the applied electric field @xmath103 and choose the gauge @xmath104 . for @xmath105 the eigenvalue of @xmath106
is given by the landau level spectrum @xmath107 where @xmath12 , @xmath108 is the band index , @xmath109 is the magnetic length , and @xmath110 denotes the index of landau level . for dirac electrons on the surface of hgte @xmath111 .
thus for usual magnetic field strength we can in general ignore the zeeman contribution in the spectrum @xcite .
nevertheless we keep this zeeman contribution in our calculation and the dominant feature that results from the zeeman term is the lack of particle hole degeneracy at @xmath112 level .
the loss of symmetry leads to anomalous features in off diagonal conductivity @xmath113 as well as thermopower . within linear response
we take @xmath114 in the hamiltonian and the field operator @xmath29 are expressed in linear combinations of energy eigenstates .
@xmath115 with @xmath116 , @xmath117 and @xmath118 is the hermite polynomial@xcite .
the sum over level index @xmath110 is cut off by the linear spectrum boundary @xmath119 . for @xmath120 , @xmath121 in the typical surface state of topological
insulator@xcite the upper cutoff obtained this way is @xmath122 . in our numerical computation
we take @xmath123 for practical computation@xcite and checked that the results do not change much by comparing with @xmath124 . in this representation @xmath106
is diagonal and the green s function in energy eigenstate is @xmath125 the impurity hamiltonian , eq.([hi ] ) , in this basis is @xmath126c^{\dagger}_{p , m,\bar{\gamma } } c_{k , n,\gamma}\big]\end{aligned}\ ] ] to include the effect of dilute random impurities we use the self consistent born approximation and the averaging procedure over impurities positions as in ref . .
the position averaged green s function is @xmath127 g(k , n,\gamma;i\omega;\vec{r}_1,\vec{r}_2,\vec{r}_3,\ldots,\vec{r}_{n_{imp}})\\&&\equiv g(k , n,\gamma;i\omega)\end{aligned}\ ] ] in the averaging procedure we first integrate @xmath128 and then @xmath129 and use the following identity @xmath130 and then sum over momentum index in the propagator by using @xmath131 the self energy within self consistent born approximation is @xmath132 in eq.([scba ] ) the self energy , identity matrix @xmath133 , and green s functions are all of dimension @xmath134 with @xmath135 as the maximal landau level cutoff by the linear spectrum
. denote @xmath136 and @xmath137 as @xmath110 by @xmath110 matrix with matrix element @xmath138 along @xmath139 , the bare self energy term written in the energy eigenvalue basis is then given by @xmath140 an important point to note is that off diagonal term appears when the zeeman effect is important .
such terms are absent in graphene and are responsible for g - factor dependent contributions to transport . by performing analytical continuation ( @xmath141 ) , the diagonal and off diagonal green s functions ( the off diagonal part describes the change from @xmath142 to @xmath143 state or particle hole transition ) are @xmath144 with diagonal self energy @xmath145 at landau level @xmath110 , self energy @xmath146 at zeroth landau level , and off diagonal self energy @xmath147 given by @xmath148 the explicit expression for internal propagators , @xmath149 , @xmath150 , and @xmath151 are given in the appendix [ a2 ] . within these internal propagator the sum over @xmath68 ( in @xmath152 direction ) gives a factor @xmath153 , which accounts for degeneracy of a given landau level per unit cell and we drop the @xmath68 dependence in the green function in the appendix [ a2 ] .
the self consistent solutions of eq.([gr ] ) and eq.([selfenergy1 ] ) gives the density of states and the renormalization of landau levels due to impurities@xcite . in the case of weak magnetic field
where the zeeman contribution can be neglected we obtain the same results , modulo a constant factor due to spin and valley degeneracies in graphene , as in ref . .
our main goal is to study the transport properties of surface states .
the particle current current correlation @xmath154 and @xmath155 from eq.([kubo ] ) are given in the appendix [ a3 ] .
before we proceed to evaluate the conductivity @xmath66 and thermal power @xmath81 numerically for general impurity strength , we take the clean system limit @xmath156 and study analytically the thermopower in the two dimensional helical metal . we show that in this clean limit we obtain universal feature for diagonal thermopower , similar to its counterpart in the conventional two dimensional electron gas . for weak impurities @xmath90 we have @xmath157 .
taking this expression into eq.([l11xx ] ) gives @xmath158 since the energy difference between landau levels is large enough to prevent the transitions between different levels without the help of impurity broadening .
also from eq.([l11xy ] ) @xmath159 with @xmath160 given by @xmath161\\\nonumber & = & \frac{-t v_f^2}{4\pi l_b^2 \omega\hbar}\big\{\tanh(\frac{\epsilon_0-\mu}{2t})\big[\frac{\epsilon_0+\omega}{(\epsilon_0+\omega)^2-\epsilon_1 ^ 2}\\\nonumber&-&\frac{\epsilon_0-\omega}{(\epsilon_0-\omega)^2-\epsilon_1 ^ 2}\big ] -\frac{\tanh(\frac{\epsilon_1-\mu}{2t})\omega}{(\epsilon_1-\epsilon_0)^2-\omega^2}-\frac{\tanh(\frac{-\epsilon_1-\mu}{2t})\omega}{(\epsilon_1+\epsilon_0)^2-\omega^2}\\\nonumber & + & \sum_n \frac{1}{2}\big ( ( \tanh(\frac{\epsilon_n-\mu}{2t})+\tanh(\frac{-\epsilon_n-\mu}{2t}))\\\nonumber&\times&\big[\frac{\epsilon_n+\omega}{(\epsilon_n+\omega)^2-\epsilon_{n+1}^2}-\frac{\epsilon_n-\omega}{(\epsilon_n-\omega)^2-\epsilon_{n+1}^2}\big ] \\\nonumber & -&(\tanh(\frac{\epsilon_{n+1}-\mu}{2t})+\tanh(\frac{-\epsilon_{n+1}-\mu}{2t}))\\\label{eq11 } & \times&\big[\frac{\epsilon_{n+1}+\omega}{(\epsilon_{n+1}+\omega)^2-\epsilon_{n}^2}-\frac{\epsilon_{n+1}-\omega}{(\epsilon_{n+1}-\omega)^2-\epsilon_{n}^2}\big]\big)\big\}\end{aligned}\ ] ] from eq.([eq11 ] ) it is easy to see that for @xmath162 we have @xmath163 as a result of particle hole symmetry .
for @xmath164 we have @xmath165\big\}\end{aligned}\ ] ] here @xmath166 is the cyclotron frequency .
eq.([eq27 ] ) suggests that the hall conductivity can be used as a probe for nonzero @xmath167 , or in other words , whether the dirac spectrum is gapped by the zeeman term under the assumption that the sample is clean . by using eq.([eq27 ] ) in eq.([l12 ] ) we get @xmath168(-2)-(\frac{\omega_c^2}{(\epsilon_1-\epsilon_0)^2}+1)[\theta(\epsilon_1-\epsilon)\\&&-\theta(\epsilon-\epsilon_1 ) ] + ( \frac{\omega_c^2}{(\epsilon_1+\epsilon_0)^2}+1)[\theta(\epsilon+\epsilon_1)-\theta(-\epsilon_1-\epsilon)]\\ & & -2\sum_{n=2}^{n_{max}}[(\theta(\epsilon_n-\epsilon)-\theta(\epsilon-\epsilon_n))-(\theta(\epsilon_n+\epsilon)\\&&-\theta(-\epsilon-\epsilon_n ) ) ] \big\}\end{aligned}\ ] ] in the clean limit the off diagonal thermopower @xmath169 . the diagonal component @xmath170 , in the low temperature limit ( i.e. @xmath171 ) with chemical potential close to landau level @xmath172 , is given by @xmath173 as a function of chemical potential @xmath95 for temperature @xmath174 ( blue , dashed ) , @xmath175 ( purple , thick line ) , and @xmath176 ( brown , dot dashed ) in the clean helical metal . @xmath95 and @xmath177 are in unit of @xmath178 and @xmath5 is in unit of @xmath179 . around @xmath96 @xmath180
becomes order of @xmath181 for all three cases .
the peaks value are in nice agreement with eq.([clean ] ) . ] as a function of chemical potential @xmath95 for temperature @xmath182 with gyromagnetic ratio @xmath183 ( blue thin line ) , @xmath184 and @xmath185 ( purple thick line ) , and @xmath184 and @xmath186 ( brown dot dashed line ) in the clean helical metal . @xmath95 and @xmath177 are in unit of @xmath178 and @xmath5 is in unit of @xmath179 .
the peak height is not universal with large zeeman effect . ]
we can approximate @xmath187 because @xmath188 is small .
under this approximation we see that for @xmath189 the thermopower becomes @xmath190 , which again serves as a measurable quantity for the zeeman term . in real material
, however , this singular behavior around @xmath167 will be masked by the broadening of zeroth landau level due to impurities or coulomb interaction . for materials with negligible zeeman interaction the thermopower reaches @xmath191 for @xmath192 , irrespective of the strength of magnetic field .
this is consistent with the berry phase argument@xcite . for weak magnetic field strength ( @xmath193 ) the thermopower of two dimensional helical metal is similar to that of the conventional 2d metal in that there exists universality for peak value of thermopower in the clean limit@xcite .
the feature observed in the helical metal , however , is different from the case of conventional 2d metal where every landau level @xmath6 has universal value @xmath194 . from eq.([clean ] )
we see that for large zeeman effect ( @xmath195 ) the peak height of @xmath5 is no longer universal and depends on the strength of zeeman interaction .
we artificially increases this zeeman effect by increasing the gyromagnetic ratio and magnetic field strength .
the results are shown in fig.[fig01 ] .
in addition to the non - universality another feature to note is the shift in the position of the peaks reflecting the zeeman contribution to the energy .
such a shift is most prominent for the lowest landau levels , increasing the spacing for positive @xmath6 .
the significant reduction in the peak values as a function of magnetic field , as compared to graphene , provides a clear signature of the surface state .
one consequence is that for the same impurity broadening , systems with larger zeeman coupling possess gaps in the spectrum near the zeroeth landau level even with the broadening from impurity interaction .
the effect of impurities are explored in the next subsection . for general impurity strength and large magnetic field strength
we expand @xmath196 . within this approximation @xmath197 .
the off - diagonal self energy @xmath147 is @xmath198 the @xmath199 self energy terms in self consistency equations eq.([gr ] ) and eq.([selfenergy1 ] ) are therefore simplified to two variables @xmath200 and @xmath201 which we evaluated numerically in this section .
we put band energy @xmath202 , fermi velocity @xmath203 , and impurity density @xmath61 in the following numerical computations .
the real and imaginary part of the self energy are shown in fig.[fig2 ] for @xmath204 . for strong impurity strength ( blue lines ) the imaginary part of the self energy shows particle hole symmetry ( @xmath96 in all self energy figures ) while for moderate impurity strength ( purple dashed lines ) the self energy shows more weight on the hole side . in clean limit , as discussed in last section , or with weak impurity strength the particle hole symmetry of the self energy is restored .
this can be understood by taking @xmath205 for vacancies and @xmath156 for weak impurity in eq.([scba ] ) . for @xmath205 , @xmath206 and thus @xmath207 , which renders the self energy to be particle hole symmetric . for @xmath156 , @xmath208 .
the first constant term can be absorbed in the shift of chemical potential and therefore the self energy obtained by second order self consistency equation is particle hole symmetric . for intermediate impurity strength , the factor @xmath133 in numerator of eq.([scba ] ) is comparable with @xmath209 .
thus for repulsive @xmath26 the self energy is shifted to the left .
( left ) and @xmath210 ( right ) vs @xmath211 .
energy is in unit of @xmath178 and @xmath204 in all figures .
top panel shows real part and bottom panel shows imaginary part .
impurity strength @xmath212 ( blue lines ) , @xmath213 ( purple dashed lines ) , and impurity concentration @xmath214.,title="fig : " ] ( left ) and @xmath210 ( right ) vs @xmath211 .
energy is in unit of @xmath178 and @xmath204 in all figures .
top panel shows real part and bottom panel shows imaginary part .
impurity strength @xmath212 ( blue lines ) , @xmath213 ( purple dashed lines ) , and impurity concentration @xmath214.,title="fig : " ] + ( left ) and @xmath210 ( right ) vs @xmath211 .
energy is in unit of @xmath178 and @xmath204 in all figures .
top panel shows real part and bottom panel shows imaginary part .
impurity strength @xmath212 ( blue lines ) , @xmath213 ( purple dashed lines ) , and impurity concentration @xmath214.,title="fig : " ] ( left ) and @xmath210 ( right ) vs @xmath211 .
energy is in unit of @xmath178 and @xmath204 in all figures .
top panel shows real part and bottom panel shows imaginary part .
impurity strength @xmath212 ( blue lines ) , @xmath213 ( purple dashed lines ) , and impurity concentration @xmath214.,title="fig : " ] vs @xmath215 for @xmath204 ( left ) and @xmath216 ( right ) .
blue lines are @xmath212 and purple dashed lines are for @xmath213 .
the brown dotted lines indicate the density of state with degeneracy factor @xmath217 and without impurities.,title="fig : " ] vs @xmath215 for @xmath204 ( left ) and @xmath216 ( right ) .
blue lines are @xmath212 and purple dashed lines are for @xmath213 .
the brown dotted lines indicate the density of state with degeneracy factor @xmath217 and without impurities.,title="fig : " ] from the self energy we obtain the density of state per unit cell as @xmath218\end{aligned}\ ] ] the numerical results are shown in fig.[fig2b ] for two different magnetic field strengths .
for @xmath213 we see clear signature of particle hole asymmetry and there are more holes than electrons when @xmath96 .
the hall conductivity @xmath113 obtained in this case does not flip its signs at @xmath96 .
rather the charge neutrality point has been shifted to positive @xmath95 .
another feature is that the oscillation due to applied magnetic field is more apparent in @xmath219 region than in @xmath220 region for @xmath213 .
this feature also shows up in the longitudinal conductivity @xmath221 v.s .
chemical potential .
the numerical results for dc conductivity @xmath221 and @xmath113 as a function of chemical potential @xmath95 are plotted in fig.[fig3 ] . as is expected from the results of self energy and density of state , both @xmath221 and @xmath113
show more oscillatory behavior / larger steps for @xmath222 region .
( left ) and @xmath113 ( right ) as a function of chemical potential @xmath95 .
conductivity is in unit of @xmath223 .
@xmath95 , @xmath224 , impurity strength @xmath225 ( blue lines ) , and @xmath226 ( purple dashed lines ) are in unit of @xmath178 .
magnetic field strength @xmath204 for upper and @xmath216 for lower figures .
the inset of upper right figure shows the blowup plot with smaller chemical potential range .
the @xmath113 in @xmath204 , @xmath213 case has been shifted down by @xmath227 and @xmath113 in @xmath216 , @xmath213 case has been shifted down by @xmath228 to match the correct charge neutral point.,title="fig : " ] ( left ) and @xmath113 ( right ) as a function of chemical potential @xmath95 .
conductivity is in unit of @xmath223 .
@xmath95 , @xmath224 , impurity strength @xmath225 ( blue lines ) , and @xmath226 ( purple dashed lines ) are in unit of @xmath178 .
magnetic field strength @xmath204 for upper and @xmath216 for lower figures .
the inset of upper right figure shows the blowup plot with smaller chemical potential range .
the @xmath113 in @xmath204 , @xmath213 case has been shifted down by @xmath227 and @xmath113 in @xmath216 , @xmath213 case has been shifted down by @xmath228 to match the correct charge neutral point.,title="fig : " ] + ( left ) and @xmath113 ( right ) as a function of chemical potential @xmath95 .
conductivity is in unit of @xmath223 .
@xmath95 , @xmath224 , impurity strength @xmath225 ( blue lines ) , and @xmath226 ( purple dashed lines ) are in unit of @xmath178 .
magnetic field strength @xmath204 for upper and @xmath216 for lower figures .
the inset of upper right figure shows the blowup plot with smaller chemical potential range .
the @xmath113 in @xmath204 , @xmath213 case has been shifted down by @xmath227 and @xmath113 in @xmath216 , @xmath213 case has been shifted down by @xmath228 to match the correct charge neutral point.,title="fig : " ] ( left ) and @xmath113 ( right ) as a function of chemical potential @xmath95 .
conductivity is in unit of @xmath223 .
@xmath95 , @xmath224 , impurity strength @xmath225 ( blue lines ) , and @xmath226 ( purple dashed lines ) are in unit of @xmath178 .
magnetic field strength @xmath204 for upper and @xmath216 for lower figures .
the inset of upper right figure shows the blowup plot with smaller chemical potential range .
the @xmath113 in @xmath204 , @xmath213 case has been shifted down by @xmath227 and @xmath113 in @xmath216 , @xmath213 case has been shifted down by @xmath228 to match the correct charge neutral point.,title="fig : " ] + the thermopower @xmath5 and @xmath16 obtained from eq.([tp ] ) are shown in fig.[fig4 ] for @xmath204 and @xmath216 with @xmath212 and @xmath213 .
the general features for @xmath5 are the peak positions corresponding to different landau levels are shifted away from the @xmath112 landau level and the width of the peaks increases due to the impurity broadening , similar to effect of temperature .
there is no universal value at the peaks of @xmath5 due to impurities as in the case of conventional two dimensional electron gas@xcite .
near the charge neutrality point @xmath16 shows a peak while @xmath5 shows singular behavior around this point .
the oscillatory behavior in @xmath5 and @xmath16 is more apparent for larger magnetic field , while the asymmetry is more prominent at lower fields .
note that the oscillations are entirely absent for negative chemical potentials for moderate impurity potentials .
these results are in qualitative agreement with available scanning tunneling microscopy(stm ) data which find oscillations in density of states only for positive gate voltages@xcite .
currently only limited experimental results for the thermopower of helical metals are available .
since our formulation is very similar to the case of graphene , we compare our thermopower results with data on graphene . in ref .
the observed peaks of magneto - thermopower @xmath5 shows the quantized @xmath229 ( @xmath6 is the nonzero landau level index ) trend but with a reduced factor in height and peak broadening , which are qualitatively consistent with our numerically results shown in fig.[fig4 ] for @xmath212 . in ref .
similar results have been reported , with the measured peak value of @xmath5 at @xmath230 and @xmath231 is @xmath232 , which is smaller than the clean limit where @xmath233 .
the thermopower shown in fig.[fig4 ] for @xmath204 , @xmath212 is similar to the case seen in ref . that @xmath234 for @xmath230 and smaller @xmath5 at @xmath235 , suggesting the influence of impurity interactions . in both experiments a large peak at @xmath11 landau level
is observed which is consistent with our numerical results .
this peak in @xmath16 can be explained by the presence of impurities in addition to the dirac spectrum .
the major difference between our results and experimental results in graphene@xcite is the behavior of @xmath5 near @xmath11 region .
our numerical results suggest a large electron / hole peak between @xmath11 and @xmath236 , showing remnant singular behavior around @xmath11 landau level in the clean limit as seen in fig.[fig0 ] . in experimental data
a hole / electron peak between @xmath11 and @xmath236 is observed instead .
this discrepancy is related to the higher @xmath221 value when @xmath95 is at @xmath11 landau level , as confirmed by the following observations . in all our numerical results
the diagonal conductivity @xmath221 is close to the universal value @xmath88 at @xmath96 . by adding a lorentzian profile to artificially increase @xmath221 around @xmath96 we can reproduce this results around @xmath11 seen in the experiments .
similar observations were reported in ref . where they performed numerical study on discrete lattice model for graphene system .
the physical reason for the increase of @xmath221 near @xmath11 landau level is still an open question and we discuss the possible mechanism for this enhancement of @xmath221 near the dirac node under magnetic field in sec .
in this article we compute linear response conductivity and thermopower of the helical metal , both with and without magnetic field , in the presence of static , short ranged , random impurities .
analytic results of thermopower in the clean limit , i.e. impurity strength @xmath156 , are given in both cases .
in the case of zero magnetic field the thermopower obtained is still susceptible to impurity interaction and does not show universal feature . in the case of finite magnetic field
the thermopower shows universal value , independent of temperature and magnetic field strength , when the chemical potential is fixed around landau levels at low temperature .
we also compute the thermopower and conductivity numerically for the case of finite impurity strength .
the effect of impurities broaden the peaks in diagonal thermopower and the height of the peaks show no universal behavior , which is similar to the case in the conventional 2d electron gas@xcite .
we also vary the impurity potential and study its affect on transport . for moderate impurity strength @xmath26
we find the particle hole symmetry is broken while for large or small @xmath26 the symmetry is restored in our perturbation formulation . due to the lack of available data on thermopower measurement of the helical metal
, we compare our thermopower results with experimental data on graphene .
our computation is on the helical metal system which differs from graphene@xcite mainly in the zeeman effect .
since the zeeman contribution is small for normal magnetic field strength ( due to @xmath237 ) in both graphene and helical metal samples such as hgte , our results are applicable to the case of graphene with short range , randomly positioned impurities modulo a numerical factor in conductivity due to valley and spin degeneracy in graphene .
we find qualitative agreement between our numerical results of thermopower and experimental results in graphene , except for the @xmath5 near the zeroth landau level .
the discrepancy is due to the enhanced diagonal conductivity nearby zeroth landau level .
the enhancement in @xmath221 seems to decrease with increasing magnetic field@xcite .
possible mechanism includes the metal to insulator transition on the zeroth landau level@xcite or electron hole puddles @xcite .
transport measurements of graphene near the minimum conductivity in ref . partly supports the latter mechanism but further study is needed to clarify this issue . whether similar behavior is reproduced in helical metals
is an open question .
the magneto - electric coupling , unique to topological insulators , leads to novel features in transport .
the dependence of the thermopower on the zeeman coupling provides a diagnostic of the state . to amplify the expected behavior we have analyzed thermopower and nernst for materials with an artificially enhanced gyromagnetic ratio .
the relative importance of the zeeman term to the landau level splitting can be enhanced by an in - plane electric field , providing a possible route to detecting the signatures proposed .
one caveat the electric field leads to additional effects such as non - linear transport , a detailed analysis of which is the subject of future investigation .
the authors wish to acknowledge jing shi and peng wei for useful discussions .
vivek aji and sung - po chao s research is supported by university of california at riverside under the initial complement .
the free electron green s functions in the energy eigenstate basis eq.([sp1 ] ) are @xmath238 in this basis , the impurity interaction in eq.([hi ] ) is @xmath239 the impurity interaction is diagonal in the spin space .
the position averaged green s function , obtained within scba@xcite , is @xmath240 g_{s , s}(\vec{k},\omega_n;\vec{r}_1,\vec{r}_2,\vec{r}_3,\ldots,\vec{r}_{n_{imp}})\end{aligned}\ ] ] thus the bare propagator spin - diagonal component ) in eq.([tmatrix ] ) is replaced by full propagator which leads to full self consistent born approximation@xcite @xmath241^{-1}}\end{aligned}\ ] ] where @xmath242 is the electron self energy .
this is independent of momenta due to spherical symmetry in the impurity potential in eq.([hi ] ) .
by using analytic continuation @xmath243 we have the self energy of electron given by @xmath244^{-1}}\end{aligned}\ ] ]
the explicit form of @xmath85 after the momentum integral is given as @xmath245 with @xmath246 and @xmath247 .
the internal propagators for self energy computation , @xmath149 , @xmath150 , and @xmath151 are given by @xmath248\\\label{selfenergy1a } & & + \frac{f_{\gamma , n}^2}{1+f_{\gamma , n}^2}g_c g(0;\omega+0^+)\\\label{selfenergy2a } & & z(0,\omega)=\sum_{m,\bar{\gamma}}[\frac{f_{\bar{\gamma},m}^2}{1+f_{\bar{\gamma},m}^2}g_c g(m,\bar{\gamma};\omega+0^+ ) -\frac{g_cf_{\bar{\gamma},m}f_{-\bar{\gamma},m}g(m,\bar{\gamma},-\bar{\gamma};\omega+0^+)}{\sqrt{(1+f_{\bar{\gamma},m}^2)(1+f_{-\bar{\gamma},m}^2)}}]+g_c g(0;\omega+0^+ ) \\\nonumber & & z(\gamma,-\gamma , n,\omega ) = \sum_{\bar{\gamma},m}[\frac{(1-f_{\gamma , n}f_{-\gamma , n}f_{\bar{\gamma},m}^2)g_c g(m,\bar{\gamma};\omega+0^+)}{(1+f_{\bar{\gamma},m}^2)\sqrt{(1+f_{\gamma , n}^2)(1+f_{-\gamma , n}^2 ) } } + \frac{(1+f_{\gamma , n}f_{-\gamma , n}f_{\bar{\gamma},m}f_{-\bar{\gamma},m})g_c g(m,\bar{\gamma},-\bar{\gamma};\omega+0^+)}{\sqrt{(1+f_{\gamma , n}^2)(1+f_{-\gamma , n}^2)(1+f_{\bar{\gamma},m}^2)(1+f_{-\bar{\gamma},m}^2)}}]\\\label{selfenergy3a } & & -\frac{f_{\gamma , n}f_{-\gamma , n}}{\sqrt{(1+f_{\gamma , n}^2)(1+f_{-\gamma , n}^2)}}g_c g(0;\omega+0^+)\end{aligned}\ ] ]
for the case of finite magnetic field in the energy eigenstates basis the particle current operators , as in eq.([particlecurrent ] ) , are @xmath251 the particle current current correlation @xmath154 and @xmath155 from eq.([kubo ] ) are given by @xmath252\big[\sum_{\gamma}\frac{1}{1+f_{\gamma,1}^2}[\im g(0;\epsilon+0^+)\im
g(1,\gamma;\epsilon+\omega+0^+)\\\nonumber&&+\im g(1,\gamma;\epsilon+0^+)\im g(0;\epsilon+\omega+0^+)]+\sum_{n,\gamma,\bar{\gamma}}\frac{f_{\bar{\gamma},n}^2}{(1+f_{\gamma ,
n+1}^2)(1+f_{\bar{\gamma},n}^2)}\im g(n+1,\gamma;\epsilon+0^+)\\\nonumber&&\times\im g(n,\bar{\gamma};\epsilon+\omega+0^+ ) + \sum_{n,\gamma,\bar{\gamma}}\frac{f_{\gamma , n}^2}{(1+f_{\bar{\gamma},n+1}^2)(1+f_{\gamma , n}^2)}\im g(n,\gamma;\epsilon+0^+)\im g(n+1,\bar{\gamma};\epsilon+\omega+0^+)\\\nonumber & & + \sum_{\gamma}\frac{\im g(0;\epsilon+0^+)\im g(1,\gamma,-\gamma;\epsilon+\omega+0^+)+\im g(1,\gamma,-\gamma;\epsilon+0^+)\im g(0;\epsilon+\omega+0^+)}{\sqrt{(1+f_{\gamma,1}^2)(1+f_{-\gamma,1}^2)}}\\\nonumber & & + \sum_{n,\gamma,\bar{\gamma}}\frac{f_{\bar{\gamma},n}^2}{(1+f_{\bar{\gamma},n}^2)\sqrt{(1+f_{\gamma , n+1}^2)(1+f_{-\gamma , n+1}^2)}}\im g(n+1,\gamma,-\gamma;\epsilon+0^+)\im g(n,\bar{\gamma};\epsilon+\omega+0^+)\\\nonumber & & -\sum_{n,\gamma,\bar{\gamma}}\frac{f_{\bar{\gamma},n}f_{-\bar{\gamma},n}}{(1+f_{\gamma , n+1}^2)\sqrt{(1+f_{\bar{\gamma},n}^2)(1+f_{-\bar{\gamma},n}^2)}}\im g(n+1,\gamma;\epsilon+0^+)\im g(n,\bar{\gamma},-\bar{\gamma};\epsilon+\omega+0^+)\\\nonumber & & + \sum_{n,\gamma,\bar{\gamma}}\frac{f_{\gamma , n}^2}{(1+f_{\gamma , n}^2)\sqrt{(1+f_{\bar{\gamma},n+1}^2)(1+f_{-\bar{\gamma},n+1}^2)}}\im g(n,\gamma;\epsilon+0^+)\im g(n+1,\bar{\gamma},-\bar{\gamma};\epsilon+\omega+0^+)\\\label{l11xx } & & -\sum_{n,\gamma,\bar{\gamma}}\frac{f_{\gamma , n}f_{-\gamma , n}}{(1+f_{\bar{\gamma},n+1}^2)\sqrt{(1+f_{\gamma , n}^2)(1+f_{-\gamma , n}^2)}}\im g(n,\gamma,-\gamma;\epsilon+0^+)\im g(n+1,\bar{\gamma};\epsilon+\omega+0^+)\big]\end{aligned}\ ] ] @xmath253+\sum_{n,\alpha=\pm1,\gamma,\bar{\gamma}}\frac{\alpha f_{\bar{\gamma},n}^2}{(1+f_{\gamma , n+1}^2)(1+f_{\bar{\gamma},n}^2)}\re g(n+1,\gamma;\epsilon+\alpha\omega+0^+)\\\nonumber & & \times\im g(n,\bar{\gamma};\epsilon+0^+ ) -\sum_{n,\alpha=\pm1,\gamma,\bar{\gamma}}\frac{\alpha f_{\gamma , n}^2}{(1+f_{\bar{\gamma},n+1}^2)(1+f_{\gamma , n}^2)}\re g(n,\gamma;\epsilon+\alpha\omega+0^+)\im g(n+1,\bar{\gamma};\epsilon+0^+)\\\nonumber & & + \sum_{\alpha=\pm 1,\gamma}\frac{\alpha[\re g(1,\gamma,-\gamma;\epsilon+\alpha\omega+0^+)\im g(0;\epsilon+0^+)-\im g(1,\gamma,-\gamma;\epsilon+0^+)\re g(0;\epsilon+\alpha\omega+0^+)]}{\sqrt{(1+f_{\gamma,1}^2)(1+f_{-\gamma,1}^2)}}\\\nonumber & & + \sum_{n,\alpha=\pm 1,\gamma,\bar{\gamma}}\frac{\alpha f_{\bar{\gamma},n}^2}{(1+f_{\bar{\gamma},n}^2)\sqrt{(1+f_{\gamma , n+1}^2)(1+f_{-\gamma , n+1}^2)}}\re g(n+1,\gamma,-\gamma;\epsilon+\alpha\omega+0^+)\im g(n,\bar{\gamma};\epsilon+0^+)\\\nonumber & & -\sum_{n,\alpha=\pm 1,\gamma,\bar{\gamma}}\frac{\alpha f_{\bar{\gamma},n}f_{-\bar{\gamma},n}}{(1+f_{\gamma , n+1}^2)\sqrt{(1+f_{\bar{\gamma},n}^2)(1+f_{-\bar{\gamma},n}^2)}}\re g(n+1,\gamma;\epsilon+\alpha\omega+0^+)\im g(n,\bar{\gamma},-\bar{\gamma};\epsilon+0^+)\\\nonumber & & -\sum_{n,\alpha=\pm 1,\gamma,\bar{\gamma}}\frac{\alpha f_{\gamma , n}^2}{(1+f_{\gamma , n}^2)\sqrt{(1+f_{\bar{\gamma},n+1}^2)(1+f_{-\bar{\gamma},n+1}^2)}}\re
g(n,\gamma;\epsilon+\alpha\omega+0^+)\im g(n+1,\bar{\gamma},-\bar{\gamma};\epsilon+0^+)\\\label{l11xy } & & + \sum_{n,\alpha=\pm 1,\gamma,\bar{\gamma}}\frac{\alpha f_{\gamma , n}f_{-\gamma , n}}{(1+f_{\bar{\gamma},n+1}^2)\sqrt{(1+f_{\gamma , n}^2)(1+f_{-\gamma , n}^2)}}\re g(n,\gamma,-\gamma;\epsilon+\alpha\omega+0^+)\im g(n+1,\bar{\gamma};\epsilon+0^+)\big]\end{aligned}\ ] ] the sum over @xmath68 ( in @xmath152 direction ) results in the normalization factor @xmath254 being replaced by @xmath255 .
j. c. y. teo , l. fu , and c. l. kane , phys . rev . *
b 78 * , 045426 ( 2008 ) ; l. fu , c. l. kane , and e. j. mele , phys .
. lett . * 98 * , 106803 ( 2007 ) .
r. roy , phys . rev . *
b 79 * , 195322 ( 2009 ) ; j. e. moore and l. balents , phys . rev . *
b 75 * , 121306(r ) ( 2007 ) .
zhang , c .- x .
qi , x .- y .
deng , x. dai , s .- c .
zhang , and z. fang , phys . rev . *
b 80 * , 085307 ( 2009 ) .
d. hsieh , d. qian , l. wray , y. xia , y. hor , r. j. cava , and . m. z. hasan , nature ( london ) * 452 * , 970 ( 2008 ) .
y. xia , l. wray , d. qian , d. hsieh , a. pal , h. lin , a. bansil , d. grauer , y.s .
hor , r.j .
cava , m.z .
hasan , arxiv:0812.2078 .
m. k@xmath256nig , s. wiedmann , c br@xmath257ne , a. roth , h. buhmann , l. w. molenkamp , x .-
qi , and s .- c .
zhang , science * 381 * , 766 ( 2007 ) ; y. xia , d. qian , d. hsieh , l. wray , a. pal , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , and m. z. hasan , nature phys .
* 5 * , 398 ( 2009 ) .
h. steinberg , d.r .
gardner , y.s .
lee , and p. jarillo - herrero , nano lett .
* 10 * , 5032 ( 2010 ) .
analytis , j - h chu , y. chen , f. corredor , r.d .
mcdonald , z.x .
shen , and i.r .
fisher , phys .
b * 81 * , 205407 ( 2010 ) j.g .
checkelsky , y.s .
hor , m .- h .
liu , d .- x .
cava , and n.p .
ong , phys .
lett . * 103 * , 246601 ( 2009 ) .
p. wei , z. wang , x. liu , v. aji , and j. shi , unpublished j. chen , h.j .
qin , f. yang , j. liu , t. guan , f.m .
zhang , j.r .
shi , x.c .
xie , c.l .
yang , k.h .
li , and l. lu , phys .
lett . * 105 * , 176602 ( 2010 ) .
qu , y. s. hor , j. xiong , r. j. cava , and n. p. ong , science * 329 * , 821 ( 2010 ) . s. mondal , d. sen , k. sengupta , and r. shankar , phys . rev . *
b 82 * , 045120 ( 2010 ) .
m. jonson and s. m. girvin , phys . rev . *
b 29 * , 1939 ( 1984 ) ; s. m. girvin and m. jonson , j. phys .
c * 15 * , l1147 ( 1982 )
. y. m. zuev , w. chang , and p. kim , phys .
lett . * 102 * , 096807 ( 2009 ) .
j. g. checkelsky and n. p. ong , phys .
b * 80 * , 081413(r ) ( 2009 ) . n. m. r. peres , f. guinea , and a. h. castro neto , phys . rev . *
b 73 * , 125411 ( 2006 ) .
y. zheng and t. ando , phys . rev . *
b 65 * , 245420 ( 2002 ) . v. ugarte , v. aji and c.m .
varma , arxiv:1007.3533 , unpublished .
e. h. hwang , e. rossi , s. das sarma , phys .
* b 80 * , 235415 ( 2009 ) .
s. das sarma , s. adam , e.h .
hwang , and e. rossi , rev .
phys . * 83 * , 407 ( 2011 ) .
x. z. yan , y. romiah , and c. s. ting , phys
b 80 * , 165423 ( 2009 ) .
p. wei , w. bao , y. pu , c. n. lau , and j. shi , phys .
lett . * 102 * , 166808 ( 2009 ) .
h. bruus and k. flesberg , _ many - body quantum theory in condensed matter physics _
( oxford university press , oxford , 2004 ) .
l. zhu , r. ma , l. sheng , m. liu , and d - n sheng , phys .
* 104 * , 076804 ( 2010 ) .
l. hao and t. k. lee , phys .
* b 82 * , 245415(2010 ) .
l. zhang , y. zhang , m. khodas , t. valla , and i. a. zaliznyak , phys .
lett . * 105 * , 046804 ( 2010 ) .
s. adam , e. h. hwang , v. m. galitski , and s. d. sarma , proc .
104 * , 18392 ( 2007 ) .
s. cho and m. s. fuhrer , phys . rev .
* b 77 * , 081402(r ) ( 2008 ) .
p. cheng , c. song , t. zhang , y. zhang , y .
wang , j .- f .
jia , j. wang , y. wang , b .-
f zhu , x. chen , x. ma , k. he , l. wang , x. dai , z. fang , x. xie , x .-
qi , c .- x .
liu , s.c .
zhang , and q .- k .
xue , phys .
lett . * 105 * , 076801 ( 2010 ) .
t. hanaguri , k. igarashi , m. kawamura , h. takagi , and t. sasagawa , phys . rev .
b * 82 * , 081305 ( 2010 ) .
|
we study electrical and thermoelectric transport properties of the surface state of the topological insulator and graphene in the presence of randomly distributed impurities . for finite impurity strength ,
the dependence of the transport coefficients as a function of gate voltage , magnetic field and impurity potential , are obtained numerically . in the limit of zero impurities ( clean limit ) ,
analytic results for the peak values of the magneto - oscillations in thermopower are derived .
analogous with the conventional two dimensional electron gas , the peak values are universal in the clean limit .
unlike graphene , in topological insulators the coupling of the electron spin to its momentum leads to a dependence of the transport coefficients on the gyromagnetic ratio ( @xmath0 ) .
we compare our results with data on graphene and identify unique signatures expected in topological insulators due to the magnetoelectric coupling .
|
scale - dependence of the spectrum of the curvature perturbation has been used to discriminate inflationary scenarios @xcite . in addition to the scalar mode , recent discovery of the b - mode polarization@xcite
has ignited studies of the tensor modes .
although the detection of the inflationary tensor mode by the bicep2 was not successful , it stimulated study of the mechanism of generating peculiar scale - dependence @xcite . besides those recent trends in inflationary cosmology
, models of particle physics ( e.g , supersymmetric models and string theory ) are expecting a large amount of scalar fields that may be dynamical in the very early stage of inflation . in the light of multi - field inflation
, those extra degrees of freedom may alter the scale - dependence of both the tensor and the scalar perturbations .
consequently , they may bring in a kind of ambiguity to the inflationary parameters . even if one is able to assume that the inflaton is the only dynamical field
, it is still hard to believe that the inflaton velocity `` before '' inflation coincides with the slow - roll velocity `` during '' inflation .
initially the inflaton velocity could deviate from the slow - roll ( attractor ) velocity and it will lead to the shift of the scale - dependence . these effects are very common and they can be responsible for the scale dependence of the spectrum .
we start with the obvious situation .
the model shown in fig.[fig1 ] is a textbook example , which helps understand why it is very easy to shift the scale - dependence of the spectrum .
we focus on the simple and common dynamics .
since the scale dependence is currently used to constrain inflation models , it is very important to consider the shift of the scale dependence which could be caused by the dynamics that does not appear in the original inflation model .
we call such dynamics the `` scalon '' , since it contributes only to the generation of the scale dependence .
our study can be discriminated from other attempts in which non - trivial interactions play the key role .
we are considering ubiquitous remnants .
we also consider the case in which the field is not negligible at the end of inflation .
the model is presented in fig.[fig2 ] .
finally we consider single - field inflation in which small deviation from the slow - roll velocity can cause significant shift of the scale dependence .
we show that in those cases the observation of the spectral index of the tensor mode will play important role in finding the scalon contribution .
more precise measurement of the higher runnings of the scalar spectrum will be helpful for the discrimination .
we briefly explain the overview before the details . we are focusing on the generality of the situation and considering simple dynamics of a scalar field .
we thus avoid introduction of specific interactions and non - trivial kinetic terms .
this point could be distinguishable from the other recent attempts @xcite .
single - field inflation `` usually '' expects @xmath0 , where @xmath1 is the tensor - to - scalar ratio and @xmath2 is the spectral index of the tensor mode .
the relation could be violated if there are multiple scalar fields during inflation .
intuitively we are expecting simple cases for multi - field inflation explained below .
* consider trajectory of fig.[fig1 ] . in this scenario
the scalon is short - lived .
the curvature perturbation evolves during inflation @xcite . *
consider trajectory of fig.[fig2 ] . in this scenario
the scalon remains dynamical until the end of inflation .
the curvature perturbation is generated at the end . in both cases the curvature perturbation at the pivot scale
is unchanged , while the scale dependence is shifted by the scalon . before explaining more details
, we point out that similar ( but opposite ) situation already appeared in the curvaton model . in the simplest curvaton model ,
in which the slow - roll parameters of the curvaton field is negligible ( i.e , the curvaton dynamics can not generate the scale dependence ) , the scale dependence is generated by the dynamics of the inflaton field whose perturbation is negligible in the curvaton model . in that case
the inflaton dynamics contributes only to the scale dependence .
note that the field ( or the dynamics ) that contributes only to the scale dependence may commonly appear in any inflationary model .
similar effect may appear without adding a scalar field , as we will describe the situation below for single - field inflation .
* inflaton velocity may have deviation from the slow - roll velocity . in that case
the curvature perturbation is known to evolve after horizon exit , since the additional degree of freedom ( i.e , the deviation ) forms the decaying mode . in that case
@xmath0 is not satisfied in single - field inflation .
deviation from the slow - roll velocity may have serious impact on the scale dependence .
we will consider this possibility in sec.[dev - slow ] .
the sign of the spectral index could be reversed .
the relation @xmath0 is useless in the curvaton scenario , since the form of the curvature perturbation is completely different .
we are not discussing those `` alternative '' models , including modulation ( e.g. modulated reheating ) .
in general , `` multi - field inflation '' includes models in which multiple perturbations contribute the curvature perturbation .
such models are extensively studied since they may have a significant evidence of multi - field inflation in the relation among non - linear parameters @xcite . to avoid confusions in this paper ,
those models are specifically called `` mixed perturbation scenario '' . in this paper
we are avoiding mixed perturbation scenarios since the non - gaussianity parameter is small .
one might notice that the current cmb data is sometimes used to distinguish multi and single - field inflation .
the major reason could be that there is the stringent upper bound on the isocurvature perturbation .
although an additional field `` can '' generate significant isocurvature perturbations , it is not mandatory . in other cases , as we have stated above , `` multi - field inflation '' sometimes means `` mixed perturbation scenario '' . to avoid those confusions
we emphasize that `` inflation with the scalon dynamics '' will not be distinguishable at this moment ; we mean that the experimental plans at hand will hardly find the distinction .
however , since what we call `` the scalon '' is the very common dynamics that could appear in the early universe , we claim that it has to be identified and removed by more future experiments .
we start with the simplest case .
we introduce the scalon field ( additional free scalar field @xmath3 ) , which is dynamical at the beginning of inflation .
one can imagine the case in which @xmath4 at horizon exit but @xmath5 at the end . here `` @xmath6 '' denotes the value at the horizon exit .
see also fig.[fig1 ] .
using the standard definition of the curvature perturbation , the scalon determine the initial adiabatic curvature perturbation .
we also consider the case with @xmath7 , which can also shift the scale dependence . .
mixing in the spectrum is negligible . ]
we are considering multi - field inflation .
however , since we are considering a `` temporal '' field in this section , the derivative of the e - foldings with respect to the extra field ( @xmath8 ) is negligible . here
the subscript with comma denotes derivative with respect to the field .
cases with significant @xmath8 has been considered in ref.@xcite .
remember that we are focusing on the simple dynamics that does not assume specific interactions . we will assume canonical kinetic term for the fields .
we are avoiding mixing of the fields in the potential .
the scalon field is separated from the inflaton @xmath9 and ceases to be dynamical before the end of inflation . in that way
our analysis does not depend on specific inflationary model .
one can add @xmath3 to any ( single - field , canonical kinetic term ) inflationary model on one s choice .
when a couple of fields @xmath9 and @xmath3 are dynamical during inflation , the spectral index of the curvature perturbation is calculated as @xcite @xmath10 where the subscripts @xmath11 and @xmath12 denote the adiabatic and entropy directions at the horizon exit .
we choose @xmath13 . here
the definitions of the slow - roll parameters are @xmath14_*\nonumber\\ \eta_{ab}&\equiv & \left[\frac{v_{,ab}}{3h^2}\right]_*,\end{aligned}\ ] ] where @xmath15 is the hubble parameter , @xmath16 is the reduced planck mass and @xmath17 denote @xmath18 and @xmath12 .
we introduce separation given by @xmath19_*=\epsilon_\phi+\epsilon_\chi$ ] .
@xmath20 is defined using the transfer matrix and is given by @xmath21 where @xmath22 is an element of the transfer matrix , which describes the evolution of the adiabatic and the isocurvature perturbations .
we define the instantaneous adiabatic and entropy perturbations as @xmath23 where @xmath24 .
also we define the curvature and entropy perturbations @xmath25 the curvature perturbation after horizon exit is expressed as @xmath26.\ ] ] here @xmath27 is the observable .
since here we are assuming `` temporal '' @xmath28 at a moment , adiabatic velocity soon reaches @xmath29 after horizon exit .
the final curvature perturbation recovers the original single - field inflation model @xcite : @xmath30_*$ ] .
let us see more details about the scalon contributions . if @xmath31 , the trajectory has a turn during inflation . initially the adiabatic direction is @xmath32 but it will be @xmath29 before the end of inflation . then the curvature perturbation becomes @xmath33_*\gg r_*\simeq \left[\frac{h}{\dot{\chi}}\delta \chi\right]_*$ ] , which is possible since @xmath34
. then one will find @xmath35 and the spectral index given by @xmath36 the opposite limit ( @xmath37 ) is the standard single - field inflation .
one will find @xmath38 and an almost straight trajectory .
the spectral index will be @xmath39 seeing the above discrepancy between ( [ eq - multiindex - bend ] ) and ( [ eq - multiindex - straight ] ) , one may find that the situation is intuitively similar to the curvaton .
more radical situation can be found in the inflating curvaton @xcite . in the above formalism one has to take into account the scale - dependence of the transfer function @xmath22 .
the latter case ( @xmath37 already at the beginning ) seems to be the same as the conventional single - field scenario .
however , as far as @xmath28 , there is the contribution caused by @xmath40 .
although in the spectral index the contribution of the scalon could be negligible , the shift in the higher runnings can be significant since @xmath41 is possible @xcite
. we will be back to this topic in sec.[running ] .
below , we will examine the relation @xmath0 when @xmath40 .
since one extra degree of freedom is added to the original scenario , we need another observable ( i.e , an independent equation ) that can fix the ambiguity .
the spectral index of the tensor mode can play the role .
the spectrum of the tensor perturbation is @xmath42 , which gives the tensor to scalar ratio @xmath43 when @xmath44 , we find conventional result @xmath45 and @xmath46 .
however , when @xmath4 , things are not so trivial . to check the consistency of the calculation
it is very useful to consider the equations related to @xmath47 .
therefore , in this paragraph we introduce @xmath47 in addition to @xmath2 .
remember that the running of @xmath1 ( i.e , @xmath47 ) can `` directly '' be evaluated using indices of the scalar and the tensor modes as @xmath48 where @xmath49 .
we use this relation to check the consistency of the calculation below .
let us see more details of the relations between parameters , and see how one can remove the ambiguity .
* we first consider @xmath37 .
using @xmath50 and the definitions of the slow - roll parameters ( [ slow - def ] ) , one will find @xmath51 which leads to @xmath52 .
of course the result is consistent with eq.([eq - index - r ] ) .
then , from the spectral index ( [ eq - multiindex - straight ] ) and @xmath49 , one can evaluate the slow - roll parameters from the observables as @xmath53 here @xmath54 will be the sign of @xmath55 . *
consider the case with @xmath4 , in which @xmath3 soon ceases to be dynamical during inflation . in this case
we are expecting @xmath56 ( i.e , the scalon is dominating @xmath57 ) . from the direct calculation ( [ eq - index - r ] )
, one will find @xmath58 .
although rather complicated , one can evaluate the same result by considering @xmath22 @xcite . to distinguish the scale dependence
, it would be useful to define @xmath59 and evaluate @xmath60 again , discrepancy @xmath61 is the sign of an extra dynamical field .
note that in both limits @xmath54 is the sign of the additional dynamical field .
although @xmath3 introduces ambiguity to the cosmological parameters , the contribution can be removed if @xmath2 is measurable .
the above scenario is a well - known example of multi - field inflation .
although in the past studies we could not find explicit argument that relates the spectral index of the tensor mode to the ambiguity in the inflationary parameters , the above results should have been known among scientists in this field .
below , we will consider less simple examples for which the shift of the scale dependence has not been discussed in detail before .
the usual multi - field model @xcite considers `` evolution '' of the curvature perturbation during inflation . in the model
represented in fig.[fig1 ] , this effect compensates the curvature perturbation .
the situation we are going to consider in this section is rather different from such scenario .
we introduce the scalon that can survive until the end of inflation .
then , the initial curvature perturbation is compensated by the curvature perturbation generated at the end .
typically the scenario of generating the curvature perturbation at the end of inflation is considered for hybrid - type potential , since the original lyth s model @xcite considers modulation of the waterfall ( @xmath62 ) . in that way
the lyth s scenario requires interaction with the waterfall field
. however , our scenario * does not require interaction * and works with many variety of inflationary models in which the potential is given by @xmath63 ( @xmath64 ) .
this includes higgs inflation .
although the result is quite conceivable in the light of the @xmath65 formalism , it is not obvious how the initial curvature perturbation can be compensated at the end . therefore , to avoid confusions we first review the discussion in ref.@xcite .
the calculation applies to other non - hybrid models .
first consider the inflaton @xmath9 and the waterfall field @xmath66 with the hybrid - type potential given by @xmath67 suppose that inflation starts with @xmath68 and the waterfall begins at @xmath69 .
the critical point @xmath70 is @xmath71 .
the number of e - foldings is given by @xmath72 remember that * the original lyth s model @xcite assumes additional field @xmath3 that couples to the waterfall field*. namely , if one introduces @xmath3 that has the same interaction as @xmath9 , the end of inflation is defined as @xmath73 then , if @xmath3 is lighter than the inflaton , the entropy perturbation @xmath74 creates the perturbation @xmath75 at the end .
consequently , the perturbation of the number of e - foldings created at the end of inflation is given by @xmath76_e \ne 0.\ ] ] this is the usual scenario of `` generating the curvature perturbation at the end '' . )
causes @xmath65 at the end . ]
_ in contrast to the original scenario , we are introducing a ubiquitous field , which is decoupled from the waterfall field .
_ therefore , we are * not * expecting @xmath77 .
just for the simplest example , we consider @xmath78 note that * unlike * the usual multi - field extension of the hybrid - type potential we are omitting interaction @xmath79 .
degeneracy of the mass term ( @xmath80 ) makes the trajectory straight and avoids the effect considered in sec.[modest - multi ] .
the adiabatic field is defined as @xmath81 , which gives @xmath82 we find the end * for the adiabatic field * is @xmath83 which is perturbed when @xmath84 .
( note that @xmath85 does not evolve during inflation when @xmath80 . ) therefore , although @xmath70 is not perturbed , the entropy perturbation ( @xmath86 ) causes @xmath87 at the end .
the curvature perturbation generated at the end of inflation is thus given by @xmath88 considering perturbation generated at the horizon exit : @xmath89_*,\ ] ] we find the ratio between `` initial '' and `` at the end '' : @xmath90 here @xmath91 is used for the calculation .
more intuitive argument is possible .
considering similar triangles in fig.[fig2 ] , we find the number of e - foldings @xmath92 therefore , intuitively the @xmath65 formalism suggests @xmath93 let us see what happens if @xmath94 .
then the final curvature perturbation is dominated by @xmath95 , which reproduces the `` conventional '' perturbation @xmath96 . here , we already know that the scalon can shift the scale - dependence of the spectrum . separating the slow - roll parameter @xmath97 , we find the spectral index given by @xcite @xmath98 in the opposite case , when @xmath99 , generation of the curvature perturbation at the end is negligible and we find @xmath100 . in both cases @xmath57 is shifted by @xmath101 .
the result is quite similar to the conventional `` multi - field '' model in fig.[fig1 ] , although the mechanism of generating the curvature perturbation is utterly different .
the property of the scalon is also different .
for the model considered in this section , @xmath102 can be significant until the end of inflation .
usually the typical hybrid - type model is excluded because its spectrum is blue ( @xmath103 ) . however , as we have seen above , one can introduce a ubiquitous field @xmath3 to change the spectral index into red ( @xmath104 ) @xcite .
again , the spectral index of the tensor mode is needed to distinguish the scalon contribution .
the situation presented in fig.[fig2 ] is quite general .
the mechanism can be applied to any inflationary model in which @xmath3 causes negligible change in @xmath70 .
imagine a potential like @xmath63 with @xmath64 , which is very common .
introduction of the `` extra degree of freedom that is responsible for the scale dependence of the curvature perturbation '' does not always require additional scalar field .
it may appear as a parameter measuring deviation from the inflationary attractor .
first , remember that at the very beginning of inflation the inflaton velocity ( @xmath105 ) may not coincide with the slow - roll velocity ( @xmath106 ) .
moreover , if the inflaton is moving fast in the opposite direction before the onset of inflation , inflaton may even stop during inflation @xcite .
although the analysis could be slightly model - dependent , the essential of the argument is quite simple .
if the inflaton velocity deviates from the slow - roll velocity defined there ( i.e , @xmath107 ) , the curvature perturbation converges to the value evaluated using the slow - roll velocity .
namely , one will find the evolution of the curvature perturbation @xmath108 before the end of inflation .
intuitively the above result is quite conceivable in the light of the @xmath109 formalism . since the time - dependence of the quantity @xmath9
is not determined by @xmath106 , the scale - dependence ( i.e , evolution of @xmath9 and @xmath15 that determines @xmath110-dependence of the curvature perturbation ) will be different from the standard scenario . unlike the multi - field model , the correction in the spectral index will appear not only from @xmath111 ( i.e , @xmath57 ) but also from @xmath112 .
namely , if one uses @xmath50 for the calculation , there will be corrections like @xmath113\dot{\epsilon}_\phi^{(0)},\ ] ] where we defined @xmath114 , and @xmath115 .
note that `` @xmath116 '' is used for quantities without deviation , except for the hubble parameter that always includes @xmath117 for the kinetic energy .
see eq.([slow - def ] ) for the definitions of the slow - roll parameters .
our definition is @xmath118 . in the same way
, we also find @xmath119 where @xmath120 comes from the time derivative of the kinetic energy in the hubble parameter . since the kinetic energy of the inflaton is not identical to the attractor solution
, @xmath121 could not be negligible .
correction to the spectral index is @xmath122 if the hubble parameter can be written as @xmath123 , we find @xmath124 where we used @xmath125 , @xmath126 , @xmath127 and @xmath128 . here
we expect @xmath129 for the deviation .
if the inflaton `` stops '' during inflation one will find @xmath130 , which gives @xmath131 .
the result is consistent with the naive intuition : scale dependence disappears when inflaton stops .
since @xmath132 changes quickly during inflation , there will be a signature of @xmath133 in the running of the spectral index .
one will find @xmath134\nonumber\\ & & -6r_{d*}^2\left[-6\epsilon_\phi^{(0)}+4\epsilon_\phi^{(0)}\epsilon_h-2\eta_\phi^{(0)}\epsilon_\phi^{(0 ) } \right].\end{aligned}\ ] ] contribution related to @xmath135 mainly appears from @xmath136 .
since @xmath137 is almost excluded by the observations , we can see that @xmath138 is not a realistic scenario .
note that the blue spectrum of hybrid - type inflation ( @xmath139 for @xmath140 ) can be turned into red when @xmath141 .
therefore the ambiguity related to the deviation is quite serious . even though @xmath1 in hybrid inflation could be very small and the tensor mode could not be seen in the observations , more precise measurement of @xmath142
can be used to exclude such possibility .
@xmath0 is violated because @xmath57 is not identical to @xmath143 . here
@xmath144 and @xmath49 .
note that according to the discussion in ref.@xcite one will find that @xmath145 is not correct here .
again , the additional parameter @xmath135 can be fixed if @xmath2 could be observed . since eq.([ehplusdelta ] ) can be rewritten as @xmath146 , we find @xmath147 . the extra dynamical field can change the scale dependence of the spectrum
. the situation could be serious if there are many scalar fields , whose effective masses are @xmath148 or less during inflation . in this section
we estimate the possible contributions from such remnants . here
we assume @xmath149 .
in that case the conventional slow - roll approximations are still valid for the estimation .
if one considers the quadratic potential @xmath150 for the fast - rolling field @xmath3 , one will find @xcite @xmath151 where @xmath152.\end{aligned}\ ] ] this gives @xmath153 where contribution from the kinetic energy has not been neglected .
just for simplicity we consider @xmath154 .
note that @xmath3 can change the spectral index and predict @xmath155 at the same time .
we find @xmath156 which could be negative if @xmath157 or simply @xmath158 @xcite .
if the universe started with a chaotic state , there could be many dynamical fields before the onset of inflation .
inflationary expansion starts when the vacuum energy of the inflaton starts to dominate . at that moment
the other dynamical fields may have energy density comparable to the inflaton vacuum energy .
if one wants to disregard their effect , these fields must be inflated away before the onset of the last 60 e - foldings .
this is not a trivial assumption since their effect could not be negligible even if their fraction is small . is larger than the planck scale , the fraction of @xmath159 is @xmath160 .
this can not be a `` small fraction '' when the slow - roll parameters of @xmath3 are not negligible .
we are not assuming such significant modification of the original inflationary model . ] in this paper we first considered a non - interacting field @xmath3 and calculated the possible shift of the scale dependence .
the first model is presented in fig.[fig1 ] , which we considered as a textbook example . in this case the additional inflaton field ( @xmath3 ) may disappear during inflation but it can significantly change the scale dependence of the spectrum . the second model is presented in fig.[fig2 ] , which is less known but is showing a very common situation . in both cases our result illuminates the importance of the observation of the spectral index of the tensor mode .
we also considered a single - field inflation model in which deviation from the slow - roll is not negligible .
we found a similar effect and showed that observation of the tensor mode is important .
note that in the last scenario the shift of the scale - dependence is considered in a single - field inflation model . in this paper
we have chosen specific scalon dynamics that are very common in the early universe , and showed that the ambiguity caused by the remnant is crucial for the scientific argument .
although our results are encouraging and seemingly suggesting the next step , unfortunately in reality the tensor mode of the inflationary perturbations has not been observed yet and the present experimental plans are not designed to determine the spectral index of the tensor mode accurately enough to distinguish the source of the scale dependence .
nevertheless , we hopefully expect that more future observations will be able to distinguish the scalon effect and our study will be useful for identifying the inflation model .
k.k . is supported in part by grant - in- aid for scientific research from the ministry of education , science , sports , and culture ( mext ) , japan , nos .
23540327 , 26105520 , 26247042 , 15h05889(k.k . )
the work of k.k .
is also supported by the center for the promotion of integrated science ( cpis ) of sokendai ( 1hb5804100 )
. 1 d. h. lyth and a. riotto , phys .
rept . * 314 * , 1 ( 1999 ) [ hep - ph/9807278 ] .
k. a. olive , phys .
rept . * 190 * , 307 ( 1990 ) p. a. r. ade _ et al .
_ [ polarbear collaboration ] , astrophys . j. * 794 * , no . 2 , 171 ( 2014 ) [ arxiv:1403.2369 [ astro-ph.co ] ]
. p. a. r. ade _ et al .
_ [ bicep2 collaboration ] , arxiv:1403.3985 [ astro-ph.co ] .
g. hinshaw _ et al .
_ [ wmap collaboration ] , astrophys .
j. suppl .
* 208 * , 19 ( 2013 ) [ arxiv:1212.5226 [ astro-ph.co ] ] .
p. a. r. ade _ et al .
_ [ planck collaboration ] , arxiv:1303.5082 [ astro-ph.co ] .
m. kawasaki , m. yamaguchi and j. i .
yokoyama , phys .
d * 68 * , 023508 ( 2003 ) [ hep - ph/0304161 ] : t. kobayashi and o. seto , arxiv:1404.3102 [ hep - ph ] : a. ashoorioon , k. dimopoulos , m. m. sheikh - jabbari and g. shiu , arxiv:1403.6099 [ hep - th ] : c. r. contaldi , m. peloso and l. sorbo , arxiv:1403.4596 [ astro-ph.co ] : j. mcdonald , arxiv:1403.6650 [ astro-ph.co ] : m. s. sloth , arxiv:1403.8051 [ hep - ph ] : m. czerny , t. kobayashi and f. takahashi , arxiv:1403.4589 [ astro-ph.co ] : d. k. hazra , a. shafieloo , g. f. smoot and a. a. starobinsky , arxiv:1404.0360 [ astro-ph.co ] . k. kohri and t. matsuda , arxiv:1405.6769 [ astro-ph.co ] .
t. matsuda , jhep * 0802 * , 099 ( 2008 ) [ arxiv:0802.3573 [ hep - th ] ] . t. matsuda , jcap * 0609 * , 003 ( 2006 ) [ hep - ph/0606137 ] .
t. fujita , m. kawasaki and s. yokoyama , jcap * 1409 * , 015 ( 2014 ) [ arxiv:1404.0951 [ astro-ph.co ] ] .
c. t. byrnes and d. wands , phys .
d * 74 * , 043529 ( 2006 ) [ astro - ph/0605679 ] .
j. o. gong , phys .
d * 79 * , 063520 ( 2009 ) [ arxiv:0710.3835 [ astro - ph ] ]
. k. dimopoulos , k. kohri , d. h. lyth and t. matsuda , jcap * 1203 * , 022 ( 2012 ) [ arxiv:1110.2951 [ astro-ph.co ] ] ; k. dimopoulos , k. kohri and t. matsuda , phys .
d * 85 * , 123541 ( 2012 ) [ arxiv:1201.6037 [ hep - ph ] ] ; s. enomoto , k. kohri and t. matsuda , phys .
d * 87 * , no .
12 , 123520 ( 2013 ) [ arxiv:1210.7118 [ hep - ph ] ] .
j. o. gong , jcap * 1407 * ( 2014 ) 022 [ arxiv:1403.5163 [ astro-ph.co ] ] .
x. gao , t. li and p. shukla , jcap * 1410 * , no .
10 , 008 ( 2014 ) [ arxiv:1403.0654 [ hep - th ] ] .
o. seto , j. i .
yokoyama and h. kodama , phys .
d * 61 * , 103504 ( 2000 ) [ astro - ph/9911119 ]
. b. a. bassett , s. tsujikawa and d. wands , rev .
phys . * 78 * , 537 ( 2006 ) [ astro - ph/0507632 ] .
y. -i .
takamizu and j. i .
yokoyama , phys . rev .
d * 83 * , 043504 ( 2011 ) [ arxiv:1011.4566 [ astro-ph.co ] ] .
d. h. lyth , jcap * 0511 * , 006 ( 2005 ) [ astro - ph/0510443 ] .
t. matsuda , jcap * 1204 * , 020 ( 2012 ) [ arxiv:1204.0303 [ hep - ph ] ] .
k. dimopoulos and m. axenides , jcap * 0506 * , 008 ( 2005 ) [ hep - ph/0310194 ] .
p. a. r. ade _ et al .
_ [ planck collaboration ] , astron .
astrophys .
* 571 * , a16 ( 2014 ) [ arxiv:1303.5076 [ astro-ph.co ] ] ; p. a. r. ade _ et al . _
[ planck collaboration ] , astron .
astrophys . * 571 * , a22 ( 2014 ) [ arxiv:1303.5082 [ astro-ph.co ] ] ; p. a. r. ade _ et al . _
[ planck collaboration ] , astron .
astrophys . * 571 * , a24 ( 2014 ) [ arxiv:1303.5084 [ astro-ph.co ] ] .
|
if the beginning of inflation is defined at the moment when the vacuum energy of the inflaton starts to dominate , the energy density of the other fields at that moment is ( by definition ) comparable to the inflaton .
although the fraction will be small at the horizon exit due to the inflationary expansion , they can alter the scale dependence of the spectrum . at the same time , velocity of the inflaton field
may not coincide with the slow - roll ( attractor ) velocity
. those dynamics could be ubiquitous but can easily alter the scale dependence of the spectrum . since the scale dependence is currently used to constrain or even exclude inflation models , it is very important to measure its shift , which is due to the dynamics that does not appear in the original inflation model .
considering typical examples , we show that the spectral index of the tensor mode is a useful measure of such effect .
precise measurement of the higher runnings of the scalar mode will be helpful in discriminating the source .
|
hm thanks mike barnett , john ellis , lance dixon , ian hinchliffe , and kam - biu luk for useful conversations .
we thank axel kwiatkowski for comments on the manuscript .
adg was supported by cnpq ( brazil ) , and hm was supported in part by the director , office of energy research , office of high energy and nuclear physics , division of high energy physics of the u.s .
department of energy under contract de - ac03 - 76sf00098 and in part by the national science foundation under grant phy-95 - 14797 .
99 for a recent review , see h. baer _
et al . _ , lbl preprint lbl-37016 , hep - ph/9503479 , to appear in `` electroweak symmetry breaking and beyond the standard model , '' eds . t. barklow , s. dawson , h. haber , and j. siegrist , world scientific .
r. muoz - tapia and w.j .
stirling , _ phys .
rev . _ * d49 * , 3763 ( 1994 ) .
the qcd color factors are defined by @xmath67 and @xmath68 for the fundamental representation , and @xmath69 for the adjoint representation . @xmath70
( @xmath71 ) are group generators in each representations .
there are several possible definitions of gluino mass which may appear in experimental constraints : the mass of glueballino @xmath72 , the constituent mass @xmath73 , the @xmath74 current mass @xmath75 , the on - shell ( pole ) mass @xmath76 , and one half the mass of the pseudo - scalar gluinoball @xmath77 .
the various definitions are not expected to differ much from each other if @xmath78 gev@xmath1 .
one may worry about this ambiguity for smaller gluino masses , but our result is insensitive to it since the confidence level in fig .
4 is extremely flat up to 2 gev@xmath1 .
a. signer and l. dixon , _ phys .
lett . _ * 78 * 811 ( 1997 ) ; + e.w.n .
glover , and d.j .
miller , dtp-96 - 66 , hep - ph/9609474 ; + z. bern , l. dixon , d. a. kosower , and s. weinzierl , slac - pub-7316 , hep - ph/9610370 .
|
we reanalyze the constraints on light gluinos ( @xmath0 gev@xmath1 ) from the hadronic @xmath2 decays into four jets .
we find that the published opal data from the 1991 and 1992 runs exclude a light quasi - stable gluino with mass @xmath3 gev@xmath1 at more than 90% confidence level .
this limit depends little on assumptions about the gluino fragmentation and the definition of the gluino mass .
the exclusion confidence level is shown as a function of the mass .
a future projection is briefly discussed .
we also discuss quantitatively how the distributions in the bengtsson zerwas and the modified nachtmann
reiter angles change due to the finite bottom quark or gluino mass .
the analysis is limited to the leading - order calculations .
we , however , give an empirical reason to why the next - to - leading - order corrections are unlikely to change our conclusions . #
1#2 c#10=0=1ex 30 # 1 lbl-39030 + ucb - pth-96/28 + .25 in * excluding light gluinos from @xmath2 decays * 0.3 in andr de gouva and hitoshi murayama + + _ department of physics , university of california + berkeley , california 94720 _ + and + _ theoretical physics group + ernest orlando lawrence berkeley national laboratory + university of california , berkeley , california 94720 _ .3 in supersymmetry is one of the primary targets of extensive searches at various collider experiments , most importantly at cern @xmath4 collider lep and fermilab @xmath5 collider tevatron @xcite .
negative searches at these and previous colliders have already put significant constraints on the parameter space of low - energy supersymmetry . however , a light gluino below the few gev mass range has surprisingly weak experimental constraints as emphasized recently by various authors @xcite ( see , however , an opposing view @xcite ) .
it is an extremely important task to verify or exclude a gluino in this light window experimentally . while the tevatron run ii is expected to extend the reach of heavy gluinos up to a few hundred gev , little effort is devoted to definitively exclude or verify the light gluino window . on the other hand
, a careful reexamination of the existent data may reveal an overlooked constraint on a light gluino ; this is our motivation to study the existent data in detail .
we reanalyzed published data on @xmath2 decays into four jets @xcite , and found that they already exclude a gluino lighter than 1.5 gev@xmath1 at more than 90% confidence level .
we assume that the gluino does not decay inside the detector .
since the published results use only 1991 and 1992 data , it is conceivable that the currently available data , if analyzed properly , could put a much more significant constraint on a light gluino .
we hope our result urges the experimental groups to analyze the whole data set including a possible light gluino .
let us briefly review the existent constraints on a light gluino ( see @xcite for more details ) .
the negative searches at beam dump experiments have excluded a light gluino which decays inside the detector into photino , which in turn interacts with the neutrino detector . however
, a gluino tends to leave the detector without decaying if the squark mass is above a few hundred gev@xmath1 @xcite .
even if the gluino decays , the photino interacts very weakly in this case and can not be detected .
if the gluino does not decay , it forms bound states such as gluinoball @xmath6 , glueballino @xmath7 or baryon - like states , especially @xmath8 @xcite .
other states are likely to decay into these neutral bound states , and searches for exotic charged hadrons may not apply unless a charged gluino bound state decays only weakly .
one the other hand , the mass region above 1.5 gev@xmath1 and below 4 gev@xmath1 is excluded from quarkonium decay @xmath9 , where @xmath10 is the pseudo - scalar gluinoball , independent of the gluino lifetime @xcite .
whether the bound extends to lower masses is controversial because of the applicability of perturbative qcd calculations @xcite .
the mass range above 4 gev@xmath1 is expected to give a shorter lifetime and is excluded by a negative search for events with missing energy at ua1 @xcite .
the authors of @xcite claim that the limit from ua1 extends down to 3 gev@xmath1 . in any case ,
the least constrained region is the mass range below 1.5 gev@xmath1 , where the gluino is relatively stable so that it does not decay inside detectors .
this is our window of interest in this letter .
we would like to emphasize that the best method to exclude the gluino mass range below 1.5 gev@xmath1 is to use inclusive processes rather than searching for specific bound states with certain decay modes .
the latter search would heavily depend on assumptions such as the mass spectrum of various gluino bound states and their decay modes and decay lifetimes .
one would have to design experiments and put constraints with all possible theoretical assumptions on gluino bound states in order to exclude the light gluino definitively .
on the other hand , the constraints would be much less sensitive to theoretical assumptions if they were based on inclusive processes where perturbative qcd is applicable .
there are several possibilities pointed out in the literature along this line .
the most popular one is to study the effect of light gluinos in the running of the qcd coupling constant @xmath11 .
it was even pointed out that the values of @xmath11 from higher energy measurements tend to be higher than those extrapolated from lower energies using qcd with the ordinary quark flavors , and the data actually prefer the existence of a light gluino to compensate the slight discrepancy @xcite .
however , this issue remains controversial @xcite . even though the discrepancy between low - energy and high - energy measurements is diminishing @xcite ,
still the data are not precise enough to exclude or verify a light gluino definitively .
the second one is its effect on the altarelli
parisi evolution of the nucleon structure functions @xcite .
unfortunately the effect is too small to be tested using the present experimental data .
it might be that the more recent hera data could improve the situation , but making a definite statement on the existence of a light gluino appears to be difficult .
the third one is to study the angular correlations in the so - called `` 3 + 1 '' jet events at hera @xcite .
however , the effect of the light gluino was found to be negligible .
the final one , which we employ in this letter , is the study of four jet correlations in @xmath12 collisions @xcite .
previous studies did not find significant constraints , but given the size of the current lep data , we find this to be the most promising direction .
the only data we use in this letter are studies of qcd color factors @xcite .
the experimental groups at lep have performed impressive analyses of the hadronic @xmath2 decays into four jets , extracting qcd color factors @xmath13 and @xmath14 @xcite from jet angular distributions , to confirm su(3 ) as the qcd gauge group and five light quark flavors .
the angular distributions of @xmath15 final state differ from those of @xmath16 , where @xmath17 refers to a generic quark and @xmath18 to a gluon .
three angles are commonly used in four - jet analyses : the bengtsson zerwas ( bz ) angle @xmath19 @xcite , the modified nachtmann reiter ( nr ) angle @xmath20 @xcite , and the opening angle of the two less energetic jets @xmath21 . if there exists a light gluino @xmath22 , the final state @xmath23 also contributes to the @xmath2 decays into four jets .
the angular distributions of @xmath23 would be identical to those of @xmath15 .
therefore , a possible light gluino would change the extracted @xmath14 but not @xmath13 .
apart from the mass effects , @xmath14 should increase by a factor of @xmath24 , because the gluino is a color - octet and counts effectively as three additional massless quarks .
note that these analyses do not use the overall rate of four - jet events since it is sensitive to the choice of @xmath25 in the absence of next - to - leading order ( nlo ) calculations .
so far the experimental analysis which used the highest statistics is the one by opal @xcite , which also briefly discussed constraints on a light gluino .
they found that the light gluino is barely outside the 68% confidence level contour and decided the data did not put a significant constraint .
however , we find the previous analyses not carefully designed to study the effect of a light gluino because of the following reason . when one discusses a possible light gluino , qcd with the color group su(3 ) should be assumed .
given overwhelming experimental evidences of qcd , it is not wise to , for instance , vary the number of colors @xmath26 when one studies the effect of a particle ( light gluino ) _ added to _ qcd .
therefore , we must fix the qcd color factor @xmath13 to be that of the su(3 ) group , 9/4 .
second , we already _ know _ that there are five quark flavors @xmath27 , @xmath28 , @xmath29 , @xmath30 and @xmath31 , which appear in @xmath2 hadronic decays .
when one puts constraints on an _ additional _ contribution from a light gluino , one should not vary the number of flavors below 5 , or equivalently , @xmath14 below 3/8 .
the only lep paper which analyzed data in a way close to this spirit , and put an upper bound on possible additional @xmath15-type final states , is the one from opal @xcite ; but it used very limited statistics .
all more recent papers @xcite varied both @xmath13 and @xmath14 without constraints . by reanalyzing data with these constraints we can put a much more significant bound on a light gluino than reported . actually , fixing the group to be su(3 ) ( @xmath32 ) has the greatest impact on the confidence level , while restricting @xmath33 has a much smaller effect ( actually it makes the significance worse ) .
we further include the finite mass of the bottom quark in the analysis which slightly improves the significance .
overall , a massless gluino is excluded already better than at 90% confidence level by the opal 1991 and 1992 data only @xcite .
let us start from the reported contour on the @xmath13 , @xmath14 plane , shown in fig . 1
. we fix @xmath34 because of the philosophy of our study stated above .
since one - dimensional @xmath35 distributions have much higher confidence levels than two - dimensional ones , this change improves the significance of the data drastically . from their @xmath35 contours , we minimized @xmath35 with fixed @xmath32 , and defined @xmath36 relative to the @xmath35 at the minimum ( @xmath37 ) .
the confidence levels are calculated using a one - dimensional @xmath35 distribution with @xmath36 defined in this manner .
this is a conservative choice because @xmath38 .
we obtain @xmath39 with fixed @xmath13 .
if one had used this central value and the standard deviation , a massless gluino would be excluded at 95% confidence level .
however , we also need to impose another constraint , @xmath40 , which can be easily taken into account .
the standard method is to use the gaussian distribution only in the physical region , and scale the normalization of the distribution so that the total probability in the physical region becomes unity . since the central value is very close to the theoretical value of the qcd , this effectively increases the probability of allowing light gluinos by a factor of two ; numerically the confidence level is 88% .
finally , we study the effect of the finite mass of the bottom quark and gluinos on the extracted @xmath14 .
the authors of @xcite studied the effect of the finite mass of quarks on the four - jet rates .
they also looked at the angular distributions and reported there were little changes .
even though it is true that the distributions do not change drastically , they gradually become similar to those of @xmath16 final state as one increases the mass of the quark , and hence the extracted @xmath14 from the fit to the distributions has a relatively large effect due to the finite mass of the bottom quark .
the papers @xcite do not take this effect into account at all . the opal experiment @xcite used parton level event generators by the authors of @xcite and @xcite to study the effect .
they have found a surprisingly large effect : the bottom quark contribution to @xmath14 was about one half of a massless quark at @xmath41 .
we confirmed their estimate in a detailed parton - level calculation based on that done in @xcite , neglecting the interference between primary and secondary quarks .
this approximation is known to be better than a few percent . on the other hand
, this approximation has the clear advantage of enabling us to distinguish primary and secondary quarks unambiguously .
our code employs helicity amplitude technique using the helas package @xcite , which made it straight - forward to incorporate finite masses in the four - jet distributions . the finite mass affects the extracted @xmath14 in two ways
first , the rate of producing secondary massive quarks is suppressed compared to the massless case as shown with the solid line in fig . 2 .
for instance , there is about 20% suppression with @xmath42 gev@xmath1 and @xmath43 .
this result is consistent with @xcite .
the mass of the primary quark has little effect on the rate : only a 6% suppression for @xmath44 gev@xmath1 .
we also checked that the distributions in bz and nr angles with a massive primary quark are indistinguishable from the massless case .
these observations are consistent with naive expectations , because the primary quarks are much more energetic than the secondary ones and hence the mass effect is suppressed by @xmath45 .
we therefore neglect the finite mass of primary quarks hereafter .
second , the nr and bz angle distributions gradually approach those of the @xmath16 final state as one increases the mass of the secondary quarks .
we are not aware of detailed analyses of these distributions with massive quarks in the literature .
the distributions are shown in fig . 3 normalized so that the total area below the curve is unity , in order for the effect on the rate and that on the distribution to be clearly separated .
we fit the distributions as linear combinations of @xmath16 and massless @xmath15 distributions to determine the _ effective _ @xmath14 , in order to mimic the experimental analyses .
the fit is surprisingly good ; we checked this for quark masses between 0 and 5 gev@xmath1 .
combined with the reduction in the rate , the net effect of the finite mass of secondary quarks is shown in fig . 2 .
with @xmath46 gev@xmath1 for secondary bottom quarks , the overall rate of @xmath47 final state is reduced to 82.5% , while the fit to angular distributions gives a @xmath14 reduced to 76.4% ( bz ) or 85.5% ( nr ) compared to that of a massless quark flavor ( 3/8 ) , on top of the reduction in the rate . in total , secondary bottom quarks contribute to @xmath14 as @xmath48 or @xmath49 , which is not a negligible suppression .
the extracted @xmath14 from the data is an average of @xmath14 from five flavors .
the reported @xmath14 in @xcite includes a correction to compensate the apparent suppression due to the finite bottom quark mass .
such a correction in turn effectively enhances the additional contribution from gluinos by a factor of @xmath50 or @xmath51 .
note that this slight enhancement effect does not change significantly even when one varies @xmath52 from 4 to 5 gev@xmath1 , as can be seen in fig . 2 .
the actual opal analysis @xcite fits the data in the three dimensional space spanned by bz , nr and @xmath21 angles after bin - by - bin systematic corrections from monte carlo simulations .
such an analysis is beyond the scope of this letter .
we assume that the total effect of the finite mass is somewhere between the effects on bz or nr angles since @xmath21 is not as effective in extracting @xmath14 . as it is clear from fig .
3 , fits to distributions of massive quarks give apparent additional contributions to @xmath16 and hence @xmath13 .
they are completely negligible , however , compared to the size of the true @xmath16 which is about one order of magnitude larger than the sum of all @xmath15 final states , and hence we will neglect such contributions hereafter . given the above considerations , we can now present the exclusion confidence levels on a light gluino for varying gluino masses in fig .
4 . for both curves
, we used @xmath53 gev@xmath1 and used the effective @xmath14 extracted from the fits to bz and nr angles .
the finite mass effect of the gluino is treated in the same manner .
first of all , it is clear that the finite mass effect which we studied depends little on the choice of bz or nr angles , and hence we believe it mimics the true experimental fits ( which use bz , nr and @xmath21 angles simultaneously in a three - dimensional fit with 295 bins ) quite well .
second , the confidence level is extremely flat up to 2 gev@xmath1 .
this implies that we do not need to worry about complication due to non - perturbative dynamics in defining the gluino mass @xcite .
the lower bound of @xmath54 gev@xmath1 at 90% confidence level is already in the perturbative region .
it is quite likely that the gluino mass relevant to this analysis is a running mass defined at the scale @xmath55 @xcite .
it is then straight - forward to convert the bound to the on - shell gluino mass : the lower bound of @xmath56 gev@xmath1 in the @xmath57 scheme corresponds to @xmath58 gev@xmath1 .
we would like to comment that the clever jet reconstruction method used in the opal analysis @xcite is particularly suited for the study of light gluinos in four - jet events .
they did not scale the measured jet energies by an overall ratio @xmath59 , as done traditionally in similar analyses , but instead used the angular information of the jets to calculate the energy of each jet using energy and momentum conservation .
this method avoids uncertainties in the gluino fragmentation .
since it is not well understood how a gluino fragments , one should use a similar method to avoid dependence on assumptions about the gluino fragmentation in future studies .
unfortunately , our analysis is limited to the leading - order ( lo ) calculations .
it is a natural question whether the nlo corrections may change our conclusion .
first of all , we expect that the corrections to the angular variables used in the analysis are presumably not large .
the nlo corrections are important when a variable involves @xmath25 , such as 3- and 4-jet rates , thrust , etc .
the variables used in our analysis are not proportional to powers of @xmath25 , and hence scale - independent at the lo approximations .
this is analogous to the case of the forward - backward asymmetry which is an ( integrated ) angular variable and is @xmath25 independent at the lo . it does receive an nlo correction of @xmath60 , where @xmath61 in the case of a massless quark @xcite . in our case
, we also expect a correction to the angular distributions of the order of @xmath62 , where @xmath63 is probably an educated guess
. then a typical size of the nlo correction is about 5 % .
however , a correction of this order of magnitude may still be of concern because of the following reason .
the @xmath16 final state is roughly an order of magnitude larger than the @xmath15 final state .
therefore , a 5% correction to @xmath16 may result in a 50% correction to @xmath15 final state , to be compared with a possible 60 % contribution from the gluino .
we argue , however , that such a higher order correction is not likely to change our conclusion .
first of all , the helicity structure and the color flow in the @xmath16 final state and @xmath15 final state are quite different . if a correction to the @xmath16 final state changes the conclusion , the following must be happening : the correction term to the @xmath16 exactly mimics an additional contribution to the @xmath15 final state in the angular distributions with a negative sign such as to mask the contribution from the @xmath23 final state .
we do not find this to be likely because they have different structures in the helicities and colors .
moreover , the data do not indicate that the nlo correction is large .
opal data @xcite are fit very well by the lo monte carlo on three - dimensional histograms of 295 bins with @xmath64 .
this excellent agreement between the matrix element calculation and the data found in @xcite supports the smallness of the nlo corrections empirically .
however , the calculations of nlo corrections are necessary to justify it .
expansion is small @xcite . ] for future studies , it is also desirable to compare different monte carlo programs , while only jetset was used in recent experimental papers @xcite . finally , it is worth emphasizing that the result in this letter is based on the 1991 and 1992 opal data with 1.1 m hadronic @xmath2 s @xcite .
the statistical and systematic uncertainties are comparable in their paper . given the current size of the lep data , which is more than an order of magnitude larger , the statistical uncertainty should reduce substantially once all of the data has been analyzed .
this change alone could drastically improve the sensitivity to the light gluino in four - jet events . on the other hand , it is not obvious how systematic uncertainties can be further reduced .
the largest systematic uncertainty originates in the bin - by - bin acceptance corrections which needed to be done before performing a fit in bz , nr , and the opening angle space .
it is not clear how this uncertainty can be reduced if one employs the same method . perhaps choosing larger values of @xmath65 reduces the uncertainty while reducing the statistics at the same time .
there could be an optimal choice of @xmath65 for this particular purpose .
some of the other large systematic uncertainties are specific to the opal experiment and could be reduced by averaging results from all four experiments . in any case , there is no doubt that we can expect a better result from the currently available data set . in summary ,
we reanalyzed the published opal 1991 and 1992 data on the qcd color factors @xcite to constrain possible additional contributions to four - jet events in @xmath2 decays due to @xmath23 final states .
the main difference from the original opal study is to fix @xmath34 as required by qcd .
we further imposed @xmath66 and treated the finite mass effects of both the bottom quark and the gluino carefully .
we find that a light gluino with a mass below 1.5 gev@xmath1 is excluded at better than 90% confidence level .
the result is insensitive to assumptions about what bound state it forms , the definition of its mass , and the gluino fragmentation provided that it does not decay inside the detectors .
we believe that the currently available data set is much more sensitive to a possible additional contribution from the light gluino .
we argued that the nlo corrections are unlikely to modify the conclusion ; still , this assertion needs to be justified by explicit calculations in the future .
as a by - product of this analysis , we discussed the effect of finite bottom quark mass on bz and nr distributions in detail , which is not negligible when extracting qcd color factors at current precisions .
|
a significant amount of time and computing resources are spent on calculating events at hadron colliders .
whether a theoretical calculation of matrix elements , or an experimental simulation of events with detector effects , one common element is the evaluation of parton distribution functions ( pdfs ) .
these functions return the probability of finding a parton ( quark or gluon ) inside of a proton , based on two parameters : the fraction of momentum carried by the parton @xmath0 , and the square of the energy scale of the process @xmath1 . because the input parameters can span several orders of magnitude , the values of these functions are stored in two - dimensional tables for a finite number of input points . an approximate result for an arbitrary input of @xmath0 and @xmath1
is derived by interpolating between the values obtained from the nearest table entries . in profiling ` ztop ` @xcite , a `
fortran`code written to simulate next - to - leading - order jet distributions in single - top - quark production , it has become apparent that much of the execution time of real production code is spent acquiring pdfs . upon close examination of the cteq4 and cteq5 pdf codes @xcite , a handful of trivial optimizations arise that can cut this time in half .
based on this success , i examine further algorithmic improvements in the typical interface functions that reduce the execution time by another factor of two or more for all cteq pdfs ( including cteq6 @xcite ) .
i provide specific recommendations that are simple to implement , but which can have large consequences for efficiency
. a ` gprof ` profile of ` ztop ` @xcite indicates that up to 90% of the execution time is spent in acquiring pdfs .
execution times in other programs appear to be dominated by the same routines . in table
[ tab : cteqdef ] i show the typical fraction of time spent evaluating pdfs for ` ztop ` , and the two most common showering event generators , ` herwig ` 6.1 @xcite , and ` pythia ` 6.2 @xcite . the results in table [ tab : cteqdef ]
were generated using the gnu ` g77 ` 3.1 compiler for linux on a 1.4 ghz pentium 4 processor with the flags ` -g -pg -o3 -march = pentium4 -msse2 ` , and were verified by commenting out the routines .
the results vary by less than 3% when changing compilers or compiler flags .
times for cteq4 and cteq5 differ from cteq6 , because the latter uses a different interpolation algorithm .
retrieving pdfs is always the most time - intensive operation in these calculations .
therefore , it behooves us investigate what options are available to speed up the pdf routines .
.fraction of time spent evaluating pdf functions using default cteq computer codes for three programs : the next - to - leading - order jet calculation ` ztop ` , and two showering event generators , ` herwig ` and ` pythia`.[tab : cteqdef ] [ cols="^,^,^",options="header " , ] we observe in sec .
[ sec : results ] that replacing ` polint ` with ` polint3/4 ` greatly reduces the dependence of program execution time on the choice of compiler flags .
this is a simple change , and can improve running times of some programs by up to a factor of 2 .
since most programs call several pdfs in a row with the same values of @xmath0 and @xmath1 , the obvious next step is to modify the cteq routines ` partonxn ` to save previous results , and calculate only what has changed .
the third level of optimization replaces the typical ` cernlib ` interface functions ` structm ` and ` pftopdg`@xcite with fully optimized versions that are specialized to the cteq parton distributions .
i recommend that these improvements be incorporated into the base cteq distribution @xcite , the ` pdflib ` routines in ` cernlib ` @xcite , and the new les houches accord compilation of pdfs ` lhapdf ` @xcite .
full versions of the routines presented here may be obtained from the author , or from ref .
@xcite .
despite these impressive gains , we see in table [ tab : fraction ] that evaluating pdfs remains the most time - consuming aspect of hadronic calculations .
this suggests two avenues of investigation that should be considered for future programs .
first , a systematic study of pdf evolution codes should be performed to determine whether there are more efficient interpolation algorithms to use with table - based pdfs .
this would allow universal improvements in code execution .
second , each monte carlo writer should be aware of the timing issues ( and potential bugs if @xmath2 is not the same as @xmath3 ) , and consider using an interface structure other than ` structm ` .
the potential savings from calling one pdf instead of eight or more could be very significant .
finally , it is interesting that the optimizations i have listed can remove the apparent need for parameterizations of the parton distribution functions . in general
, any calculation that relies heavily on interpolation , or multiple evaluations of a function in which some pieces do not vary , should see similar improvements in performance by applying these same techniques .
this work was supported by universities research association inc . under contract
de - ac02 - 76ch03000 with the united states department of energy .
this is a `` fast '' version of ` polint3 ` , which has been optimized for the special case of 3-point fitting , and no possibility of divisions by zero .
an error estimate is not returned , since it is never used in the cteq evolution codes .
this code should be used to replace the version of ` polint ` called by the cteq4 and cteq5 pdfs . ....
c this is a specialized recoding of neville 's c algorithm based on the polint routine from c " numerical recipes " , but assuming n=3 , and c ignoring the error estimation .
c written by z. sullivan , may 2004 c this file uses a minimal number of c instructions to do 3-point fitting .
c subroutine polint ( xa , ya,3,x , y , ignored ) subroutine polint3 ( xa , ya , n , x , y , dy ) implicit none double precision xa(3),ya(3),x , y , dy , den double precision c1,ho , hp , hp2,w , d1,d2 integer n b. w. harris , e. laenen , l. phaf , z. sullivan , and s. weinzierl , phys .
d * 66 * , 054024 ( 2002 ) .
zack sullivan , phys .
d * 70 * , 114012 ( 2004 ) .
cteq collaboration , h. l. lai _ et al .
_ , phys .
d * 55 * , 1280 ( 1997 ) ; z. phys .
c * 74 * , 463 ( 1997 ) ; eur .
j. c * 12 * , 375 ( 2000 ) .
cteq collaboration , j. pumplin _ et al .
_ , j. high energy phys . * 07 * , 012 ( 2002 ) . g. corcella _ et al .
_ , j. high energy phys . * 01 * , 010 ( 2001 ) . t. sjstrand _ et al .
. phys .
commun . *
135 * , 238 ( 2001 ) . w. press , s. teukolsky , w. vetterling , and b. flannery , _ numerical recipes in fortran : the art of scientific computing _
( cambridge university press , cambridge , 1986 ) , chap . 3.1 .
zack sullivan , _ faster parton distribution evaluation in monte carlos _ , fermilab - fn-755 , hep - ph/0403055 .
_ intel architecture optimization reference manual _ , intel corporation document 245127 - 001 , 1999 ; _ intel fortran compiler for linux systems user s guide , volume ii : optimizing applications _
, intel corporation document 2553260 - 002 , 2003 .
j. s. conway _ et al .
_ , _ report of the tevatron higgs working group _ , in _ physics at run ii : the supersymmetry / higgs workshop _ , fermilab , 1998 , edited by m. carena and j. lykken ( fermilab , batavia , 2002 ) , p. 39 , hep - ph/0010338 . w. giele _ et al .
_ , _ the qcd / sm working group : summary report _ , in _ les houches 2001 : physics at tev colliders _ , les houches , france , 2001 , edited by p. aurenche _ et al .
_ , p. 275
, hep - ph/0204316 .
|
a few changes to the routines that calculate cteq parton distribution functions allow modern compilers to optimize the evaluations , while having no quantitative effect on the results .
overall computation time is reduced by a factor of 45 in matrix - element calculations , and by 1.32.5 in showering monte carlo event generators .
similar improvements in performance may be expected in any calculations relying heavily on interpolation or multiple calls to functions .
|
we present results of a detailed comparison between the parameters characterizing the structural and dynamical properties of a sample of 29 simulated disk - like objects ( dlos ) and those measured in observed spiral galaxies .
the properties we focus on are the bulge and disk structural parameters and the rotation curves , as they can be constrained with available data for observed spiral galaxies .
these data are taken principally from the compilation of galaxy structural parameters of broeils ( 1992 ) , de jong ( 1996 ) and courteau ( 1996 , 1997 ) .
the dlos have been identified in ap3m - sph fully consistent hierarchical hydrodynamical simulations , realizations of a cdm flat model , with @xmath0 , @xmath1 and @xmath2 , made using 64@xmath3 particles in a periodic box of comoving side 10 mpc ( h@xmath4 km s@xmath5 mpc@xmath5 ) , and where an _ inefficient _ schmidt law - like algorithm to model the stellar formation processes has been implemented ( see tissera , lambas , & abadi 1997 ; silk 1999 ) .
a surface density bulge - disk decomposition was performed on the dlos , using a double - exponential profile , @xmath6 + \sigma_{\rm d}(0)\exp[-r / r_{\rm d}].\ ] ] the resulting bulge and disk scale lengths , @xmath7 and @xmath8 , and their ratio @xmath9 , are consistent with available data ( courteau , de jong , & broeils 1996 ; de jong 1996 ; moriondo , giovanelli , & haynes 1999 ) .
see fig .
[ bd](a ) .
the rotation curves of the simulated objects can be constructed by adding up in quadrature the contributions to circular rotation from the bulge and the disk , both formed by baryonic particles , and the dark matter halo , @xmath10 our rotation curve analysis and shape modeling uses a parameterization based on the spatial scale @xmath11 and the velocity ( mass ) estimate @xmath12 .
these parameters were compared with those measured for the courteau
faber sample of bright sb
sc field spirals with long - slit h@xmath13 rotation curves , and the broeils ( 1992 ) compilation of late - type spirals with extended rotation curves .
deep surface photometry is available for both samples .
in contrast to findings in other fully consistent hydrodynamical simulations ( navarro & steinmetz 2000 and references therein ) , we find dlo @xmath14 values that are consistent with the observational data . this is a consequence of disk formation with conservation of the specific angular momentum @xmath15 .
see fig .
[ bd](b ) .
in real spirals , the relative contributions of baryons and dark matter to the rotation curve can not be inferred uniquely from the observations .
one has to postulate some hypothesis to infer the mass distribution of the halo .
the so - called ` maximum disk ( or maximal light ) hypothesis ' ( van albada & sancisi 1986 ) assumes that the halo dark mass component needed to fit the rotation curve should be minimum , or , equivalently , that the contribution of the disk and bulge to the inner parts of the rotation curves should be maximum . considering this
, one has @xmath16 , at the 95% confidence level ( sackett 1997 ) .
the other common constraint used is that of sub - maximal disks , obtained either by matching the vertical velocity dispersion , scale height and scale length of a thin exponential disk which yields the maximum disk rotation ( bottema 1993 ) , or , independently , showing that the tully fisher relation of bright spirals is independent of surface brightness ( courteau & rix 1999 )
. then one has @xmath17 .
the average relative contribution of baryons to @xmath14 in our sample of dlos can be measured directly .
we find @xmath18 , in very good agreement with bottema ( 1993 , 1997 ) and courteau & rix ( 1999 ) .
see fig .
[ vbvt ] .
most dlos have been found to have sub - maximal disks .
this means that the amount of baryon mass that has ended up inside @xmath19 is not excessive , again as a consequence of @xmath15 conservation .
in conclusion , the comparison between the dlos produced in our simulations and observational data allows us to affirm that the dlos have counterparts in today spiral galaxies .
this agreement suggests that the process operating in fall & efstathiou ( 1980 ) standard model for disk formation ( i.e. , gas cooling and collapse with @xmath15 conservation ) is also at work in the _ quiescent _ phases of the dlo formation in these simulations .
however , _
violent _ episodes ( i.e. , interactions and merger events ) also occur and play an important role in the dlo assembly ( domnguez - tenreiro , tissera , & siz 1998 ; siz , tissera , & domnguez - tenreiro 1999 ) .
to provide the right conditions for disk regeneration after the last violent episode of the dlo assembly , a compact central stellar bulge is needed ; this will ensure the axisymmetric character of the gravitational potential well at scales of some kpcs at all times , avoiding excessive @xmath15 losses in violent events .
a second condition is necessary : the availability of gas at low @xmath20 , to form the disk .
the good match of our dlo parameters with observational data suggests that our _ inefficient _ star formation algorithm meets both requirements .
this global agreement with observations also represents an important step towards making numerical approaches more widely used for the study of galaxy formation and evolution in a cosmological framework , i.e. , from primordial fluctuations .
bottema r. , 1993 , a&a , 275 , 16 ; 1997 , a&a , 328 , 517 broeils a. h. , 1992 , ph.d . thesis , univ . groningen casertano s. , van gorkom j. h. , 1991 , aj , 101 , 1231 courteau s. , 1996 , apjs , 103 , 363 ; 1997 , aj , 114 , 2402 courteau s. , rix h .- w . , 1999 , apj , 513 , 561 courteau s. , de jong r. s. , broeils a.h .
, 1996 , apj , 457 , l73 de jong r. s. , 1996 , a&a , 313 , 45 domnguez - tenreiro r. , tissera p. b. , siz a. , 1998 , apj , 508 , l123 fall s. m. , efstathiou g. , 1980 , mnras , 193 , 189 moriondo g. , giovanelli r. , haynes m. p. , 1999 ,
a&a , 346 , 415 navarro j. f. , steinmetz m. , 2000 , apj , 538 , 477 sackett p. d. , 1997 , apj , 483 , 103 siz a. , tissera p. b. , domnguez - tenreiro r. , 1998 , ap&ss , 263 , 43 silk j. , 1999 , in _ proceedings of the xixth moriond meeting _ , eds . f. hammer , t. x. thuan , v. cayatte , b. guiderdoni , & j. tran thanh van , ed .
frontires , 439 tissera p. b. , lambas d. g. , abadi m. g. , 1997 , mnras , 286 , 384 van albada t. s. , sancisi r. , 1986 , phil .
london ser .
a , 320 , 447
|
we analyze the structural and dynamical properties of disk - like objects formed in fully consistent cosmological simulations which include inefficient star formation .
comparison with data of similar observable properties of spiral galaxies gives satisfactory agreement , in contrast with previous findings using other codes .
this suggests that the stellar formation implementation used has allowed the formation of disks as well as guaranteed their stability .
|
radiative corrections to jet production in @xmath6 annihilation have been known for a long time @xcite .
previous calculations were , however , performed for massless quarks . in most practical applications
this is sufficient , since at relatively low energy , the @xmath7 fraction is strongly suppressed , and at high energy ( i.e. on the @xmath5 peek and beyond ) mass effects are presumably suppressed .
nevertheless there are several reasons why a next - to - leading - order calculation is desirable .
first of all , at sufficiently high energies , top pairs will be produced , and mass effects there are very likely to be important .
a second important reason is to understand the relevance of mass corrections due to bottom production to the determination of @xmath8 from event shape variables . as a third point ,
quantities such as the heavy flavour momentum correlation @xcite , although well defined in the massless limit , can not be computed using the massless results of refs .
@xcite .
we have recently completed a next - to - leading - order calculation of the heavy - flavour production cross section in @xmath6 collisions , including quark mass effects . as an illustration , in fig .
[ fig : graphs ] we show the feynman diagrams for a born term ( a ) , a virtual correction term ( b ) , and two real next - to - leading contributions ( c , d ) .
next - to - leading corrections arise from the interference of the virtual graphs with the born graphs , and from the square of the real graphs .
observe that we always deal with the cross section for the production of the heavy quark pair plus the emission of at least one extra particle ( i.e. a gluon or a quark ) .
the inclusion of virtual graphs with only a @xmath9 pair in the final state is not needed if one only computes 3-jet related quantities .
furthermore , by using available results on the total heavy flavour cross section at order @xmath10 ( see @xcite and references therein ) , one can avoid altogether to compute this kind of virtual graphs .
virtual graphs , besides the usual ultraviolet divergences , which are removed by renormalization , have also infrared and collinear divergences .
these cancel when suitable infrared safe final - state variables are considered .
our treatment of the infrared cancellation is such that the final result is expressed as a partonic event generator , in which pairs of weighted correlated events are produced .
any shape variables can be computed for each generated event .
only infrared - safe variables give rise to a finite distribution .
no arbitrary cutoff is needed in this calculation in order to implement the cancellation of virtual and real infrared divergences .
therefore , one does not have to worry about taking the limit for a vanishing soft cutoff .
thus , this method is similar to the one of ref .
@xcite , which was used there to compute a large class of shape variable distributions for the lep experiments .
we have verified that our calculation has the correct massless limit , by comparing our result with that of the massless calculation of ref .
@xcite for all the shape variables listed there , and for the moments of the energy
energy correlation , proposed as a benchmark in ref .
@xcite .
very recently , two calculations have appeared in the literature that address the same problem @xcite .
they both use a slicing method in order to deal with infrared divergences .
we were able to perform a partial comparison of our result with that of ref .
@xcite , and found satisfactory agreement . in the older work of ref .
@xcite , a calculation of the process @xmath11 has been given , but virtual corrections to the process @xmath12 were not included . in ref .
@xcite , the nlo corrections to the production of a heavy quark pair plus a photon are given , including both real and virtual contributions .
we will give full details of our calculation in a future publication .
in the present letter we would like to present a first application , which could not be dealt with using the massless results .
this is the average momentum correlation @xmath13 defined as [ defr ] r= , where @xmath14 are the feynman @xmath15 of the produced @xmath16 mesons , @xmath17 , and @xmath18 . in ref .
@xcite @xmath13 is computed at leading order in the strong coupling constant .
the interest in this quantity lies in the fact that it may affect the determination of @xmath4 @xcite , when the tagging efficiency for @xmath19 mesons is determined by looking at both the single - tag and the double - tag events .
a simplified description of how this works is given in ref .
it is convenient to rewrite @xmath13 in the following way [ betterr ] r=. we notice that , thanks to the above form , the contribution of virtual graphs with only a @xmath9 pair in the final state is not needed for this calculation . defining now & = & b + ( ) ^2 c + o(^3 ) , + & = & a+o(^2 ) , where @xmath8 is the @xmath20 , 5-flavour coupling constant evaluated at a scale @xmath21 equal to the annihilation energy @xmath1 .
we have r= b + ( ) ^2 ( c+2ab - a^2)+ o(^3 ) . in the particular case of the computation of the average momentum correlation , we have treated as massive the heavy quark directly produced from the electroweak vertex .
secondary heavy quarks produced by gluon splitting have been treated as light .
similarly , we have not included graphs in which a light flavour is produced by the electroweak vertex , and the heavy quark pair is produced by a gluon .
although the inclusion of these contributions does not pose any particular problem , we believe that this quantity is the closest to what is usually defined to be the correlation induced by hard radiation . other dynamical correlation mechanisms , arising for example from the production of four heavy flavours , or from the indirect ( e.g. via gluon splitting ) production of a heavy flavour pair in an event with primary production of light quarks can be examined separately .
their influence upon the determination of @xmath22 depends upon a more detailed specification of the experimental setup , such as , for example , on whether cuts are applied that remove soft heavy quarks .
.mass dependence of the coefficients @xmath23 , @xmath7 , @xmath24 and @xmath13 .
the digit in parenthesis ( to be taken as zero if not given ) represents the accuracy of the last digit .
[ cols="^,^,^,^,^",options="header " , ] as far as its perturbative expansion in powers of @xmath8 is concerned , the average momentum correlation is a quantity that is well behaved in perturbation theory , and it is also quite small . since its effect is typically of the order of 1% , one may worry that non - perturbative effects , of order @xmath28 ( where @xmath29 is a typical hadronic scale ) may compete with the perturbative result .
this is a very delicate problem , since we know very little about the hadronization mechanism in qcd . in ref .
@xcite , this problem was addressed in the context of the renormalon approach to power corrections .
it was shown there that ( in the renormalon approach ) corrections to the momentum correlations are at least of order @xmath30 , and thus negligible at lep energies .
although this result can not be considered as a definitive answer to the problem , it is at least an indication that power corrections to this quantity are small .
ellis , d.a .
ross and a.e .
terrano , . j.a.m .
vermaseren , k.j.f .
gaemers and s.j .
k. fabricius , g. kramer , g. schierholz and i. schmitt , .
p. nason and c. oleari , , hep - ph/9607347
chetyrkin , r. harlander , j.h .
khn and m. steinhauser , preprint mpi - pht-97 - 012 , hep - ph/9704222 . z. kunszt and p. nason , `` qcd '' , in `` z physics at lep 1 '' , eds .
g. altarelli , r. kleiss and c. verzagnassi ( report cern 89 - 08 , geneva , 1989 ) .
p. nason and b.w .
webber , `` qcd '' , in `` physics at lep2 '' , eds .
g. altarelli , t. sjstrand and f. zwirner ( report cern96 - 01 , geneva 1996 ) .
g. rodrigo , _ nucl .
suppl . _ * 54a*(1997)60 , hep - ph/9609213 ; g. rodrigo , ph .
d. thesis , univ . of valncia , 1996 , hep - ph/9703359;g .
rodrigo , a. santamaria and m. bilenkii , preprint ftuv-96 - 80 , mar 1997 , hep - ph/9703358 .
w. bernreuther , a. brandenburg and p. uwer , preprint pitha-97 - 5 , mar 1997 , hep - ph/9703305 .
a. ballestrero , e. maina and s. moretti , ; .
l. magnea and e. maina , , hep - ph/9604385 .
aleph collab . ,
r. barate , preprint cern - ppe-97 - 018 , feb 1997;aleph collab . ,
r. barate , preprint cern - ppe-97 - 017 , feb 1997;aleph collab .
, d. buskulic et al .
, ; aleph collab . , d. buskulic et al . , ; delphi collab .
, p. abreu et al . , ; delphi collab . , p. abreu et al . , ; delphi collab . , p. abreu et al . , ; l3 collab . , o. adriani et al . , ; opal collab .
, k. ackerstaff et al . , ;
opal collab .
, r. akers et al .
, ; sld collab . , k. abe et al . , .
particle data group , r.m .
barnett et al . , .
p. nason and b.r .
webber , , hep - ph/9612353 .
|
we recently completed a calculation of the process @xmath0 , where @xmath1 is a heavy quark , at order @xmath2 .
as a first application of this calculation we compute the momentum correlations of @xmath3 pairs at next - to - leading - order .
this quantity is interesting since it may affect the determination of @xmath4 as measured in @xmath5 decays .
we find that the next - to - leading corrections are of moderate size , thus confirming the conclusions that can be drawn from the leading - order calculation .
.5 cm * paolo nason , * + 0.1 cm cern , th division , geneva , switzerland + .5 cm * carlo oleari * + .1 cm dipartimento di fisica , universit di milano and infn , milan , italy cern - th/97 - 92 + may 1997
|
the origin of the electroweak symmetry breaking ( ewsb ) remains a mystery . in the standard model ( sm ) , the ewsb requires a negative mass squared for the higgs doublet scalar field , whose magnitude is set by hand .
we expect a fundamental theory which naturally gives the negative mass squared with the suitable value . in a model of supersymmetric extension of the sm
, the ewsb can be realized by so - called radiative breaking @xcite .
however , the supersymmetry breaking scale must be high because of no signal of super - particle at any experiments so far . in technicolor ( tc )
model @xcite , the higgs doublet field is no longer an elementary scalar field , and the ewsb is triggered by the techni - fermion condensation under strongly coupled tc gauge interaction .
however , the naive tc model , which is just scale up of qcd , has already been excluded by the electroweak precision measurements .
recently , there are a lot of studies of other possibilities to solve the gauge hierarchy problem by imposing a classically scale invariance with an additional @xmath4 gauge symmetry @xcite-@xcite . from the viewpoint of bardeen
s argument @xcite , we can only focus on logarithmic divergences , and the scale invariance protects large higgs mass corrections . under the classically scale invariance in terms of the cutoff regularization
, the quadratic divergence itself can be subtracted by a boundary condition of the uv complete theory @xcite .
once we subtract the quadratic divergence from the theory , it never appears in the observables . in the model with an additional @xmath4 gauge symmetry ,
the scale invariance is broken by the coleman - weinberg mechanism @xcite , and if the breaking scale is not so far from the electroweak ( ew ) scale , there is no gauge hierarchy problem . on the other hand
, a strong coupling dynamics can also realize such an ewsb with classically scale invariance @xcite , where an additional singlet scalar mediates dimensional transmutation in the strong coupling sector to the sm sector .
however , the sign of the coupling between the higgs doublet and the additional scalar is assumed to be negative , so that the negative mass squared of the higgs doublet is realized .
therefore , the origin of the ewsb is not necessary and inevitable in this scenario , and we are going to try the dynamical realization of negative mass squared by the bosonic seesaw mechanism @xcite . in this paper , we expand the sm gauge group by @xmath5 technicolor gauge symmetry with the classically scale invariant framework .
the techni - fermions , which belong to vector - like representations under tc gauge symmetry as well as electroweak gauge symmetry , are introduced .
though the chiral symmetry breaking happens by techni - fermion condensations , the ewsb does not happen by this strong coupling tc dynamics itself .
we show that the ewsb dynamically occurs in an inevitable way by the bosonic seesaw mechanism between the elementary higgs scalar field and a composite scalar field . to avoid massless nambu - goldstone ( ng ) bosons by the chiral symmetry breaking in strong coupling sector , we introduce a real pseudo - scalar singlet field and its interactions with techni - fermions and higgs doublet .
we analyze the mass spectrum of the pseudo - ng ( png ) bosons and estimate their decay rates .
we show that the png bosons can decay fast enough to avoid cosmological problems .
we further show that our model can be regarded as the first model of the flatland scenario with strong coupling dynamics .
all three coupling constants in scalar potential can vanish at the planck scale . similarly to the conventional flatland model with coleman - weinberg mechanism , the ew vacuum in our model is meta - stable .
by imposing classically scale invariance , the mass term of the higgs potential is forbidden and the higgs potential becomes @xmath6 the ewsb does not occur by this potential , and we try to use the dimensional transmutation in the strong coupling sector , where there are two vector - like techni - fermions as shown in tab . [
tab : model ] .
.charge assignments of techni - fermions and the higgs doublet .
[ cols="^,^,^,^",options="header " , ] due to the classically scale invariance , vector - like fermion masses are also forbidden . in the model , the chiral symmetry in the strong coupling sector @xmath7 is explicitly broken by the sm gauge symmetry , @xmath8 , and the remaining symmetry is @xmath9 .
there is also @xmath10 , which is similar to the baryon number symmetry .
this vector - like symmetry is expected to be unbroken by the strong - coupling technicolor dynamics due to the vafa - witten s theorem @xcite .
the chiral symmetry should be broken as preserving @xmath8 symmetry , then we expect @xmath11 and @xmath12 , @xmath13 .
they cause chiral symmetry breaking @xmath14 .
there are five ng bosons ; two massive png bosons of anomalous @xmath15 and @xmath16 breakings , and three massless ng bosons corresponding to the breaking @xmath17 symmetry .
if we neglect @xmath8 , the chiral symmetry breaking of @xmath18 occurs .
there are nine ng bosons ; one massive png boson of @xmath19 breaking , and eight massless ng bosons .
the techni - fermions interact with higgs doublet @xmath20 through the yukawa interactions , @xmath21 after the techni - fermion condensation , @xmath22 and @xmath23 are confined by non - perturbative effects , and @xmath24 and @xmath25 couple to a `` meson '' state , that is just a composite higgs doublet , @xmath26 . when @xmath27 and @xmath28 are real , there is the charge conjugation invariance . here
, we assume @xmath29 for simplicity . the yukawa interactions in eq.([eq : yukawa ] )
are cp invariant in this case .
the composite higgs doublet mixes with the elementary higgs doublet , and the mass matrix becomes @xmath30 where @xmath31 is a dimensionless positive coefficient of @xmath32 .
here , @xmath33 is assumed , since the chiral symmetry breaking terms should be small to be treated perturbatively .
as we will see later , the small yukawa coupling @xmath34 is also necessary for the hierarchy between the ew and tc condensation scales . as a result , the lighter ( heavier ) mass eigenstate @xmath35 ( @xmath36 ) is almost @xmath20 ( @xmath37 ) . the field @xmath35 is regarded as the sm - like higgs doublet , and the negative mass squared is dynamically obtained through the bosonic seesaw mechanism .
the field @xmath36 has mass of @xmath38 . there are massless ng bosons in the present stage . to avoid the massless ng bosons , we introduce real pseudo - scalar field , @xmath39 , which has interactions , @xmath40 where
@xmath41 and @xmath42 are taken to be real to keep the cp invariance . since we can expect @xmath43 and @xmath44 in vector - like technicolor dynamics , the no tadpole term of @xmath39 is not generated .
now the potential in eq.([eq : pot_ori ] ) is modified as @xmath45 where fourth and fifth terms are obtained from eq .
( [ eq : higgs_mass ] ) . since we have assumed the hierarchy between the light and heavy mass eigenstates , the heavier mass eigenstate @xmath36 is decoupled at low energies .
therefore , the effective potential at low energy is @xmath46 where we include the mass term of @xmath39 which is generated by bosonic seesaw mechanism again .
we will give an analysis about this issue shortly in the next section .
the vacuum expectation values ( vevs ) of @xmath35 and @xmath39 can be evaluated by the effective potential eq .
( [ eq : veff ] ) .
the stationary conditions are @xmath47 where @xmath48 and @xmath49 .
note that @xmath50 should corresponds to the ew scale ( @xmath51 ) , and nonzero @xmath52 causes spontaneous cp violation .
except for a trivial solution @xmath53 , there are three possibilities of solutions as follows .
* @xmath54 and @xmath55 + in this case the ew symmetry is not broken , while @xmath52 can be estimated as @xmath56 where @xmath57 must be positive .
to satisfy @xmath58 , i.e. , to realize the positive mass squared of @xmath35 , the following condition must be satisfied : @xmath59 thus , a certain large value of @xmath60 is required when @xmath52 is @xmath61 .
anyway , we do not consider this case , since the ew symmetry is unbroken .
* @xmath62 and @xmath63 + in this case the ewsb occurs and its scale is given by @xmath64 this is really a solution if the mass squared of @xmath39 is positive , that is , @xmath65 this condition is always satisfied for a sufficiently large @xmath60 .
since we would like to treat @xmath60 perturbatively in good approximation , we do not adopt this case also .
* @xmath62 and @xmath55 + this case leads a suitable result .
the stationary conditions give @xmath66 since the squared vevs must be positive , a certain small value of @xmath60 are required . in the limit of @xmath67 , the vevs
are approximately given by @xmath68 where @xmath57 must be positive . since @xmath39 obtains a nonzero vev , a mixing term with higgs doublet affects the higgs mass through @xmath69 .
however , it is negligible because @xmath60 is assumed to be sufficiently small .
( in the case of @xmath70 , we can treat @xmath35 and @xmath39 independently . ) from now on , we adopt this case with taking sufficiently small value of @xmath60 .
now let us investigate the mass spectra and decay rates of the png bosons . actually , all nine ng bosons should become massive due to the introduction of @xmath39 , since the chiral symmetries @xmath9 is explicitly broken down into @xmath71 by the interactions of eq .
( [ lag : pseudo ] ) .
the results of the mass spectra are summarized in tab . [
tab : tp ] .
masses + @xmath72 & @xmath73 & @xmath74 & @xmath75 & @xmath76 + ' '' '' @xmath77 & @xmath78 & @xmath74 & @xmath75 & @xmath76 + @xmath79 ( @xmath80 ) & @xmath81 & @xmath82 & @xmath75 & @xmath83 + ' '' '' @xmath84 & @xmath85 & @xmath86 & @xmath87 & @xmath88 + ' '' '' @xmath89 & @xmath90 & @xmath86 & @xmath91 & @xmath88 + first , we investigate png boson mass spectra . the sm singlet png bosons ( @xmath72 and @xmath77 ) mix with @xmath39 , and the mass matrix is written by @xmath92 where @xmath93 and @xmath94 are dimensionless positive coefficients of @xmath32 .
all off - diagonal elements are induced from eq .
( [ lag : pseudo ] ) .
the determinant of this mass matrix is @xmath95 , thus @xmath39 has a negative mass term .
taking @xmath96 , for simplicity , mass eigenvalues of @xmath39 , @xmath72 , and @xmath77 can be estimated as @xmath97 respectively .
the smallness of @xmath41 is natural , since it is expected to break the chiral symmetry perturbatively .
note that @xmath39 has the negative mass term by the bosonic seesaw mechanism again ( see eq .
( [ eq : veff ] ) . using dashen s formula @xcite and vevs of @xmath35 and @xmath39 in eq.([eq : vev_ew ] ) ,
the masses of @xmath98 and @xmath99 are estimated as @xmath100 \right ] | 0 \rangle \simeq \frac{8 \pi^2 g_s^4}{\lambda_s \beta^2 } \lambda_{\rm tc}^4\,,\\ m_{\sigma}^2 f_\sigma^2 { & \hspace{-0.6em}=\hspace{-0.6em } & } \langle 0 | \left [ q,\ , \left [ q,\ , \mathcal{h}_s \right ] \right ] | 0 \rangle \simeq \frac{g_s^4}{\lambda_s \beta^2 } \lambda_{\rm tc}^4\,,\end{aligned}\ ] ] where @xmath101 from eq .
( [ lag : pseudo ] ) and , @xmath102 and @xmath103 are decay constants of @xmath98 and @xmath99 , respectively .
both decay constants are evaluated by naive dimensional analysis @xcite as @xmath104 by analogy with qcd .
therefore , the masses of @xmath98 and @xmath99 are estimated as @xmath105 in the following , we take @xmath106 and @xmath107 for an explicit example . then , the coupling @xmath34 is evaluated as @xmath108 from eq .
( [ eq : vev_ew ] ) .
if we also take @xmath109 and @xmath110 , @xmath111 and the mass of @xmath98 and @xmath99 are evaluated as @xmath112 , @xmath113 and @xmath114 , respectively .
next , we estimate decay rates of the png bosons by analogy with light mesons in qcd
. a charged components of @xmath99 and @xmath98 can decay into their neutral components and the sm fermions through the weak interactions . @xmath72 and the neutral component of @xmath98 ( @xmath115 ) decay into two photons by analogy with @xmath116 decay in the sm .
the decay rate of @xmath72 is evaluated as @xmath117 where we have used @xmath118 and @xmath119 .
when we take @xmath120 for example , the decay rate is estimated by @xmath121 .
the neutral component of @xmath99 ( @xmath122 ) also decays into two photons via a mixing with @xmath72 .
the effective @xmath99-@xmath72 mixing is evaluated by dashen s formula as @xmath123 and hence , the magnitude of @xmath99-@xmath72 mixing is given by @xmath124 thus , we obtain @xmath125 by using the same numerical values as above . as a result , we find @xmath126 . the decay mode of the lightest neutral png boson @xmath77 is a little bit tricky .
the decay process is @xmath127 through mass mixings . therefore
, the lifetime of @xmath77 would be the longest among the png bosons . since @xmath39-@xmath77 and @xmath39-@xmath72 effective mixing couplings can be evaluated from eq .
( [ eq : light - mix ] ) as @xmath128 and @xmath129 , respectively , the decay rate can be estimated as @xmath130 thus , @xmath131 is around @xmath132 using the same numerical values as above .
even the lightest png boson can decay much faster than the qcd neutral pion ( @xmath133 ) . as a result
, we can expect that all the png bosons decay into the sm particles fast enough without cosmological problems .
our model is classically scale invariant . now , as an interesting possibility ,
let us consider whether the scalar potential vanishes at the planck scale .
this constraint is severer than the scale invariant condition , and all the scalar quartic couplings must be zero at the planck scale .
this situation has been studied in so - called flatland scenario @xcite , where a singlet scalar field has an interaction with the sm higgs doublet , and its vev induces negative mass squared of the higgs @xcite .
the mass scale is determined by the singlet vev , which is generated by coleman - weinberg mechanism .
this scenario can induce not only the vev of the scalar field but also the negative mass squared of the higgs doublet . on the other hand , another dimensional transmutation mechanism by using strong coupling dynamics has been proposed in @xcite .
these models do not follow the flatland scenario , since scalar couplings do not vanish at the planck scale . in these models
the ewsb is not necessary and inevitable , because we need to choose the sign of the coupling appropriately .
it is worth noting that our model can inevitably induce the negative mass squared of the higgs doublet by the strong coupling dynamics , and also all the scalar couplings can vanish at the planck scale simultaneously .
these are remarkable points .
the running of the scalar quartic couplings in our model are shown in fig .
[ fig_quartics ] . and the planck scale which are denoted by the black lines .
we take @xmath134 .
all quartic couplings are zero at the planck scale .
, width=302 ] here we have taken the top quark mass as @xmath135 as a reference value , which is slightly heavier than the central value from the collider experiment @xcite .
the curve of @xmath136 is concave down between @xmath137 gev and just below the planck scale , where @xmath138 in actually .
this means that there is a lower - energy vacuum than the ew one .
thus , imposing the vanishing potential at the planck scale makes the higgs quartic coupling negative below the planck scale , and the ew vacuum is meta - stable .
this situation is the same as the flatland model @xcite .
it is because the magnitude and energy scale dependence of @xmath139 around the planck scale are almost same as the sm .
although there seem to be some degrees of freedom in parameters of the bosonic seesaw mechanism , there is no room of adjustment , since the value of negative mass squared at the ew scale must be fixed to reproduce the sm .
thus our model can not escape from the meta - stable vacuum , but the ew vacuum is stable enough compared with the age of our universe @xcite .
our model is the first one which realizes the flatland scenario through the dimensional transmutation of the strong coupling dynamics , in which we can evaluate the energy dependence of the scalar couplings perturbatively .
note that the larger @xmath34 could make the vanishing potential realized with the central value of the top quark mass .
it leads smaller @xmath140 and lighter the png bosons .
the origin of the ewsb is not established yet , although the sm - like higgs boson has been discovered . in this paper
, we have investigated the dynamical origin of the ewsb via the bosonic seesaw mechanism in a classically scale invariant version of the sm .
we have introduced the @xmath5 technicolor gauge symmetry for the dimensional transmutation by the techni - fermion condensations . in this model
, the mixing between the elementary and composite higgs doublets becomes the origin of ewsb .
an extra real pseudo - scalar singlet field has also been introduced to avoid massless ng bosons .
we have estimated mass spectra and decay rates of the png bosons .
we have checked that all of the png bosons can decay fast enough without cosmological problems .
our model is the first model which realizes the flatland scenario through the dimensional transmutation of the strong coupling dynamics , in which we can evaluate the energy dependence of the scalar couplings perturbatively .
similarly to the conventional flatland model with coleman - weinberg mechanism , the ew vacuum in our model is meta - stable . finally , we comment on the collider phenomenology .
when the singlet pseudo - scalar is light enough to be produced at the collider , some vestiges could be searched in the future collider experiments .
in addition , since there can exist light new mesons depending on the parameters , they might be detectable at collider experiments .
this work is partially supported by scientific grant by the ministry of education , culture , sports , science and technology , nos .
24540272 , 26400253 , and 26247038 .
the works of n.h . and y.y .
are supported by scientific grant by the ministry of education , culture , sports , science and technology ( no .
15h01037 ) and research fellowships of the japan society for the promotion of science for young scientists ( no .
26@xmath1412428 ) .
s. weinberg , phys .
d * 13 * ( 1976 ) 974 , l. susskind , phys .
d * 20 * ( 1979 ) 2619 .
k. inoue , a. kakuto , h. komatsu and s. takeshita , prog .
* 68 * ( 1982 )
927 [ prog . theor .
* 70 * ( 1983 ) 330 ] .
r. hempfling , phys . lett .
b * 379 * ( 1996 ) 153 [ hep - ph/9604278 ] .
w. f. chang , j. n. ng and j. m. s. wu , phys .
d * 75 * ( 2007 ) 115016 [ hep - ph/0701254 [ hep - ph ] ] . s. iso , n. okada and y. orikasa , phys . rev .
d * 80 * ( 2009 ) 115007 [ arxiv:0909.0128 [ hep - ph ] ] .
s. iso , n. okada and y. orikasa , phys .
b * 676 * ( 2009 ) 81 [ arxiv:0902.4050 [ hep - ph ] ] .
n. okada and y. orikasa , phys .
d * 85 * ( 2012 ) 115006 [ arxiv:1202.1405 [ hep - ph ] ] .
s. iso and y. orikasa , ptep * 2013 * ( 2013 ) 023b08 [ arxiv:1210.2848 [ hep - ph ] ] .
c. englert , j. jaeckel , v. v. khoze and m. spannowsky , jhep * 1304 * ( 2013 ) 060 [ arxiv:1301.4224 [ hep - ph ] ] .
e. j. chun , s. jung and h. m. lee , phys .
b * 725 * ( 2013 ) 158 [ phys . lett .
b * 730 * ( 2014 ) 357 ] [ arxiv:1304.5815 [ hep - ph ] ] .
i. oda , phys .
b * 724 * ( 2013 ) 160 [ arxiv:1305.0884 [ hep - ph ] ] .
v. v. khoze and g. ro , jhep * 1310 * ( 2013 ) 075 [ arxiv:1307.3764 [ hep - ph ] ] .
m. hashimoto , s. iso and y. orikasa , phys .
d * 89 * ( 2014 ) 1 , 016019 [ arxiv:1310.4304 [ hep - ph ] ] .
m. lindner , d. schmidt and a. watanabe , phys .
d * 89 * ( 2014 ) 1 , 013007 [ arxiv:1310.6582 [ hep - ph ] ] . m. hashimoto , s. iso and y. orikasa , phys .
d * 89 * ( 2014 ) 5 , 056010 [ arxiv:1401.5944 [ hep - ph ] ] . s. benic and b. radovcic , phys .
b * 732 * ( 2014 ) 91 [ arxiv:1401.8183 [ hep - ph ] ] . v. v. khoze , c. mccabe and g. ro , jhep * 1408 * ( 2014 ) 026 [ arxiv:1403.4953 [ hep - ph ] ] .
s. benic and b. radovcic , jhep * 1501 * ( 2015 ) 143 [ arxiv:1409.5776 [ hep - ph ] ] .
h. okada and y. orikasa , arxiv:1412.3616 [ hep - ph ] .
j. guo , z. kang , p. ko and y. orikasa , phys .
d * 91 * ( 2015 ) 11 , 115017 [ arxiv:1502.00508 [ hep - ph ] ] .
p. humbert , m. lindner and j. smirnov , jhep * 1506 * ( 2015 ) 035 [ arxiv:1503.03066 [ hep - ph ] ] .
n. haba and y. yamaguchi , arxiv:1504.05669 [ hep - ph ] .
s. oda , n. okada and d. s. takahashi , phys .
d * 92 * ( 2015 ) 1 , 015026 [ arxiv:1504.06291 [ hep - ph ] ] .
humbert , m. lindner , s. patra and j. smirnov , arxiv:1505.07453 [ hep - ph ] .
a. d. plascencia , arxiv:1507.04996 [ hep - ph ] .
n. haba , h. ishida , n. okada and y. yamaguchi , arxiv:1508.06828 [ hep - ph ] .
w. a. bardeen , fermilab - conf-95 - 391-t , c95 - 08 - 27.3 . s. r. coleman and e. j. weinberg , phys . rev .
d * 7 * ( 1973 ) 1888 .
t. hur , d. w. jung , p. ko and j. y. lee , phys .
b * 696 * ( 2011 ) 262 [ arxiv:0709.1218 [ hep - ph ] ] ; t. hur and p. ko , phys . rev . lett .
* 106 * ( 2011 ) 141802 [ arxiv:1103.2571 [ hep - ph ] ] ; m. heikinheimo , a. racioppi , m. raidal , c. spethmann and k. tuominen , mod .
lett . a * 29 * ( 2014 ) 1450077 [ arxiv:1304.7006 [ hep - ph ] ] .
m. holthausen , j. kubo , k. s. lim and m. lindner , jhep * 1312 * ( 2013 ) 076 [ arxiv:1310.4423 [ hep - ph ] ] .
x. calmet , eur .
j. c * 28 * ( 2003 ) 451 [ hep - ph/0206091 ] , h. d. kim , phys .
d * 72 * ( 2005 ) 055015 [ hep - ph/0501059 ] , n. haba , n. kitazawa and n. okada , acta phys .
b * 40 * ( 2009 ) 67 [ hep - ph/0504279 ] .
o. antipin , m. redi , a. strumia and e. vigiani , jhep * 1507 * ( 2015 ) 039 [ arxiv:1503.08749 [ hep - ph ] ] . c. vafa and e. witten , phys .
* 53 * ( 1984 ) 535 .
r. f. dashen , phys .
* 183 * ( 1969 ) 1245 .
a. manohar and h. georgi , nucl .
b * 234 * ( 1984 ) 189 .
h. georgi and l. randall , nucl .
b * 276 * ( 1986 ) 241 .
[ atlas and cdf and cms and d0 collaborations ] , arxiv:1403.4427 [ hep - ex ] .
|
we propose a new dynamics of the electroweak symmetry breaking in a classically scale invariant version of the standard model .
the scale invariance is broken by the condensations of additional fermions under a strong coupling dynamics .
the electroweak symmetry breaking is triggered by negative mass squared of the elementary higgs doublet , which is dynamically generated through the bosonic seesaw mechanism .
we introduce a real pseudo - scalar singlet field interacting with additional fermions and higgs doublet in order to avoid massless nambu - goldstone bosons from the chiral symmetry breaking in a strong coupling sector .
we investigate the mass spectra and decay rates of these pseudo - nambu - goldstone bosons , and show they can decay fast enough without cosmological problems .
we further evaluate the energy dependences of the couplings between elementary fields perturbatively , and find that our model is the first one which realizes the flatland scenario with the dimensional transmutation by the strong coupling dynamics .
similarly to the conventional flatland model with coleman - weinberg mechanism , the electroweak vacuum in our model is meta - stable
. 0.6 cm su - het-13 - 2015 * a new dynamics of electroweak symmetry breaking with classically scale invariance * 1.2 cm naoyuki haba@xmath0 , hiroyuki ishida@xmath0 , noriaki kitazawa@xmath1 , and + yuya yamaguchi@xmath2 0.4 cm @xmath0graduate school of science and engineering , shimane university , + matsue 690 - 8504 , japan + @xmath1department of physics , tokyo metropolitan university , + hachioji , tokyo 192 - 0397 , japan + @xmath3department of physics , faculty of science , hokkaido university , + sapporo 060 - 0810 , japan 0.2 cm .5 in
|
the past decade has witnessed a tremendous resurgence of interest in the physics of quantum many - body systems out of equilibrium .
this growth has been partly motivated by remarkable developments in experiments , especially in experiments on cold atoms , which made the explicit observation of time evolution of many - body quantum systems possible @xcite .
in addition to the experimental motivation , fundamental conceptual issues , such as thermalization in isolated systems and the nature of adiabaticity , have also played a strong role in driving the development of this field .
a central paradigm in the study of non - equilibrium physics is the _ quantum quench_. a quantum quench involves a protocol starting from the ground state ( or a thermal state or another eigenstate ) of an ` initial ' hamiltonian @xmath0 , and then rapidly changing the hamiltonian so that nontrivial time evolution occurs under a different ( ` final ' ) hamiltonian @xmath1 .
many aspects of the dynamics induced by quantum quenches have been considered in recent years ; the literature is rapidly growing and already too vast to review here .
a quantum quench can lead to nontrivial dynamics as the initial state generally has overlap with many eigenstates of the final hamiltonian .
if @xmath2 and @xmath3 are the eigenstates and eigenvalues of the final hamiltonian , i.e. , @xmath4 , an initial state @xmath5 will lead to the following time dependence of the system wavefunction : @xmath6 if a large number of final eigenstates have non - negligible overlap @xmath7 , this sum can lead to highly nontrivial time dependence .
clearly , both the overlaps @xmath8 and the corresponding eigenenergies @xmath3 are important in determining the subsequent dynamics .
the object of study of the present paper is the distribution of overlaps , i.e. , the overlap magnitudes @xmath7 as a function of energy @xmath3 .
we will refer to this as the `` overlap distribution '' .
note that , if the squares @xmath9 were considered , this could be regarded as a discrete probability distribution , as @xmath10 and the squares @xmath9 represent probabilities .
since @xmath7 and @xmath9 provide equivalent information , we will focus on @xmath7 .
also note that the distribution of phases of @xmath11 may provide useful information but are not individually well - defined , since the phases of individual eigenstates @xmath2 can be chosen arbitrarily .
this work presents an investigation of the features of the overlap distribution for various quenches in a particular system . in this introductory section ,
we provide motivation for the study of the overlap distribution , by reviewing its connections to topics that have attracted widespread attention in recent years .
the overlap distribution is closely related to the statistics of the work performed in a quench , which is characterized by the so - called work distribution : @xmath12 where @xmath13 is the initial energy .
the work distribution is thus simply a continuous version of the distribution of squared overlaps , with the zero of the energy shifted by @xmath13 .
the work statistics in non - equilibrium processes in various quantum many - body systems is a topic of much recent interest @xcite .
another quantity of wide current interest , closely related to the overlap distribution and work distribution , is the return probability , @xmath14 this is a special case of the loschmidt echo where the reverse time evolution occurs under the hamiltonian @xmath0 , for which the initial state @xmath5 is an eigenstate .
following common practice , we will refer to this quantity simply as the loschmidt echo .
the behavior of the loschmidt echo after a quantum quench has been studied in a large body of recent literature , see , e.g. , @xcite . in particular , the loschmidt echo is central to the study of so - called dynamical phase transitions @xcite . by expanding @xmath5 in terms of the eigenstates @xmath2 of the final hamiltonian @xmath1
, we find @xmath15 knowledge of the overlap distribution , i.e. , the @xmath7 s and corresponding @xmath3 s , thus leads directly to the loschmidt echo .
for example , the energy width of the distribution determines the coefficient of the initial quadratic decay of @xmath16 , as it can be seen by expanding @xmath17 in eq .
to second order in @xmath18 @xcite . after a quench ,
a generic observable @xmath19 has the following time evolution : @xmath20 the long - time behavior of such expectation values , @xmath21 , is of fundamental interest since this is closely related to the question of whether an isolated quantum system thermalizes in some sense . as outlined above , this long - time limit is determined solely by the overlap magnitudes and the diagonal matrix elements .
this is often called the `` diagonal ensemble '' value of the long - time limit @xcite .
the eigenstate thermalization hypothesis ( eth ) , which is expected to hold for non - integrable systems , postulates that the diagonal matrix elements @xmath22 are smooth functions of eigenenergy , for large enough systems @xcite .
the question of thermalization can then reduce to the question of how narrow in energy the overlap distribution is .
recent work ( such as the `` quench action '' approach @xcite and an approach based on linked - cluster expansions @xcite ) has formulated calculations of long - time expectation values directly in the thermodynamic limit , thus avoiding the explicit calculation of overlaps in finite - size systems . the diagonal ensemble formulation of eq . for finite sizes , involving the overlap distribution @xmath7 , will play an essential role in the discussion of such calculations for long - time stationary state values of observables , in particular for comparing the ordering of the limiting procedures ( limits of large sizes and large times ) .
we also note that the overlaps @xmath11 appear not only in the long - time limit , but also in the arbitrary - time expression for @xmath23 ( cf .
eq . ) , although in this case the phases of @xmath11 are also important . from the discussion above ,
it is clear that the overlap distribution plays a central role in many aspects of quench dynamics .
numerical data on overlap distributions have appeared in many different non - equilibrium studies for specific quenches , e.g. , in refs .
@xcite where the overlap distributions are called `` local density of states '' , in refs .
@xcite in the context of thermalization , in ref .
@xcite in connection with intermediate - time behaviors of the loschmidt echo , and in ref .
@xcite for a spin spiral initial state .
there have been , however , relatively few systematic studies of the overlap distribution so far .
the present work is a step towards addressing this gap in the literature .
we provide a detailed study of the overlap distribution for a well - known model system , namely the one - dimensional anisotropic spin-1/2 heisenberg model ( xxz chain ) at zero magnetization .
we consider quenches of the anisotropy parameter @xmath24 , see eq . , and provide an overview of the overlap distributions obtained in quenches from various initial @xmath25 values to various final @xmath26 values , using full numerical diagonalization .
for the special case of @xmath27 and @xmath28 , i.e. , quenching from the nel state to the so - called xx point , we present a full analytical study of the overlap distribution .
the xx point is mappable to a chain of free fermions , which allows this detailed analysis .
we find that all nonzero overlaps are exactly equal and derive expressions for them , using the slater determinant structure of the xx eigenfunctions , as well as using the bethe ansatz .
we derive the selection rules determining the eigenstates that have nonzero overlaps for this quench , both in the language of fermionic momenta and using the language of bethe roots .
we show how these results on the overlap distribution can be used to derive explicit expressions for the loschmidt echo @xmath16 and for equal - time correlators after the quench . for quenches ending at small but nonzero anisotropy , @xmath29
, a splitting structure appears in the overlap distribution ; we analyze the main features of this structure using perturbation theory .
similarly , for quenches to the xx point starting from large but finite initial value @xmath30 , a perturbative expansion in @xmath31 explains the structure of the overlap distribution . after introducing the model and some notation in sec .
[ sec : model ] , we provide our numerical overview of overlap distributions in quenches from various @xmath25 to various @xmath26 in sec .
[ sec : numerical ] .
section [ sec : b ] is devoted to a detailed analysis of the @xmath32 case .
we also use the results for the overlap distribution to derive analytical expressions for time - dependent quantities , in particular , for the loschmidt echo and for longitudinal and transverse two - point correlators .
section [ sec : a ] outlines a perturbative treatment of the @xmath33 case ( perturbation in @xmath26 ) , while sec .
[ sec : c ] describes a perturbative treatment of the @xmath34 case ( perturbation in @xmath31 ) .
this work focuses on the one - dimensional anisotropic spin-1/2 heisenberg model ( xxz chain ) with nearest - neighbor interactions .
the hamiltonian is given by @xmath35 where @xmath36 , @xmath37 , with @xmath38 being the usual pauli matrices , and @xmath39 .
we will refer to the @xmath40 part of the hamiltonian as the xx hamiltonian : @xmath41 . throughout this manuscript
we choose the number @xmath42 of lattice sites to be even .
we also distinguish between two different types of boundary conditions . for an open chain we set @xmath43 , for a periodic chain @xmath44 , @xmath45 .
the xxz chain is a natural generalization of the isotropic heisenberg chain @xcite .
it is one of the most widely studied models in quantum magnetism in particular , and condensed matter physics in general .
it is absolutely central to the study of integrability , having a relatively simple integrable structure , yet possessing rich physics and displaying a phase transition . in the current era of non - equilibrium physics
, it has been widely used as a model system to explore and exemplify non - equilibrium phenomena , including studies of the loschmidt echo ( e.g. @xcite ) , studies of quenches of the anisotropy parameter @xmath24 ( e.g. @xcite ) , and of quenches starting from a nel state ( e.g. @xcite ) .
dynamics starting from various non - uniform states ( such as domain wall initial states ) , leading to propagation or transmission along the xxz chain , is also a rapidly growing topic of investigation ( see , e.g. , @xcite ) .
the xxz chain is also being increasingly used to study real - time thermal transport and real - time dynamics at finite temperatures .
it is a primary testbed for developments and applications of bethe ansatz techniques for addressing non - equilibrium issues ( e.g. @xcite ) .
we therefore find the xxz chain to be an appropriate model for a systematic study of the work distribution in quantum quenches . in this work
, we restrict to positive anisotropies @xmath46 .
the isotropic critical point @xmath47 separates two regimes with different behaviors ; in terms of low - energy physics , the @xmath48 regime is gapless , while the @xmath49 regime is gapped .
the hamiltonian commutes with the @xmath50-component @xmath51 of the total spin operator . for
the zero magnetization sector contains the ground state for all @xmath49 .
our initial states are ground states of the xxz chain and , hence , lie in the zero magnetization sector .
thus , we can restrict to the subspace spanned by energy eigenstates @xmath2 with @xmath52 . at @xmath40
the chain can be mapped onto a gas of free fermions via the jordan - wigner mapping , and subsequently solved using free - fermion techniques . for generic values of the anisotropy parameter
@xmath24 the xxz model is integrable and can be solved by means of bethe ansatz techniques @xcite . in the fermionic language
the anisotropy @xmath24 acts as the strength of nearest - neighbor interactions
. a particular focus of this work will be on quenches starting from the purely antiferromagnetic regime ( @xmath53 ) . in the limit @xmath53
the initial hamiltonian effectively becomes the ising hamiltonian , @xmath54 .
the ground state of @xmath55 is exactly degenerate , with both symmetric and antisymmetric combinations of the nel states having the same energy .
for @xmath25 large but finite this degeneracy is lifted .
we denote the two nel states as @xmath56 for @xmath27 , any linear combination of @xmath57 and @xmath58 is a valid ground state and , in principle , any such linear combination could be chosen as the initial state with which to calculate overlaps .
physically , we find it most reasonable to choose the symmetric combination for @xmath59 even and the antisymmetric combination for @xmath59 odd , which is adiabatically connected to the ( unique ) ground state for @xmath25 large but finite , @xmath60 for the periodic chain we set @xmath61 with @xmath62 . since @xmath63 , where @xmath64 is the translation operator by one lattice site , @xmath65 is nothing but the momentum of the initial state . for the open chain , @xmath66
is related to the symmetry of @xmath67 under reflection , @xmath68 maps lattice site @xmath69 to @xmath70 for all @xmath71 .
recently , some progress has been made in calculating overlaps using the bethe ansatz @xcite .
there are now exact formulas for the overlap between a nel state and any eigenstate in terms of the bethe roots describing the eigenstate @xcite . by taking the thermodynamic limit of the results of ref .
@xcite , it was possible to investigate the long - time limit in quenches to the antiferromagnetic gapped phase @xmath72 @xcite and also to the isotropic point @xmath73 @xcite , using the quench action approach @xcite . however , extracting all the overlaps for large but finite chains by means of the exact formulas of refs .
@xcite is technically quite challenging .
for this reason , in this work we will use these forumlas only in the limit @xmath74 , where the structure of bethe roots is well understood .
in this section we present an overview of the features of overlap distributions for quenches from various @xmath25 to various @xmath26 , obtained using numerical exact diagonalization .
we present overlaps with only those eigenstates of the post - quench hamiltonian @xmath1 that lie in the zero magnetization sector .
since the initial states are always in this sector , eigenstates of @xmath1 in other magnetization sectors have zero overlap due to symmetry . in cases where eigenstates of @xmath1 are ( nearly or exactly ) degenerate
, numerical diagonalization might give overlaps with arbitrary linear combinations of the ( nearly ) degenerate eigenstates .
in all cases where we discuss visible features of the overlap distribution , we have taken care to resolve such degeneracies .
there are more degeneracies for periodic chains , so we show mainly open - chain data .
we have checked in each case that the qualitative features are very similar for open and periodic chains . as explained in sec .
[ sec : model ] , for quenches starting from @xmath27 we choose as the initial state @xmath75 the combination @xmath76 of nel states as defined in eq . .
overlap distributions for @xmath27 to @xmath28 quenches are shown in fig .
[ fig : infty2zero ] , for both open and periodic chains .
the remarkable feature of these distributions is that the overlaps are either zero or have exactly equal nonzero value .
there is no reason to expect that the ground state at @xmath77 should have any particular preference between low - energy and high - energy eigenstates of the @xmath40 hamiltonian .
nevertheless , such a physical argument does not suffice to predict that the nonzero overlaps should all be exactly equal .
this is a particular feature of the nel - to - xx quench in the xxz chain . .
data are shown for both open and periodic chains , for two different system sizes .
there are @xmath78 nonzero overlaps , all exactly equal to @xmath79 which is @xmath80 for @xmath81 and @xmath82 for @xmath83 . in the periodic cases some of the nonzero overlaps are not visually distinguishable because of degeneracies , e.g. , in panel ( c )
there are four degenerate nonzero overlaps at @xmath84.,scaledwidth=60.0% ] among the @xmath85 final eigenstates in the zero magnetization sector , only @xmath78 have nonzero @xmath11 s .
the number of nonzero overlaps fixes the nonzero values to be @xmath86 , by virtue of normalization and the fact that they are all equal .
the counting can sometimes be tricky due to degeneracies .
for example , there appears to be only five instead of eight nonzero overlaps in the @xmath81 periodic case in fig .
[ fig : infty2zero](c ) .
this is because the symbol at the center represents four exactly degenerate eigenstates , which can not be distinguished .
there are also few eigenstates with the same eigenenergy but zero overlap .
of course , any linear combination of the degenerate eigenstates is also a valid eigenstate .
so , there is some ambiguity in the overlaps plotted . by choosing the eigenstates to be the appropriate ` slater determinants ' in fermionic language ( see sec .
[ sec : b ] ) , the overlaps turn out to be either zero or @xmath86 . the overlap distribution for quenches from @xmath27 to @xmath28 will be analyzed in detail in sec .
[ sec : b ] , in particular we will derive the selection rules , which determine whether or not an eigenstate has nonzero overlap .
we will also exploit this analysis to obtain closed expressions for time - dependent quantities .
we now consider quenches starting from the @xmath76 state ( @xmath27 ) to various nonzero values of @xmath26 .
( initial state @xmath87 ) in an open chain with @xmath83 sites .
the splitting that can be observed for @xmath29 disappears as @xmath26 grows .
( b ) magnitude of the average normalized shifts for the two branches as function of @xmath26 ( double logarithmic scale ) and as function of system size @xmath42 ( linear scale ) .
the normalized shift for small @xmath26 is observed to be linear in @xmath26 and @xmath42.,scaledwidth=80.0% ] in fig .
[ fig : smalldelta](a , c , d ) , we show overlap distributions for quenches to @xmath26 values in the gapless region , @xmath88 . when @xmath26 is very small , the distribution is only slightly distorted from the @xmath28 case , as expected from perturbative arguments .
however , the pattern of distortion is rather remarkable .
the @xmath78 nonzero overlaps now split into two well - defined branches ; one branch is shifted upwards and one branch is shifted downwards , see fig .
[ fig : smalldelta](a ) .
the upward- ( resp .
downward- ) shifted branch is biased toward lower ( resp .
higher ) energies . in particular , among the eigenstates satisfying the selection rule ,
the lowest few all have positive shift , the highest few always have negative shift , and the eigenstates with intermediate energies can have either positive or negative shift .
in addition , many of the overlaps that were zero for the @xmath28 case , now have nonzero ( albeit tiny ) @xmath7 values
. we will refer to the difference @xmath89 of the absolute value of the nel overlap with an eigenstate of the @xmath90 hamiltonian and the nel overlap at @xmath28 , i.e. , @xmath79 , as the `` shift '' of the overlap .
the splitting structure is well - defined as long as the shift is much smaller than @xmath79 ; otherwise the lower branch is difficult to distinguish from the states with near - zero overlap .
it is useful to normalize the shift with respect to the @xmath28 value , i.e. , for each eigenstate connected to one having nonzero overlap at @xmath28 , we define @xmath91 in fig .
[ fig : smalldelta](b ) , the magnitudes of the normalized shifts are averaged over the eigenstates in each branch .
we see this to be linear with @xmath26 at fixed @xmath42 , and linear with @xmath42 at fixed @xmath26 , i.e. , proportional to @xmath92 .
this indicates that the shift can be calculated perturbatively in @xmath26 for any finite chain .
however as the chain size gets larger the perturbative region of @xmath26 shrinks . in sec .
[ sec : a ] we will analyze the splitting and shift using first - order perturbation theory .
the branches have internal structures , e.g. , the overlap values roughly increase with energy within each branch . for a fixed @xmath42 , as @xmath26 increases further , the splitting structures gets diffuse , and the separation between the @xmath93 larger values and the rest smaller values also gets obscured , see fig .
[ fig : smalldelta](c ) . eventually , there is no visible remnant of the characteristic structure at @xmath28 , as seen for @xmath94 in fig .
[ fig : smalldelta](d ) . for @xmath94
the overlap distribution has no striking pattern , and no obvious remnant of the flat distribution at @xmath28 .
a slight bias toward smaller energies is visible .
this may be expected because , for any finite @xmath95 , the @xmath27 ground state @xmath76 should have more weight in the lower part of the spectrum than in the upper part .
this trend gets more prominent at larger @xmath26 . in fig .
[ fig : largedelta ] we show some examples for @xmath72 .
the bias toward lower - energy states gets progressively stronger .
this indicates that for quenches from @xmath27 to @xmath72 , the dynamical evolution will be determined mostly by low - energy states .
for @xmath96 the ground state dominates the overlap distribution very strongly . and various @xmath26 in an open chain with @xmath83 sites .
in the last panel we see how for large values of @xmath26 the energy spectrum is organized in ` bands'.,scaledwidth=80.0% ] and various @xmath26 in an open chain with @xmath83 sites . in the last panel
we see how for large values of @xmath26 the energy spectrum is organized in ` bands'.,scaledwidth=99.0% ] and various @xmath26 in an open chain with @xmath83 sites . in the last panel
we see how for large values of @xmath26 the energy spectrum is organized in ` bands'.,scaledwidth=90.0% ] we now consider quenches to @xmath28 , starting from @xmath30 large but finite .
the initial state @xmath5 is now no longer exactly the combination @xmath76 of nel states . from perturbative intuition
, we expect that as @xmath25 decreases toward smaller values the overlap distribution will deviate gradually from the characteristic shape of fig .
[ fig : infty2zero ] .
this is seen to be true in fig . [
fig : fromlarge ] , where we have plotted overlap distributions for an open chain with @xmath97 , for three different @xmath30 .
the @xmath78 nonzero values exhibit a roughly linear dependence on @xmath3 as @xmath25 is decreased from @xmath98 .
one expects the ground state at any finite @xmath25 to be closer to the low - energy eigenstates of the @xmath28 than to the higher - energy eigenstates .
hence , it is not surprising that the slope of the line is negative .
the set of overlaps fall on a very nearly linear curve for large @xmath25 . for moderate @xmath25 , see fig .
[ fig : fromlarge](c ) , the overlaps follow a more curved line and show more scatter .
in addition the near - zero overlap values start to be noticeable .
the shift of the ground state overlap from @xmath79 can be seen numerically to scale as @xmath99 for a given @xmath42 and large @xmath25 , see fig .
[ fig : fromlarge](d ) .
the overlap of the highest - energy eigenstate obeying the selection rules scales in the same way .
also , the normalized shift for the ground state scales linearly with chain length @xmath42 for a given @xmath25 . in sec .
[ sec : c ] , treating the xx part of the hamiltonian as a perturbation to the ising hamiltonian , we will show that the shift of the overlaps is exactly a linear function of energy at first order in @xmath31 .
in fact , the normalized shift for the eigenstates satisfying the selection rule will be shown to be @xmath100 at first order ( exact for periodic chains , approximate up to order @xmath101 for open chains ) , consistent with the numerical data .
we now consider quenches starting from @xmath102 .
thus , the initial state is the ground state of a free fermionic hamiltonian .
examples for three @xmath26 values are shown in fig .
[ fig : zero2various ] for an @xmath83 open chain . for any finite value of @xmath26
, we expect a bias toward low - energy states .
this is seen in all three cases .
the bias is naturally stronger for smaller @xmath26 .
we see that up to @xmath103 the ground state overlap dominates the overlap distribution .
other than the ground state , most of the weight for @xmath104 goes to a few final eigenstates at relatively low energies .
we also see from the logarithmic plots in the insets of fig .
[ fig : zero2various](a , b ) that there is a characteristic decaying shape .
the upper envelope of this scatter - plot decays roughly exponentially , although a sharp qualitative statement about the decay rate is difficult to make , due to finite - size effects .
the @xmath105 case has some bunching structure reflecting the fact that the energy spectrum near the ising limit is grouped into ` bands ' .
each band is characterized by the number of domain walls in the configurations that dominate the eingenstates of that band .
accordingly , the overlap distribution is broken into groups .
interestingly , within each band the overlap distribution is dominated by the lower - energy part of that band . moreover , each band of the distribution has roughly similar shape as the characteristic shape of the full distribution for the @xmath104 case ( see figs . [
fig : zero2various](a , b ) ) .
the inset to fig .
[ fig : zero2various](c ) exemplifies this by zooming into one of the bands .
in this section , we will present a series of analytical results on the quench from @xmath27 to @xmath28 , i.e. , the quench from the combination @xmath76 of nel states to the xx chain .
this is the case addressed numerically in sec .
[ sec : infty_to_zero ] and fig .
[ fig : infty2zero ] . in sec .
[ sec : freefermions ] , we set notation by providing a brief reminder of how the xx hamiltonian is diagonlized by mapping to jordan - wigner fermions . in sec .
[ sec : selection ] , we use the fermionic language to derive selection rules for an xx eigenstate to have nonzero overlap with @xmath76 .
the rules are expressed in terms of the momenta ( periodic chain ) or pseudo - momenta ( open chain ) of the fermions .
physically , one rule arises from the structure of the nel states @xmath106 , and a second rule arises from the well - defined symmetry of the combination @xmath76 .
we also express the overlap magnitudes as a sum of determinants ( one for each @xmath106 ) .
this is evaluated as a function of the chain length @xmath42 , both by exploiting determinant properties , and from normalization , using the fact that all the nonzero overalps are equal for this quench . in sec .
[ sec : ba ] , we use the overlap formulas of refs .
@xcite ( valid for all @xmath26 ) and take the limit @xmath107 . in this way we formulate selection rules in the language of bethe roots .
we show how these rules are equivalent to the rules in terms of fermionic ( pseudo-)momenta . finally , we provide two applications of our detailed understanding of the overlap distribution in this particular case . in sec . [
sec : loschech ] , we derive closed expressions for the loschmidt echo after the quench . in sec .
[ sec : correlators ] , we present closed expressions for the full time dependence of equal - time two - point correlators .
we consider both longidudinal ( @xmath108 ) and transverse ( @xmath109 ) correlators .
the expressions for the transverse correlators appear here for the first time , to the best of our knowledge . to diagonalize @xmath110 , i.e. , the @xmath40 part of the hamiltonian , one uses the jordan - wigner ( jw ) transformation @xmath111 here @xmath112 and @xmath113 are usual fermionic operators , satisfying anticommutation relations @xmath114 , @xmath115 . for periodic boundary conditions
, the jw transformation turns @xmath110 to the free fermionic hamiltonian @xmath116 , with the boundary condition @xmath117 , where @xmath118 is the number of fermions in the state . using a fourier transformed set of operators ,
@xmath119 @xmath120 can be written in diagonal form @xmath121 the boundary conditions enforce quantization of the momenta @xmath122 : @xmath123 where @xmath124 for @xmath125 even , and @xmath126 for @xmath125 odd .
for the open chain , the jw transformation leads to @xmath127 .
this hamiltonian is diagonalized be introducing the canonical operators @xmath128 leading to @xmath129 provided that the numbers @xmath130 are quantized as @xmath131 since there is no translational invariance in the open chain , we call @xmath130 `` pseudo - momenta '' . in terms of the jordan - wigner fermions introduced above
, our initial state can be written as @xmath132 where @xmath133 is the vacuum state ( all spins down ) .
the eigenstates of the xx hamiltonian in the zero - magnetization ( @xmath134 fermion ) sector can be written for the periodic ( resp .
open ) chain as @xmath135 with @xmath136 ( resp .
@xmath137 ) for all @xmath138 , and @xmath139 ( resp .
@xmath140 ) for @xmath141 .
the overlap between an eigenstate of the form and the initial state @xmath76 from eq . can now be computed using wick s theorem .
for example , with periodic boundary conditions , we obtain @xmath142 \\ \label{eq : wick2 } & = \frac{1}{\sqrt{2}}\left[\det_{1\leq j,{\alpha}\leq l/2}\left(\langle0|c_{2j-1}^{\phantom{\dag}}\tilde{c}_{k_{\alpha}}^\dag|0\rangle\right ) + \sigma\det_{1\leq j,{\alpha}\leq l/2}\left(\langle 0|c_{2j}^{\phantom{\dag}}\tilde{c}_{k_{\alpha}}^\dag|0\rangle\right)\right]{\,.}\end{aligned}\ ] ] equation follows from eq . by separate application of wick s theorem on the two correlators . using definition this
can be simplified to @xmath143 { \,.}\label{eq : overlappbc } \intertext{for the open chain we similarly obtain } \label{eq : overlapobc } \langle \psi_\sigma|\{q_{\alpha}\}_{{\alpha}=1}^{l/2}\rangle & = \frac{1}{\sqrt{2}}\left(\frac{2}{l+1}\right)^{l/4}\left[\det_{1\leq j,{\alpha}\leq l/2}\big(\sin\big(q_{\alpha}(2j-1)\big)\big ) + \sigma\det_{1\leq j,{\alpha}\leq l/2}\big(\sin\big(2q_{\alpha}j\big)\big)\right ] { \,.}\end{aligned}\ ] ] the overlaps and can be zero if both determinants vanish , or if their sum exactly cancels .
as discussed below , it turns out that both cases can occur .
we first analyze the formula for the periodic chain .
we set @xmath61 and simplify the square bracket in eq . , @xmath144 now , if there is a pair of the form @xmath145 , the determinant in eq .
is zero as two columns of the matrix are equal , @xmath146 for all @xmath69 .
the set @xmath147 of all possible momenta , defined in eq . , decomposes into @xmath59 many of such pairs : @xmath148 with @xmath124 for @xmath59 even and @xmath126 for @xmath59 odd .
in order to get a non - vanishing determinant , each pair in this decomposition has to be occupied by exactly one momentum .
this shows that there are only @xmath93 states for which the determinant can be nonzero .
if we flip one of the momenta by @xmath149 , i.e. , choosing the other member of the pair ( ` pair flip ' ) , the determinant does not change , again due to @xmath150 .
therefore , the determinant has always the same non - vanishing value for all of these @xmath93 states .
note that the overlaps might differ by a factor @xmath151 due to the different order of momenta in pair - flipped states . the other momentum - dependent factor in eq .
is @xmath152 , which is zero if the total momentum @xmath153 of the eigenstate @xmath154 and the momentum @xmath65 of the initial state @xmath67 differ by @xmath155 ( modulo @xmath156 ) .
this means that only an even number of pair flips are allowed .
hence , the number of states that have nonzero overlap gets reduced by a factor of @xmath157 . thus , there are in total @xmath78 eigenstates @xmath154 , all with the same overlap ( up to a possible global minus sign as dicussed above ) . since the initial state @xmath76 is normalized , this common value is @xmath158 the derivation above did not require any explicit determinant calculation .
we also checked that the result can be recovered by directly computing the determinant in eq . .
the vandermonde determinant formula @xmath159 allows to express it as a double product that can be further simplified .
crucial in the evaluation is the formula @xmath160 , which can be established by noticing that @xmath161 and @xmath162 .
the analysis of the overlap formula for the open chain is similar .
if there is a pair of the form @xmath163 , both determinants are zero as two columns are proportional to each other : @xmath164 or @xmath165 , respectively .
the set @xmath166 of all possible pseudo - momenta , see eq .
, decomposes into @xmath59 many of such pairs : @xmath167 again , in order to get non - vanishing determinants , each pair in this decomposition has to be occupied by exactly one pseudo - momentum .
if we flip one of them by @xmath149 , i.e. , doing a pair flip , the second determinant does not change , while the first one picks up a minus sign .
this shows that there are again @xmath93 states for which the determinants are nonzero and that they all have the same absolute value .
the two determinants are equal for the state where all pseudo - momenta are smaller than @xmath168 and also for all states that can be obtained from this state by an even number of pair flips .
in contrast , for states obtained by an odd number of pair flips they differ by a factor @xmath151 .
if @xmath169 denotes the number of pair flips , i.e. , the number of pseudo - momenta larger than @xmath168 , the overlap is proportional to @xmath170 , which is zero for @xmath171 and @xmath169 odd or @xmath172 and @xmath169 even , and it is @xmath173 otherwise .
this supplemental restriction also follows from the reflection symmetry of the initial state .
since the reflection operator @xmath174 commutes with the hamiltonian , the energy eigenstates are also eigenstates of @xmath174 , i.e. @xmath175 .
hence , @xmath176 and the overlap with @xmath67 can only be nonzero if @xmath177 .
( in the fermionic language the action of @xmath174 on a many particle state can be deduced from the formula @xmath178 , where @xmath179 is the total fermion number . applying this to the nel state we recover the same rule . ) therefore , as in the periodic case , the number of states that have nonzero overlap gets reduced by a factor of @xmath157 due to the reflection symmetry .
again , there are @xmath78 eigenstates with overlap @xmath180 note that this result is in agreement with the ground - state overlap of ref .
@xcite , obtained by an explicit determinant calculation .
all other eigenstates ( those with a pair @xmath181 or those with the wrong number of momenta larger than @xmath168 ) have zero overlap with @xmath67 .
let us summarize our findings .
we have a simple algorithm that allows us to determine which post - quench eigenstates have nonzero overlap with the initial state @xmath76 and , therefore , contribute to the time evolution and to the non - equilibrium dynamics after the quench . for that to happen two conditions need to be fulfilled : 1 .
[ rule1]the first condition ( ` forbidden pairs ' ) selects the states for which the determinants are individually nonzero , i.e. states with momenta or pseudo - momenta satisfying @xmath182 2 . [ rule2]the second condition ( ` number of pair flips ' ) selects the states for which the two terms in eqs . and do not cancel .
this corresponds to selecting eigenstates with the same symmetry ( translation or reflection ) as the initial state . in the periodic case
this restriction can be written as @xmath183 , which relates the total momentum of the eigenstate to the momentum @xmath65 of the initial state .
for the open chain the restriction is imposed by the reflection symmetry of the initial state , @xmath184 , via @xmath185 , where @xmath169 is the number of pseudo - momenta larger than @xmath168 .
note that , if we use @xmath186 or @xmath187 as initial state , only the first rule [ rule1 ] above applies and there are @xmath93 nonzero overlaps .
the second rule [ rule2 ] arises because @xmath76 respects the translation / reflection symmetry of the hamiltonian . in this subsection
, we treat the overlap distribution for the @xmath188 to @xmath189 quench using the bethe ansatz formalism .
recent work has provided expressions for the overlap between nel states and bethe eigenstates at arbitrary @xmath190 @xcite .
we specialize the formulas of ref .
@xcite to the xx limit ( @xmath191 , i.e. @xmath192 ) .
we will show that the condition of having nonzero overlap with an xx bethe state can be visualized by adjacent corners of certain rectangles in the pattern of possible xx bethe roots .
we further relate this condition to the selection rule of single particle momenta .
the description here is mostly self - contained , but the interested reader may wish to consult refs .
@xcite for background and details on related matters . for simplicity
we focus on the case @xmath59 even .
we start our analysis by presenting the solutions of xx bethe equations with twist and constant inhomogeneity ( see , e.g. , @xcite ) .
the twist deformation is related to twisted boundary conditions of the spin chain , @xmath193 , @xmath194 .
it is necessary to introduce a nonzero twist to resolve issues with singular terms in the norm formula for bethe states in the limit @xmath192 .
we then use ` twist - deformed ' solutions to calculate overlaps by means of the aforementioned formulas .
the bethe equations for anisotropy @xmath195 with twist @xmath196 and constant inhomogeneity @xmath197 read @xmath198^l = -{\mathrm e}^{{\mathrm i}\phi } \prod_{j'=1}^{m}\frac{\sinh(\lambda_j -\lambda_{j'}+{\mathrm i}\gamma)}{\sinh(\lambda_j-\lambda_{j'}-{\mathrm i}\gamma)}{\,,}\qquad j = 1,\ldots , m { \,.}\label{eq : betwist}\ ] ] we consider @xmath199 small . for @xmath200 and @xmath201
even , the right hand side simplifies to @xmath202 , which eventually leads to @xmath203 where @xmath204 are integers . for small @xmath196 and @xmath197
, we have @xmath205 for @xmath206 , and @xmath207 or @xmath208 for @xmath209 .
here , @xmath210 and @xmath211 are the numbers of bethe roots with imaginary part zero and @xmath168 , respectively .
we consider the set @xmath212 of four possible bethe roots , corresponding to the numbers @xmath169 , @xmath213 , @xmath214 , and @xmath215 with @xmath216 . for small @xmath196 we can expand them in the small parameter @xmath217 . setting @xmath218\right)$ ]
we obtain @xmath219 \lambda_4 = -\lambda - \delta + \epsilon { \ , , } & \lambda_1 = \lambda - \delta + \epsilon { \,.}\end{array}\ ] ] these four approximate solutions can be considered as a deformation of the rectangle with corners @xmath220 . in the formulas for the overlaps of ref .
@xcite valid also for the so - called ( unnormalized ) off - shell bethe states we first send @xmath221 and afterwards @xmath222 .
it can be shown that overlaps of the nel state with normalized bethe states identified by a set of bethe roots @xmath223 containing an opposite pair of the form @xmath224 and/or @xmath225 are of order @xmath197 .
thus , if there are three or four bethe roots in one rectangle , there is at least one opposite pair and the overlap is zero in the limit @xmath226 .
therefore , to have non - vanishing overlap , there can be at most two xx bethe roots in each rectangle .
since there are in total @xmath59 bethe roots and @xmath227 rectangles , all rectangles are occupied exactly by two bethe roots sitting at two adjacent corners ( ` rectangle condition ' ) . in fig .
[ fig : rectangles ] we show an example for which the rectangle condition is fulfilled . the overlap formula for bethe states @xcite consists of two parts , a prefactor and a ratio of two determinants .
it can be shown that the latter is exactly one in the xx limit , if the rectangle condition is fulfilled , zero otherwise .
the remaining part is the prefactor that simplifies in the xx limit to @xmath228 { \,.}\ ] ] here , the bethe roots @xmath229 , @xmath230 , belong to a solution of the untwisted homogeneous bethe ansatz equations , each of them representing one of the @xmath227 rectangles . due to the @xmath155-periodicity of the tangent function we have @xmath231 = 1 $ ] , and
finally obtain @xmath232 here , ` otherwise ' means that all @xmath227 rectangles are occupied exactly by two roots sitting at adjacent corners , see fig . [
fig : rectangles ] for an example .
the rectangle condition allows for four possibilities for each rectangle and hence , in total , for @xmath233 many states , each contributing to the sum of absolute squares of overlaps with a constant @xmath234 .
this can be translated into the language of momenta of xx bethe states .
the momentum assigned to a bethe root , which can be interpreted as a single particle momentum , is given by @xmath235{\,.}\label{eq : momentum}\ ] ] the rule that eigenstates with bethe roots in opposite corners are forbidden translates into the rule that the corresponding momenta can not differ by @xmath149 , which is exactly the selection rule . .
the four ` rectangles ' are indicated by the dashed lines .
the ` rectangle condition ' is fulfilled , leading to a nonzero overlap with the nel state .
each collection of four possible roots with equal @xmath236 contains exactly two roots sitting on two adjacent corners of the rectangle.,scaledwidth=60.0% ] in this section we use our findings for the overlap distribution in the quench @xmath237 to compute the loschmidt echo ( le ) defined in eq . .
we start from the definition of the loschmidt echo and exploit the selection rules of sec .
[ sec : selection ] to obtain an explicit product formula . for the sake of simplicity
we only discuss open chains , but the periodic case is very similar .
we start by considering the case with only one of the nel states @xmath186 or @xmath187 , before deriving the result for the initial state @xmath238 .
recall that the le is defined as @xmath239 where the hamiltonian @xmath240 of the time evolution operator is given in this subsection by the open chain hamiltonian of the free system .
expanding @xmath241 in a basis of energy eigenstates @xmath2 , we obtain @xmath242 where @xmath3 is the many - body energy and @xmath243 . as in sec .
[ sec : selection ] , we use the sets of single particle momenta @xmath244 to label the eigenstates @xmath2 . from secs .
[ sec : selection ] and [ sec : ba ] we know that the overlap distribution is flat and that if we take @xmath245 or @xmath246 as initial state , only selection rule [ rule1 ] of sec .
[ sec : selection ] must be fulfilled , namely @xmath247 for all indices @xmath248 and @xmath249 . in terms of the single particle energies
@xmath250 , the previous condition for the momenta can be interpreted in the following way : @xmath251 since the total energy of the state is @xmath252 , eq . can be rewritten as @xmath253 where @xmath254 is a sum over all possible states @xmath255 , i.e. , a sum over all subsets @xmath256 , restricted to the states / subsets for which eq .
since half of the possible single particle energies are positive , while the others are their negative counterparts , we can rewrite eq .
as @xmath257 in this way condition is automatically satisfied , since it is impossible , when expanding the product , to have in a single term the same energies with opposite signs .
therefore , for the open chain the le corresponding to one of the nel states @xmath106 is given by the simple formula @xmath258 { \,.}\label{eq : loschech3}\ ] ] we note that similar formulas have been derived for periodic chains in refs .
our result for the open chain can be also obtained by expressing the le as a determinant and then computing it explicitly .
this is however unnecessary .
as we have seen it can be obtained from simple counting arguments based on the selection rules of sec .
[ sec : selection ] .
we now consider @xmath76 as initial state .
the le is still given by a formula similar to eq .
, @xmath259 where the prime at the sum now means that both selection rules [ rule1 ] and [ rule2 ] of sec .
[ sec : selection ] must be fulfilled .
in addition to eq . , the set of pseudo - momenta have to satisfy @xmath260 , where @xmath169 is the number of pseudo - momenta with negative energy .
equation can then be recast as @xmath261 a simple way to prove this is to expand the two products in eq . , and check that all states that do not satisfy the selection rules [ rule1 ] and [ rule2 ] cancel .
therefore , the le corresponding to @xmath76 is given by @xmath262 \;+\ ; { \mathrm i}^{l/2}\prod_{m=1}^{l/2 } \sin^2\left[t\cos\left(\frac{\pi m}{l+1}\right)\right ] \right|^2\ ] ] both results and are plotted in fig .
[ fig : loschech ] for several system sizes @xmath42 . corresponding to the nel initial state @xmath106 ( left ) , see eq . and to the initial state @xmath76 ( right ) , see eq . .
for the two smaller sizes , a comparison with loschmidt echoes numerically calculated by means of exact diagonalization shows exact agreement with the derived closed formulas.,scaledwidth=82.0% ] we now provide a second application of the results on overlap distributions in sec . [
sec : selection ] for the quench @xmath237 .
we will compute the full time - dependence of finite distance two - point correlation functions after such a quench . for simplicity we will restrict to the periodic chain and list results which are derived in the large-@xmath42 limit .
we provide only a skeleton outline of the derivation , pointing out how the selection rules are used .
the time evolution occurs under the @xmath263 hamiltonian @xmath110 .
the equal - time two - point correlators are defined as @xmath264 we refer to them as the longitudinal and transverse two - point correlators , respectively .
we consider as initial state @xmath5 the four different cases @xmath265 , @xmath266 and the two nel states @xmath57 and @xmath58 themselves . since the state @xmath267 is an eigenstate of the translation operator @xmath64 , the two - point correlators only depend on the distance between the two points , @xmath268 . applying the jordan wigner mapping we obtain @xmath269 c_{j+n}^{\phantom\dag}\rangle{\,,}\end{aligned}\ ] ]
where from now on the time dependence is implicit , i.e. , the expectation values are taken between the time evolved states , @xmath270 with @xmath271 .
in the first line we used that the initial state @xmath5 lies in the zero magnetization sector and therefore @xmath272 for all times @xmath18 . in the second line
we wrote the ` jordan - wigner string ' operators as @xmath273 , which can be written as @xmath274 it turns out that the first two terms cancel up to order @xmath101 , so that the last term dominates in the large @xmath42 limit . thus , the transverse correlator will involve correlators of products like @xmath275\left[\prod_{m=1}^{n}c_{q_m}^{\phantom\dag}\right]$ ] , where @xmath276 and @xmath277 are sets of @xmath169 distinct single - particle momenta . as an example , the correlator @xmath278 is a momentum sum over products of phase factors , which can be rewritten as @xmath279\left[\prod_{m=1}^{n}c_{p_{a(m)+\pi}}^{\phantom\dag}\right]|\psi_\sigma\rangle & = \sum_{\{k\},\{k'\}}\langle \psi_\sigma|\{k\}\rangle\langle \{k'\}|\psi_\sigma\rangle\;\ ; \langle \{k'\}|\left[\prod_{m=1}^{n}c_{p_m}^{\dag}\right]\left[\prod_{m=1}^{n}c_{q_{m}}^{\phantom\dag}\right]|\{k\}\rangle \notag\\ & = ( -1)^{\frac{n(n-1)}{2 } } \left(\frac{1+(-1)^n}{2}\right ) \frac{(-1)^{|a|}}{2^n}{\,,}\label{eq : matrix_elements_transverse}\end{aligned}\ ] ] the second line has been obtained using the selection rules of sec . [
sec : selection ] .
since @xmath169 momenta of the eigenstates @xmath154 that obey condition are fixed , there are @xmath280 states with overlap square @xmath281 .
this gives the factor @xmath282 .
the factor @xmath283 comes from the additional selection rule ( ` number of pair flips ' ) for the total momentum of the eigenstates @xmath284 and @xmath285 .
hence , the transverse two - point correlator at points @xmath69 and @xmath286 is only nonzero for @xmath169 even . here , @xmath287 is the mapping matrix from the set @xmath276 to the set @xmath277 , given by an element of the symmetric group @xmath288 .
the factor @xmath289 implies that the result is a determinant . carrying all factors and simplifying ,
we eventually obtain ( @xmath290 ) @xmath291 here , we have defined the quantity @xmath292 similarly , one obtains for the nel states themselves @xmath293 in the derivation of the formulas for @xmath294 some @xmath101 terms were dropped .
however , we have found that formulas and for @xmath278 are exact for @xmath295 and for any finite even @xmath42 .
the longitudinal correlators are simpler to derive , since they have no jordan - wigner strings .
the result is @xmath296 for all four initial states @xmath297 . in the thermodynamic limit , the summation in eq .
becomes the integral that defines the @xmath298-th order bessel function , @xmath299 thus , the above formulas for @xmath294 and @xmath300 are also valid in the thermodynamic limit , with the substitution @xmath301 . previously , ref .
@xcite presented ( connected versions of ) longitudinal correlators in the thermodynamic limit . to the best of our knowledge
, the determinant formulas for the more involved transverse correlators have not appeared previously .
quenches from the initial state @xmath76 ( @xmath27 ) to small but nonzero values of @xmath26 display a ` splitting ' of the overlap distribution , as discussed in sec .
[ sec : numeric_infty2various ] and shown in figs .
[ fig : smalldelta](a , b ) . in this section
, we will use first order perturbation theory in @xmath26 to characterize this splitting . for the final hamiltonian @xmath302
, we consider @xmath303 as a perturbative potential @xmath304 .
unfortunately , the unperturbed hamiltonian @xmath305 is highly degenerate .
in particular , among the states which obey the selection rules , there are many which are degenerate .
hence , a full calculation requires the use of degenerate perturbation theory for many of the eigenstates .
however , our purpose here is to provide a basic understanding of the splitting and the general form of the overlap distribution for small nonzero @xmath26 .
hence , it will suffice to concentrate on those states which have non - degenerate energy eigenvalues , and use the much simpler non - degenerate perturbation theory .
we will show that this treatment recovers the main features observed numerically . for the sake of simplicity we consider a periodic chain with @xmath59 even , which means that in the zero magnetization sector there is an even number @xmath134 of fermions in the system .
all the calculations can be trivially extended to the odd particle case . in terms of the jordan - wigner fermions of eq . , the perturbative potential @xmath306 can be written as @xmath307 where @xmath308 .
we can drop the constant term and the terms proportional to the particle number @xmath309 since they represent just shifts of the energy and are irrelevant for a perturbative expansion . using again momentum operators defined in eq .
the only relevant contribution can be written as @xmath310 carrying out the sum over the spatial indices we find the condition @xmath311 , which is nothing but the momentum conservation .
after these considerations we can rewrite the perturbative potential for the periodic chain as @xmath312 let us now formally write the expression for the corrected overlap . given @xmath313 the ` corrected eigenstate ' with @xmath314 , the latter being an eigenstate of the xx chain that obeys the selection rules ( ` pair condition ' and zero total momentum ) , we obtain @xmath315 where the states @xmath316 and @xmath317 are eigenstates of the free fermionic hamiltonian . here , we only choose states @xmath284 for which the energy value @xmath318 is non - degenerate , such that the energy difference in the denominator is never zero .
normalizing with @xmath319 we obtain @xmath320 where we used that the determinant under the sum is either zero or always given by the same absolute value @xmath321 .
the prime at the sum means that it is restricted to the ` nonzero overlap states ' .
the perturbative potential connects states @xmath284 and @xmath285 in which just two momenta are exchanged . due to the selection rule we have that @xmath322
therefore , the energy difference in the denominator can be written as @xmath323 furthermore , for the numerator , we obtain after a short calculation @xmath324 the sign of @xmath325 is fixed by the order of momenta of the states @xmath284 and @xmath285 .
it exactly cancels the sign coming from the determinant that we pulled out of the sum in eq . .
the relative minus sign in front of the sum in eq .
is due to the factor @xmath326 in eq . and comes from the fact that the momenta @xmath327 , @xmath328 and @xmath329 , @xmath330 differ by @xmath149 . in summary , we find @xmath331 follwoing the definition in sec .
[ sec : numeric_infty2various ] , we call the quantity in eq . the `` normalized shift '' and the numerator of the left - hand side just the `` shift '' . in fig .
[ fig : perturbative_shift ] we plot the perturbative results for the nonzero overlaps for @xmath332 and @xmath333 .
( we plot the predictions for actual overlaps instead of the normalized shifts . )
the perturbative treatment only covers non - degenerate states .
hence , a quantitative comparison with the numerical data of fig .
[ fig : smalldelta](a ) is not appropriate .
however , non - degenerate perturbation theory clearly reproduces the main features of the data .
figure [ fig : perturbative_shift ] shows the overall shape to be very similar to that seen in the inset to fig . [ fig : smalldelta](a ) , with two branches each having an overall upward slope .
the perturbative approach explains the shift being proportional to @xmath26 ( by providing a non - vanishing contribution at first order in @xmath26 , see above ) .
in addition , we will show in the subsection below that the result implies that the smallest and largest normalized shifts grow linearly with @xmath42 , so that the normalized average branch shifts grow linearly with @xmath42 , as found numerically , see fig .
[ fig : smalldelta](b ) . the perturbative prediction in fig .
[ fig : perturbative_shift ] also shows some substructures , e.g. , each branch is futher split ( recursively ) into sub - branches .
it is unclear whether such substructures might survive when the degenerate states are taken into account , or when higher - order corrections are included .
in addition , substructures in the xxz spectrum are often heavily determined by boundary conditions @xcite .
so , they can be expected to be different for periodic and open boundary conditions .
a detailed study of such substructures is beyond the scope of the current work . to small @xmath26 ,
obtained from first - order non - degenerate perturbation theory .
the specific values plotted here are for @xmath332 and @xmath333 .
only non - degenerate states are included .
the dashed line depicts the unperturbed ( @xmath28 ) value @xmath79 .
the four horizontal lines ( dots and crosses ) indicate the extremal values of the shift acoording to eqs . and .,scaledwidth=65.0% ]
in this section we use a perturbative approach to explain one of the features of the overlap distribution that can be observed in the quench protocol from @xmath25 large but finite to the free fermion point @xmath28 . as it can be seen in fig .
[ fig : fromlarge ] , the nontrivial part of the overlap distribution is rather smooth as a function of @xmath3 , and decreases roughly linearly with @xmath3
. the slope increases with decreasing @xmath25 .
this behavior is in sharp contrast to the splitting structures of the @xmath33 case . in the following
we will use perturbation theory in @xmath31 to show that the normalized shift is proportional to @xmath3 at first order ( for periodic chains ) , which explains the smooth behavior of the @xmath7 s as a function of @xmath3 .
we will also compute the quadratic correction by making use of second order perturbation theory . for simplicity
we consider a periodic chain . in the case of an open chain
the main steps of the following calculation can be similarly performed , but there are complications because of the edges .
the initial state @xmath75 , which is the ground state at large but finite @xmath25 , is adiabatically connected to the state @xmath76 , which we have used as the initial state for @xmath27 : @xmath351 in order to make the perturbative calculation mathematically stringent , we divide the initial hamiltonian ( and henceforth all corresponding energies ) by @xmath25 .
we consider the second term in @xmath352 as a small perturbation . in order to analyze the overlaps @xmath353 , where latexmath:[$ final hamiltonian with @xmath28 , perturbatively , we expand the ground state @xmath75 of @xmath0 in the small parameter @xmath31 , @xmath355 the zeroth - order ( unperturbed ) wavefunction is @xmath356 .
the energy eigenvalue @xmath357 of @xmath76 is two - fold degenerate at zeroth order and the corresponding other state is simply @xmath358 . since the two states @xmath358 and @xmath5 ( with @xmath66 fixed ) are both eigenstates of the translation operator with opposite eigenvalues @xmath359 and @xmath66 , we deduce @xmath360 .
this means that @xmath358 does not appear in any expansion of @xmath361 .
we can therefore apply non - degenerate perturbation theory .
using @xmath362 the higher order corrections read @xmath363 where @xmath364 is the eigenvalue of @xmath365 with eigenstate @xmath366 .
we first compute the matrix elements in the numerator of eq .
: @xmath367 where @xmath368 are the spaces spanned by the states @xmath369 , @xmath71 , i.e. , the space spanned by the spin configurations obtained via a single neighboring spin pair flip from a nel state .
the energies @xmath364 are all the same , independent of the choice of @xmath370 and given by @xmath371 .
therefore , @xmath372 for all @xmath373 and , henceforth , @xmath374 thus , the first - order term is obtained by application of @xmath375 on the initial state ( a similar trick was used in ref .
@xcite ) . using that the states @xmath316 are eigenstates of the hamiltonian @xmath305 with energy @xmath376 we obtain up to first order @xmath377 the second order term is a bit more involved .
the first term in eq . can be computed as @xmath378 for each fixed state @xmath379 there are of order @xmath380 many terms for which the flipped spins do not connect to each other and for which the energy difference is simply @xmath381 .
in contrast , there are only of order @xmath42 many terms for which the energy difference is not given by this value .
we drop these terms because they are subleading in @xmath42 . since the second and the third term in eq .
are also negligible with respect to the leading contribution , we obtain @xmath382 the normalized overlap therefore reads , up to second order in @xmath31 and by neglecting the @xmath101 contributions , @xmath383 this provides a second order correction for the nontrivial part of the overlap distribution in the quench protocol from @xmath30 to @xmath28 . note
that the energy @xmath376 is extensive and bounded by @xmath384 .
so , if the condition @xmath385 is fulfilled , we expect the perturbation series to converge .
the data for open chains from numerical exact diagonalization , see fig .
[ fig : fromlarge ] , is described reasonably well by the large-@xmath42 perturbative results derived in this section for periodic chains .
motivated by the observation that the overlap distribution is important for many aspects of quench dynamics , in this work we have considered a prominent paradigm model , the xxz chain , and undertaken a systematic study of the properties of the overlap distribution for quenches of the anisotropy parameter @xmath24 . using numerical exact diagonalization ,
we have provided a broad overview of the features of the overlap distribution , for quenches from various values @xmath25 to various values of @xmath26 .
in particular , we have thoroughly analyzed the quench from @xmath27 to @xmath28 , i.e. , when the initial state is of the form ( combination of nel states ) and the post - quench hamiltonian is described by a free - fermion theory through jordan - wigner transformation .
the overlap distribution in this case has the remarkable feature of being ` flat ' : all nonzero overlaps have the same value . using the free - fermion structure of the post - quench eigenstates , it was possible to analyze the overlap distribution analytically , to express the overlaps using determinant formulas , and to explicitly calculate their values ( see sec .
[ sec : b ] ) .
the states that have nonzero overlap can be chosen using certain ` selection rules ' .
one rule arises from the spatial structure of the nel states , while a second rule arises from the symmetry of the initial state , in terms of translation ( periodic boundary conditions ) or reflection ( open boundary conditions ) .
the ` selection rules ' for the periodic chain have been compared with results that we derived from recently obtained bethe ansatz formulas @xcite .
we found that these rules can be equivalently expressed in terms of adjacent corners of rectangles in the pattern of possible xx bethe roots .
we also provided two example applications of our detailed understanding of the overlap distribution for the @xmath237 quench . using the selection rules
, we were able to provide concise expressions for the loschmidt echo ( return probability ) as well as equal - time two - point correlators after the quench .
in particular , time - dependent transverse correlators , @xmath386 , are somewhat nontrivial due to jordan - wigner strings .
we provide determinant formulas for such correlators , derived by making use of the selection rules . for quenches from @xmath27 to small @xmath95
, the overlap distribution shows a splitting structure which we analyzed perturbatively ( perturbation in @xmath26 , see sec .
[ sec : a ] ) . for quenches from large @xmath387 to @xmath28 ,
the overlap distribution acquires a simple dependence on energy which we analyzed using perturbation theory in @xmath31 , see sec .
[ sec : c ] .
the present work is related in various ways to recent advances in non - equilibrium physics .
many recent publications have provided overlap distributions for specific quenches .
this motivates systematic studies like the present one . using the bethe ansatz ,
formulas are now available @xcite for the overlap between nel states and bethe eingestates at any @xmath24 .
since the use of these results requires knowledge of the bethe roots describing all eigenstates , it is technically challenging to explore overlap distributions using this approach .
our numerical overview and detailed explicit analysis of specific cases thus provides specific contexts for these general technical results .
our applications to loschmidt echoes and correlators can be compared with existing calculations of the loschmidt echo @xcite and of longitudinal correlators @xcite ) .
in contrast to these approaches , we have used the structure of the overlap distribution to derive time - dependent quantities .
this has allowed us to obtain some new expressions .
finally , certain overlaps with low energy eigenstates in the gapless phase @xmath388 can be studied using field - theoretical methods . in particular , subleading terms in such overlaps @xcite allow to access universal data about the underlying conformal field theory .
our focus here , in contrast , has been on the full spectrum , not just low - energy physics .
this work opens up many new questions .
first of all , the present study is by no means exhaustive .
there are features of the overlap distribution in the xxz chain which remain to be analyzed in detail .
for example , the physical nature of the eigenstates which have large overlap for @xmath102 to @xmath26 of order one quenches ( or vice versa ) , are unclear and might deserve further study .
second , in addition to the applications we have provided for the cleanest case ( @xmath237 ) . it might be possible to exploit the observations on overlap distributions in other quench protocols ( other @xmath25 and @xmath26 ) to extract properties of the time dependence of loschmidt echoes and correlation functions . finally , it would certainly be interesting to analyze overlap distributions in a series of other models .
one could ask , for example , whether integrability plays a general role in determining the features of the overlap distribution .
we thank jacopo viti for useful discussions .
v.a . acknowledges support by the erc under the starting grant 279391 edeqs .
|
the dynamics after a quantum quench is determined by the weights of the initial state in the eigenspectrum of the final hamiltonian , i.e. , by the distribution of overlaps in the energy spectrum .
we present an analysis of such overlap distributions for quenches of the anisotropy parameter in the one - dimensional anisotropic spin-1/2 heisenberg model ( xxz chain ) .
we provide an overview of the form of the overlap distribution for quenches from various initial anisotropies to various final ones , using numerical exact diagonalization .
we show that if the system is prepared in the antiferromagnetic nel state ( infinite anisotropy ) and released into a non - interacting setup ( zero anisotropy , xx point ) only a small fraction of the final eigenstates gives contributions to the post - quench dynamics , and that these eigenstates have identical overlap magnitudes .
we derive expressions for the overlaps , and present the selection rules that determine the final eigenstates having nonzero overlap .
we use these results to derive concise expressions for time - dependent quantities ( loschmidt echo , longitudinal and transverse correlators ) after the quench .
we use perturbative analyses to understand the overlap distribution for quenches from infinite to small nonzero anisotropies , and for quenches from large to zero anisotropy .
|
the investigation of resonant light - matter interaction has received considerable attention during the last decades and remarkable experimental progresses in this field has opened the door to whole new experimental opportunities @xcite .
these developments are not only interesting from a fundamental point of view but also from practical points of view for possible applications in the area of quantum information processing , for example . in this
latter context methods for achieving an optimal transfer of quantum information between single photons ( flying qubits ) and elementary material systems ( stationary qubits ) over large distances by using the resonant strong coupling , as required for the realization of a quantum repeater , are of particular interest .
therefore , recently considerable efforts @xcite have been devoted to investigate the interaction of matter qubits with few modes of the radiation field within the framework of jaynes - cummings - paul models @xcite .
recent developments stimulated extensions of this work to extreme multimode scenarios or to scenarios involving structured continua of electromagnetic field modes which are characteristic for half open cavities of a parabolic shape @xcite , for example . in the following we review some of these recent research activities focusing on the resonant matter - field interaction of two - level atoms located around the focal point of a parabolic shaped cavity and
generalize them to prolate ellipsoidal cavities which include parabolic cavities as limiting cases .
a particularly interesting elementary quantum electrodynamical situation arises in a prolate ellipsoidal cavity if two two - level atoms are trapped in the two foci of this cavity and exchange a single photon which is emitted spontaneously by one of these two - level atoms .
in our model we investigate the dynamics of @xmath0 identical two - level atoms with transition frequencies @xmath1 and dipole matrix elements @xmath2 .
these two - level atoms are assumed to be situated in the focal points of a prolate elliptical cavity @xmath3 and to interact with the quantized radiation field inside this cavity .
furthermore , we assume that the sizes of the relevant electronic states two - level atoms are small compared to the wavelengths @xmath4 of the atomic transition so that the dipole approximation is applicable ( @xmath5 is the speed of light in vacuum . ) . moreover , as we are considering atomic transitions in the optical frequency regime the relevant coupling ( rabi- ) frequencies are small in comparison with @xmath1 so that the rotating - wave approximation is applicable .
thus , the quantum electrodynamical interaction between @xmath6 two - level atoms and the quantized electromagnetic field inside the cavity can be described by the hamiltonian @xmath7 with @xmath8 , @xmath9 @xmath10 and with the dipole operator @xmath11 of the two - level atom @xmath12 .
thereby , the energies of the atomic ground states @xmath13 have been set equal to zero . in the schrdinger picture the positive frequency part of the ( transversal ) electromagnetic field operator
is given by @xmath14 and @xmath15 denote the normalized transversal electric mode functions which fulfill the boundary conditions for an ideally conducting cavity .
for arbitrary boundary conditions the calculation of the mode functions @xmath15 is generally difficult .
however , in cases in which the wave lengths of relevant field modes are small in comparison with all other length scales of the problems this task may be facilitated by semiclassical methods . in order to put the problem into perspective
let us consider cases in which the dipole matrix elements @xmath16 of the atomic transitions are oriented along the symmetry axis , i.e. the z - axis , of a parabolic or ellipsoidal cavity .
thus , we can write @xmath17 .
this symmetry can be exploited as it is possible to show that in these cases the only field modes which couple to the two - level atoms in the dipole approximation are of the form @xcite @xmath18 thereby , @xmath19 are the symmetry adapted parabolic coordinates in the case of a parabolic cavity and the prolate ellipsoidal coordinates in the case of an ellipsoidal cavity . for a parabolic cavity the functions @xmath20 and
@xmath21 turn out to be coulomb functions and for an ellipsoidal cavity they turn out to be prolate spheroidal wave functions . in both cases
we can use semiclassical methods to express the quantization conditions resulting from the boundary conditions of an ideally conducting wall , for example , semiclassically in a simple way by taking advantage of the separability of the helmholtz equation in these coordinates .
this way one obtains a single quantization function for one of the coordinates , say @xmath22 , in the case of a parabolic cavity and two quantization conditions in the case of an ellipsoidal cavity .
furthermore , these semiclassical methods are also well suited for normalizing these mode functions appropriately .
we can now use these results to calculate the time evolution of the system .
our hamiltonian has the property that the number of excitations in the system ( so @xmath23 ) is a conserved quantity .
if we assume that initially only one two - level atom is in an excited state and that the field is in the vacuum state we know that for all later times the number of excitations in the system equals unity .
this leads to the following ansatz for the wave function @xmath24 @xmath25 in the case of two two - level atoms in the cavity .
( the superscripts @xmath26 and @xmath27 refer to atoms and photons , respectively . )
it is a straight forward task to generalize the following calculation to an arbitrary number of atoms .
the schrdinger equation leads to a coupled system of linear differential equations . in order to solve this system
we apply the laplace transform in order to obtain a system of linear algebraic equations . due to basic properties of our hamiltonian
it is straight forward to eliminate the laplace transforms of the photonic excitations @xmath28 ( @xmath29 denotes the parameter of the laplace transform ) in order to obtain a linear system of equation for @xmath30 and @xmath31 .
this system of equations is given by @xmath32\left(\begin{array}{c } \widetilde{b^{1}}(s)\\ \widetilde{b^{2}}(s ) \end{array}\right)\ ] ] with @xmath33 the functions @xmath34 encode all the properties of the cavity .
it is a non trivial task to evaluate these functions because they are defined by an infinite sum over all modes @xmath35 which couple to the atomic dipoles .
however , we can evaluate these functions with the help of the semiclassical methods developed previously . with these functions at hand we can solve the linear system of equations and apply the inverse laplace transform in order to obtain expressions for the time evolution of the probability amplitudes of observing atoms @xmath36 and @xmath37 in the excited states , i.e. @xmath38 and @xmath39 .
however , evaluating the inverse laplace transform is still a non trivial task . for this purpose
it is convenient to solve the linear system of equations by applying the neumann series in order to obtain an infinite sum of simple expressions for @xmath40 and @xmath41 and then apply the inverse laplace transform to each of these terms .
the neumann series is given by @xmath42^{n}}{(s+\frac{\gamma_{\text{free}}}{2}+{{\ensuremath}{\mathrm{i}}}\omega_{eg})^{n+1}}\cdot\left(\begin{array}{c } b^{1}(0)\\ b^{2}(0 ) \end{array}\right).\ ] ] thereby , we have defined @xmath43 and @xmath44 denotes the spontaneous decay rate of each two - level system originating from spontaneous emission of a photon in free space , i.e. in the absence of any boundary conditions for the radiation field . of course
, at first sight it may appear as a major complication to apply the neumann series in order to invert a 2 by 2 matrix .
however , this expansion leads to a semiclassical photon path representation of the relevant probability amplitudes @xcite so that we are able to interpret each term of this expansion as a contribution of a photon path .
thereby , each photon path involves a sequence of spontaneous emission and absorption processes . in this expansion the functions @xmath34 encode the possible paths a photon can take to reach atom @xmath45 starting from atom @xmath12 . in case of a parabolic cavity we can perform the same calculations but instead of a two by two matrix of the functions @xmath34 we only have to consider the single term @xmath46 .
we can use our results of the preceding chapter to calculate the influence of the purcell effect @xcite on the spontaneous decay rate . in the pole or weisskopf - wigner @xcite approximation the spontaneous decay rate of atom @xmath12
is given by @xmath47\bigr\vert_{s=-{{\ensuremath}{\mathrm{i}}}\omega_{eg}}\;.$ ] thus , for a parabolic cavity we obtain the result @xcite @xmath48\\ \times\frac{2m\mathcal{s}(u)\coth\left[2m\mathcal{s}(u)\right]-1}{\sinh^{2}\left[2m\mathcal{s}(u)\right]}\bigr\vert_{u=2\pi\frac{f}{\lambda_{eg}}}\end{aligned}\ ] ] with @xmath49 and with @xmath50 denoting the focal length of the parabola .
each summand in the expression for @xmath51 is connected with a possible photon path inside the cavity , i.e. with bouncing back and forth of a photon ( counted by @xmath52 ) between the atom and the vertex of the parabola .
these contributions to the decay rate are connected with diffraction effects which are suppressed for large values of @xmath53 .
similar but more complicated results are obtained for an ellipsoidal cavity .
the main difference between both cases is that for an ellipsoidal cavity we obtain a double sum which can be traced back to the existence of two quantization conditions .
of course , the results for a parabolic cavity can be obtained from the more general results for an ellipsoidal cavity in the limit in which the distance between the two focal points tends to infinity . in figure
[ fig : purcell ] the semiclassical analytical results are compared with exact numerical results .
this numerical comparison demonstrates the quality of the semiclassical results even in cases in which the wave length of the spontaneously emitted photon becomes comparable to the characteristic length scales inside the cavity .
thus , the previously discussed semiclassical approximation also enables us to investigate diffraction effects which become relevant for @xmath54 . as a function of the scaled focal length @xmath55 of the parabola : numerically exact results ( dots ) , semiclassical results ( solid ) .
we also indicate the free - space value as a horizontal line .
, title="fig:",width=207 ] + as a function of the scaled focal length @xmath55 of the parabola : numerically exact results ( dots ) , semiclassical results ( solid ) .
we also indicate the free - space value as a horizontal line . , title="fig:",width=207 ] and @xmath56 ( a ) and @xmath57 ( b ) . the excitation probability of atom @xmath36 ( @xmath37 ) is represented by the blue ( red ) curve.,title="fig:",width=207 ] + and @xmath56 ( a ) and @xmath57 ( b ) .
the excitation probability of atom @xmath36 ( @xmath37 ) is represented by the blue ( red ) curve.,title="fig:",width=207 ]
we can now investigate the time evolution of atomic excitations . in this chapter
we are going to present results for a prolate ellipsoidal cavity . for this purpose
we need some additional quantities for describing the cavity , namely the distance @xmath58 between the focal points and the shortest distance @xmath50 between one of the vertices and its closest focal point . in the following we concentrate on the regime of short wavelengths , i.e. @xmath59 , in order to obtain a simple picture of the dynamics .
if we start with an initial excitation of atom 1 we observe a bouncing back and forth of the atomic excitation between the two atoms inside the cavity by exchange of a spontaneously emitted photon .
the time @xmath60 the excitation needs to travel from one atom to another is given by @xmath61 .
the corresponding behavior of the atomic excitation probabilities is depicted in figures [ fig : atomic_excitation_probabilities ] a and b. we observe that the probability of a successful transfer of the excitation to the second atom depends significantly on the shape of the cavity . if @xmath62 , i.e. the cavity is almost spherically symmetric , the excitation probabilities associated with repeated returns of a photon to an atom are significantly larger than for cases with @xmath63 .
this can be observed by comparing the excitation probabilities depicted in figures [ fig : atomic_excitation_probabilities ] a ) and [ fig : atomic_excitation_probabilities ] b ) .
this phenomenon may be traced back to the fact that both atomic dipole matrix elements are oriented along the z - axis . in the case of @xmath63 the radiation field of the photon emitted spontaneously by the first atom
has to propagate almost along the z - axis in order to approach atom @xmath37 .
therefore , its electrical field is almost perpendicular to the z - axis and its coupling to the dipole of atom 2 is small . as a consequence
the excitation probability is also small . in the opposite limit of an almost spherically symmetric cavity such a confinement of the photonic wave packet along the symmetry axis due to the shape of the cavity
does not occur and the excitation probabilities are larger .
we can also calculate the energy density of the electromagnetic radiation field in the case @xmath64 by using multidimensional semiclassical methods @xcite .
let us investigate the dynamics of a photonic wave packet which carries the excitation from one atom to the other inside a prolate ellipsoidal cavity .
similar to the bouncing back and forth in the time evolution of the atomic excitation probabilities we can observe a bouncing back and forth of the photonic wave packet in the cavity .
the dynamics of repeated absorption and spontaneous emission events also leads to a change of the shape of the photonic wave packet .
this process is illustrated in figures [ fig : energydensity ] a - d for the electromagnetic energy density of a single photon .
the corresponding atomic excitation probabilities are depicted in figure [ fig : atomic_excitation_probabilities ] b. figure [ fig : energydensity ] a illustrates the energy density produced by the spontaneous emission of a single photon .
the cavity guides the one - photon wave packet to the second focus as illustrated in figure [ fig : energydensity ] b. figure [ fig : energydensity ] c shows the wave packet directly after the interaction with the second atom .
figure [ fig : energydensity ] d illustrates the wave packet after its interaction with the first atom .
we have investigated the dynamics of spontaneous photon emission of two - level systems in parabolic and prolate ellipsoidal cavities assuming that the atoms are situated in the focal points of these cavities .
our theoretical approach is based on treating the multimode aspects of this problem with the help of one - dimensional semiclassical methods by exploiting the separability of the relevant helmholtz equation . this way we are able to improve straight forward multidimensional semiclassical approaches considerably . with the help of semiclassical photon path representations of relevant probability amplitudes
we have explored the intricate dynamical interplay between spontaneous photon emission and absorption processes on the one hand and reflections of the photon wave packet at the boundary of the cavity on the other hand .
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 , ed .
, @noop _ _ ( , , ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , )
|
we investigate spontaneous photon emission processes of two - level atoms in parabolic and ellipsoidal cavities thereby taking into account the full multimode scenario .
in particular , we calculate the excitation probabilities of the atoms and the energy density of the resulting few - photon electromagnetic radiation field by using semiclassical methods for the description of the multimode scenario .
based on this approach photon path representations are developed for relevant transition probability amplitudes which are valid in the optical frequency regime where the dipole and the rotating - wave approximations apply .
comparisons with numerical results demonstrate the quality of these semiclassical results even in cases in which the wave length of a spontaneously emitted photon becomes comparable or even larger than characteristic length scales of the cavity .
this is the dynamical regime in which diffraction effects become important so that geometric optical considerations are typically not applicable .
|
the theoretical problem of polymer chain collapse in poor solvents has received much attention over the past decades , both from analytic @xcite and numerical approaches @xcite .
this generic collapse , often called the `` coil - globule transition '' , is well - understood as a local version of the polymer demixing flory has originally formulated for polymer solutions @xcite .
the coil - globule transition is closely associated with the initial stages of protein folding @xcite .
collapse takes place as a result of effective pair interaction between chain monomers becoming attractive , e.g. due to changing solvent quality .
recent studies of the collapse dynamics have shown that it is a non - trivial kinetic process , with competing interactions leading to inhomogeneous intermediate structures forming and persisting on the way to the final globular state @xcite . in this work ,
we look at the collective behaviour of several chains in a confined configuration known as `` polymer brush '' . by definition ,
a polymer brush is a layer of polymers attached with one end to a surface at a sufficiently high grafting density ( i.e. a sufficiently close distance between individual polymer chains ) . in practice ,
the attachment can take place via a covalent bond , a specific ligand , physical adsorption or self - assembly . in our simulations , the chains have their first monomer fixed on a flat surface , resulting in a constant grafting density throughout the simulation .
chains themselves can vary in length , composition and individual properties , from very primitive ideal chains to complex models incorporating block copolymers , electric charge and hydrodynamics .
the subclass of charged ( polyelectrolyte ) brushes is even more interesting in biological context .
there are numerous examples of polymer brushes in biological systems such as protein micelles , planar brushes of polysaccharides , cylindrical polymer brushes in microtubules and neurofilaments . in terms of useful physical , chemical and biomedical applications
, brushes can be used to stabilise colloids , lubricate surfaces , deliver drugs by biodegradable micelles , in dna microarrays for diagnostic analysis of mutations and to reduce friction in artificial joints , to name but a few .
@xcite here we present a simple , planar brush of self - avoiding bead - and - spring polymer chains without additional electrostatic interactions or hydrodynamics , which makes the analysis tractable and comparable with well - established theories but we also examine the case of chains with a long persistence length as a crude attempt to account for polyelectrolyte stiffening .
the simple picture of polymer brushes may serve as the basis for models in diverse interfacial systems in biophysics and polymer science , such as polymeric surfactants , stabilised suspensions of colloid particles , and many structures formed by block copolymers .
the recent theory of benetatos et al .
@xcite considered stiff and semiflexible chains in a brush configuration , and predicted a bundling instability when an attractive interaction between monomers exceeds a critical value . in their work ,
benetatos et al .
established that there is a competition between permanent grafting favouring a homogeneous state of the polymer brush and the non - localised attraction , which tends to induce in - plane collapse of the aligned polymers , giving rise to an instability of the homogeneous phase to a bundled state . in the bundled state ,
the density in the plane is modulated by a length scale that depends on the strength of the attractive interaction . in this work
we carry out coarse - grained brownian dynamics simulations of a collection of chains uniformly grafted to a flat repulsive surface , to find their equilibrium swollen conformation and then follow their behavior after instant quenching .
we use the parameters setup of rosa and everaers @xcite within the large scale atomic / molecular massively parallel simulator ( lammps ) package @xcite .
the initial structure of the polymer chains was configured to be linear , and the first step was to equilibrate the linear brush into a constricted coil structure in good solvent . due to the high density of grafting , in a good solvent their individual conformation is quite stretched out , and we find a good agreement with classical results on the parabolic density profile @xcite .
after equilibration , the solvent conditions were instantly switched to poor solvent , resulting in chains collapsed onto their flat substrate surface . here
we find an unexpected effect of spinodal decomposition in the plane of collapsing brush , and study it quantitatively through density profile- , potential energy- and characteristic length - scale measurements , as well as the kinetics of collapse . in all cases , we compared the behaviour of flexible and semiflexible polymer brushes upon collapse over time .
brownian dynamics simulations allowed us to follow the density evolution of brushes over time , as it evolves from the equilibrium parabolic profile in the swollen brush to the densely packed layer on the surface .
the characteristic length scale of bundled regions in the collapsed brush appear to agree with the theoretical prediction of the square root dependence of this length scale on the attractive potential well depth @xcite .
-@xmath0 plane was on a uniform square lattice with the spacing @xmath1 ( b ) the equilibrated brush in a good solvent .
( c ) comparison of the initial structure and the equilibrated coil . ]
the model used for numerical simulations in this work is based on the bead - spring model described in the molecular dynamics context by kremer and grest @xcite and plimpton @xcite .
we take an individual polymer chain composed of connected monomeric units consisting of @xmath2 monomers , and 900 of these chains formed a brush on a 30x30 square lattice as shown in fig .
model parameters are similar to the ones described by lappala et al.@xcite : each monomer has a diameter of @xmath3 in reduced units , which is chosen to be 0.3 nm in real units ( a typical size of an amino acid residue in proteins ) , @xmath4 and @xmath5 of the fene potential were set so as to avoid any bond crossing . for our simulations ,
the values were used as in ref.@xcite : the maximum bond length @xmath6 and the spring constant @xmath7 .
the spring constant was specifically studied by kremer and grest @xcite and was found to be strong enough such that the maximum extension of the bond was always less than 1.2@xmath3 for @xmath8 , making the covalent bond breaking energetically unfeasible .
the important ratio measuring the local bending modulus of the chain , @xmath9/@xmath10 ( the chain persistence length @xmath11 ) was set to @xmath12 ( or kuhn s length of 0.6 nm ) for a case of flexible chains , and to @xmath13 for a semiflexible brush with the kuhn s length of 4.8 nm .
integration time step was chosen to be @xmath14 where @xmath15 is the reduced ( lennard - jones ) time , defined as @xmath16 ( that is , determined by the ratio of particle mass and the depth of the lennard - jones ( lj ) potential , @xmath17 ) as discussed in refs.@xcite .
taking the value of @xmath3 above , @xmath18 , and the typical mass of an aminoacid residue ( molecular weight @xmath19 ) , we obtain the estimate of @xmath20ps .
hence , with @xmath21 time - steps of this coarse - grained simulation we are able to follow the dynamics of a polymer brushes for 196 nanoseconds .
interactions between monomers are described by the standard potentials : fene for connected pairs of particles , bending elasticity @xmath22 with @xmath23 the angle formed between two consecutive bonds , and the lj potential representing a longer - range interaction between any two particles . in good solvent the lj potential was set to zero at @xmath24 , retaining only the soft repulsion between monomers . in order to simulate poor solvent conditions when there is an effective attraction between monomers , the cut - off value was set to @xmath25 , while retaining the depth of @xmath17 .
the initial parameters of the set up play an important role in the physics of the simulated brush collapse , and therefore the parameters used to build a brush had to be chosen carefully . considering a box in the cartesian frame , the brush was created using a @xmath26 square plane lattice at @xmath27 , with lattice spacing of @xmath28 .
the length of the brush has a direct impact on the grafting density , and hence the optimal length of the brush was defined to be @xmath2 if the chains were longer , the 900 chains would make the simulation much heavier and acquisition of the required statistics more difficult .
the lattice spacing ( or grafting density ) of @xmath28 had to be chosen such that it results in ample interactions between chains in good solvent
if this number was higher , the chains would simply behave as individual coils , also described as `` mushrooms '' @xcite with few inter - chain interactions .
the grafting density was chosen on the basis of the cut - off radius used for the attractive potential , and also based on the volume of a collapsed globule , which for an @xmath2 chain should be @xmath29 giving a linear distance of @xmath30 , so we are assured of the overlap in the collapsed brush .
the first atom of each chain was bound to the @xmath27 plane which was given a repulsive lennard - jones potential to keep the chains restricted to the positive @xmath31 direction . in order to confirm that the @xmath2 chains of a polymer brush were initially equilibrated in a good solvent
, we simulated a single grafted chain of 1001 atoms , relaxing from an initial straight configuration to a random self - avoiding coil .
this equilibrium state was confirmed visually and by density measurements for the standard random coil .
it is important that commonly used measurements of potential energy relaxation as a function of time are not suitable for judging equilibration , because the potential energy of a chain quickly reaches a very stable plateau for a ` crumbled ' extended chain conformation , which then takes a very long time to equilibrate into a proper self - avoiding coil . after the equilibrium of a long chain was confirmed after 10 million timesteps , we used this time for a polymer brush with much shorter chains on a planar surface to assure ourselves that it was properly equilibrated .
once an equilibrium of grafted chains was achieved , an attractive lj potential between all particles ( poor solvent conditions ) was switched on and the coil collapsed into a globule . among other parameters , we monitored the collapse by the average areal density .
density of the brush @xmath32 can be described from the condition that an integral @xmath33 is equal to the total number of particles , which in our simulations is @xmath34 .
we calculated the density along the @xmath31-direction , at each time step .
the entire box was divided into parallel bins ( in @xmath35-plane ) of unit thickness along the @xmath31-direction .
the total number of particles in each bin forms a one - dimensional array @xmath36 , which we present as the non - normalised density .
this gives a constant density profile at @xmath37 for @xmath38 for the initial straight configuration of chains .
for averaging of equilibrium structures , each @xmath31-bin was averaged over 5 brush configurations separated by 500,000 ts . according to the classical theory of expanded brush @xcite , in a good solvent the density
@xmath36 will have a parabolic profile , which we accurately record in our simulations ( see fig . [ fig5 ] below ) . in the dynamic process of collapse we has to present an instantaneous @xmath36 without time - averaging , resulting in a certain noise in corresponding profiles .
in order to simulate the collapse of a polymer brush , the initial linear chains have to be fully equilibrated in conditions that imitate good solvent , which will result in a realistic simulation of collapse if the chains are not equilibrated in such a way , the collapse will happen by rolling of straight segments into a globular structure , which is not a realistic response .
after equilibration , the solvent conditions are rapidly changed to poor a process known as quenching allowing the coils to collapse in an entangled , irregular manner .
this more realistic collapse results in an unexpected highly inhomogeneous structure in the plane , fully resembling spinodal decomposition morphology .
these structures were highly stable at equilibrium and their structural properties , such as the characteristic length scale of the formed collapsed structures , were dependent on the strength ( the well - depth @xmath17 ) of the attractive potential .
we followed the same principle as with verifying equilibration in an extended brush in good solvent .
we first recorded the potential energy relaxation of the system until it reached an apparent plateau and then waited for another 26 million time - steps for further equilibration ( it was only 10 mts needed for the @xmath39 chain to equilibrate into a coil ) .
in contrast to the annealed case of polymer coils , here we have no structural criteria to verify equilibration ; in fact , one might argue that in the dense collapsed mesh of mutually attracting chains the full equilibrium is unattainable on any reasonable time scales ; although the system is not a glass , the internal mobility must be very low ) .
however , we were certain that neither the morphological features of spinodal decomposition patterns , nor the average density profiles have not changed between @xmath40 kts and @xmath41 mts .
hence we declare these states equilibrium with our accuracy .
as the polymer brush collapses , we see the initially homogeneous distribution of monomers in the @xmath35-plane breaking down and a phenomenon that appears indistinguishable from the classical spinodal decomposition taking place .
this clustering of polymer chains depends on the strength of the attractive potential between individual chains of the brush , characterised by a ratio @xmath42 .
in flexible chains , we found distinctive coarsening and large clusters at small potential well depths , and increasingly smaller - sized clusters of polymers as the potential well depth increases , as shown in fig .
although the difference in characteristic length scale of the structure , @xmath43 , is apparent between different panels , it was not easy for us to determine it quantitatively .
due to the relatively small area of the @xmath35-map , the fourier transformation methods ( so useful in classical studies of spinodal decomposition , in cahn - hilliard model and beyond ) were not possible in our case .
in the end we have resorted to manually extracting many independent measurements of ` thickness ' of the dense - polymer regions ( a distance perpendicular to both interfaces ) and averaging the result .
the final panel in fig .
[ fig2 ] presents such an average characteristic length scale , as a function of increasing strength of the attractive lj potential , which we fitted to a square - root dependence @xmath44 .
[ t ] ) at varying attractive potential well - depths , @xmath45 , labelled on each panel .
the plot of the average characteristic length scale of the spinodal pattern is fitted to the square - root dependence : @xmath44 .
, title="fig : " ] for semiflexible brushes with persistence length @xmath13 , there is a similar trend in morphological properties as a function of attractive potential strength as the potential increases , the number of chain clusters also increases and these become smaller in size , as demonstrated in fig .
the similar quantitative analysis of the average characteristic length scale of the frozen spinodal pattern follows a similar scaling dependence @xmath44 ( acknowledging the large errors of its calculation and only a few data points available for the fitting ) . comparing with the flexible brush ( @xmath46 ) the values of this length are uniformly lower by a constant factor of @xmath47 in this case of semiflexible chains .
[ t ] ) at varying attractive potential well - depths , @xmath45 , labelled on each panel .
the plot of the average characteristic length scale of the spinodal pattern is fitted to the square - root dependence : @xmath44 .
, title="fig : " ] figure [ fig4 ] shows the ` early time ' relaxation of the potential energy of each brush after the instantaneous quench at @xmath48 .
the potential energy experiences a sudden drop at short times , when large gains could be made by all attracting particles coming closer together ; the amplitude of this drop is obviously proportional to the depth of the attractive lj potential well .
after the system becomes compact ( the density plots after @xmath49 kts do not have much of the further variation , fig .
[ fig5 ] below ) , the approach to equilibrium becomes very slow .
this is due to structural changes in chain configurations in the dense state .
all structures were equilibrated for 26 million time steps and for the most of this long time the total potential energy of the 900 chains in our brushes did not vary at all .
[ t ] ; ( b ) semiflexible chains , @xmath50 .
different curves correspond to increasing values of effective attraction measured by @xmath45 , labelled on plots .
, title="fig : " ] figure [ fig5 ] presents the density profiles of our brushes in different conditions and at different times of their collapse , as always , comparing the flexible and the semiflexible chain cases .
the first interesting result is for the flexible brush in good solvent : the density @xmath36 is parabolic to a good approximation , as the fitted line in fig .
[ fig5](a ) indicates therefore our results are in full compliance with the classical theory of brushes @xcite , according to which @xmath51 and the effective potential acting on the chains is @xmath52 , with @xmath53 the excluded volume parameter and @xmath54 the brush height .
this also confirms that our initial brush is equilibrated .
in contrast , the brush of semiflexible chains in fig .
[ fig5](b ) is much more extended and its density does not fit a parabolic function in any approximation .
the second fact one can notice from the density @xmath36 evolution on brush collapse is the fact that there is no further change in density past a time @xmath55 kts ( compared with the @xmath56 mts in the plots ) .
this corresponds to a completed relaxation of the potential energy of the brush , cf .
[ fig4 ] one has to assume that the further re - arrangements of the microstructure in the dense system are mainly entropy - driven .
it is evident from both the potential energy relaxation , as well as the density profile evolution , that the time that it takes for a chain to collapse is a function of the attractive potential strength .
we declared the polymer brush ` collapsed ' ( and recorded the corresponding time it took ) when the density profile @xmath36 have reached the shape that have no longer evolved with time ( cf . fig .
[ fig5 ] ) .
hence we found that the stronger the attractive potential ( deeper the quench ) , the faster this collapse takes place : see fig .
the lines in the plot are a guide to an eye fitted dependence of @xmath57 . at the same time
, the semiflexible brush ( more extended in good solvent ) is much slower to reach the fully collapsed state on the substrate .
[ t ] , calculated as the average number of monomers in a given @xmath35-plane , at various time - points of brush collapse at @xmath58 : ( a ) flexible chains , @xmath59 ; ( b ) semiflexible chains , @xmath50 .
the initial density profile in good solvent ( ` equilibrium coil ' ) shows a good approximation to the classical parabolic profile in the flexible brush @xcite , fitted with a solid blue line ; the corresponding semiflexible chains are much more extended near the grafting plane and only show the expected drop in density from about half - way along the brush height .
, title="fig : " ] . ]
there is one example in the literature , where a competition between the attractive forces between all chain segments and the constraint of remaining uniformly grafted on a flat surface has been studied theoretically @xcite .
the direct analogy with our simulation work is difficult to establish , because the authors there have only consider the so - called directed polymers the chains never allowed to fold back upon themselves by either the constraint of a fixed height of the brush ( the second plane fixing the ends of the chains as well ) , or by having a very high persistence length .
both of these limits have only marginal correspondence to our system , which is collapsing into a highly folded dense melt ( our semiflexible chain case might be the closest to what has been studied in @xcite .
nevertheless , the fundamental physics of a spatial instability developing in the @xmath35-plane in order to resolve the mentioned competition has to be the same in both works , which is why we reflect on this comparison .
in essence , the final answer of that theoretical study @xcite is a prediction that an initially homogeneous areal density @xmath60 will develop an instability when the following condition is first met ( written in an approximate form , in the limit of the range of the effective attractive lj interaction less than @xmath61 , following @xcite ) : @xmath62 where @xmath63 is the separation of the grafting points in the plane ( i.e. @xmath64 is the grafting density ) , and the non - dimensional parameter @xmath65 is proportional to the fourier wave vector of the polymer density in the @xmath35-plane .
the minimum of the universal function in the left - hand side of ( [ beneta ] ) is at @xmath66 , which means the spinodal - decomposition instability will first occur with a length scale @xmath67 when the strength of attraction potential @xmath45 exceeds a critical value @xmath68 . as mentioned earlier
, we can not hope to achieve quantitative agreement with this theory : in our systems the chains can fold into a densely packed state where the approximations made to carry out their calculations fail . besides , we only examine the final state of the arrested spinodal decomposition not the initial point of instability .
nevertheless , we do see that the characteristic length scale of the inhomogeneous structure @xmath43 is smaller in a system with a larger persistence length ( fig .
[ fig3 ] ) , and the dependence on the quenching depth is also in full correspondence with the physics of spinodal decomposition away from the critical point . we feel that these physical ideas and considerations should initiate a more focused theoretical and simulation effort to establish the laws and the morphology of polymer brushes collapsed in poor solvent now that we know their density in the plane can be highly non - uniform .
we also believe that this analysis can be used to finely tune polymer brushes for experimental purposes to regulate the surface properties of materials .
our results can facilitate the physical understanding of polymer brushes in experimental applications such as directed drug - delivery systems , biosensors and polymer brush modified membranes for protein analysis .
this work was funded by the osk .
huttunen foundation ( finland ) , and the cambridge theory of condensed matter grant from epsrc .
simulations were performed using the darwin supercomputer of the university of cambridge high performance computing service ( http://www.hpc.cam.ac.uk/ ) , provided by dell inc . using strategic research infrastructure funding from the higher education funding council for england .
lifshitz , e. m. _ sov .
. jetp _ * 1969 * , _ 28 _ , 12801286 moore , m. _ j. phys .
* 1977 * , _ 10 _ , 305314 kholodenko , a. l. ; freed , k. f. _ j. chem .
phys . _ * 1984 * , _ 80 _ , 900911 muthukumar , j. ; nickel , b. g. _ j. chem . phys . _ * 1984 * , _ 80 _ , 58395847 des cloiseaux , j. ; jannink , g. oxford university press , oxford , 1990 baumgartner , a. ; muthukumar , j. _ j. chem .
phys . _ * 1987 * , _ 87 _ , 30823095 kremer , k. ; binder , k. _ comput . phys .
_ * 1988 * , _ 7 _ , 259275 flory , p. _ j. chem .
phys . _ * 1942 * , _ 10 _ , 5163 banavar , j. r. ; hoang , t. x. _ phys rev e _ * 2004 * , _ 70 _ , 041905 de gennes , p. g. _ j. phy . lett .
_ * 1985 * , _ 46 _ , 639642 halperin , a. ; goldbart , p. m. _ phys rev e _ * 2000 * , _ 1 _ , 565573 lappala , a. ; terentjev , e. m. _ macromolecules _ * 2013 * , _ 46 _ , 12391247 ayres , n. _ polym .
chem . _ * 2010 * , _ 1 _ , 769777 benetatos , p. ; terentjev , e. m. ; zippelius , a. _ phys .
e _ * 2013 * , _ 88 _ , 042601 rosa , a. ; everaers , r. _ plos comput .
* 2008 * , _ 4 _ , e1000153 plimpton , s. j. _ j. comput .
phys . _ * 1995 * , _ 117 _ , 119 milner , s. t. ; witten , t. a. ; cates , m. e. _ macromolecules _ * 1988 * , _ 21 _ , 26102619 kremer , k. ; grest , g. s. _ j. chem . phys . _ * 1990 * , _ 8 _ , 50575086 auhl , r. ; everaers , r. ; grest , g. s. ; kremer , k. ; plimpton , s. j. _ j. chem . phys . _ * 2003 * , _ 119 _ , 1271812728
|
we study the brownian dynamics of flexible and semiflexible polymer chains densely grafted on a flat substrate , upon rapid quenching of the system when the quality of solvent becomes poor and chains attempt collapse into a globular state .
the collapse process of such a polymer brush differs from individual chains , both in its kinetics and its structural morphology .
we find that the resulting collapsed brush does not form a homogeneous dense layer , in spite of all chain monomers equally attracting each other via a model lennard - jones potential . instead , a very distinct inhomogeneous density distribution in the plane forms , with a characteristic length scale dependent on the quenching depth ( or equivalently , the strength of monomer attraction ) and the geometric parameters of the brush .
this structure is identical to the spinodal - decomposition structure , however , due to the grafting constraint we find no subsequent coarsening : the established random bundling with characteristic periodicity remains as the apparently equilibrium structure .
we compare this finding with a recent field - theoretical model of bundling in a semiflexible polymer brush .
|
supernovae of type ibn ( sne ibn ) are considered a rare group of stripped - envelope core - collapse ( cc ) events which interact with h - depleted circumstellar material ( csm ) .
the spectra of sne ibn are characterized by relatively narrow lines of he i in emission ( hence the designation as ibn , * ? ? ?
* ) , with full - width at half maximum ( fwhm ) velocities ranging from several hundreds to a few thousands km s@xmath7 .
these features are thought to arise in the interactions between the sn ejecta and he - rich ( and h - poor ) csm .
however , weak h lines have been occasionally detected in the spectra of a few type ibn sne , suggesting the presence of residual h in the csm of -at least- a sub - sample of sne ibn @xcite .
the prototype of this family , sn 2006jc , has been associated with a precursor eruptive episode which occurred a couple of years before the sn explosion and reached a peak absolute magnitude @xmath12 @xmath13 -14 @xcite .
sn 2006jc , although discovered after the maximum light , was widely studied . the brightest magnitude registered for
this sn was @xmath12 = -18.3 @xcite , and its optical light curve experienced a rapid post - peak decline ( about 3.5 mag during the first 50 days after discovery ) . at later phases ,
the increasing decline rate of the optical light curve and the simultaneous near - infrared ( nir ) excess were explained as the consequence of dust forming in a cool dense shell ( e.g. * ? ? ?
* ; * ? ? ?
@xcite and @xcite attempted to reproduce the evolution of sn 2006jc with radioactively powered models .
they both favoured models with a high kinetic energy ( e@xmath14 10@xmath15 erg ) , and moderate masses of ejecta ( m@xmath16 @xmath13 4.5 - 5 m@xmath17 ) and synthesized @xmath8ni ( m@xmath18 0.2 - 0.25 m@xmath17 ) .
nonetheless , radioactively powered models with even lower e@xmath19 and m@xmath16 provided decent matches with observations ( e.g. , models a and b in * ? ? ?
* were obtained adopting the following parameters : e@xmath14 10@xmath20 erg , m@xmath21 1 m@xmath17 and m@xmath22 0.25 - 0.40 m@xmath17 ) . on the other hand
, @xcite showed that radioactive models with modest m@xmath16/m@xmath23 ratio were unrealistic for a type ib / c sn explosion , and proposed that the observed parameters of sn 2006jc were best explained with strong ejecta - csm interaction occurring already at early phases .
in fact , the interaction with a circumstellar shell with mass of a few hundredths m@xmath17 would fairly well explain the properties of sn 2006jc .
it is clear , however , that for an accurate modelling good data coverage is essential along the duration of the entire sn evolution , including the very early phases , not available for sn 2006jc .
the first opportunity to study a sn ibn soon after its explosion was offered by sn 2010al , which initially showed an almost featureless spectrum with the shape of a hot blackbody ( the data of this object are presented in * ? ? ?
the proliferation of well - organized imaging surveys , with different depths and strategies , has significantly increased the number of transients discovered and , hence , the number of type ibn candidates . the _ optical gravitational lensing experiment _
( @xmath24 , * ? ? ?
* ) is a long - term project carried out initially with the 1-m swope telescope , and later with the 1.3-m warsaw university telescope , both located at the las campanas observatory ( chile ) .
the fields monitored in the course of the different seasons of the search included the two magellanic clouds and the magellanic bridge . in the galactic bulge and disk . ] in the current fourth season ( @xmath0 , * ? ? ?
* ) , which started in 2009 , the survey is making use of a 32 ccd mosaic camera covering a field of 1.4 square degrees and equipped with @xmath25 and @xmath4 filters .
originally conceived to study microlensing events @xcite , a by - product of the project is the discovery of a huge number of transients and variable stars of all types , including sne @xcite .
the study of new types of stellar explosions is the main goal of the _ `` public eso spectroscopic survey of transient objects '' _
( @xmath1 , see * ? ? ?
* ) , which is a public spectroscopic survey running at the 3.58-m new technology telescope ( ntt ) of the european southern observatory ( eso ) in la silla ( chile ) .
the survey is planned for four years ( starting from april 2012 ) and is operative 90 nights per year .
the survey aims to classify a large number of transients and to monitor the most interesting objects with efosc2 ( optical ) and sofi ( near - infrared ) .
priority @xmath1 targets are the most unusual sne , sn impostors or very nearby transients , for which detailed high - quality , multi - wavelength observational campaigns can be arranged .
a specific subproject has been defined within _
pessto _ in order to study sne ibn @xcite .
two type ibn sne discovered by the @xmath26 variability survey and followed by @xmath1 ( lsq12btw and lsq13ccw ) , are studied in a companion paper to the present work @xcite .
a few months after lsq12btw , a new sn ibn candidate , ogle-2012-sn-006 , was discovered in the course of the @xmath0 survey @xcite .
the object has coordinates @xmath27 and @xmath28 ( j2000 ) , and lies in a faint , anonymous host galaxy ( figure [ fig1 ] ) .
according to @xcite , the object had a magnitude @xmath4 = 17.62 at maximum light .
ogle-2012-sn-006 was spectroscopically classified on 2013 january 10.2 ut by @xcite as a type ibn sn , similar to sn 2006jc .
the host galaxy redshift , as estimated from the position of the narrow sn features , is z = 0.057@xmath3 0.001 . adopting a hubble constant value of h@xmath29 = 73 @xmath3 5 km s@xmath7 mpc@xmath7 ( @xmath30 = 0.27 , and @xmath31 = 0.73 )
, we obtain a luminosity distance of 244.5 @xmath3 20.0 mpc .
@xcite , on the basis of a very small sample of objects , proposed that sne ibn might form a relatively homogeneous group of mildly interacting stripped - envelope sne .
however , recent discoveries suggest that sne ibn span a much wider range of properties than initially believed @xcite . in this context , because of its slow evolving light curve ( see section [ lc ] ) , ogle-2012-sn-006 can be considered as an unprecedented addition to the type ibn sn zoo .
our photometric and spectroscopic data have been reduced using standard procedures in the iraf environment .
optical imaging frames were first pre - reduced ( i.e. overscan , bias and flat - field corrected , and trimmed in order to remove the unexposed regions of the image ) . in the near - infrared ( nir ) photometry images we also removed the contribution of the bright nir background from the science images .
sky images were obtained by median - combining a number of dithered science frames , and were finally subtracted from each science image . in order to improve the signal to noise , the sky - subtracted science images were spatially registered and finally combined .
photometric measurements in the optical and nir bands were obtained through the psf - fitting technique .
a template psf was built using stars in the sn field . with this psf model along with a low - order polynomial surface , we finally performed a fit to the sn and the underlying background .
@xmath0 photometry was obtained using the difference imaging analysis ( dia ) , which is a template subtraction method adapted to the @xmath24 data and detailed in @xcite ( see also * ? ? ?
the magnitudes of several stars in the sn field ( see figure [ fig1 ] , left ) were calibrated using observations of standard fields of the @xcite catalogue .
the magnitudes obtained in photometric nights were used to estimate reliable average magnitudes for the entire local stellar sequence ( 23 stars ) , which was finally used for the calibration of the sn photometry obtained during non - photometric nights .
the nir sn photometry was calibrated through a comparison with 14 stars of the optical sequence for which nir magnitudes were available in the the two micron all sky survey ( 2mass ) catalogue @xcite .
the optical and nir magnitudes of the sn are reported in tables [ taba1 ] and [ taba2 ] , while those of the local sequence stars are in table [ taba3 ] .
the pre - reduction process of optical spectroscopy frames followed the steps described above for photometric frames .
then , one - dimensional spectra were optimally extracted in iraf , and wavelength calibrated using reference spectra of arc lamps . finally , sensitivity curves obtained using spectra of spectro - photometric standards allowed us to flux - calibrate the sn spectra .
telluric absorption features were removed using normalized telluric absorption profiles derived from the spectrum of the standard star . for the nir spectra ,
two additional steps were required : firstly , we had to remove the intense background emission .
this was achieved through the subtraction of two consecutive exposures ( one from the other ) , since during the observation the source was dithered along the slit direction .
secondly , we used spectra of telluric standard stars to remove the very broad ( typically saturated in the nir ) atmospheric absorption bands .
finally , the consistency of the spectroscopic flux calibration in the optical and nir spectra was checked with available sn magnitudes and , in case of a discrepancy , the spectral fluxes were rescaled to the photometry .
the log of spectroscopic observations of ogle-2012-sn-006 is reported in table [ tab1 ] .
the spectroscopic data of ogle-2012-sn-006 are available from the eso archive as pessto phase 3 data products ( see _ http://www.pessto.org_ ) and are also available via wiserep @xcite .
spectral data products ( ssdr1 ) .
this is because the analysis was performed before @xmath1 ssdr1 , and the faintness of the object required a careful treatment . ]
routine @xmath32band observations of the @xmath0 fields resulted in an excellent monitoring of the region where ogle-2012-sn-006 exploded .
@xmath0 observations @xcite of the sn field were collected since early july 2012 , and the last non - detection was registered on 2012 september 29.26 ut .
the first detection of ogle-2012-sn-006 was on 2012 october 7.34 ut , so we can constrain the explosion epoch with a relatively small uncertainty to 2012 october 3.30 ut ( @xmath2 = 2456203.8 @xmath3 4.0 ) . after discovery , the object showed a relatively fast light curve rise to @xmath32band maximum , lasting @xmath33 2 weeks from the adopted explosion time .
using a low - order polynomial fit to the data , we estimated the magnitude of the @xmath32band light curve maximum to be @xmath34 ( on 2012 october 17.6 ut , i.e. @xmath2 = 2456218.1 @xmath3 1.8 ) .
adopting a distance modulus of @xmath35 = 36.94 @xmath3 0.19 mag for the galaxy hosting ogle-2012-sn-006 ( see section [ intro ] ) , assuming negligible host galaxy reddening ) , we do not see any clear signature of the narrow interstellar na i doublet ( na i d ) at the parent galaxy redshift , suggesting little ( if any ) contribution of the host galaxy to the total reddening of ogle-2012-sn-006 . ] and a galactic contribution to the total extinction of @xmath36 = 0.12 mag ( obtained adopting the dust map calibration of * ? ? ?
* ) , we obtain an absolute magnitude at peak of @xmath37 = -19.65 @xmath3 0.19 . after maximum ,
the @xmath32band light curve experienced a relatively fast decline ( with a slope of @xmath38 4 mag 100d@xmath7 ) lasting about 25 days , followed by a flattening from + 25 days to + 70 days past maximum ( @xmath39 0.40 mag 100d@xmath7 ) .
such an optical / nir flattening , making the magnitude decline rate slower than expected for a sn powered by radioactivity of @xmath8co , has never been observed in any sn ibn . after that , the @xmath4-band light curve becomes slightly steeper ( @xmath40 0.74 mag 100d@xmath7 ) , showing a slope which is closer to @xmath8co decay ( 0.98 mag 100d@xmath7 ) , but still slightly flatter than that slope .
a similar decline rate in the late - time luminosity is measured in all optical bands .
we note that the relatively slow late - time decline of ogle-2012-sn-006 looks also alike that observed in some h - rich interacting sne , such as the type iin sn 2010jl ( see below ) .
the increased slope of the @xmath4-band light curve at very late phases is confirmed by a very late detection on 2013 august 1st , and by deep detection limits at later epochs .
this is possibly an indication of dust formation at very late phases .
we note that spectroscopic indicators also support this claim ( see section [ spec ] ) .
-band absolute light curves ( top ) , @xmath41 ( middle ) and @xmath42 ( bottom ) colour curves of sne ogle-2012-sn-006 , 2010al , 2006jc and the type iin sn 2010jl .
data of sn 2010al are from @xcite , those of sn 2006jc are from @xcite and those of sn 2010jl are from @xcite . in the top panel , together with the @xmath4-band light curves of the sn sample , the well - followed @xmath43-band light curve of sn 2010al has been included.,width=340 ]
|
we present optical observations of the peculiar type ibn supernova ( sn ibn ) ogle-2012-sn-006 , discovered and monitored by the @xmath0 survey , and spectroscopically followed by @xmath1 at late phases .
stringent pre - discovery limits constrain the explosion epoch with fair precision to @xmath2 = 2456203.8 @xmath3 4.0 .
the rise time to the @xmath4-band light curve maximum is about two weeks .
the object reaches the peak absolute magnitude @xmath5 on @xmath6 .
after maximum , the light curve declines for about 25 days with a rate of 4 mag 100d@xmath7 .
the symmetric @xmath4-band peak resembles that of canonical type ib / c supernovae ( sne ) , whereas sne ibn usually exhibit asymmetric and narrower early - time light curves . since
25 days past maximum , the light curve flattens with a decline rate slower than that of the @xmath8co to @xmath8fe decay , although at very late phases it steepens to approach that rate .
however , other observables suggest that the match with the @xmath8co decay rate is a mere coincidence , and the radioactive decay is not the main mechanism powering the light curve of ogle-2012-sn-006 . an early - time spectrum is dominated by a blue continuum , with only a marginal evidence for the presence of he i lines marking this sn type .
this spectrum shows broad absorptions bluewards than 5000 , likely o ii lines , which are similar to spectral features observed in super - luminous sne at early epochs .
the object has been spectroscopically monitored by @xmath1 from 90 to 180 days after peak , and these spectra show the typical features observed in a number of sn 2006jc - like events , including a blue spectral energy distribution and prominent and narrow ( @xmath9
km s@xmath7 ) he i emission lines .
this suggests that the ejecta are interacting with he - rich circumstellar material .
the detection of broad ( @xmath10 km s@xmath7 ) o i and ca ii features likely produced in the sn ejecta ( including the [ o i ] @xmath116300,6364 doublet in the latest spectra ) lends support to the interpretation of ogle-2012-sn-006 as a core - collapse event .
[ firstpage ] supernovae : general supernovae : individual ( ogle-2012-sn-006 , sn 2006jc , sn 2010al )
|
the realization of manifolds as branched coverings is a classical long - standing problem in topology . a well - known theorem of alexander @xcite states that every oriented , closed , smooth @xmath2-dimensional manifold is a branched covering of @xmath3 .
strong restrictions for the existence of branched coverings were found by berstein and edmonds @xcite . branched coverings
have been investigated in many different contexts and they turned out to be a useful tool for the study of several problems in geometry , such as the minimal genus problem in four dimensions .
recall that a branched @xmath4-fold covering is a smooth proper map @xmath5 with a codimension two subcomplex @xmath6 , called the branch locus of @xmath7 , such that @xmath8 is a @xmath4-fold covering in the usual sense and for each @xmath9 the map @xmath7 is given by @xmath10 , for some charts of @xmath11 and @xmath12 and some positive integer @xmath13 .
the point @xmath11 is called singular and its image @xmath12 is called a branch point . in dimensions two and three
, edmonds showed that a dominant map is quite often homotopic to a branched covering @xcite .
more precisely , edmonds proved that every non - zero degree map between two closed surfaces is homotopic to the composition of a pinch map followed by a branched covering @xcite .
a pinch map in dimension two is a map which collapses 2-handles , i.e. is a quotient map @xmath14 , where a submanifold @xmath15 , with circle boundary in the interior of @xmath16 , is identified to a point . in dimension three ,
edmonds result is that every @xmath17-surjective map of degree at least three is homotopic to a branched covering @xcite .
the existence of branched coverings in low dimensions has been explored by several other people , including fox , hilden , hirsch and montesinos .
our aim in this paper is to show that every simply connected , closed four - manifold , resp .
five - manifold , admits a branched double covering by a product @xmath18 , composed with a certain pinch map , resp . a map whose degree depends on the torsion of the integral homology of the target .
the main result is the following general statement : [ t : main ] for @xmath19 and every @xmath20 there is a branched double covering @xmath21 where @xmath22 denotes the total space of the non - trivial @xmath23-bundle over @xmath24 with structure group @xmath25 .
we note that the statement of theorem [ t : main ] does not contain any reference on orientations , because the targets are simply connected and therefore always orientable .
the existence of dominant maps , not necessarily branched coverings , where the domain is a non - trivial product has an independent interest and its study is partially motivated by gromov s work on functorial semi - norms on homology @xcite .
kotschick and lh @xcite investigated such maps and showed that many targets with suitably large fundamental groups are not dominated by products . on the other hand ,
the fundamental group conditions given in @xcite are not always sufficient to deduce domination by products , cf .
@xcite .
however , as pointed out in @xcite , it is natural to ask whether every connected , oriented , closed manifold with finite fundamental group is dominated by a non - trivial product . in order to study this question
, it obviously suffices to obtain an answer for simply connected targets . in dimensions two and three ,
the answer is easy and affirmative , because @xmath24 and @xmath26 respectively represent the only simply connected manifolds in those two dimensions . in this paper
, we combine our constructions of branched coverings , with some classification results for simply connected manifolds in dimensions four and five ( cf .
@xcite and @xcite respectively ) to obtain the following : [ t:4and5-mfds ] every connected , oriented , closed manifold with finite fundamental group in dimensions four and five is dominated by a non - trivial product . in dimension four , kotschick and lh @xcite
have previously obtained a non - constructive proof for the above statement based on a result of duan and wang @xcite .
our proof here is independent of those earlier works and it moreover gives an explicit construction of a degree two map from a product of type @xmath27 to every simply connected four - manifold , obtained as the composition of a branched double covering with a certain degree one map ; cf . theorem [ t:4-mfds ] . in section [ s : construction ] , we first prove theorem [ t : main ] and then we discuss briefly the existence of branched double coverings for connected sums of direct products with a sphere factor . in sections [ s:4-mfds ] and [ s:5-mfds ] , we apply the constructions of section [ s : construction ] to simply connected four- and five - manifolds respectively in order to prove theorem [ t:4and5-mfds ] . in the final section [ s : final ] ,
we give another application in higher dimensions and we make further remarks .
we begin this section by proving theorem [ t : main ] . after that , we give a branched double covering for every connected sum of copies of a direct product @xmath28 , where @xmath29 is any oriented , closed @xmath2-dimensional manifold ( cf .
theorem [ t : general ] ) . at 235 149 at -5 91 at 82 91 at 82 -5 at -5 -5 at 355 137 at 386 137 at 391 23 at 357 23 at 235 87 ,
branched along four points @xmath30 , @xmath31 , and so that @xmath32 , where @xmath33 is an annulus in @xmath34 containing two singular points.,title="fig:",width=302 ] by assumption , we consider oriented @xmath23-bundles over @xmath24 with structure group @xmath25 , where @xmath19 .
these bundles are classified by @xmath35 ( cf .
steenrod @xcite ) , therefore there exist only two ; the product @xmath36 and the twisted bundle @xmath37 .
let @xmath38 denote the twisted bundle .
there is a branched double covering with four branch points , @xmath39 , given by the quotient for an involution on @xmath34 ; see figure [ f : branchcovtor ] ( this map is known as `` pillowcase '' ) .
we pull back @xmath40 by @xmath41 to obtain a branched double covering @xmath42 .
now @xmath43 is the total space of an oriented @xmath23-bundle over @xmath34 with structure group @xmath25 .
again , there exist only two such bundles and since @xmath41 has even degree we deduce that @xmath43 is the trivial bundle , i.e. the product @xmath44
. therefore , @xmath45 is a branched double covering of @xmath46 , with branch locus four copies of the @xmath23-fiber of @xmath46 , given by the preimages under @xmath40 of the four branch points of @xmath41 .
this proves the statement of theorem [ t : main ] for @xmath47 .
next , we prove the claim for @xmath48 .
let the branched covering @xmath49 constructed above .
we can think of @xmath45 as a trivial @xmath50-bundle over @xmath51 .
we thicken an @xmath50-fiber of this bundle to an annulus @xmath33 in @xmath34 so that @xmath32 , as in figure [ f : branchcovtor ] .
a fibered neighborhood of this @xmath50-fiber is a product @xmath52 in @xmath45 and @xmath53 is an @xmath2-disk @xmath54 in @xmath55 .
we now remove the fibered neighborhood @xmath56 from two copies of @xmath45 and perform a fiber sum by gluing the @xmath51-boundaries . since @xmath44 is a trivial @xmath50-bundle this fiber sum will produce another trivial @xmath50-bundle , namely a product @xmath57 . at the same time
, we connected sum two copies of @xmath55 along @xmath54 , so that the branch loci fit together .
we have now obtained a branched double covering @xmath58 proving the claim for @xmath59 .
for @xmath60 we iterate the above construction .
the statement of theorem [ t : main ] is the most general possible concerning oriented @xmath23-bundles over @xmath24 , but it moreover presupposes the structure group of these bundles to be linear .
this assumption is not necessary in dimensions four and five ( cf .
@xcite and @xcite respectively ) and so we will be able to construct many branched coverings in those two dimensions ; cf .
sections [ s:4-mfds ] and [ s:5-mfds ] .
the inspiration for the construction of the branched covering of theorem [ t : main ] stems from a previous joint work with kotschick on domination for three - manifolds by circle bundles @xcite .
more precisely , we proved there that for every @xmath20 there is a @xmath17-surjective branched double covering @xmath61 .
a direct generalization of that construction is achieved if we replace @xmath50 in @xmath62 by any oriented , closed @xmath2-dimensional manifold @xmath29 .
the steps of the proof follow those of @xcite and they are left to the reader : [ t : general ] let @xmath29 be a connected , oriented , closed @xmath2-dimensional manifold . for every @xmath20
there is a @xmath17-surjective branched double covering @xmath63 .
the constructions given in theorems [ t : main ] and [ t : general ] are of different nature , because the connected summands of the target in theorem [ t : main ] are twisted products , while in theorem [ t : general ] the summands are direct products .
for this reason the proof of theorem [ t : main ] can not be deduced using theorem [ t : general ] .
( nevertheless , theorem [ t : general ] could be seen as a special case of theorem [ t : main ] for @xmath64 . ) a further generalization of theorem [ t : main ] would be to consider targets that are connected sums of fiber bundles over @xmath24 , where the fiber is an arbitrary oriented , closed manifold @xmath29
. however , the comprehension of arbitrary twisted products @xmath65 ( and of connected sums built out of such summands ) seems , in general , not sufficiently enough to produce further generalization of theorem [ t : main ] .
[ r : generalpillow ] the two - dimensional pillowcase @xmath66 can be generalized for any sphere @xmath67 .
namely , for every @xmath68 , there is a branched double covering @xmath69 .
the branched locus of @xmath41 is @xmath70 and there is an @xmath2-ball @xmath71 such that @xmath72 .
therefore , in place of @xmath24 in theorem [ t : general ] we may take any sphere of dimension at least two .
we refer to chapter 3 of @xcite for further details .
we now apply theorem [ t : main ] to simply connected four - manifolds .
first , we show that every four - manifold diffeomorphic to a connected sum @xmath73 admits a branched double covering by a product @xmath74 .
we then deduce by the classification results of wall @xcite and freedman @xcite , that every simply connected , closed four - manifold admits a branched double covering by a product of the circle with a connected sum of copies of @xmath0 , followed by a collapsing map . by a result of smale @xcite ,
the inclusion @xmath75 is a homotopy equivalence and so there exist only two oriented @xmath24-bundles over @xmath24 .
moreover , @xmath76 is diffeomorphic to @xmath77 ( cf .
wall @xcite ) , therefore theorem [ t : main ] implies the following for @xmath78 : [ c:4-mfds ] for every @xmath20 there is a branched double covering @xmath79 . given a connected sum @xmath80 , we define a pinch map @xmath81 , by collapsing the gluing sphere and then mapping @xmath82 to a point .
this degree one map is called collapsing map . for a connected sum @xmath83 , instead of the whole summand @xmath84
, we may collapse the exceptional embedded sphere @xmath85 to obtain again a degree one map @xmath86 .
the collapsed @xmath87 has self - intersection number @xmath88 and is called @xmath88-sphere . if @xmath41 is smooth , then this is the usual blow - down operation . similarly , for a connected sum @xmath89
we obtain a degree one map @xmath90 by collapsing the embedded @xmath91-sphere @xmath92 . in the smooth category , this operation is known as the antiblow - down of @xmath41 . in the light of corollary [ c:4-mfds ] , we now rely on results of wall @xcite and freedman @xcite on the classification of simply connected , closed four - manifolds to obtain the following statement which moreover proves theorem [ t:4and5-mfds ] in dimension four : [ t:4-mfds ] every simply connected , closed four - manifold @xmath29 admits a degree two map by a product @xmath93 , which is given by the composition of a branched double covering @xmath94 with a collapsing map @xmath95 . by corollary [ c:4-mfds ] , it suffices to show that for every simply connected , closed four - manifold @xmath29 there exists a @xmath20 and a collapsing map @xmath95 . first , if necessary , we perform connected sums of @xmath29 with copies of @xmath96 or @xmath84 ( or both ) to obtain a manifold @xmath97 , whose intersection form is odd and indefinite ( and therefore diagonal ; cf .
@xcite ) .
if @xmath29 is smooth , then wall s stable diffeomorphism classification in dimension four @xcite implies that @xmath98 is stably diffeomorphic to a connected sum @xmath99 .
the connected summing with @xmath84 , resp .
@xmath96 , is the blow - up , resp .
antiblow - up , operation . for @xmath29
non - smooth , we may first assume that the homotopy type of @xmath29 has trivial kirby - siebenmann invariant and then deduce that @xmath98 is homeomorphic to a connected sum @xmath100 , by freedman s topological classification of simply connected four - manifolds @xcite .
in particular , @xmath101 inherits a smooth structure .
we can now assume that , in both cases , @xmath98 is ( stably ) diffeomorphic to a connected sum @xmath102 , after connected summing with more copies of @xmath96 or @xmath84 , if necessary . finally , a degree one collapsing map @xmath103 is obtained by collapsing the @xmath104 embedded exceptional spheres @xmath85 and the @xmath105 embedded spheres @xmath106 . if @xmath29 is smooth , then the collapsing map is also smooth .
we have now completed the proof of theorem [ t:4-mfds ] and therefore of theorem [ t:4and5-mfds ] in dimension four .
[ r : wall1 ] wall s stable diffeomorphism classification @xcite is obtained after adding summands @xmath107 .
however , for simply connected , closed , smooth four - manifolds with odd intersection form ( as in the proof of theorem [ t:4-mfds ] ) , this is equivalent to adding summands @xmath108 , again by a result of wall @xcite ; see also remark [ r : wall2 ] . in dimension two ,
edmonds @xcite proved that every non - zero degree map between two closed surfaces is homotopic to the composition of a pinch map followed by a branched covering .
we observe that the maps constructed in theorem [ t:4-mfds ] for every simply connected , closed four - manifold have an analogy to edmonds result , where however the order between the pinch map and the branched covering is reversed .
moreover , edmonds @xcite showed that a non - zero degree map between two closed surfaces , @xmath109 , is homotopic to a branched covering if and only if either @xmath7 is @xmath17-injective or @xmath110 $ ] .
this implies that every non - zero degree map between two closed surfaces can be lifted to a ( @xmath17-surjective ) map which is homotopic either to a pinch map ( absolute degree one ) or to a branched covering . in dimension three , every @xmath17-surjective map of degree greater than two between two closed three - manifolds is homotopic to a branched covering , again by a result of edmonds @xcite .
this result fails in dimension four , by a recent work of pankka and souto @xcite , where it is shown that @xmath111 is not a branched covering of @xmath112 , while every integer can be realized as the degree for a map from @xmath111 to @xmath112 ( by a criterion of duan and wang @xcite ; see section [ ss : final2 ] ) .
in this section we show that every closed five - manifold with finite fundamental group is dominated by products .
we first recall the classification of simply connected , closed five - manifolds by smale @xcite and barden @xcite and show that every five - manifold diffeomorphic to a connected sum of copies of the two @xmath26-bundles over @xmath24 admits a branched double covering by a product @xmath113 .
we then give some existence results of dominant maps between simply connected five - manifolds and using these results we prove theorem [ t:4and5-mfds ] for five - manifolds . given two @xmath2-dimensional manifolds @xmath29 and @xmath114 with boundaries @xmath115 and @xmath116 respectively , we form a new manifold @xmath117 , where @xmath7 is an orientation reversing diffeomorphism of any @xmath118-dimensional submanifold of @xmath116 with one of @xmath115 .
smale @xcite classified simply connected , spin five - manifolds and a few years later barden @xcite completed that classification by including non - spin manifolds as well .
the following constructions are given in @xcite : let @xmath119 , @xmath120 be the two @xmath26-bundles over @xmath24 and @xmath121 , @xmath122 be the two @xmath123-bundles over @xmath24 with boundaries @xmath124 and @xmath125 respectively . as in dimension four , we do nt need to assume that the structure group of oriented @xmath26-bundles over @xmath24 is linear , because the inclusion @xmath126 is a homotopy equivalence , by the proof of the smale conjecture ( cf .
hatcher @xcite ) .
a prime , simply connected , closed , spin five - manifold is either @xmath127 or @xmath128 , if its integral homology groups have no torsion .
if its second homology group is torsion , then is @xmath129 where @xmath130 denotes the boundary connected sum of two copies of @xmath33 and @xmath131 is an orientation preserving diffeomorphism on @xmath132 , realizing a certain isomorphism @xmath133 the second integral homology of @xmath134 is given by @xmath135 , @xmath136 ; cf .
@xcite . for details on the @xmath131 s construction see @xcite .
a prime , simply connected , closed , non - spin five - manifold with torsion - free integral homology is the non - trivial @xmath26-bundle over @xmath24 , denoted by @xmath137 .
a simply connected , closed , non - spin five - manifold with torsion second integral homology is either * @xmath138 , where @xmath139 is an orientation preserving diffeomorphism on @xmath140 , realizing a certain isomorphism @xmath141 or * @xmath142 , @xmath143 , where @xmath144 denotes the boundary connected sum of two copies of @xmath145 and @xmath146 is an orientation preserving diffeomorphism on @xmath147 , realizing a certain isomorphism @xmath148 except @xmath149 , each @xmath150 is prime .
their second integral homology group is given by @xmath151 , @xmath152 and @xmath153 ; cf .
@xcite . also
, @xmath154 is the wu manifold @xmath155 ; cf . @xcite .
finally , we set @xmath156 which is spin and has torsion - free homology . according to the above data , we have the following classification theorem for simply connected five - manifolds .
[ t : barden ] every simply connected , closed five - manifold @xmath29 is diffeomorphic to a connected sum @xmath157 , where @xmath158 , @xmath159 , @xmath160 and @xmath161 divides @xmath162 or @xmath163 .
since every summand , except @xmath164 , is prime , we may refer to the above decomposition of @xmath29 as prime decomposition .
moreover , a summand @xmath165 exists if and only if @xmath29 is not spin . for a simply connected
, closed five - manifold with torsion - free homology the following particular classification result holds : [ t : basm ] every simply connected , closed five - manifold @xmath29 with torsion - free second homology group @xmath166 is ( up to diffeomorphism ) 1 . either a connected sum @xmath167 , for @xmath29 spin , or 2 . a connected sum @xmath168 , for @xmath29 non - spin . [ r : wall2 ] by a theorem of wall @xcite , a connected sum @xmath168 is diffeomorphic to @xmath169 . in remark [ r : wall1 ]
, we refer to the corresponding statement in dimension four .
this is also true for all oriented @xmath23-bundles over @xmath24 , @xmath170 , with structure group @xmath25 .
the above classification results will be our guide to prove theorem [ t:4and5-mfds ] for five - manifolds .
first , we can prove this theorem for every simply connected five - manifold with torsion - free second homology group : [ p : torsionfree ] let @xmath29 be a simply connected , closed five - manifold with torsion - free second homology group @xmath166
. then @xmath29 admits a branched double covering by the product @xmath113 .
we know by theorem [ t : basm ] and by remark [ r : wall2 ] that such @xmath29 is diffeomorphic to a connected sum of copies of the twisted product @xmath171 or of the trivial bundle @xmath119 . now theorem [ t : main ] and theorem [ t : general ] imply the proof , for @xmath172 and @xmath167 respectively . in the remainder of this section we complete the proof of theorem [ t:4and5-mfds ] for five - manifolds .
we observe that @xmath173 is a branched double covering of the other two prime , simply connected , closed five - manifolds with torsion - free homology , namely of @xmath174 and @xmath175 . indeed : * a branched double covering @xmath176 is obtained by pulling back the @xmath26-bundle map @xmath177 by a branched double covering @xmath178 ( recall that @xmath26-bundles over @xmath24 are classified by @xmath179 ) .
* @xmath173 is a branched double covering of @xmath175 , by pulling back the @xmath50-bundle map @xmath180 by a branched double covering @xmath181 ; cf .
( a branched double covering @xmath181 is obtained as the quotient map for the involution @xmath182 of @xmath183 . ) as we shall see below , every simply connected , closed five - manifold @xmath29 with torsion second integral homology group admits a non - zero degree map by @xmath173 , given by the composition of a branched double covering @xmath184 with a dominant map @xmath185 whose degree depends on @xmath186 .
the latter map will be obtained by applying the hurewicz theorem modulo a serre class of groups .
a serre class of abelian groups is a non - empty class @xmath187 of abelian groups such that for every exact sequence @xmath188 , where @xmath189 , then @xmath190 .
a serre class @xmath187 is called a ring of abelian groups if it is closed under the tensor and the torsion product operations .
moreover , @xmath187 is said to be acyclic if for every aspherical space @xmath191 with @xmath192 , then the homology groups @xmath193 , for all @xmath194 .
we say that @xmath195 are isomorphic modulo @xmath187 if there is a homomorphism between @xmath33 and @xmath145 whose kernel and cokernel belong to @xmath187 .
the hurewicz theorem modulo a serre class states the following : let @xmath191 be a simply connected space and @xmath187 be an acyclic ring of abelian groups .
then the following are equivalent : * @xmath196 , for all @xmath197 , * @xmath198 , for all @xmath197 .
moreover , each of the above statements implies that the hurewicz homomorphism @xmath199 is an isomorphism modulo @xmath187 for all @xmath200 . as a consequence of this version of the hurewicz theorem ,
every simply connected , closed @xmath2-dimensional manifold @xmath29 whose homology groups @xmath201 are all @xmath20-torsion , for @xmath202 , is minimal with respect to the domination relation ( i.e. it is dominated by every other manifold ) .
[ c : ru ] let @xmath29 be a simply connected , closed @xmath2-dimensional manifold with @xmath20-torsion homology groups @xmath203 , for @xmath202 and some integer @xmath20 .
then the image of the hurewicz homomorphism @xmath204 is given by @xmath205 for some @xmath206 .
in particular , there is a map @xmath207 of degree @xmath208 .
we have now shown that every five - manifold @xmath29 which is a connected sum of copies of @xmath134 and @xmath150 , where @xmath209 and @xmath210 , admits a branched double covering by the product @xmath173 , followed by a map @xmath185 whose degree is determined by @xmath186 , being a power of the least common multiple of the torsion second integral homology groups @xmath211 and @xmath212 . as we have seen above , the non - trivial @xmath26-bundle over @xmath24 admits a branched double covering by the product @xmath173 .
we now want to combine these maps , together with our constructions from section [ s : construction ] , to obtain domination by products for every simply connected , closed five - manifold .
we remark that it is not generally possible to obtain a non - zero degree map @xmath213 by connected summing any two non - zero degree maps between closed @xmath2-dimensional manifolds , @xmath214 , @xmath215 .
a first obstruction is that the preimage @xmath216 of a removed @xmath2-disk @xmath217 from @xmath218 , is not generally an @xmath2-disk in @xmath219 .
moreover , even if we can find such a disk , we need the degrees of @xmath220 and @xmath221 to be equal in order to connected sum the domains and the targets , otherwise we can not paste those maps along the gluing sphere ( recall that maps between spheres are classified by their degrees ) .
if the above two constraints are satisfied , then we can paste those @xmath222 together to obtain a new map @xmath213 of the same non - zero degree .
the @xmath17-surjectivity of the @xmath222 is a sufficient condition to overcome the first obstacle ( cf . @xcite ) .
[ l : consum ] let @xmath219 , @xmath218 be connected , oriented , closed @xmath2-dimensional manifolds , @xmath223 , and assume that for @xmath224 there exist @xmath17-surjective maps @xmath225 of non - zero degree @xmath4 .
then there is a @xmath17-surjective map @xmath226 of degree @xmath4 . for simply connected targets the @xmath17-surjectivity condition is automatically satisfied .
moreover , @xmath173 and @xmath175 admit self - maps of any degree , which implies that the domination for every minimal summand by @xmath175 and for every @xmath26-bundle over @xmath24 by @xmath173 can be done by maps of the same degree .
we therefore obtain the following statement which also completes the proof of theorem [ t:4and5-mfds ] for five - manifolds : [ t:5-mfds ] let @xmath29 be a simply connected , closed five - manifold so that @xmath186 has rank @xmath20 .
then @xmath29 admits a non - zero degree map by the product @xmath227 , which is given by the composition of a branched double covering @xmath228 with a map @xmath229 whose degree is determined by the torsion of @xmath186 . by theorem [ t : barden ] ,
a simply connected , closed five - manifold @xmath29 is diffeomorphic to a connected sum @xmath230 , where the summands @xmath231 , @xmath150 are described at the beginning of this section .
clearly , the rank @xmath20 of @xmath186 is equal to the number of @xmath232 and @xmath137 , i.e. the number of @xmath26-bundles over @xmath24 .
furthermore , we may assume that the torsion of @xmath186 is not trivial , otherwise we appeal to proposition [ p : torsionfree ] to deduce that @xmath29 admits a branched double covering by a product @xmath233 . by corollary
[ c : ru ] ( and the comments following that ) we deduce that the part of @xmath230 not containing @xmath234 or @xmath137 admits a dominant map by @xmath175 whose degree is a power of the least common multiple of the torsion second integral homology groups @xmath235 and @xmath212 , @xmath236 . now lemma [ l : consum ] implies that @xmath29 is dominated by a manifold diffeomorphic to @xmath237 ( if @xmath238 , we additionally use the fact that @xmath232 is a branched double covering of @xmath137 , as we have seen above in this section ) .
the degree of @xmath229 is clearly determined ( up to multiplication by two ) by the torsion of @xmath186 .
finally , theorem [ t : general ] implies that @xmath237 admits a branched double covering by @xmath113 , completing the proof of theorem [ t:5-mfds ] . by remark [ r : generalpillow ]
, we also have that @xmath237 admits branched double coverings by @xmath239 and @xmath240 , and branched four - fold coverings by @xmath241 and @xmath242 , where @xmath243 is an oriented , closed surface of genus @xmath20 .
[ r : mappingdegrees5-mfds ] in contrast to theorem [ t:4-mfds ] , the statement of theorem [ t:5-mfds ] does not provide absolute control of the degree of the map @xmath244 following the branched double covering @xmath228 .
this is related to the problem of determining the sets of ( self-)mapping degrees of simply connected five - manifolds , which is essentially still unsolved .
however , an aside result of this paper is that the sets of degrees of maps between the summands @xmath134 and @xmath150 ( for @xmath245 ) are infinite , because every multiple of a power of the torsion second homology of @xmath150 ( resp .
@xmath134 ) can be realized as a degree for a map @xmath246 ( resp .
@xmath247 ) ; recall that @xmath175 admits self - maps of any degree and it is minimal for the domination relation .
in particular , we deduce , without using formality ( see @xcite ) , that the set of self - mapping degrees for every simply connected five - manifold is infinite .
by the above results of wall @xcite and smale @xcite , every 2-connected , closed six - manifold @xmath29 is diffeomorphic to @xmath254 , for some non - negative integer @xmath255 .
the proof now follows by the generalized pillowcase map in remark [ r : generalpillow ] .
we note that @xmath29 admits a degree two map by @xmath260 as well . for take
the branched double covering @xmath261 of corollary [ c:6-manifolds ] and precompose it with a degree one map @xmath262 , using lemma [ l : consum ] and the fact that every sphere is minimal admitting maps of any degree . every dominant map onto a simply connected four - manifold @xmath263 is a priori @xmath17-surjective because the target has trivial fundamental group . however , the branched double covering @xmath264 obtained in theorem [ t : general ] ( for @xmath265 ) is @xmath17-surjective also by construction . in a previous joint work with kotschick
@xcite we have shown that there is a branched double covering @xmath266 which is again @xmath17-surjective ( now the target @xmath267 is not simply connected ) .
we multiply ( [ eq.2 ] ) with the identity map on @xmath50 to obtain a @xmath17-surjective branched double covering @xmath268 now we compose ( [ eq.3 ] ) with ( [ eq.1 ] ) to obtain a dominant map @xmath269 . this map is not a branched covering , not even in the case where @xmath114 is diffeomorphic to @xmath102 , because the branch locus of ( [ eq.3 ] ) intersects the preimage of the branch locus of ( [ eq.1 ] ) in a codimension four subset of @xmath93 .
nevertheless , we note that @xmath270 is a branched four - fold cover of @xmath271 , by remark [ r : generalpillow ] . the existence of dominant maps from products @xmath270 to every simply connected
four - manifold has also been shown by kotschick and lh @xcite , using a result of duan and wang @xcite .
that result states that if @xmath191 and @xmath272 are oriented , closed four - manifolds and @xmath272 is simply connected , then a degree @xmath273 map @xmath5 exists if and only if the intersection form of @xmath272 , multiplied by @xmath4 , is embedded into the intersection form of @xmath191 , where the embedding is given by @xmath274 ( the `` only if '' part is obvious ) .
as we mention in section [ s:4-mfds ] , the above existence criterion by duan and wang implies that every integer can be realized as the degree of a map from @xmath111 to @xmath112 , although no such map can be deformed to a branched covering by a recent result of pankka and souto @xcite .
however , according to theorem [ t:4-mfds ] , it is natural to ask when a @xmath17-surjective non - zero degree map between two connected , oriented , closed four - manifolds is homotopic to the composition of a branched covering with a pinch map .
another natural question in this context is the following ( suggested to me by j. souto ) : let @xmath114 be a connected , oriented , closed four - manifold and @xmath275 a @xmath17-surjective map of non - zero degree .
is there a @xmath276 so that the composition of the pinch map @xmath277 with @xmath7 is homotopic to a branched covering ?
kotschick and lh @xcite asked whether every closed manifold with finite fundamental group is dominated by a product . in this paper , we have answered in positive that question in dimensions four and five and in certain higher dimensions . for all our dominant maps onto a simply connected @xmath2-dimensional manifold ,
the domain can be taken to be a product of the circle with a connected sum of copies of @xmath278 . according to this
, one could ask whether every closed manifold with finite fundamental group is dominated by a product of type @xmath18 , where @xmath114 is an @xmath118-dimensional manifold ( e.g. @xmath279 for some @xmath276 ) , or more generally , by a product of type @xmath280 . this is a considerably stronger question than that of @xcite
, however , it seems less likely to be true .
for instance , simply connected manifolds which admit self - maps of absolute degree at most one might not be dominated by products ( at least ) of type @xmath280 .
such manifolds exist and are called inflexible ; see for example @xcite . in this direction , and since the set of self - mapping degrees of @xmath280 is infinite , we ask the following ( at least in the simply connected case ) : are there examples of inflexible manifolds that are dominated by flexible ones ( i.e. by manifolds whose sets of self - mapping degrees are infinite ) ?
d. ruberman , _ null - homotopic embedded spheres of codimension one _ , math .
publ . , * 32 * , 229232 , in _ tight and taut submanifolds _ , edited by t. e. cecil and s .- s .
chern , cambridge univ . press , cambridge 1997 .
a. v. zhubr , _ closed simply connected six - dimensional manifolds : proofs of classification theorems _ , ( russian ) , algebra i analiz * 12 * no . 4 ( 2000 ) , 126230 ; translation in st . petersburg math .
j. * 12 * no .
4 ( 2001 ) , 605680 .
|
we construct branched double coverings by certain direct products of manifolds for connected sums of copies of sphere bundles over the 2-sphere . as an application we answer a question of kotschick and
lh up to dimension five .
more precisely , we deduce that every simply connected , closed four - manifold admits a branched double covering by a product of the circle with a connected sum of copies of @xmath0 , followed by a collapsing map .
furthermore , every simply connected , closed five - manifold admits a branched double covering by a product of the circle with a connected sum of copies of @xmath1 , followed by a map whose degree is determined by the torsion of the second integral homology group of the target .
|
non - relativistic quantum chromodynamics(nrqcd ) factorization formalism@xcite is an useful tool for analyzing the inclusive production of the heavy quarkonium . according to it
, the cross section is expressed as a sum of products of short - distance coefficients and nrqcd matix elements .
the short - distance coefficients can be calculated as perturbation series in coupling constant @xmath5 at the scale of the heavy quark mass .
the matrix elements scale in a definite way with the typical relative velocity @xmath6 of heavy quark in the quarkonium state .
thus the production cross sections can be expressed as double expansions both in @xmath5 and in @xmath6 to any desired order .
recently , the cross section of the exclusive process @xmath7 at the center - of - momentum(cm ) energy @xmath8gev was computed by chung , lee , and yu@xcite , where @xmath9 represents a @xmath1wave or @xmath2wave heavy quarkonium state with charge - conjugation parity @xmath10 , to the leading order(lo ) both in @xmath5 and in @xmath6 in the nrqcd factorization formalism .
their predicted cross sections are about @xmath11 , @xmath12 , @xmath13 , @xmath14 and @xmath15 for @xmath16 , @xmath17 , @xmath18 , @xmath19 and @xmath20 production , respectively
. some early calculations of similar processes like @xmath21 were presented in @xcite in the color - singlet model .
these relatively large cross sections make the processes be measurable at the b factories . to have more precise predictions
, it is necessary to carry out the calculations to higher order both in @xmath5 and in @xmath6
. one may expect that those correction terms are considerable .
especially , for charmonium production , both @xmath5 and @xmath22 are not small . in this paper , in the framework of nrqcd factorization formalism , we compute the short - distance coefficients of the leading nrqcd matrix element to the nlo qcd corrections both for @xmath1wave and @xmath2wave heavy quarkonium production .
for the @xmath16 , we also calculate the tree level short - distance coefficient of the @xmath22 relativistic correction term .
we will not perform the calculation for the nlo relativistic corrections to @xmath23 and @xmath2wave quarkonium production since the nlo nrqcd matrix elements for these states are still not determined preciously by fitting data .
the nlo qcd correction to the @xmath24 was calculated first in @xcite .
however , we found that their original expressions given in eqs .
( 23 ) , ( 25 ) in their paper were incorrect .
probably there were some typos in their equations .
the nrqcd factorization formulism separates the short - distance effects happened at energy scale @xmath25 or higher , where @xmath25 is the heavy quark mass , and the long - distance effects happened at long - distance @xmath26 or larger which describes the formation of the heavy quarkonium .
the short - distance effects are described by the short - distance coefficients while the long - distance ones are described by the nrqcd long - distance matrix elements . in the process @xmath27 ,
the emission of the photon can be either short - distance effects or long - distance ones depending on the emitted photon being hard or soft comparing to the quark mass @xmath25 .
thus when the nrqcd factorization approach is used to analyze the process of @xmath28 at the b factories , the factorization formulism for the cross sections takes different forms in the cases of @xmath9 being @xmath29 or @xmath30 state . in the former case ,
the emitted photon is so hard that the short - distance process is @xmath31 .
however , in the later one , the energy of the emitted photon is much less than the @xmath32 quark mass so that the emission of the photon is a long - distance effect and should be described by the long - distance matrix as an electric @xmath33 or a magnetic @xmath34 transition .
the rest of this article is organized as follows . in section ii
, we apply the nrqcd factorization formulism to analyze @xmath35 at the b factories where we distinguish two different cases of @xmath29 or @xmath30 , corresponding to the emitted photon being hard or soft . in section iii , we determine the short - distance coefficient for the cross section of the process @xmath36 up to the nlo radiative qcd correction for the lo matrix element and the lo short - distance coefficient for the relativistic correction term .
we also discuss various limitations of those short - distance coefficients . in section iv , we determine the short - distance coefficients for the cross sections of the processes of @xmath37 up to the nlo qcd radiative corrections and discuss various limitations .
section v contributes to discussions of the factorization form for the processes of @xmath38 with the emitted photon being soft . in section
vi , we apply the obtained formulae to carry out numerical estimations for the cross sections of those processes .
we also summarize our results in this section .
we present the analytical nlo short - distance coefficients of the leading order nrqcd matrix elements in @xmath6 for the cross sections of @xmath36 and @xmath37 in appendix a and appendix b , respectively .
in the process of @xmath40 , where @xmath9 represents a @xmath1wave or @xmath2wave heavy quarkonium state with c - parity even , an almost on - shell and near threshold @xmath41 pair is created at the short - distance scale of @xmath42 or smaller , then constitutes a heavy quarkonium @xmath9 that happened at the distance scale of @xmath26 .
the production cross section can be analyzed in the nrqcd factorization formulism@xcite as double expansions both in velocity @xmath6 and in coupling @xmath5 .
the factorization takes two different forms in two different cases corresponding to the emitted photon being hard and soft . for charmonium production at the b factories via process @xmath43 , the momentum of the emitted photon satisfies @xmath44 .
the photon is so hard that the short - distance effects arise from the process @xmath45 .
we assign @xmath46 with @xmath47 in the collection of @xmath48 , @xmath49 , @xmath50 and @xmath51 as follows@xcite : @xmath52 the nrqcd factorization formulism for the cross section of @xmath53 ( @xmath54 , @xmath17 , @xmath55 ) takes the following uniform form : @xmath56 where @xmath57 and @xmath58 are nrqcd matrix elements and @xmath59 and @xmath60 are the short - distance coefficients corresponding to the matrix elements .
the leading contributions to the matrix elements arise from these states , which possess the same quantum numbers as @xmath47 . in ( [ eq : formular - hard ] ) , the nrqcd matrix element in the second term is suppressed by @xmath22 compared to that in the first term .
the second term is usually called as the relativistic correction term .
the short - distance coefficients can be expanded as power series of @xmath61 at energy scale of @xmath25 or higher .
they can be determined by matching the cross section of a free quark pair production process @xmath62 with the color - singlet on - shell @xmath29 pair near the threshold .
given the total and the relative momenta of the @xmath29 pair are @xmath63 and @xmath64 , respectively , we then have @xmath65 then the s - matrix element for this process is given by @xmath66 where @xmath67 is the electromagnetic coupling , @xmath68 is the polarization four - vector of the photon and @xmath69 is the square of the center - of - mass(cm ) frame energy .
the leptonic current @xmath70 in ( [ treelevelamp ] ) is defined by @xmath71 and the amplitude @xmath72 is given by @xmath73 , \label{aaa}\end{aligned}\ ] ] where @xmath74 is the projector operator of spin - single ( @xmath75 ) or spin - triplet ( @xmath76 ) state and @xmath77 is the amplitude of @xmath78 with the wave - functions of the external quark lines removed . for the spin - singlet state , @xmath79 reads@xcite @xmath80 and for the spin - triplet state with spin - polarization vector @xmath81 , @xmath82 reads@xcite @xmath83 where @xmath84 . . , width=566 ] at tree level , there are two feynman diagrams contributing to the process as illustrated in fig .
@xmath77 is written as @xmath85,\end{aligned}\ ] ] where @xmath86 is the electric charge of the heavy quark . by expanding the amplitude in terms of @xmath87 ,
the contributions arising from @xmath1wave and @xmath2wave can be picked out .
the cross section in the qcd can then be evaluated by squaring the amplitude and integrating over the phase space .
the cross section of the same process can also be evaluated in the nrqcd factorization formulism with the hadron in the nrqcd matrix elements replaced with the free quark pair being of the same quantum number as @xmath9 .
the short - distance coefficients @xmath88 and @xmath89 can then be inferred from equaling the cross section calculated both in the qcd and in the nrqcd factorization formulism . in our paper , we determine the @xmath90 and @xmath91 to the nlo qcd corrections in @xmath5 and @xmath92 to the lo in @xmath5 . on the other hand , for bottomonium production at the b factories via process @xmath93 , the momentum of the emitted photon yields @xmath94 .
the photon is so soft that the wave - length of the emitted photon is compatible with the size of the bottomonium .
hence the emission process is sensitive to the long - distance effects .
its effects should be absorbed into the long - distance nrqcd matrix elements .
the short - distance effects arise from the contributions of the process @xmath95 with @xmath30 in a spin - triplet and a color - singlet state .
so , the nrqcd factorization formulism for the cross section of @xmath96 takes the following form : @xmath97 where @xmath98 and @xmath99 are the short - distance coefficients .
again , those short - distance coefficients can be inferred from perturbative qcd .
the long - distance matrix elements are non - perturbative in nature .
the short - distance coefficients in this case can be determined by matching the cross section of the free quark pair production process @xmath100 , with the @xmath30 pair near the threshold .
the cross section of this process can be evaluated both in the qcd and in the nrqcd factorization formulism ( [ eq : formular - soft ] ) .
the short - distance coefficients can then be determined by equaling the cross section inferred from both theories .
the short - distance coefficients @xmath98 and @xmath101 in ( [ eq : formular - soft ] ) have been determined to the nnlo@xcite@xcite and to the nlo@xcite in the expansion of @xmath5 in literatures .
in this section , we apply nrqcd factorization formulism ( [ eq : formular - hard ] ) to the process @xmath103 and determine the short - distance coefficients @xmath104 and @xmath105 to the nlo and to the lo in @xmath5 respectively . as mentioned in previous section , they can be determined by matching the process @xmath106 with @xmath29 in the color - singlet @xmath107 state near the threshold .
since the @xmath29 is in the @xmath107 state , lorentz and cpt invariance implies that the tensor @xmath108 can only be expressed as @xmath109 where @xmath110 is a lorentz invariant quantity and can be inferred from perturbative qcd .
consequently , the differential and the total cross section read : @xmath111 here , we use a notation @xmath112 , where @xmath113 is the angle between the photon and the beam line at cm frame .
the cross section for the same process can also be derived in the nrqcd factorization formulism .
thus , we can determine the short - distance coefficients by equaling both results . in this subsection , we determine the short - distance coefficients @xmath114 and @xmath115 to the lo in @xmath5
. they can be obtained by matching the cross section of @xmath116 at tree level both in the qcd and in the nrqcd factorization formulism .
to gain @xmath115 which is the short - distance coefficient of the relativistic correction matrix element , we need to expand the cross section of this process to order @xmath117 . imposing the on - shell condition of the external quarks on ( [ a - mn - tilde ] ) ,
@xmath72 in ( [ aaa ] ) is simplified as @xmath118\;.\end{aligned}\ ] ] the lorentz structure of the projection operator @xmath79 can be written in a general form @xmath119 where @xmath120 , @xmath32 , @xmath121 are lorentz invariant and can be easily inferred from ( [ projector-1 ] ) .
one see that only the first term in ( [ projector-1 - 1 ] ) gives contribution to ( [ a - mn ] )
. inserting this term into ( [ a - mn ] ) , we find @xmath122 it readily follows that @xmath123 where we have performed a substitution@xcite @xmath124 meanwhile , expand @xmath120 with respect to @xmath117 , to obtain @xmath125 combining ( [ a - a ] ) and ( [ a ] ) , we present the total cross section ( [ sigma - tot ] ) in the expansion of @xmath117 as : @xmath126 where @xmath127 and superscript ( 0 ) denotes the contribution arising from the lo in @xmath5 .
next , we turn to calculate the production cross section of @xmath128 in the nrqcd factorization formulism ( [ eq : formular - hard ] ) . at tree level ,
the matrix elements give : @xmath129 inserting them into ( [ eq : formular - hard ] ) and then expanding ( 2 ) in terms of @xmath130 up to order @xmath130 , we obtain @xmath131 by equaling ( [ eq : formular - hard-2 ] ) with ( [ sigma - tot-1 ] ) , we determine the short distance coefficients @xmath104 and @xmath105 to the lo in @xmath5 : @xmath132 finally , the cross section of @xmath24 to the lo in @xmath5 and the nlo in @xmath22 is given as : @xmath133.\end{aligned}\ ] ] here the two hadron matrix elements are normalized relativistically , which is related to the standard non - relativistic normalized hadron matrix elements@xcite by @xmath134),{\nonumber}\\ \langle0|\chi^\dagger(-\frac{i}{2}\tensor{{\bf d}})^2\psi|\eta_c\rangle&=&\sqrt{4m}\langle0|\chi^\dagger(-\frac{i}{2}\tensor{{\bf d}})^2\psi|\eta_c\rangle_{bbl}.\end{aligned}\ ] ] inserting ( [ transfer ] ) into ( [ nrqcdsigma ] ) , we obtain the cross section with the hadron matrix elements in non - relativistic normalization as @xmath135.\end{aligned}\ ] ] . ,
width=529 ] in this subsection , we determine the short - distance coefficients @xmath104 to the nlo in @xmath5 . to this end , we need to match the cross section of the process @xmath136 to the one - loop level but the lo in @xmath22 both in the qcd and in the nrqcd factorization formulism .
the feynman diagrams responsible for the one - loop qcd correction are illustrated in fig .
we perform the renormalization in on - mass - shell(os ) scheme and take dimensional regularization to regulate both the ultraviolet(uv ) and the infrared(ir ) divergences . in os scheme ,
external quark lines do not receive any qcd corrections , and the counterterms from the renormalization constants of the heavy quark wavefunction and the heavy quark mass are given by@xcite : @xmath137 where @xmath138 is the renormalization scale , @xmath139 is the euler s constant , and @xmath140 for @xmath141 . among the eight diagrams in fig .
[ fig2 ] , there are two self - energy diagrams , four triangle diagrams and two box diagrams . the divergences of these diagrams can be analyzed as follows : the self - energy and the triangle diagrams contain uv divergences , while the box diagrams possess ir divergences .
in addition , coulomb singularities may arise from the box diagrams , due to the exchange of longitudinal gluon between @xmath121 and @xmath142 .
our calculations indicate that the uv divergences from the self - energy and the triangle diagrams are canceled by that from the renormalization constants of the quark field and the mass and ir divergences from the box diagrams are canceled by that in @xmath143 .
as one will see , the remaining coulomb singularities will be canceled by coulomb singularities in nrqcd matrix elements by matching condition . in our practical calculations ,
the ` feycalc ' package@xcite has been employed to take the trace of the @xmath144matrix and to reduce tensor integrals into scalar ones .
furthermore , we carry out the ir - safe scalar integrals by using g.t hooft and m. veltman s method@xcite . for the ir - divergent scalar integrals ,
we take the results from ref . @xcite . after canceling all these uv and ir divergences ,
we present the nlo cross section of @xmath145 : @xmath146 where @xmath147 is the lo cross section ( [ sigma - tot-1 ] ) with @xmath148 and @xmath149 is a function of @xmath150 and is both uv and ir finite .
the explicit expression for @xmath149 is given in appendix [ appendix1 ] .
subsequently , we turn to deal with the same process in the nrqcd factorization formulism .
the main ingredient of this calculation is to obtain the nrqcd matrix element to the nlo in @xmath5 , which reads@xcite , @xmath151 where the coulomb singularity term emerges from the coulomb gluon exchanging between the quark and the anti - quark .
the total cross section of @xmath152 in the nrqcd factorization formulism can be readily expressed as @xmath153 by equaling ( [ includec ] ) with ( [ nlonrqcd matrixelement ] ) , we determine the nlo short - distance coefficient : @xmath154 from which we see that the coulomb singularity cancels in the short - distance coefficient as expected .
@xmath155 is usually referred to the @xmath156factor . inserting this coefficient into ( [ eq : formular - hard ] )
, we reach the final result for the cross section to the nlo qcd correction in @xmath5 and the lo in @xmath22 : @xmath157 where @xmath158 is given by ( [ nrqcdsigma ] ) with @xmath159 .
it is instructive to look at the behaviors of @xmath149 in two limits of @xmath160 and @xmath161 .
the limit of @xmath160 corresponds to a very high energy production of @xmath16 . in this limit
, @xmath149 reduces to @xmath162.\end{aligned}\ ] ] the coefficient of @xmath163 term agrees with that obtained in ref . @xcite .
the limit of @xmath161 corresponds to the nlo qcd corrections to the process of @xmath16 decay into two photons . to achieve this limit
, one needs to recalculate the @xmath156factor for @xmath164 .
we define @xmath165 in this region via @xmath166 whose expression is given in appendix [ appendix1 ] . with the expression of @xmath167 in hand ,
we take the limit of @xmath161 for @xmath167 , which reads @xmath168 it is encouraging that ( [ limit2 ] ) is consistent with the qcd correction to the cross section of @xmath169 . comparing our results with
that given in ref .
@xcite , we found that the expressions given in their eqs .
( 23 ) ( 25 ) , from which they obtained the two limits of @xmath160 and @xmath170 , were incorrect .
actually , it is easy to check that the results given in their eqs .
( 23 ) ( 25 ) go to infinity by taking these two limits .
therefore we claim that there might be some typos in their results .
in this section , we apply the nrqcd factorization formulism ( [ eq : formular - hard ] ) to the processes @xmath172 with @xmath173 and determine the short - distance coefficients @xmath174 to the nlo in @xmath5 . as mentioned in sec .
[ sec:2 ] , these coefficients can be determined by matching the processes @xmath175 with the @xmath29 near the threshold and in the color - singlet @xmath176 states .
we first determine the lo coefficients in @xmath5 by calculating the cross sections of @xmath177 both in the qcd and in the nrqcd factorization formulism .
two tree - level diagrams contributing to the processes @xmath178 are illustrated in fig .
[ fig1 ] . to carry out the calculations , we need to construct the polarization vectors and tensors for @xmath179 and @xmath180 respectively .
three polarization vectors @xmath181 for @xmath179 can be explicitly expressed as(provided the hadron moving along @xmath182 axis . )
@xmath183 then five polarization tensors for @xmath180 are readily written in terms of @xmath184 as @xmath185 the amplitudes of @xmath186 can be picked out from ( [ a - mn ] ) by extracting the contributions corresponding to the @xmath187 production . to this end
, one needs to expand the amplitudes to the lo in @xmath87 and to project it into the @xmath188 states with helicity @xmath189
. it can be done by using the formulae@xcite@xcite : @xmath190 { \cal p}^{\mu\nu}_j \,,\end{aligned}\ ] ] where @xmath191 is defined via @xmath192 and the projection operators @xmath193 are defined by @xmath194 with these formulae , it is straightforward to calculate the cross sections of the processes @xmath195 with a specific helicity @xmath189 for @xmath29 .
the differential cross sections of these processes in the lo in @xmath5 yield @xmath196 where the functions @xmath197 are @xmath198 since relativistic corrections of @xmath188 production will not be calculated , we have set @xmath199 .
our formulae ( [ coefficient1 ] ) agree with that obtained in ref .
@xcite@xcite .
we now turn to evaluate the cross sections of the same processes in the nrqcd factorization formulism ( [ eq : formular - hard ] ) .
the matrix elements for @xmath200 production are given by @xmath201 thus , the differential cross sections over @xmath202 within the framework of nrqcd factorization formulism read @xmath203 where @xmath204 denote the lo short - distance coefficients . by equaling the cross sections on the qcd side with that on the nrqcd side , we determine the short - distance coefficients : @xmath205
consequently , the differential cross sections of @xmath206 in the lo in @xmath6 and @xmath5 are written as : @xmath207 where the @xmath208 states in the matrix element are normalized relativistically . to express our formulae as the product of short - distance coefficients and standard non - relativistic normalized matrix elements
, we can use the relations between relativistic normalized matrix elements and non - relativistic normalized matrix elements@xcite @xmath209 where @xmath210 is the non - relativistic normalized matrix element used in ref .
@xcite and @xmath211 is the mass of @xmath9 . in the lo in @xmath6 , we set @xmath212 .
the differential cross sections of @xmath213 can be reexpressed as : @xmath214 the total cross sections are achieved by integrating over @xmath202 and summing over all the helicity states @xmath189 of the @xmath215 .
we present the final results : @xmath216 where @xmath217 for @xmath173 are given by @xmath218 in this subsection , we determine the short - distance coefficients in the nrqcd factorization formulism for processes @xmath219 up to the nlo in @xmath5 .
as done in the case of @xmath16 production , we calculate the cross sections of the processes @xmath220 both in the qcd and in the nrqcd factorization formulism to one - loop level .
feynman diagrams responsible for the processes are the same as that of @xmath107 production as illustrated in fig .
we still carry the renormalization in on - mass - shell(os ) scheme .
the wave function renormalization constant and mass renormalization constant have been given in ( [ z - q])([z - mq ] ) .
the uv and ir divergences of each diagram are similar with that of @xmath16 production .
the self - energy and triangle diagrams contain ultraviolet(uv ) divergences .
the box diagrams have both infrared(ir ) divergences and coulomb singularities . combining contributions from the eight diagrams and the counterterms ,
all divergences cancel except coulomb singularities which will be canceled by coulomb singularities in nrqcd matrix elements .
the cross sections of the processes @xmath221 in the nlo in @xmath5 are given by @xmath222 where the index @xmath189 denotes the helicity of @xmath55 , @xmath197 have been given in previous subsection and the expressions for @xmath223 are listed in appendix [ appendix2 ] .
one can immediately get the total cross sections by integrating over @xmath202 and summing over the polarizations of @xmath55 . as a result , the cross sections yield @xmath224 where @xmath225 for @xmath173 are given by @xmath226 we now calculate the cross sections of the same processes in the nrqcd factorization formulism .
the nrqcd matrix elements of @xmath63-wave operators in the lo in @xmath6 and nlo in @xmath5 are given by @xmath227 with it , one readily obtains the cross sections of @xmath228 on the nrqcd side : @xmath229.\end{aligned}\ ] ] the short - distance coefficients @xmath174 can be determined by equaling ( [ pwaveqcd1 ] ) with ( [ pwavenrqcd ] ) , @xmath230.\end{aligned}\ ] ] finally , the cross sections of the processes @xmath231 in the nlo in @xmath5 can be expressed as @xmath232 where @xmath233 denote the lo cross sections given in ( [ lochi ] ) .
it is interesting to look at the behaviors of @xmath225 in two limits of @xmath234 and @xmath161 .
the limit of @xmath160 is related to the @xmath55 production in a very high energy . in this limit ,
we find @xmath235,{\nonumber}\\ \lim_{r\to 0 } c_1(r ) & = & -\frac{2}{9 } \left[(9 - 6 \log2 ) \log r+9\log^2 2 - 15 \log2+\pi ^2 + 21\right],{\nonumber}\\ \lim_{r\to 0 } c_2(r ) & = & -\frac{2}{9 } \left[(3 - 6\log2 ) \log
r+9 \log^2 2 + 3\log2+\pi ^2 + 18\right].\end{aligned}\ ] ] the limit of @xmath161 is related to the nlo qcd corrections to the cross sections of @xmath236 .
as in the case of @xmath16 , one needs to recalculate the @xmath156factor for @xmath237 .
we define @xmath238 via @xmath239 explicit expressions of @xmath240 are listed in appendix [ appendix2 ] . in the limit of @xmath161 , @xmath241 go to : @xmath242 the formulae([plimit2 ] ) are desirable , since these are nothing but the qcd corrections to the processes @xmath243 . in ( [ plimit2 ] ) , we did not take the limit of @xmath161 for @xmath244 , since the cross section of @xmath19 decay into @xmath245 vanishes due to the yang s theory .
as discussed in sec .
ii , for bottomonium production at the b factories through process @xmath247 , the emitted photon is soft comparing to the @xmath32 quark mass .
the cross section of the process is expressed as ( [ eq : formular - soft ] ) in the framework of the nrqcd factorization formulism . the short - distance coefficients @xmath98 and @xmath101 in ( [ eq : formular - soft ] ) can be determined by matching the cross section of the process @xmath248 with the @xmath29 in the color - singlet @xmath249 state predicted by qcd and by the nrqcd factorization formulism .
it has been calculated up to order @xmath250 . here
we list only the @xmath98 result@xcite@xcite @xcite @xcite@xcite : @xmath251\frac{\alpha_s^2}{\pi^2 } \right\}{\nonumber}\\\end{aligned}\ ] ] the matrix elements in ( [ eq : formular - soft ] ) describe probabilities of a point - like @xmath249 @xmath30 state transiting into the @xmath107 or the @xmath188 state through a magnetic @xmath34 transition or an electric @xmath33 transition by emitting one soft photon .
although they are qed processes , the matrix elements are non - perturbative in nature .
one may evaluate those @xmath34 and @xmath33 transition processes by perturbative qcd and relate the matrix elements appearing in ( [ eq : formular - soft ] ) to those appearing in ( [ eq : formular - hard ] ) .
however , this kind of calculations is not reliable . some additional coulomb singularities may arise from gluon exchange between almost on - shell @xmath32 and @xmath252 quarks
terms like @xmath253 appearing in radiative corrections spoil fixed order perturbation calculations on those matrix elements . even at tree level , it is proportional to a factor @xmath254 for the @xmath255 production and @xmath256 for the @xmath257 production
. thus the results are very sensitive to the quark mass . to gain a reliable prediction on the matrix in ( [ eq : formular - soft ] ) , one need to invoke some nonperturbative methods such as lattice qcd .
in this section , we carry our numerical calculations on the production cross sections of @xmath258 at the b factories using the results derived in section iii and section iv . to this end , we need to determine some relevant parameters such as quark mass , @xmath259 , and those nrqcd matrix elements .
we will estimate them below .
[ sec : input parameters ] at the b factories , we take @xmath260 and @xmath261 . the quark mass is taken as @xmath262 .
there is an ambiguity in choosing the value of @xmath138 in @xmath5 since there are several energy scales involved , such as @xmath263 and heavy quark mass . in the limit of @xmath160
, one can use the leading log approximation ( lla ) to resume those large logarithmic terms such as @xmath264@xcite .
this accounts for the effects of the running of the strong coupling constant @xmath5 from energy scale @xmath263 to @xmath25 .
however , in doing this approximation , one has omitted constant terms and terms suppressed by power of @xmath150 . to recover these contributions
, one needs to carry out calculations to higher loops and to include contributions from higher dimensional operators .
those are much more complicated . in our full one - loop results ,
we keep all these terms but leave an ambiguity in choosing the value of @xmath138 in @xmath259 . to estimate the uncertainty caused by it , we take three different values of @xmath138 .
they are @xmath263 , @xmath265 , and @xmath266 .
the corresponding values of @xmath259 are @xmath267 , @xmath268 , and @xmath269 . as for the hadron matrix elements
, we will take them as same as those used in @xcite@xcite@xcite , @xmath270 for the @xmath16 production , in order to assess the contribution from the relativistic correction , we will evaluate the nrqcd matrix element corresponding to the relativistic correction by using the gremm - kapustin@xcite relation , @xmath271 we take @xmath272 for @xmath16 as that in ref .
@xcite and the correspondent nlo nrqcd matrix element yields : @xmath273 for the @xmath17 production , the lo nrqcd matrix element , which is decided by fitting data from @xmath274 $ ] based on heavy - quark spin symmetry , is accurate only in the lo in @xmath6 as pointed out in ref .
therefore we will not consider the relativistic correction to @xmath17 production for consistence . in the following subsection , we will use these input parameters to calculate the production cross sections . with the parameters given above
, we can evaluate the cross sections for the production of @xmath16 , @xmath17 , and @xmath55 by using eqs .
( [ nrqcdsigma ] ) , ( [ nloeta ] ) , ( [ lochi ] ) , and ( [ nlochi ] ) . for comparison , we list the results of the lo and the nlo in @xmath5 in table [ table1 ] . for the nlo results , we have taken @xmath138 in @xmath259 to be @xmath263 , @xmath265 , and @xmath275 . in tabel [ table1 ] ,
we have omitted relativistic corrections . from the table
, we see that most of the qcd radiative corrections are negative and considerable . for some processes like @xmath180 production , the corrections are unexpectedly large
. then we can estimate the relativistic corrections by eq .
( [ nrqcdsigma ] ) . after accounting for the nlo radiative
qcd correction(we take @xmath276 in @xmath259 ) and the relativistic correction , the total cross section of the @xmath16 production decrease from 83.3fb to 58.2fb , where the nlo radiative qcd correction and relativistic correction contribute -15.3fb and -9.81fb respectively .
we see that both of them give considerably negative contributions .
.[table1 ] predicted cross sections ( in fb ) for the production of @xmath16 , @xmath17 , and @xmath55 with various @xmath138 . @xmath277 and @xmath278 are the cross sections with the lo and the nlo short - distance coefficients , repectively .
@xmath279 is relative corrections caused by the nlo corrections . [
cols="<,^,^,^,^,^,^,^ " , ] in @xmath280 collisions , the heavy quarkonium with @xmath281parity even can also be produced via the double photon processes .
at very high energy limit , this process can be described by the equivalent photon approximation(epa ) . comparing to the processes @xmath282 ,
the total production cross sections of the double photon processes are suppressed by additional power of @xmath283 but enhanced by @xmath284@xcite .
the cross sections of the @xmath285 channel processes @xmath258 are suppressed by @xmath286 . combining all these factors , at the b factories ,
the production cross sections of the heavy quarkonium with @xmath281parity even via the double photon process are larger than that via associated production with photon . however , the signals in detectors from these two production mechanisms are very different .
especially , the produced heavy quarkonia via double photon processes are predominated by small @xmath287 events at high energy which are hard to be detected from their decay products . on the other hand ,
the produced heavy quarkonia via associated production with photon are insensitive to the @xmath287 .
moreover , the events can be reconstructed by measuring the photon assuming the photon detector is good enough .
a thorough analysis on these processes in the framework of the nrqcd factorization formalism will be presented elsewhere .
after this work was finished , we were told that d. li , z.g .
he , and k.t .
chao had done a similar work@xcite .
one of the authors , wen - long sang would like to thank g. t. bodwin and yu jia for useful discussion and to thank zhi - guo he for checking results .
this work was supported partly by the national natural science foundation of china ( nnsfc ) under no.10875156 .
[ appendix1 ] the definition for @xmath149 appeared in eq .
( [ includec ] ) : @xmath288}{9(2-r)(1-r)}+\frac{8(2r-3)}{3(2-r)^2}\log(\frac{2}{r}-2)-\frac{4}{b}\log(\frac{1-b}{1+b}){\nonumber}\\ & + & \frac{2}{3(r-1)}[(1+\frac{r}{2})\log^2(\frac{1-b}{1+b})-\log^2(\frac{2}{r}-1)]+\frac{4}{3(1-r)}{\rm li_2}(\frac{r}{2-r}),\end{aligned}\ ] ] where we use @xmath289 , @xmath290 and @xmath291 in eq .
( [ includea1 ] ) .
in addition , dilogarithm(@xmath292 ) is used in the appendices [ appendix1 ] and [ appendix2 ] , which is defined as @xmath293 where @xmath294 .
the definition for @xmath295(r ) appeared in eq .
( [ etacprime ] ) : @xmath296\},\end{aligned}\ ] ]
[ appendix2 ] @xmath298{\nonumber}\\ & + & \frac{1}{3 ( r-1)^2(3r-1)}\left\{\frac{2(4r^4 - 7r^3 + 11r^2-r+1)\log^2(\frac{2}{r}-1)}{1+r^2}\right.{\nonumber}\\ & + & ( 3r^3 - 6r^2 - 3r+2)\log^2(\frac{1-b}{1+b})+\frac{4(1-r)}{b}(3r^2 - 12r+5)\log(\frac{1-b}{1+b}){\nonumber}\\ & -&\frac{8(2r^4 - 5r^3 + 6r^2 - 2r+1)\log(\frac{2}{r}-2)\log(\frac{2}{r}-1)}{(1+r^2)}{\nonumber}\\ & + & \left.\frac{8(8r^5 - 50r^4 + 125r^3 - 152r^2 + 87r-18)\log(\frac{2}{r}-2)}{(r-2)^3}\right\}.{\nonumber}\\ & + & \frac{4 \left[\left(4 r^4 - 7 r^3 + 11 r^2-r+1\right ) \text{li}_2\left(\frac{r}{2-r}\right)-2 \left(2 r^4 - 5 r^3 + 6 r^2 - 2 r+1\right ) \text{li}_2\left(2-\frac{2}{r}\right)\right]}{3 ( 3r-1)(1-r)^2(1+r^2)}.\end{aligned}\ ] ] @xmath299}{3b(r-2)(r-1)}{\nonumber}\\ & + & \frac{[\left(4 r^2 - 19 r+5\right ) \log \left(\frac{2}{r}-1\right)-2 \left(r^2 - 16 r+3\right ) \log \left(\frac{2}{r}-2\right)]\log \left(\frac{2}{r}-1\right)}{3(r-1)^2}{\nonumber}\\ & + & \frac{2 \left(4 r^2 - 19 r+5\right ) \text{li}_2\left(\frac{r}{2-r}\right)-2 \left(r^2 - 16 r+3\right ) \text{li}_2\left(2-\frac{2}{r}\right)}{3 ( r-1)^2}.{\nonumber}\end{aligned}\ ] ] @xmath300}{9(r-1)(r-2 ) } + \frac{4(r^2 - 2r-1)\log \left(\frac{1-b}{1+b}\right)}{3b(1-r)}{\nonumber}\\ & + & \frac{8(3r^3 - 10r^2 + 9r-1)\log(\frac{2}{r}-2)}{3(r-2)^2(r-1)}+\frac{(r-2)(r^2 + 1)\log^2(\frac{1-b}{1+b})}{3 ( r-1)^2}{\nonumber}\\ & -&\frac{2 \log \left(\frac{2}{r}-1\right ) [ 2(1 - 2r ) \log(\frac{2}{r}-2)+(r-1)(r^2-r+2)\log(\frac{2}{r}-1)]}{3 ( r-1)^2}{\nonumber}\\ & -&\frac{4[(1 - 2 r ) { \rm li}_2(2-\frac{2}{r})+(r-1)(r^2-r+2){\rm li}_2(\frac{r}{2-r})]}{3 ( r-1)^2}.\end{aligned}\ ] ] @xmath301}{9 ( r-1)^2}+\frac{8r^2[4 b \log \left(\frac{2}{r}-2\right)-(r-5 ) \log \left(\frac{1-b}{b+1}\right)]}{3b(r-1)}{\nonumber}\\ & + & \frac{2r^2\{\log \left(\frac{2}{r}-1\right)[(r+3 ) \log \left(\frac{2}{r}-1\right)-2 ( r-1 ) \log
\left(\frac{2}{r}-2\right)]-4 \log ^2\left(\frac{1-b}{b+1}\right)\}}{3 ( r-1)^2}{\nonumber}\\ & + & \frac{4 r^2[(r+3 ) \text{li}_2\left(\frac{r}{2-r}\right)-(r-1 ) \text{li}_2\left(2-\frac{2}{r}\right)]}{3 ( r-1)^2},{\nonumber}\end{aligned}\ ] ] @xmath302}{3 ( r-1)^2}{\nonumber}\\ & -&\frac{4[(19 r+3)(r-2)^2\log(\frac{1-b}{1+b})+b(10 r^3 - 49 r^2 + 51 r+10 ) \log(\frac{2}{r}-2)]}{3b(r-2)^2(r-1)}{\nonumber}\\ & -&\frac{2[\left(5 r^2+r+1\right ) \text{li}_2\left(2-\frac{2}{r}\right)-(2 r^2 - 20 r+3 ) \text{li}_2\left(\frac{r}{2-r}\right)]}{3 ( r-1)^2},{\nonumber}\end{aligned}\ ] ] for @xmath305 : @xmath306{\nonumber}\\ & + & \frac{2 c_f(2r-3)(4r^3 - 15r^2 + 19r-6)\log(2-\frac{2}{r})}{(r-2)^3(r-1)(3r-1)}-\frac{c_f \text{li}_2\left(\frac{2}{r}-1\right)}{(r-1)^2}{\nonumber}\\ & + & \frac{c_f}{(r-1)^2(3r-1)}\tan^{-1}(\sqrt{r-1})[2\sqrt{r-1}(3r^2 - 12r+5){\nonumber}\\ & + & \pi(3 r^3 - 6 r^2 - 3 r+2)-(3 r^3 - 6 r^2 - 3 r+2 ) \tan ^{-1}\left(\sqrt{r-1}\right)].{\nonumber}\\\end{aligned}\ ] ] for @xmath307 : @xmath308{\nonumber}\\ & + & \frac{c_f
\left(2 r^4 - 3 r^3 - 9 r^2 + 16 r-2\right ) \log
\left(2-\frac{2}{r}\right)}{(r-2)^2 \left(r^2 - 1\right)}-\frac{c_f \left(r^3+r^2 + 2\right ) \text{li}_2\left(\frac{2}{r}-1\right)}{2 ( r-1)^2 ( r+1)}{\nonumber}\\ & + & \frac{c_f}{(r-1)^2(r+1)}\tan^{-1}(\sqrt{r-1})[-2\sqrt{r-1}(2r^3 - 2r^2 + 3r+1){\nonumber}\\ & -&\pi(2 r^3-r^2+r+2)+(2r^3-r^2+r+2)\tan^{-1}(\sqrt{r-1})].\end{aligned}\ ] ] for @xmath309 : @xmath310{\nonumber}\\ & + & \frac{c_f \left(30 r^5 - 157 r^4 + 249 r^3 - 76 r^2 - 78 r+12\right ) \log \left(2-\frac{2}{r}\right)}{(r-2)^3(r-1)(6r^2 + 3r+1)}{\nonumber}\\ & + & \frac{c_f \left(9 r^3 - 21 r^2 - 6 r-2\right ) \text{li}_2\left(\frac{2}{r}-1\right)}{2 ( r-1)^2 \left(6 r^2 + 3 r+1\right)}{\nonumber}\\ & -&\frac{c_f}{(r-1)^2(6 r^2 + 3 r+1)}\tan^{-1}(\sqrt{r-1})[-2\sqrt{r-1}(12r^3 - 24r^2 - 9r+1){\nonumber}\\ & + & \pi(-6 r^3 + 15 r^2 + 9 r+2)+(6 r^3 - 15 r^2 - 9 r-2 ) \tan ^{-1}(\sqrt{r-1})]{\nonumber}\\\end{aligned}\ ] ] g. t. bodwin , e. braaten , and g. p. lepage , phys . rev .
d * 51 * , 1125 ( 1995 ) [ erratum - ibid .
d * 55 * , 5853 ( 1997 ) ] [ arxiv : hep - ph/9407339 ] .
h. s. chung , j. lee , and c. yu , phys .
d*78 * , 074022,(2008 ) [ arxiv : hep - ph/08081625 ] b. guberina , j. h. kuhn , r. d. peccei and r. ruckl , nucl .
b * 174 * , 317 ( 1980 ) .
m. a. shifman and m. i. vysotsky , nucl .
b * 186 * , 475 ( 1981 ) .
g. t. bodwin and a. petrelli , phys .
d * 66 * , 094011 ( 2002 ) [ arxiv : hep - ph/0205210 ] .
a. czarnecki and k. melnikov , phys .
lett * 80 * , 2531 ( 1998 ) [ arxiv : hep - ph/9712222 ] .
m. beneke , a. signer and v. a. smirnov , phys .
lett * 80 * , 2535 ( 1998 ) [ arxiv : hep - ph/9712302 ] .
g. t. bodwin , h. s. chung , j. lee and c. yu , phys .
d * 79 * , 014007 ( 2009 ) [ arxiv:0807.2634 ] .
g. t. bodwin and j. lee , phys .
d * 69 * , 054003(2004 ) .
e. braaten and y. q .chen , phys .
d * 57 * , 4236 - 4253 ( 1998 ) [ erratum - ibid .
d * 59 * , 079901 ( 1999 ) ] [ arxiv : hep - ph/9710357 ] .
y. j. zhang and k. t. chao , phys .
d * 78 * , 094017 ( 2008 ) [ arxiv:0808.2985 ] .
r. mertig , m. b@xmath311hm and a. denner , comput .
. commun . * 64 * , 345 ( 1991 ) e. braaten and j. lee , phys .
d * 67 * , 054007 ( 2003 ) [ erratum - ibid .
d * 72 * , 099901 ( 2005 ) ] [ arxiv : hep - ph/0211085 ] .
e. braaten and y. q. chen , phys .
lett * 76 * , 730 - 733 ( 1996 ) [ arxiv : hep - ph/9508373 ] .
r. barbieri , r. gatto , r. k@xmath311gerler and z. kunszt , phys .
* 57b * , 455 ( 1975 ) . w. celmaster , phys .
d * 19 * , 1517 ( 1979 ) .
g. t. bodwin , h. s. chung , d. kang , j. lee , and c. yu , phys .
d * 77 * , 094017 ( 2008 ) [ arxiv:0710.0994 ] .
g. t. bodwin , d. kang and j. lee , phys .
d * 74 * , 014014 ( 2006 ) [ arxiv : hep - ph/0603186 ] .
m. gremm and a. kapustin , phys .
b * 407 * , 323 ( 1997 ) [ arxiv : hep - ph/9701353 ] .
|
the cross sections of @xmath0 to @xmath1wave and @xmath2wave quarkonia with @xmath3-parity even associated with a photon are analyzed in the framework of non - relativistic quantum chromodynamics(nrqcd ) factorization formulism .
the short - distance coefficients are analytically determined up to the next - to - leading order(nlo ) qcd radiative corrections to @xmath1wave and @xmath2wave quarkonium production and nlo relativistic correction to @xmath4 production .
we also invoke the analytical expressions to estimate the cross sections .
our numerical results indicate that both the qcd and the relativistic corrections are considerable .
|
two - dimensional ( 2d ) carbon crystals are hosts for dirac - type electrons , whose unusual properties have been studied extensively in graphene monolayers produced by mechanical exfoliation from graphite.@xcite a close relative of graphene , a 2d honeycomb lattice of si atoms called
_ silicene_,@xcite does not occur in nature , but nanoribbons of silicene have been synthesized on metal surfaces.@xcite due to the similarity of the lattice structures , the band structure of silicene resembles that of graphene , featuring dirac - type electron dispersion in the vicinity of the corners of its hexagonal brillouin zone ( bz).@xcite moreover , silicene has been shown theoretically to be metastable as a free - standing 2d crystal,@xcite implying that it is possible to transfer silicene onto an insulating substrate and gate it electrically . in this work
we predict the properties of this 2d crystal .
the similarity between graphene and silicene arises from the fact that c and si belong to the same group in the periodic table of elements .
however , si has a larger ionic radius , which promotes sp@xmath0 hybridization , whereas sp@xmath1 hybridization is energetically more favorable in c. as a result , in a 2d layer of si atoms , the bonding is formed by mixed sp@xmath1 and sp@xmath0 hybridization .
hence silicene is slightly buckled , with one of the two sublattices of the honeycomb lattice being displaced vertically with respect to the other , as shown in fig .
[ fig : silicene_in_capacitor ] .
such buckling creates new possibilities for manipulating the dispersion of electrons in silicene and opening an electrically controlled sublattice - asymmetry band gap.@xcite in this article we report density functional theory ( dft ) calculations of the band gap @xmath2 for dirac - type electrons in silicene opened by a perpendicular electric field using a combination of top and bottom gates .
we show that @xmath2 can reach tens of mev before the 2d crystal transforms into a semimetal and then , at still higher fields , loses structural stability .
we also determine the weak electric field at which electrons in silicene experience a transition from a topological insulator regime@xcite caused by the kane - mele spin - orbit ( so ) coupling@xcite for electrons on a honeycomb lattice into a conventional band insulator regime .
, scaledwidth=45.0% ] the rest of this article is arranged as follows . in sec .
[ sec : struct_electronic ] we report our results for the structural and electronic properties of free - standing silicene , and compare them with other theoretical and experimental results in the literature . in sec .
[ sec : transverse_field ] we analyze the effects of a transverse electric field on the structural and electronic properties of silicene , and in sec . [ sec : spin_orbit ] we discuss the effects of so coupling on the electronic structure , arguing that a crossover from topological insulating behavior to band insulating behavior must take place as the transverse field increases in strength . in sec .
[ sec : comp_meth ] we give the technical details of our computational methodology and demonstrate the convergence of our results with respect to simulation parameters . finally , we draw our conclusions in sec . [
sec : conclusions ] .
the lattice constant and the @xmath3 ( out - of - plane ) coordinates of the si atoms lying on the 2d honeycomb lattice were both fully relaxed using dft ( i ) in the local density approximation ( lda ) , ( ii ) with the perdew - burke - ernzerhof ( pbe ) exchange - correlation functional,@xcite and ( iii ) with the screened heyd - scuseria - ernzerhof 06 ( hse06 ) hybrid functional.@xcite our dft calculations were performed using the castep@xcite and vasp@xcite plane - wave - basis codes , using ultrasoft pseudopotentials and the projector - augmented - wave ( paw ) method , respectively .
the @xmath3-coordinates of the two si atoms in the unit cell ( the @xmath4 and @xmath5 sublattices ) differ by a finite distance @xmath6 .
our results for a free silicene monolayer are shown in table [ table : parameters_summary ] . the metastable lattice that we find is the same as the `` low - buckled '' structure found by cahangirov _ _ et al.__@xcite .silicene structural and electronic parameters : lattice constant @xmath7 , sublattice buckling @xmath6 ( the difference between the @xmath3 coordinates of the @xmath4 and @xmath5 sublattices ) , cohesive energy @xmath8 , and fermi velocity @xmath9 . the calculated cohesive energy of silicene includes the dft - pbe zero - point energy , which we found to be @xmath10 ev per atom .
the theoretical results are for free - standing silicene ; the experimental results are for silicene nanoribbons on ag substrates . [ table : parameters_summary ] [ cols="<,^,^,^,^",options="header " , ] the experimental results for the lattice parameter depend on the choice of substrate on which the silicene is grown.@xcite the extent to which theoretical results obtained for free - standing silicene are applicable to the silicene samples that have been produced to date is therefore unclear .
the cohesive energy of bulk si ( including a correction for the zero - point energy ) has been calculated within dft - lda as 5.34 ev.@xcite comparing this with our dft - lda cohesive energy of silicene reported in table [ table : parameters_summary ] shows that bulk si is substantially ( 0.22 ev per atom ) more stable than silicene , implying that silicene would not grow naturally as a layered bulk crystal like graphite . however , by calculating the dft phonon dispersion it has been verified both here and in ref . that the structure is dynamically stable
: no imaginary frequencies appear anywhere in the bz .
the results of such an analysis are summarized in fig .
[ fig : phonons_in_field ] .
this convinces us that , as a metastable 2d crystal , silicene can be transferred onto an insulating substrate , where its electronic properties can be studied and manipulated as suggested below .
v@xmath11 . in both cases the calculations
were performed using the method of finite displacements , with the atomic displacements being 0.0423 , in a supercell consisting of @xmath12 primitive cells with a @xmath13 @xmath14-point grid in the primitive cell .
[ fig : phonons_in_field],scaledwidth=45.0% ] the calculated band structure of a `` free '' silicene layer is shown in fig .
[ fig : bs_v_ez ] .
as expected , it resembles the band structure of graphene ; in particular it shows the linear dirac - type dispersion of electrons near the k points , where we find the fermi level in undoped silicene .
the fermi velocity @xmath9 of electrons in silicene is lower than that in graphene ( see table [ table : parameters_summary ] ) .
although the lattice parameters and sublattice buckling found in the different dft calculations are in good agreement , our results for the fermi velocity are very much smaller than the fermi velocity reported in ref . .
with a plane - wave cutoff energy of 816 ev and a @xmath15 @xmath14-point grid : ( a ) in zero external electric field , ( b ) with @xmath16 v@xmath11 , and ( c ) with @xmath17 v@xmath11 ( shown both with and without the relaxation of the atomic coordinates in the electric field ) .
the zero of the external potential is in the center of the silicene layer .
the dashed line shows the fermi energy in each case and the insets show the spectrum near the fermi level in the vicinity of the k point .
[ fig : bs_v_ez],title="fig:",scaledwidth=45.0% ] + with a plane - wave cutoff energy of 816 ev and a @xmath15 @xmath14-point grid : ( a ) in zero external electric field , ( b ) with @xmath16 v@xmath11 , and ( c ) with @xmath17 v@xmath11 ( shown both with and without the relaxation of the atomic coordinates in the electric field ) .
the zero of the external potential is in the center of the silicene layer .
the dashed line shows the fermi energy in each case and the insets show the spectrum near the fermi level in the vicinity of the k point .
[ fig : bs_v_ez],title="fig:",scaledwidth=45.0% ] + with a plane - wave cutoff energy of 816 ev and a @xmath15 @xmath14-point grid : ( a ) in zero external electric field , ( b ) with @xmath16 v@xmath11 , and ( c ) with @xmath17 v@xmath11 ( shown both with and without the relaxation of the atomic coordinates in the electric field ) .
the zero of the external potential is in the center of the silicene layer .
the dashed line shows the fermi energy in each case and the insets show the spectrum near the fermi level in the vicinity of the k point .
[ fig : bs_v_ez],title="fig:",scaledwidth=45.0% ]
to exploit the weak buckling of silicene , we consider its behavior in an external electric field @xmath18 applied in the @xmath3-direction , as shown in fig .
[ fig : silicene_in_capacitor ] .
the main effect of such an electric field is to break the symmetry between the @xmath4 and @xmath5 sublattices of silicene s honeycomb structure and hence to open a gap @xmath2 in the band structure at the hexagonal bz points k and k@xmath19 . in the framework of a simple nearest - neighbor tight - binding model ,
this manifests itself in the form of an energy correction to the on - site energies that is positive for sublattice @xmath4 and negative for @xmath5 .
this difference in on - site energies @xmath20 leads to a spectrum with a gap for electrons in the vicinity of the corners of the bz : @xmath21 , where @xmath22 is the electron `` valley '' momentum relative to the bz corner . opening a gap in graphene by these means
would be impossible because the @xmath4 and @xmath5 sublattices lie in the same plane .
a nave estimate of the electric - field - induced gap in silicene can be made using first - order perturbation theory by diagonalizing a @xmath23 hamiltonian matrix at @xmath24 , @xmath25.\ ] ] here , @xmath26 are the degenerate lowest unoccupied and highest occupied kohn - sham orbitals at the k point at @xmath27 , and @xmath28 corresponds to the mid - plane of the buckled lattice .
this suggests a band gap which opens linearly with the electric field at a rate @xmath29 and 0.573 e for the wave functions @xmath30 found using the lda and pbe functionals , respectively .
the estimate given in sec .
[ sec : perturbation_theory ] is in fact only an upper limit for the rate at which the band gap opens , since it neglects screening by the polarization of the @xmath4 and @xmath5 sublattices . in order to obtain an accurate value of the rate at which a band gap can be opened with an electric field , we have performed fully self - consistent calculations of the dft band structure in the presence of an electric field
. a typical result of such a calculation is shown in fig .
[ fig : bs_v_ez](b ) . at small electric fields ,
relaxing the structure in the presence of the field does not have a significant effect on the band gap , but the screening of the electric potential by the sublattice polarization of the electron states makes a substantial difference .
the dft - calculated gaps are gathered in fig .
[ fig : gap_v_ez ] .
the variation of the band gap @xmath2 at k with electric field @xmath18 is almost perfectly linear for fields up to @xmath31 v@xmath11 .
the results for the rate @xmath32 at which a gap is opened are shown in the table inset in fig .
[ fig : gap_v_ez ] .
the eightfold difference between the self - consistent and the unscreened values of @xmath32 indicates that the system exhibits a strong sublattice polarizability . for silicene with a plane - wave cutoff energy of 816 ev and a @xmath33 @xmath14-point grid .
unless otherwise stated , the pbe functional was used .
the box length in the @xmath3 direction was varied from @xmath34 to 26.46 .
the results have been extrapolated to the limit @xmath35 of infinite box length ( solid lines ) as described in sec .
[ sec : boxlength ] .
unscreened band gaps calculated using perturbation theory are also shown .
the inset table shows the calculated rate at which the band gap opens .
[ fig : gap_v_ez],scaledwidth=45.0% ] our value for the rate at which the band gap opens within dft - pbe is 0.0742 e .
this is substantially lower than the result obtained by ni _ et al._,@xcite which is 0.157 e .
part of the reason for the discrepancy is that we extrapolated our results to infinite box length , whereas ni _ et al . _
used a fixed amount of vacuum between the periodic images of the layers .
another possible reason for the difference is that we used a plane - wave basis set , whereas ni _
et al . _
used a localized basis set . an incomplete localized basis set would tend to undermine the extent to which the electrons can adjust to screen the electric field .
the narrow - gap silicene band structure shown in fig . [ fig : bs_v_ez ] persists over a broad range of electric fields @xmath18 .
however , for electric fields of more than @xmath36 v@xmath11 , the band gap starts to close due to an overlap of the conduction band at @xmath37 and the valence band at k , and silicene becomes a semimetal , as shown in fig .
[ fig : bs_v_ez](c ) . according to our calculations , the buckled honeycomb crystal is still metastable at this electric field , as can be seen in fig .
[ fig : phonons_in_field ] .
the main effects of the electric field on the phonon dispersion curve are ( i ) to lift some degeneracies at k and m and ( ii ) to soften one of the acoustic branches , but without making the frequency imaginary . under much higher electric fields ,
the honeycomb structure of silicene becomes unstable .
we found that @xmath38 v@xmath11 causes the lattice parameter to increase without bound when the structure is relaxed .
we have also performed a study of the effects of so coupling ( which is more pronounced in si than in c ) on the band structure .
the so coupling term is explicitly included in the hamiltonian in the dft calculations . the results obtained with the lda and pbe functionals are shown in fig .
[ fig : so_gap ] .
both functionals predict an so gap of the order of a few mev at the k point , while the rest of the band structure barely differs from the nonrelativistic case .
our calculated lda and pbe so gaps are 1.4 mev and 1.5 mev , respectively , in agreement with the recent literature.@xcite of the @xmath37k line .
[ fig : so_gap],scaledwidth=45.0% ] in the theory of dirac electrons on the honeycomb lattice , the so gap is accounted for by the kane - mele term describing , e.g. , intrinsic so coupling in graphene.@xcite the kane - mele so coupling and the electric - field induced @xmath4-@xmath5 sublattice asymmetry for electrons in the vicinity of the bz corners @xmath39 in silicene can be incorporated in the hamiltonian @xmath40 where @xmath41 distinguishes between the two valleys , @xmath42 and @xmath43 , in silicene s spectrum . here , the pauli matrices @xmath44 , @xmath45 , and @xmath46 act in the space of the electrons amplitudes on orbitals attributed to the @xmath4 and @xmath5 sublattices , @xmath47 for the valley at @xmath42 and @xmath48 for the valley at @xmath43 . in eq .
( [ eqn : so_hamiltonian ] ) , @xmath49 is the electron spin operator normal to the silicene plane , and @xmath50 and @xmath51 are the dft - calculated so - coupling and electric - field induced gaps . the hamiltonian of eq .
( [ eqn : so_hamiltonian ] ) generically describes the transition between the 2d topological and band - gap insulators .
its spectrum , @xmath52 includes two gapped branches , one with a larger gap @xmath53 and another with a smaller gap @xmath54 . at a critical external electric field @xmath55 mv@xmath11 , @xmath56 , and the smaller gap closes , marking a transition from a topological insulator@xcite at @xmath57 to a simple band insulator at @xmath58 .
the difference between these two states of silicene is that the topological insulator state supports a gapless spectrum of edge states for the electrons , in contrast to a simple insulator , where the existence of gapless edge states is not protected by topology .
however , one may expect something reminiscent of the topological properties of dirac electrons to show up even in the band insulator state of silicene : an interface between two differently gated regions , with electric fields @xmath18 and @xmath59 ( where @xmath60 ) , should support a one - dimensional gapless band with an almost linear dispersion of electrons.@xcite
all our plane - wave dft total energies were corrected for finite - basis error@xcite and it was verified that the residual dependence of the total energy on the plane - wave cutoff energy is negligible .
we used ultrasoft pseudopotentials throughout , except where otherwise stated .
the silicene system was made artificially periodic in the @xmath3-direction ( normal to the silicene layer ) in our calculations .
the atomic structure was obtained by relaxing the lattice parameter and atom positions within dft , subject to the symmetry constraints and at fixed box length @xmath61 in the @xmath3 direction .
the cohesive energy was then evaluated using this optimized structure .
the energy of an isolated si atom ( needed when evaluating the cohesive energy ) was obtained in a cubic box of side - length @xmath62 subject to periodic boundary conditions .
we extrapolated the energy of the isolated atom to the limit of infinite box size by fitting @xmath63 to the dft energies @xmath64 obtained in a range of box sizes , where @xmath65 and @xmath66 were parameters determined by fitting .
equation ( [ eq : boxlength ] ) gave a very good fit to our data .
we have also calculated the dft zero - point correction to the energy of silicene .
this is expected to be largely independent of the exchange - correlation functional used .
indeed , our calculations show that the zero - point correction is 0.103 ev within the lda and 0.101 ev with the pbe functional.@xcite we used the pbe result in our final calculations of the cohesive energy reported in table [ table : parameters_summary ] . to evaluate the fermi velocity shown in table [ table : parameters_summary ] we evaluated the dft band structure using a @xmath15 @xmath14-point grid and a plane - wave cutoff energy of 816 ev in a cell of length @xmath67 .
we then fitted eq .
( 17 ) of ref . to the highest occupied and lowest unoccupied bands within a circular region around the k point ; the fermi velocity is one of the fitting parameters .
the radius of the circular region was 6% of the length of the reciprocal lattice vectors ; we verified that the fermi velocity was converged with respect to this radius .
the phonon dispersion curves shown in sec . [ sec : phonon_results ] were calculated using the method of finite displacements , with atom displacements of 0.042 , in a supercell consisting of @xmath12 primitive cells with a @xmath13 @xmath14-point grid in the primitive cell . in the results with the external electric field , the box length was @xmath68 and
the plane - wave cutoff energy was 435 ev .
in the results without the field , the box length was @xmath34 and the plane - wave cutoff was 816 ev .
this choice was made because the error due to a finite box length @xmath61 is potentially much larger in the presence of a transverse electric field .
the geometry optimization and band - structure calculations at zero external field were performed with both the castep@xcite and vasp@xcite codes , to verify that the results are in good agreement .
this check was necessary because it was only possible to perform the electric - field calculations with castep , while for the so calculations we had to use vasp . in principle
the only difference between the calculations performed using the two codes arises from the si pseudopotentials used .
the paw method@xcite was used in the vasp calculations , whereas ultrasoft pseudopotentials were used in the castep calculations .
as can be seen in table [ table : parameters_summary ] , the geometries predicted by the two codes agree well .
we have also verified that the band structures are in good agreement . finally , in fig .
[ fig : mono_phonons ] we show that the phonon dispersions obtained with the two codes are virtually identical when the same parameters are used .
supercell were obtained with a box length @xmath34 , a @xmath69 @xmath70-point grid in the primitive cell , and a plane - wave cutoff energy of 816 ev .
the matrix of force constants was evaluated using the method of finite displacements , with the displacements being 0.042 .
the results for a @xmath71 supercell were obtained with a box length @xmath72 , a @xmath73 @xmath14-point grid in the supercell , and a plane - wave cutoff energy of 500 ev .
the matrix of force constants was evaluated using the method of finite displacements , with displacements of 0.09 .
[ fig : mono_phonons],scaledwidth=45.0% ] figure [ fig : mono_phonons ] also demonstrates that our phonon dispersion curves are converged with respect to supercell size . the convergence of the calculated band gap with respect to the plane - wave cutoff energy for a particular applied field is shown in fig .
[ fig : gap_v_ecut ] .
the gap converges extremely rapidly .
with a @xmath74 @xmath14-point grid including k. [ fig : gap_v_ecut],scaledwidth=45.0% ] the convergence of the calculated band gap at bz point k with respect to the @xmath14-point grid used in the self - consistent field calculations is shown in fig .
[ fig : gap_v_nk ] .
the finite - sampling error falls off as the reciprocal of the total number of @xmath14 points .
the prefactor of the finite - sampling error is vastly greater when k or k@xmath19 is included in the grid of @xmath14-points for the self - consistent field calculations . of @xmath75-points for silicene
subject to an electric field of 0.257 v / in a cell of length @xmath34 with a plane - wave cutoff energy of 816 ev .
[ fig : gap_v_nk],scaledwidth=45.0% ] the dependence of the calculated gap on the exchange - correlation functional and pseudopotential is shown in fig .
[ fig : gap_v_psp ] .
the difference between the results obtained with different pseudopotentials is much smaller than the gap , but is not wholly negligible .
the on - the - fly ultrasoft pseudopotential is believed to be more accurate than the norm - conserving pseudopotential,@xcite and hence we have used the former in our final calculations . for silicene subject to an electric field of 0.257 v / in a cell of length @xmath34 with a @xmath74 @xmath14-point grid and a plane - wave cutoff energy of 816 ev .
[ fig : gap_v_psp],scaledwidth=45.0% ] the dependence of the calculated gap on the length of the simulation box is shown in fig .
[ fig : gap_v_l ] .
it is clear that this is potentially a large source of error .
however , the @xmath61-dependence is reasonably well approximated by a quadratic in @xmath76 , allowing the dft gaps to be extrapolated to infinite cell size if results at three or more different cell sizes are available . for silicene subject to an electric field of 0.257 v / with a @xmath74 @xmath14-point grid and a plane - wave cutoff energy of 816 ev .
the solid line is a fitted quadratic in @xmath76 .
[ fig : gap_v_l],scaledwidth=45.0% ] our final results for the field - induced gap were obtained using @xmath33 @xmath14 points , not including k or k@xmath19 , a plane - wave cutoff energy of @xmath77 ev , and box lengths @xmath34 , 18.521 , and 26.459 ; the results were then extrapolated to infinite box length by fitting a quadratic in @xmath76 . from the magnitudes of the variations shown in figs . [
fig : gap_v_ecut][fig : gap_v_l ] , we estimate the uncertainty in our final results for the rate @xmath32 at which the band gap opens when an electric field is applied to be less than about 0.01 e . to evaluate the field - induced band gap using perturbation theory we used norm - conserving pseudopotentials.@xcite we used a @xmath78 @xmath79-point mesh including the k point , and a cell length of @xmath67 .
it was verified that the perturbation - theory - induced rate of gap opening @xmath32 was converged to within 0.00002 e with respect to @xmath14-point mesh and @xmath61 .
the finite - basis error in @xmath32 was found to fall off approximately exponentially with respect to the plane - wave cutoff energy , and hence we extrapolated our results to basis - set completeness .
the so calculations were performed with a plane - wave cutoff of 500 ev and a @xmath80 @xmath14-point grid .
we checked that the length of the simulation box has negligible influence on the so gap : the gap is the same with simulation box lengths of 15 and 30 up to numerical accuracy .
in summary , we have shown that a 2d layer of si atoms silicene is a versatile material in which a band gap can be tuned ( in a broad range of tens of mev ) using a transverse electric field @xmath18 , while silicene remains metastable . at the low field
@xmath81 mv@xmath11 , we expect silicene to undergo a transition between a topological and a simple band insulator , whereas at much higher field @xmath82 v@xmath11 it will undergo a transition from a band insulator into a semimetal .
we acknowledge financial support from the epsrc through a science and innovation award , the eu through the grants concept graphene and carbotron , the royal society , and lancaster university through the early career small grant scheme .
computational resources were provided by lancaster university s high - end computing facility .
we thank h .- j .
gao and g. le lay for useful discussions , and j. r. wallbank for providing us with fig .
[ fig : silicene_in_capacitor ] .
|
we report calculations of the electronic structure of silicene and the stability of its weakly buckled honeycomb lattice in an external electric field oriented perpendicular to the monolayer of si atoms .
the electric field produces a tunable band gap in the dirac - type electronic spectrum , the gap being suppressed by a factor of about eight by the high polarizability of the system . at low electric fields , the interplay between this tunable band gap , which is specific to electrons on a honeycomb lattice , and the kane - mele spin - orbit coupling induces a transition from a topological to a band insulator , whereas at much higher electric fields silicene becomes a semimetal .
|
in recent years , significant efforts were dedicated to realize artificial micro and nano swimmers that locomote at low reynolds numbers , and that are further able to perform tasks , like cargo delivery @xcite .
swimmers driven by periodic conformational changes have to break the time - reversal symmetry to achieve a net propulsion at small reynolds numbers @xcite .
however , swimmers driven by a continuous propulsion mechanism need to break spatial symmetry and display a permanent polarity . despite the theoretical possibility to observe a spontaneous symmetry breaking for homogeneous spherical particles @xcite , all solid autophoretic swimmers realized in experiments so far exhibit an ingrained polarity .
spherical janus particles , for instance , exploit a difference in chemical surface composition of the two opposing hemispheres @xcite .
recent studies have shown that janus nanorods can not only self propel but also deliver drugs during self
propulsion @xcite , repair cracks in electrical circuits @xcite , or even halt blood hemorrhages @xcite . beside their desired properties , is the nature of these colloidal ( janus ) particles and the way to trigger their self propulsion often incompatible with biological applications @xcite . in contrast
, emulsion droplets can be formed with bio compatible liquids .
it was shown that emulsion droplets can self
propel by marangoni flow @xcite when a surface tension gradient is either maintained by chemical reactions @xcite , micellar solubilization @xcite , or liquid liquid phase separations @xcite .
however , the realization of self propelled janus droplets has been proven difficult @xcite and swimming droplets could so far not perform tasks comparable to solid self propelling particles . in this article , we report time evolutive self propelled ethanol / water droplets in a surrounding squalane / monoolein solution , which are able to deliver cargo .
the evolution of the emulsion droplets is caused by a continuous release of ethanol from the droplet into the continuous phase and simultaneous uptake of monoolein from the continuous phase into the droplet phase .
this leads , after a certain time , to a spontaneous phase separation between water and ethanol in the presence of the surfactant monoolein , see fig .
[ fig : exp1]a . depending on initial ethanol / water mixing ratio , the droplets evolve in up to three stages and potentially form janus droplets with a water rich leading droplet and an ethanol rich trailing droplet .
we describe and characterize the conditions of emergence and evolution of the different stages , and their corresponding propulsion mechanism .
furthermore , we demonstrate that the droplets can extract and precipitate dna during the phase separation process , which can is delivered via the ethanol rich droplet to a target location . because the emergence of phase separation is controlled by the initial chemical composition of the droplets , several different cargo delivery processes can be achieved in a programmable manner .
the propulsion mechanism is universal and could be realized with several other liquid mixtures or lipidic surfactant , including a wide range of alcoholic beverages ( if having sufficiently high alcohol content ) .
the generation of evolutive self propelling droplets is achieved using a four component system consisting of water , a water miscible solvent , which is less polar than water and partially soluble in the surrounding phase , an oily phase , and an oil soluble surfactant .
the surfactant needs a chemical affinity for the solvent and the oil should be immiscible with water .
a combination of chemical substances satisfying all these conditions is water / ethanol / monoolein and squalane ( as 4% of ethanol is soluble in squalane ) . for simplicity we will restrict our following description to this particular system , but the reported concept is universal and similar results
could be obtained for seven other chemical compositions fulfilling the above mentioned requirements , see @xcite .
droplets with a diameter of about 200 @xmath0 m were generated by injecting an ethanol / water mixture through a glass capillary ( inner diameter @xmath1 m ) into a flat observation chamber that is pre filled with a continuous squalane / monoolein mixture .
the bottom of the observation chamber is rendered hydrophobic with a self assembled monolayer of octadecyltrichlorosilane ( ots ) to prevent adhesion and subsequent spreading of the droplets at the bottom surface of the chamber .
depending on ethanol concentration , the generated water / ethanol droplets evolve with time assuming up to three different stages , as displayed fig .
[ fig : exp1 ] .
we start our description for a medium ethanol concentration , where all three stages can be observed ( fig .
[ fig : exp1]a ) , and subsequently extend our discussion to smaller and respectively to larger initial ethanol concentrations . in the following discussion the monoolein concentration in squalane
is always fixed to 15 mm , i.e. 7 times above the critical micellar composition ( cmc ) , if not explicitly stated otherwise .
for intermediate ethanol concentrations , between @xmath2 and @xmath3 , the droplets start to self
propel in the oil phase just after their production . during self propulsion
every droplet looses a few percent of its volume , and takes up monoolein molecules from the surrounding phase .
the influx of monoolein to the swimming droplet is monitored by adding fluorescent lipids to the initial water / ethanol mixture .
similar to monoolein , the fluorescent lipids display a strong affinity to ethanol and follows the repartition of monoolein molecules between the fluid phases , cf .
[ fig : exp1]c . after @xmath4 minutes in _
stage 1 _ , a nucleation process occurs inside the self propelling droplet ( _ stage 2 _ ) which lasts for about @xmath5 minutes and leads to the formation of an ethanol rich droplet inside a water
rich droplet .
the ethanol rich droplet is finally pushed out of the water
rich droplet when touching its interface , generating an ethanol rich droplet trailing behind the leading water
rich droplet .
such a droplet pair formed by the coexistence of two phase separated droplets constitutes a long lived janus droplet that is self propelling for about @xmath6 min ( _ stage 3 _ ) . at the end of _ stage 3
_ , the two droplets forming the janus droplet break up , fig . [
fig : exp1]a .
after separation the ethanol
rich trailing droplet typically spreads on the hydrophobic substrate and the water
rich leading droplet stops its motion suddenly , or in rare cases , continues its self propulsion by a few droplet radii . when initially producing water / ethanol droplets with smaller ethanol concentration of just @xmath7 , the demixing starts almost immediately after droplet production , cf . fig .
[ fig : exp1]e . droplets with this low ethanol concentration are directly starting to self propelled in _ stage 2_. the demixing in _
stage 2 _ is very fast ( @xmath8 min ) and the self propelling behavior of the formed janus droplets can effectively be observed only in _
stage 3 _ for around @xmath9 min .
the qualitative behavior in these stages is as described previously for intermediate ethanol concentrations .
reducing the initial ethanol concentration in the droplets even below @xmath10 , phase separation is not always observed and self propelled motion is not occurring reliably . droplets produced with an ethanol concentration between @xmath11 to @xmath12 show all three evolution stages ( fig .
[ fig : exp1]b ) , whereas the duration of _ stage 1 _ is extended to @xmath13 min before the phase separation occurs in _ stage 2_. the duration of the subsequent demixing in _ stage 2 _ is around @xmath14 min .
in contrast to the intermediate and low ethanol concentrations , the water rich phase is now clearly the minority phase and several water
rich droplets are formed inside the ethanol rich droplet .
accordingly , the water
rich droplets are finally pushed out of the larger ethanol rich droplet and may continue to locomote . the ethanol
rich mother droplet finally spreads on the ots - coated bottom of the device and stops moving .
generating droplets with ethanol concentrations well above @xmath15 is not possible as those droplets directly wet the ots - coated bottom of the observation chamber and do not self propel .
having introduced the the appearance of the evolution stages as function of ethanol concentration , we will shed some light on their characteristics .
a self propelling droplet in _ stage 1 _ collects a large amount of monoolein from the oil phase as can be seen from the fluorescence activity after a short propulsion time . in case of large ethanol concentration ( @xmath16 ) , additional dark clouds can be observed around the droplet , which we suspect to be marangoni rolls @xcite , together with a dark trail ( fig .
[ fig : exp1]c ) .
the droplet is releasing more material with increasing ethanol concentration , visibly by the reduction of droplet volume ranging from only a few percent at @xmath10 ethanol concentration to about @xmath17 for an ethanol concentration of @xmath15 .
obviously , the released material is ethanol which is soluble by @xmath18 in the surrounding squalane . because we neither observed self propelling nor volume loss in the case of pure water droplets within the duration of the experiment irrespectively of the monoolein concentration , we further conclude that the fraction of water leaving the droplets can be neglected with respect to the ethanol loss
the absence of self propelled motion of pure water droplets agrees with the findings of thutupalli et al .
@xcite but is at variance with the observations of izri et al .
@xcite . based on the above findings we conclude that the solubilization of ethanol into the squalane solution is driving the droplet motion .
this effect is described for the case of liquid
crystal droplets @xcite , where liquid crystal molecules are solubilized in a monoolein / squalane solution by filling surfactant micelles @xcite , leading to a variation of the surface tension at the droplet surface and so to a marangoni flow @xcite .
however our situation is more complex than in refs . @xcite as ethanol can be solubilized not only in micelles , but also at the molecular lever .
we demonstrate this point by observing the self propulsion of droplets below the critical micellar concentration ( cmc ) , cf .
[ fig : exp2]a , in contrast to all the other reported examples of self propelling droplets @xcite .
the droplets velocities in _ stage 1 _ are constant in average with strong fluctuations , similar to other reported systems displaying self propelled motion due to solubilization @xcite .
the droplet velocity has no clear dependence on the initial ethanol concentration provided it is in a range that droplets self propel ( @xmath19 ) and can be stabilized against coalescence ( @xmath20 ) , see fig [ fig : exp2]b .
but the self
propulsion running time and with it the cruising range in _ stage 1 _
is significantly increased with increasing ethanol concentration which is the fuel for this propulsion process , cf .
[ fig : exp2]b
. measurements of the surrounding flow field by micro - particle image velocimetry ( @xmath21 ) , shown in fig [ fig : exp3]a , characterize this type of self propelling droplets in _
stage1 _ as squirmer in a weak pusher mode .
note that this flow field is different from flow fields reported for comparable systems which reveal a more symmetric flow field , as expected of neutral squirmers @xcite upon self propulsion , ethanol solubilization and surfactant absorption are working together until the condition of phase separation is eventually reached @xcite , leading to the observed water / ethanol phase separation in _ stage 2_. _ stage 2 _ persists only a short time with a continuous increase of the droplet velocity while nucleation and coarsening processes are ongoing , see fig .
[ fig : exp1]c .
according to the measured flow field around a droplet in _
stage 2 _ , see fig .
[ fig : exp3]b , the droplets can be characterized as neutral squirmers .
the demixing of water / ethanol in the presence of monoolein is reported in literature for the corresponding ternary mixture ( water / ethanol / monoolein ) @xcite . inspecting these phase diagrams , sketched in fig .
[ fig : exp1]e , it is obvious why droplets with larger initial ethanol concentration have to loose more ethanol and are sustained longer in _
stage1 _ before reaching the decomposition line that is leading to the phase separation .
the small ethanol
rich droplets nucleating in the bulk of the mother droplet coarsen quickly until a janus state is reached .
due to the swirl inside the mother droplet the coarsening droplets are preferentially transported towards the rear side of the mother droplet where they coalesce to even bigger droplets leading to one large ethanol
rich droplet inside the mother droplet . when this coalesced ethanol
rich droplet touches the inner surface of the mother droplet it forms a three phase contact line and finally grows out of the mother droplet forming a stable janus - droplet when nucleation and coarsening are completed , see fig . [
fig : exp1]c .
after the completed formation of self propelling janus droplets , they consist of a water
rich leading droplet in contact with an ethanol rich trailing droplet that contains the majority of the monoolein molecules , which were previously taken up by the mother droplet . a water - rich droplet looses some of its volume ( up to a few percent ) , while the janus droplet velocity decreases exponentially until reaching a constant cruising speed .
no further nucleation in visible droplets or coarsening events could be observed .
however , estimating , the volume of both droplets and comparing them to the droplet volume at the beginning of _ stage 2 _ and the initial water / ethanol composition indicates that a non
negligible amount of ethanol is still present in the water rich phase .
based on @xmath22piv measurements of the flow field , the stable janus droplets can be identified as chains of neutral swimmer @xcite , cf .
[ fig : exp3]c . a surface flow at the water
rich droplet can be directly observed in _
stage 3 _ by secondary nucleated small ethanol
rich droplets at the surface of the water rich leading droplet .
these small droplets are continuously transported towards the ethanol
rich trailing droplet and merge with it @xcite . during transport
the small droplets eventually grow in size , reaching @xmath23 m , when reaching the trailing ethanol droplet @xcite .
interestingly , the distribution of the secondary droplet surface indicate the future macroscopic direction of motion of a janus drop : each small droplet nucleated at the interface of the water rich droplet is a local sink for monoolein molecules in its near surrounding and is thus influencing the symmetry of the marangoni flow pattern around the janus droplet . if the distribution of the small ethanol rich droplets is homogeneous , the flow is symmetric with respect to the symmetry axis of the janus droplet leading to a relatively straight trajectory .
in contrast , an inhomogeneous distribution of the secondary nucleated droplets induces an asymmetric flow field and triggers a rotation of the water
rich droplet away from the small droplets @xcite .
based on the above findings , we already concluded that the droplets in _
stage 1 _ are driven by the solubilization of ethanol into the surrounding oil phase . in view of the accumulation of monoolein inside the trailing droplet in _
stage 3 _ and the continuing formation of a monoolein depleted layer in the squalane around the leading droplet , it suggests itself to model the propulsion in _
stage 2 _ and _ 3 _ from a balance between the work dissipated in viscous flows and the chemical energy released by monoolein molecules on their path from the ambient squalane phase into the trailing droplet .
this release of chemical energy drives a marangoni flow that continuously propels the janus droplet .
since the small solvent molecules ethanol and water diffuse much faster than the large monoolein molecules , we assume a chemical equilibrium of water and ethanol between the two bulk phases . under these assumptions
the concentration of solvents to the bulk phases follows the binodal line of the phase diagram , while the relative partition of ethanol and water to the leading and trailing droplet is solely controlled by the total amount of monoolein in the janus droplets .
the flux of monoolein molecules into the trailing droplet in stage _ 3 _ can be expressed by the droplet velocity and the surface coverage of monoolein on the interface of the leading droplet .
close the final chemical equilibrium of the janus droplet with the squalane phase , it is justified to approximate the surface coverage @xmath24 by the its corresponding equilibrium value . by virtue of the gibbs adsorption isotherm
, we can relate the gradient of the interfacial free energy to the gradient of the chemical potential of monoolein molecules .
the thin depletion layer of monoolein in the squalane solution around the janus droplet with thickness @xmath25 indicates large peclet numbers @xmath26 , and we can set the chemical potential of the adsorbed monoolein molecules at the tip of the leading droplet to the value @xmath27 in the squalane solution .
the chemical potential at the three phase contact line can be related to the concentration @xmath28 of monoolein molecules in the trailing droplet .
but as more and more monoolein accumulate in the trailing droplet , the difference of chemical potentials of the monoolein molecules in the squalane phase , @xmath27 and in the ethanol rich trailing droplet , @xmath22 , decreases , and the janus droplets slow down . in return , this deceleration reduces the flux of monoolein molecules into the trailing droplet and we predict an exponential decrease of the droplet velocity @xmath29 to zero in stage _ 3_. the time constant of this exponential decrease can be expressed as @xmath30 where @xmath31 is the dynamic viscosity of the squalane phase , @xmath32 the volume of the trailing droplet and @xmath33 a factor of order unity that contains the dependence on the geometry of the janus droplet and the viscosity ratios of the bulk fluids @xcite .
spontaneous de mixing of the mother droplet into an ethanol rich and a water rich phase marking the end of _ stage 1 _ starts once the composition of the janus droplet reaches the boundary of the miscibility gap .
provided a low initial ethanol concentration of the mother droplet at the beginning of _ stage 1 _ we observe the nucleation of small ethanol
rich minority phase droplets that quickly coarsen . with a certain delay after the onset of nucleation , the first minority phase droplets attach to the interface between the squalane and the water
rich phase of the mother droplet .
the thus formed three
phase contact line allows the monoolein molecules on the interface between the water
rich majority phase and the squalane phase start to solubilize into the bulk of the ethanol
rich trailing droplet .
as outlined above for the droplet propulsion in _ stage 3 _ the resulting gradient in the chemical potential of monoolein molecules causes a marangoni flow .
since the trailing droplet is initially very small , it is quickly saturated with monoolein unless larger , ` fresh ' minority phase droplets coalesce with the trailing droplet and dilute the monoolein .
the increase of concentration caused by the influx of monoolein molecules from the interface of the leading droplet is quickly surpassed by the dilution with minority phase droplets that coalesce with the trailing droplet . as a result , the difference of the chemical potential between the monoolein molecules in the ambient squalane and the ethanol rich trailing droplet increases in _ stage 2 _ , and thus the propulsion velocity . to estimate the balance of monooleine accumulation and its dilution in _ stage 2 _
, we employ the average radius of the minority phase droplets measured in our experiments .
a quantitative analysis of the experimental data shows that both the radius of the minority phase droplets in the bulk of the mother droplet and the radius of the trailing droplet in _ stage 2 _ conforms to a power law @xmath34 in time @xmath35 @xcite .
a linear growth of the typical domain size after a quench into the miscibility gap of a binary liquid was already predicted from dimensional analysis of the transport equations @xcite and observed in experiments @xcite . setting the growth exponent of the radius to @xmath36 ,
dimensionalized undersaturation of monoolein in the trailing droplet in _ stage 2 _ , and the propuslion velocity @xmath37 are described by the equation @xmath38 solutions @xmath39 of eqn .
( [ eq : early_time ] ) and , hence , the velocity @xmath29 , displays a nearly linear rise in _ stage 2 _ , in agreement with experimental findings @xcite . since the final volume of the trailing droplet scales as @xmath40 , eqn .
( [ eq : time_constant ] ) predicts a relaxation time @xmath41 in _ stage 3 _ whereas our experimental data show a relaxation time that is independent on the dimensions of the droplet @xcite . in agreement with the experiments , the peak velocity of self propulsion
is observed at a time @xmath42 marking the transition between _ stage 2 _ and _ stage 3_. numerical solutions of eqn .
( [ eq : early_time ] ) show that the peak velocity is monotoneously decreasing with an increasing ratio @xmath43 of the coarsening time to the final relaxation time . only in the limit of small ratios
@xmath44 we find a constant peak velocity as in the experiments @xcite .
the model introduced above is based on the assumption that the water and ethanol molecules of two fluid phases of the janus droplet are chemically equilibrated at any instance in time .
however , we know from experimental observations that this assumption can not be completely fulfilled . in _ stage 3 _ , in particular , we observe the nucleation and growth of small ethanol
rich droplets at the interface between the water
rich droplet with the squalane phase .
these droplets are advected to and merge with the trailing droplet .
a secondary nucleation indicates that the de mixing of ethanol and water is still ongoing in _ stage 3_. droplets of the minority phase created in this secondary nucleation could be responsible for an increased influx of monoolein molecules into the trailing droplet that allows larger droplets to reach a chemical equilibrium on the same time scale as droplets with a smaller radius .
the discrepancy between the perdictions of the model and the experimental results could also arise from an unjustified assumption of chemcial equilibration of water and ethanol in the janus droplet .
it is well possible that ethanol and water molecules in two bulk phases of the janus droplet are not close to a chemcial equilibrium , and the gradients in the composition of the solvents also contribute to the propelling marangoni stresses .
moreover , the quick uptake of monoolein in _ stage 1 _ , and the synchroneous appearance of monodisperse minority phase droplets suggests a spinodal decomposition of an unstable phase at the beginning of _ stage 2 _ rather than a homogeneous nucleation . in view of the complexity of the many competing kinetic effects
, it does not come as a surprise that the maximum velocity and the scaling of the exponential decay time recorded in experiments do not conform to the prediction of our simple simple model . in this section
, we will use the particular properties of our evolutive self propelling droplets for controlled cargo delivery in a programmable manner . for that purpose
, we used a water / ethanol solution containing dna ( @xmath45 mg / ml ) and @xmath46 m of sodium acetate as droplet phase in a surrounding monoolein / squalane solution . even in the presence of dna ,
the droplets undergo the previously described evolution stages ending with a janus droplet consisting of a water
rich leading droplet and an ethanol rich trailing droplet .
in contrast to similar bulk methods @xcite , dna extraction in self - propelling and phase separating droplets occurs in a single step during less than @xmath47 minutes with sample volumes on the picoliter scale , cf .
[ fig : exp4]a , c . to achieve a controlled cargo delivery , we use the fact that the ethanol rich droplet , which is formed during the evolutive self propulsion spreads easily on a polydimethylsyloxane ( pdms ) surface upon contact . using this strategy ,
three different scenarios of cargo delivery can be achieved simply by varying the initial ethanol / water concentration of the droplet . for intermediate ethanol concentration between @xmath2 and @xmath3
the droplet is self propelling @xmath48 minutes before starting its phase separation .
after the janus fromtion , a _ drop - off _ cargo delivery is realized ; i.e. the swimmer delivers its cargo and leaves the target area . if such a droplet self propels around an obstacle in the janus _
stage 2 _ or @xmath49 , the ethanol
rich trailing droplet is spreading on the surface when touching the obstacle , see fig .
[ fig : exp4]b .
this touching is inevitable when the janus droplet tries to swim away from the obstacle .
the remaining water
rich droplet is continuing its self propelled motion moving away from the obstacle .
however , the cargo can also be delivered if the droplet approaches the pdms obstacle already in _ stage 1 _ , or _
stage 2 _ , i.e. in a stage before the trailing droplet is formed . in these stages
a self - propelling droplet will either stay in contact to the pdms surface , or self propel in a bound orbit around it depending on its specific squirmer mode @xcite .
therefore it is just a matter of time until the janus droplet is formed leading to the above described _ drop - off _ cargo delivery . for low ethanol concentration ( @xmath50 ) ,
the formation of janus droplets occurs almost immediately after droplet production ; such a droplet has a short total running time . as consequence , those water
rich droplets will stop to move soon after cargo delivery and stay close to the cargo delivery area shielding it after a certain number of delivery events . for large ethanol concentration ( @xmath51 ) , the droplet is self - propelling 1015 minutes before starting it phase separation .
several small water
rich droplets are formed inside an ethanol rich droplet that are finally pushed out of the larger ethanol rich droplet . the ethanol
rich droplets spread on the surface after separation , while the water rich droplets typically continue to locomote , cf .
[ fig : exp1]d . these self propelling water
rich droplets are still containing a non - negligible quantity of ethanol being able to deliver the cargo .
however this type of multi - droplet cargo delivery is hardly controllable .
all these variants can be combined with selective dna extraction , as the dna extraction is only enabled in the presence of @xmath46 m sodium acetate . because the presence of salt reduced the dna solubility in water , and therefore the dna precipitates and extracts into the ethanol
rich trailing droplet during the phase separation .
it results , that the final cargo content can be selectively controlled by the initial chemical composition of the droplet .
in particular , the different value of delay time before the phase separation , are controlling the brownian distance travel by the droplet before starting the process of cargo - delivery . in total
, three different types of cargo delivery scenarios could be time controlled and programmed just depending on the initial chemical composition @xcite . and
@xmath52 ( a ) and for ethanol concentrations @xmath53 ( b ) .
( top rows ) display optical microscopies and ( bottom rows ) sketches ( bottom rows ) , the intensity of the blue and yellow colors indicate increasing concentration of water , respectively of ethanol .
c ) optical fluorescence micrographs observed for the different droplet stages insert . _ stage 1 _ , ( _ left _ ) : droplet appears bright due to the fluorescent lipid molecules taken up from the continuous phase .
for the particular example of @xmath52 ethanol concentration additionally dark regions are visible around the drop .
_ stage 2 _ , ( _ middle _ ) : the fluorescent lipids indicate the ethanol rich phase during nucleation and coarsening .
_ stage 3 _ , ( _ right _ ) : the fluorescent lipids indicate that the trailing droplet contains the ethanol rich phase while the leading drop contains the water
rich phase in the janus regime .
d ) droplet velocity as function of time for an initial ethanol concentration of @xmath54 .
e ) phase diagram of the water / ethanol / monoolein ternary mixture .
two associated pathways for two mixtures , one above @xmath52 of ethanol and the other below @xmath52 , as function of the three evolution stages.,width=604 ] ; the data for variable ethanol concentration correspond to a monoolein concentration of 7 times the cmc.,width=566 ] piv in the lab frame .
the droplets move close to the bottom surface of the device , that is at around @xmath55 times higher than the average droplet diameter .
the flow fields were determined parallel to the bottom walls and are similar to the rotational symmetric flow fields of ( a ) a weak pusher in _ stage 1 _ , ( b ) a neutral swimmer in _ stage 2 _ , and ( c ) a chain of neutral swimmers in _ stage 3 _ : . , width=566 ]
to fabricate the observation chambers , glass slides were cut into squares of about ( @xmath56 ) @xmath57 and were pre - cleaned by sonication in ethanol , acetone and toluene for 10 min each .
the glass squares were blow - dried with nitrogen gas after each cleaning step . after this pre - cleaning
they are immersed for 30 min in piranha etch ( 50 vol.% sulfuric acid ( 98 % ) and 50 vol.% hydrogen peroxide ( 30 vol.% ) .
after that , the glass squares were thoroughly rinsed with hot ( 8090)@xmath58 c ultrapure water and immersed in hot water for another 15 min .
subsequently , the glass squares were again blow - dried with nitrogen .
the thus cleaned glass squares were coated with octadecyl - trichlorosilane ( ots ) by immersing them for 12 min in a solution consisting of 50 ml bicyclohexane , 2040 drops of carbon tetrachloride , and 520 drops of ots @xcite . remains from the ots solution were removed from the ots - coated glass by rinsing it with chloroform and blow - drying with nitrogen .
the ots coated glass squares were stacked with a cover slide ( thickness 150 @xmath22 m ) as a spacer and glued together ( epoxy glue ) . to conduct an experiment the observation chamber is prefilled with an oil / surfactant mixture and ethanol / water mixture
were carefully injected into the reservoir using a glass capillary ( 20 @xmath22 m inner diameter , connected to a syringe pump .
as water / ethanol were leaving the glass capillary , droplets were formed spontaneously .
the diameter of the droplets could be controlled between 100 and 150 @xmath22 m .
the density of the aqueous phase was always higher than that of the oily phase ( 0.81 g / ml ) , even for the highest ethanol concentration ( 80 vol.% ) used .
the concentration of monoolein is fixed at 3.5 mg / ml ( i.e. 15 mm ) if not explicitly mentioned otherwise . in experiments with varying monoolein concentration , the ethanol / water concentration
was fixed to 50 vol.% .
all chemicals were purchased from sigma - aldrich , except the fluorescent lipid that was purchase from avanti polar lipids .
we reported a general method to produce artificial self propelling janus droplets , which evolve from monoolein stabilized water / ethanol emulsion droplets .
these type of active matter is universal and could be achieved also for other surfactants and organic solvents including liqueurs and brandies . during ethanol solubilization , in _ stage 1
_ , the droplet velocity fluctuates around a constant value revealing a flow field of a weak pusher .
the droplet takes up monoolein molecules from the surrounding continuous phase finally leading to a phase separation in _
stage 2 _ , where the droplet velocity increases and the droplet moves as a neutral squirmer .
when the phase separation is completed , a janus droplet is formed in _ stage 3 _ with a leading water
rich droplet and a trailing ethanol rich droplet .
the flow field around this janus droplet is that of a pair of neutral squirmers and the droplet velocity decays exponentially with time .
the chemical potential difference of the monoolein molecules in the bulk of the oil phase and the ethanol phase drives the motion in _
stage 2 _ and _ stage 3_. we introduced a theoretical model that captures the relevant driving mechanism and describes qualitatively the characteristic velocity behavior during the _ stage 2 _ and _ stage 3 _ , despite the complexity of the propulsion mechanism . due to the phase separation occurring in _ stage 2 _ ,
these droplets could be used as a quasi - programmable smart carrier .
macromolecules could be selectively precipitated and delivered at a target location in a single step within a short time , as was demonstrated for dna delivery at pdms pillars . because the droplets self propelling properties are only controlled by the initial chemical composition
, we can realized three different cargo delivery modes in a programmable manner .
we expect that the simple strategy reported here allows to design new classes of active emulsions that could be used as smart self propelling carriers or operators .
the combination of this controlled dna cargo released with microfluidic devices may enable new strategies for droplet digital pcr @xcite , massive droplet dna barcoding @xcite , and gene , or drugs , delivery into cells @xcite .
|
we report the existence of self propelled janus droplets driven by phase separation , which are able to deliver cargo in a programmable manner .
the self propelling droplets are initially formed by a water / ethanol mixture in a squalane / monoolein solution , and evolve in up to three stages depending on ethanol concentration . in the first stage ,
the droplet propulsion is generated by marangoni flow originating from the solubilization of ethanol in the oily phase . during this process
the droplets absorb surfactant molecules ; in combination with the continuous loss of ethanol this leads to a phase separation of the water / ethanol / monoolein mixture and the formation of janus droplets , i.e. a water
rich droplet connected to an ethanol rich droplet that is able to deliver cargo .
we characterize the different evolution stages of self - propulsion by the flow field around the droplet that evolves from a weak pusher , over a neutral swimmer , to a dimer of neutral swimmers .
finally , we utilize this active system to deliver dna as a cargo . tuning the delay time before phase separation , by varying the chemical composition of the droplets ,
several different cargo delivery processes can be programmed .
|
in the almost 25 years since the pioneering work of polyakov @xcite and susskind @xcite much effort has been devoted to attempts to understand both the basic physics and quantitave features of the deconfining phase transition of qcd . the high temperature phase , i.e. for temperatures well above the critical temperature @xmath0 , is becoming well understood , and is widely believed to resemble a plasma of almost free quarks and gluons .
however , the transition region , @xmath1 , is very poorly understood .
this region is the most interesting one since it is there that the transition between ` hadronic ' and ` partonic ' degrees of freedom occurs .
the study of the transition region is a complicated and inherently non - perturbative problem which has mostly evaded treatment by analytical methods .
the variational method introduced several years by kogan and kovner @xcite appears well suited to a discussion of such a problem .
recently , the method was applied to the study of the phase transition in @xmath2 gluodynamics . as a result of these studies @xcite
, we identified a first order phase transition at a transition temperature , in units of the lighest state in the theory , of @xmath3 , in complete agreement with the lattice estimate for @xmath4 @xcite .
we minimize the relevant thermodynamic potential at finite temperature , i.e. the helmholtz free energy , on a set of suitably chosen trial density matrix functionals .
as with any variational calculation there is an element of guesswork involved .
however , the choice of the trial states is dictated by rather general principles .
as in the quantum mechanical case , the trial state ought to be simultaneously rich enough to span the interesting physics and to lead to calculable results .
this last requirement is particularly stringent , since only functional integrals of the gaussian form can be explicitly evaluated .
specific to the quantum field case are the requirements of gauge invariance and of correct uv behaviour . to understand this last statement
, it suffices to mention first , that the free energy is dominated by uv modes while we are interested in the ir physics and , second , that in an asymptotically free theory such as gluodynamics , low and high momentum modes are , in general , non - separable .
the method of @xcite to be discussed below accommodates all these requirements , leading to what is , to the author s best knowledge , the only first principle analytical analysis of the deconfinement phase transition in gluodynamics .
we consider density matrices which , in the field basis , have gaussian matrix elements and where gauge invariance is explicitly imposed by projection onto the gauge - invariant sector of the hilbert space @xmath5 = \nonumber\\ & = \int du \ ; \exp\bigg\{-\frac{1}{2}\int_{x , y } \big [ a_{i}^{a } ( x ) g^{-1 ab}_{ij}(x , y ) a_{j}^{b}(y ) \nonumber\\ & \hspace{2.5cm}+ a^{'ua}_{i } ( x ) g^{-1 ab}_{ij}(x ,
y ) a^{'ub}_{j}(y)\nonumber\\ & \hspace{2.5cm}- 2 a_{i}^{a } ( x ) h^{ab}_{ij}(x , y ) a^{'ub}_{j } ( y ) \big ] \bigg\}\ , , \end{aligned}\ ] ] where @xmath6 , @xmath7 is the @xmath2 group - invariant measure , and under an @xmath2 gauge transformation @xmath8 @xmath9 with @xmath10 and @xmath11 form an @xmath12 hermitian representation of @xmath2 : @xmath13 = i f^{abc } \frac{\tau^c}{2}$ ] with normalization @xmath14 .
we take the variational functions diagonal in both colour and lorentz indices , and translationally invariant @xmath15 further , we split the momenta into high and low modes with @xmath16 and restrict the kernels @xmath17 and @xmath18 to the one parameter momentum space forms @xmath19 the logic behind this choice of ansatz is the following . at finite temperature
we expect @xmath20 to be roughly proportional to the bolzmann factor @xmath21 . in our ansatz ,
the role of one particle energy is played by the variational function @xmath22 and its form is motivated by the propagator of a massive scalar field , i.e. @xmath23 .
we will be interested only in temperatures close to the phase transition , and those we anticipate to be small , @xmath24 .
for those temperatures one particle modes with momenta @xmath25 are not populated , and we thus put @xmath26 . for @xmath27
the bolzmann factor is non - vanishing , but small .
further , it depends only very weakly on the value of the momentum . with the above restrictions on the kernels , only two variational parameters , @xmath28 and @xmath18 , remain .
importantly , the density matrix functional in eq .
( [ qcdans ] ) describes , for @xmath29 , a pure state @xmath30><\psi [ a]|$ ] where @xmath31 $ ] are gaussian wave functionals . for @xmath32 , eq .
( [ qcdans ] ) describes a mixed state with @xmath33 proportional to the entropy of the trial density matrix . the expectation value of a gauge invariant operator in the variational state eq .
( [ qcdans ] ) is then given by @xmath34\bigg\}\bigg|_{a'=a}\ , , \end{aligned}\ ] ] where @xmath35 is the normalization of the trial density matrix @xmath36 , i.e. @xmath37\bigg\ } \end{aligned}\ ] ] to evaluate the above expressions we first perform , for fixed @xmath38 , the gaussian integration over the vector potential @xmath39 .
for @xmath35 we get , in leading order in @xmath18 , @xmath40 we now integrate out the high momentum , @xmath41 , modes of @xmath8 perturbatively to one - loop order .
this effects a renormalization group transformation on the low modes , replacing the bare coupling @xmath42 by the running coupling @xmath43 @xcite . to one - loop accuracy , the coupling @xmath43 runs identically to the yang - mills coupling @xcite .
the normalization @xmath35 can be then interpreted as the generating functional @xmath45 for an effective non - linear @xmath44-model for the low momentum modes ( @xmath46 ) defined by the action @xmath47\nonumber\\ & \quad- \frac{1}{4\pi^2 } h m^2 { { \mathrm{tr}}}u^\dagger { { \mathrm{tr}}}u\ , , \end{aligned}\ ] ] where @xmath8 independent pieces have been dropped .
the matrix @xmath8 plays the same role as polyakov s loop @xmath48 at finite temperature the functional integration over @xmath8 projects out the physical subspace of the large hilbert space on which the hamiltonian of gluodynamics is defined .
this @xmath44-model has a phase transition @xcite at the critical point ( for @xmath49mev , @xmath50 and with the one - loop yang - mills @xmath51 function ) @xmath52 for @xmath53 , the @xmath44-model is in a disordered , @xmath54 symmetric , phase with massive excitations and where @xmath55 .
since @xmath8 is the polyakov loop , this corresponds to a confined state .
when @xmath56 , the @xmath44-model is in an ordered , @xmath57 symmetric , phase with massless goldstone bosons for which @xmath58 , corresponding to a deconfined state . with this analysis
we have established a correspondence between the @xmath44-model phase transition and the deconfinement transition in @xmath2 gluodynamics .
in fact , this correspondence can be argued in rather general terms . instead of restricting ab initio the density matrix to the form eq .
( [ qcdans ] ) , imagine that we take some arbitrary gauge - invariant density matrix ansatz that depends on the @xmath39 fields and is integrated over the @xmath8 fields .
we allow this new ansatz ( and whatever kernels it may contain ) to remain arbitrary until we have no choice but to restrict it
. then we integrate out the a fields to obtain a partition function of @xmath8 fields with respect to some action .
next we introduce a separation of momenta into high and low modes with @xmath16 and integrate out the high mode @xmath8 fields as before .
this effects a renormalisation group transformation on the low modes , replacing the bare coupling @xmath42 which is not arbitrary , since it is defined by the gauge transformations eq .
( [ gt ] ) by the running coupling @xmath43 .
now provided our ansatz is sufficiently close to the correct density matrix for @xmath2 , the theory will be asymptotically free .
we are thus left with an action for the low modes which is , once again , a complicated @xmath44-model , with a renormalised coupling @xmath43 which we expect to be small provided @xmath28 is large , and vice versa . now consider this model as a statistical mechanical model at ` temperature ' @xmath43 .
we make the plausible assumption that this @xmath44-model will , as @xmath28 is varied , undergo a symmetry - breaking transition at ` temperature ' @xmath59 from a ` thermally disordered ' ( symmetric ) phase at large @xmath59 to an ordered phase at small @xmath59 .
further , it is clear since the polyakov loop @xmath60 is zero in the former phase and non - zero in the latter that this @xmath44-model phase transition corresponds directly to the deconfinement transition in the @xmath2 theory . on review ,
it is clear that our only assumptions are that the ansatz is sufficiently close to @xmath2 and that the low mode @xmath44-model undergoes a symmetry - breaking phase transition . in particular , let the ansatz , which is arbitrary and need not be gaussian , be the _ correct _ density matrix for @xmath2 .
the first assumption is certainly true .
if the second assumption is also true , then we have constructed an exact argument that the deconfinement transition in @xmath2 corresponds to the phase transition in the low mode @xmath44-model .
thus , in order to study deconfinement in @xmath2 , our aim should be to model the physics of each @xmath44-model phase as accurately as possible and calculate the transition scale @xmath61 .
we then calculate the free energy of @xmath2 in each phase , including any possible contribution from the high modes , at temperature @xmath62 and extract the minimal free energy .
the deconfinement transition occurs at the temperature for which the free energies calculated in the ordered and disordered phases of the low mode @xmath44-model coincide .
although we will take eq .
( [ qcdans ] ) as the ansatz for the density matrix , we shall keep the kernels @xmath17 and @xmath18 arbitrary until we have no choice but to restrict them .
the helmholtz free energy @xmath63 of the density matrix @xmath36 is given by @xmath64 where @xmath65 is the standard yang - mills hamiltonian @xmath66\ , , \label{ham}\ ] ] with @xmath67 @xmath68 is the entropy , and @xmath69 is the temperature .
thus @xmath70 in the disordered phase , no progress seems possible without restricting the arbitrary kernels .
following @xcite , we adopt the forms eq .
( [ rkers ] ) . for small @xmath18
, we consider only the first non - trivial order in @xmath18 , that is a term of @xmath71 in the entropy .
this term can be written as a product of left @xmath2 and right @xmath2 currents and does , therefore , vanish in the disordered , @xmath54 symmetric , phase @xcite .
the remaining contribution to the free energy , the average of the hamiltonian , is evaluated in the mean field approximation @xcite .
the free energy is minimized for @xmath72 @xmath73 the simplest option to evaluate the free energy in the disordered phase is to use perturbation theory .
perturbation is certainly appropriate for large enough values of @xmath28 , where the expectation value of the @xmath8 field is close to unity . from numerical studies
@xcite it is known that the transition occurs when the expectation value of @xmath8 is greater than @xmath74 .
we can thus expect perturbation theory to be qualitatively reliable down to the transition point . in the leading order perturbation theory approximation to the ordered phase of the @xmath44-model , however , minimisation with respect to arbitrary kernels @xmath17 and @xmath18 for both high and low modes is possible .
further , the analysis can be carried out to all orders in the thermal disorder kernel @xmath18 . in this approximation
, the @xmath8 matrices can be parameterised in the standard exponential form and expanded in the coupling @xmath75 @xmath76 hence at leading order one can take @xmath77 thus , the gauge transformations ( [ gt ] ) reduce to @xmath78 and the hamiltonian ( [ ham ] ) reduces to @xmath79\ , .
\end{gathered}\ ] ] these last two equations describe the theory @xmath80 : in the leading order of @xmath44-model perturbation theory , the @xmath2 yang mills theory reduces to the @xmath80 free theory .
moreover , the density matrix eq .
( [ qcdans ] ) becomes gaussian again , because the gauge transformations are linear .
one has @xmath81
= \int d\varphi \ ; \exp\bigg\{-\frac{1}{2 } \big [ a g^{-1 } a \\ + ( a ' - \partial \varphi ) g^{-1 } ( a ' - \partial \varphi ) \\ - 2 a h ( a ' - \partial \varphi ) \big ] \bigg\}\ , . \end{gathered}\ ] ] now the theory of @xmath82 @xmath83 free fields in @xmath84 dimensions is completely tractable ; the variational analysis for the @xmath83 theory ( with gaussian ansatz ( [ gans ] ) ) was discussed in @xcite .
the free energy in momentum space in terms of the arbitrary kernels @xmath17 and @xmath18 is @xmath85 \nonumber\\ & \qquad- \log \big [ \frac{1- ( 1-(gh)^2)^{1/2}}{gh}\big ] \cdot\nonumber\\ & \qquad\quad\cdot\big ( \frac{1- ( 1-(gh)^2)^{1/2}}{(1-(gh)^2)^{1/2}-(1-gh ) } \big ) \big ) \bigg]\ , .\end{aligned}\ ] ] the kernels which minimise the free energy are @xmath86 and the minimal value of the free energy at temperature @xmath62 is @xmath87 so we see that the free energy of @xmath2 is minimised with @xmath88 in the disordered phase of the @xmath44-model for temperatures from zero up to a temperature @xmath0 where @xmath89 which in turn implies @xmath90
we find that the deconfinement phase transition is strongly first order with a transition temperature of @xmath91 .
the transition is due to an abrupt jump in the entropy . in the disordered phase it is zero
since glueballs are heavy and , therefore , their contribution is suppressed by the boltzmann factor @xmath92 . in the ordered phase ,
the entropy is non - vanishing and proportional to the number of coloured states .
the major uncertainty in the calculation is due to the fact that the scale @xmath61 is very sensitive to the mean field approximation .
it is , therefore , only meaningful to compare the results of our approach with lattice estimates for @xmath61 independent quantities .
for one such quantity , the transition temperature in units of the mass of the lightest glueball in the model , i.e. @xmath93 @xcite , we find @xmath94 which is in agreement with the lattice result .
this talk is based upon work done in collaboration with ben gripaios , ian kogan and alex kovner .
this work and my participation in the workshop were funded by the european commission ihp programme under contract hprn - ct-2000 - 00130 .
|
the deconfinement transition in 3 + 1 dimensional gluodynamics is studied using the gauge invariant variational method introduced by kogan and kovner a few years ago .
we identify a first order phase transition , characterized by a discontinuous jump in the entropy of the system , resulting in a transparent picture of the mechanism of deconfinement .
the calculation of the ratio of the transition temperature to the mass of the lightest glueball in the model yields 0.18 in complete agreement with the lattice estimate .
|
the advances in fabrication technologies have made it possible to experimentally fabricate one dimensional electronic structures@xcite@xcite .
this has naturally resulted in an increasing interest in theoretical investigations of such structures@xcite@xcite .
the resulting one dimensional electron gas has also been subject to computational investigation@xcite .
the extensive investigations in the three dimensional electron gas have clearly shown the importance of the short range correlations at lower densities in determining the pair distribution function @xmath0 at small @xmath1@xcite . in the high density limit , the long range correlations can be described well by the random phase approximation ( rpa ) .
one way to include the short range correlation effects beyond rpa is to use the powerful approach developed by stls@xcite .
it is well known that stls approach suffers from a compressibility inconsistency .
the compressibility calculated from the small @xmath2 limit of the dielectric function @xmath3 does not agree with that calculated from the ground state energy .
there have been several attempts to overcome this inconsistency@xcite .
sstl@xcite included the effect of screening on the effective interaction potential as explained in the next section . in this work
, we aim to check the performance of the sstl approach and compare it with the stls in a one dimensional electron gas . to the best of our knowledge ,
this is the first application of the sstl approach in lower dimensions .
the organization of the paper is as follows ; the stls and sstl formalisms are given in section 2 .
the results and discussion concentrating on the comparison between stls and sstl performance on the compressibility issue are presented in section 3 .
in the mean field approximation , a key component of the electron gas is the usual static structure factor s(q ) which is related to density
density response function @xmath4 through the fluctuation
dissipation theorem as @xmath5 the density
density response function @xmath4 is defined as ( q,)= , where @xmath6 is the zero temperature susceptibility of a noninteracting electron gas and is given in one dimension by _ 0(q,)=_01(q,)+i _ 02(q , ) , with _ 01(q,)=| | , and @xmath7 where @xmath8 are the boundaries for particle hole excitations .
the fermi wave vector @xmath9 is related to the one dimensional electron density @xmath10 via @xmath11 .
the dimensionless electron density parameter is defined as @xmath12 , where @xmath13 is the effective bohr radius .
@xmath14 in eq .
( 2 ) is the self consistent effective potential related to the static structure factor through v_eff(q)=v(q)+_0^ \{(q+k)v(q+k)+(q - k)v(q - k ) } , where @xmath15 is the fourier transform of the coulomb interaction between two electrons in the lowest subband in the harmonic confinement in one dimension and is given by v(q)= f(q ) , where @xmath16 is the background dielectric constant , @xmath2 is the wave vector along the wire and @xmath17 is the form factor which takes into account the finite thickness of the wire .
the form factor reads f(q)= exp ( ) k_0 ( ) , here @xmath18 is the zeroth order modified bessel function of the second kind and @xmath19 is the characteristic length of the harmonic potential .
indeed , it is a measure of the effective radius of the quantum wire@xcite . in the stls approximation , ( q,)= , where the local field correction @xmath20 is g(q)=-_-^ [ s(q - k)-1 ] .
the set of eqs .
( 1 ) , ( 8) and ( 9 ) have to be solved self
consistently for @xmath20 , @xmath21 and @xmath22 within the stls approximation . in rpa , @xmath23 .
the sstl approximation is different from the stls approximation in that the potential under the integral sign in eq .
( 5 ) is screened by the static dielectric function @xmath3 which is given by , ( q)=1- . this is originally done to better satisfy the compressibility sum rule in three dimensional electron gas .
the ground state energy per particle in one dimensional electron gas may be written as@xcite _ g=_kin+_ex+_cor , where @xmath24 is the kinetic energy per particle , and simply @xmath25 in units of effective rydberg @xmath26 .
@xmath27 is the exchange energy per particle _
ex=-_0^k_f dq _ -(k_f+q)^k_f - q dk v(k ) , and @xmath28 is the correlation energy per particle _
cor=_0^r_s dr_s^ _ 0^ dq v(q ) .
the compressibility @xmath29 may be calculated either by using the ground state energy per particle @xmath30 = n^2 (
n _ g ) , or by using the @xmath31 limit of the static dielectric function @xmath3 _ q 0 ( q)=1+v(q ) n^2 k. the compressibility sum rule is that the compressibilities calculated by using eq . ( 14 ) and eq . (
15 ) are the same .
in fig . 1 the local field correction @xmath20 calculated self consistently using stls and sstl approximations are shown for fixed @xmath32 and @xmath33 .
the most striking difference between the two curves is their large @xmath2 limits . as stls
@xmath20 is approaching 1 , sstl @xmath20 approaches a limit bigger than 1 .
it may be noted that the relation@xcite @xmath34 is satisfied in our calculations .
this , of course , is going to lead to different results when applied to physical properties of the one dimensional electron gas .
the different large @xmath2 behaviour of @xmath20 is expected to lead to different small @xmath1 behaviour of the pair correlation function @xmath0 .
this is shown in fig .
2 . as radius of the wire gets smaller the difference between stls and sstl results become larger , as may be seen in fig .
as the density becomes smaller ( i.e. , @xmath35 becomes larger ) the sstl @xmath0 becomes negative for small @xmath1 .
this limit is where the correlation effects become important .
the compressibility calculated by using the ground state energy in three different approaches is presented in fig .
6 as a function of @xmath35 .
the similarity in energies is also reflected in these curves .
the negative compressibility values at higher @xmath35 show that the system becomes unstable .
the compressibility calculated by using @xmath31 limit of the static dielectric function is shown in fig .
the stls and sstl compressibilities calculated by two routes are as different as in rpa .
this surprising result is a natural consequence of the small @xmath2 behaviour of the local field correction which is very small for sstl at small @xmath2 values .
the compressibility in a one dimensional system is studied previously by gold and calmels@xcite using a three - sum - rule approach for the local field correction .
the compressibility sum rule is better satisfied within this variant of stls in one dimension as can be seen from their figure 4 .
we find the same result by the present full stls calculation .
we can conclude that the short range correlations are calculated using the sstl approximation in a one dimensional electron gas , for the first time .
the performance of the sstl approximation is compared with the more widely used stls approximation .
it is shown that sstl compares reasonably well with stls when the pair correlation function , dielectric function and the ground state energy per particle are considered but fails totally when the compressibility sum rule is checked .
|
we use the sstl ( singwi , sjlander , tosi , land ) approximation to investigate the short range correlations in a one dimensional electron gas , for the first time .
although sstl is introduced to better satisfy the compressibility sum rule in three dimensions , the widely used stls ( singwi , tosi , land , sjlander ) approximation turns out to be more successful in the case of the one dimensional electron gas .
16.3 true cm 23.0 true cm -0.8 true in 0.00 true in keywords : quantum wires ; screening ; compressibility
|
the existence of broad absorption line quasars and blue shifted absorption lines in seyferts provides clear evidence that high speed outflows of gas are common in agn .
the basic questions one would like to have answered by a theoretical model for these outflows are : * what is the origin of the gas flowing out ? * what is its geometry ? *
what force accelerates it ? * what determines its ionization state ?
while answering these questions , a good model should satisfy these observational constraints : * it should produce lines similar to the observed ones .
significant column densities ( @xmath0 to @xmath1 @xmath2 ) of both high ionization potential ions ( o vi , n v , civ etc . ) and low ionization potential ions ( o iii , si iv ) should be produced .
* material has to be accelerated up to at least 30,000 km sec@xmath3 in qsos and 5000 km sec@xmath3 in seyferts * the ionization parameter should not vary drastically for gas with different speeds along the line of sight * the outflow should not produce emission at speeds higher than observed in the emission lines , which presumably implies it has a covering factor of less than 20 % ( hamann et al . 1993 ) * the geometry should be such that high speed bal material occurs outside the ly@xmath4 broad emission line region , since ly@xmath4 is seen to be heavily absorbed by high speed material containing n v ions .
* it should explain why there is such a wide range in absorption line morphology ( turnshek 1988 ) * it should be consistent with the lack of variability in the structure of the absorption lines in balqsos over timescales of 10 years ( barlow 1994 ) in this paper i will give a critical review of the theoretical models that have been proposed so far , and discuss to what extent they fulfil the criteria above .
the fact that the strongest absorption lines formed in the outflow are of species like c iv , o vi , o iii , si iv and n v suggest that if the absorbing gas is irradiated by a `` standard '' agn spectrum ( e.g. mathews and ferland 1987 ) the ionization parameter @xmath5 ( density of photons below the lyman limit divided by the hydrogen density ) is of the order @xmath6 .
this leads immediately to an estimate of the density of the absorbing gas of @xmath7 with @xmath8 the luminosity of the central engine in units of @xmath9 ergs / sec , @xmath10 the size of the region in which the absorption lines are formed in units of @xmath11 cm , and @xmath12 the ionization parameter normalized to 0.1 .
estimates of the total column density of the absorbing gas are difficult to obtain due to saturation of the lines , but are typically in the range @xmath13 , so that the total thickness of the absorbing layer is @xmath14 from the fact that both the continuum and the broad emission lines are seen to be absorbed , we know that the absorbing region lies outside or is cospatial with the bel region , we have @xmath15 .
thus , comparing the size of the absorbing region with the thickness of the absorbing layer we arrive at a very small filling factor for the absorbing gas , @xmath16 if we require additionally that the flow gives rise to the smooth absorption troughs observed in some balqsos that extend over hundreds to thousands of thermal line widths , it is clear that we will need hundreds to thousands of clouds in our line of sight , implying very small sizes ( @xmath17 ) for individual clouds .
what are the consequences of this low filling factor ?
it is clear that there has to be a confining medium around the clouds to keep them from expanding and becoming too highly ionized , since the sound crossing time of the clouds is very short .
there are two obvious choices for this confining medium : hot gas or magnetic field .
confinement by hot gas would be the simplest explanation , but it leads to a large number of problems ( weymann et al .
1985 , begelman et al . 1991 ) .
if there is any velocity difference in the outflow between the clouds and the hot gas , the timescale for destruction of the clouds by hydrodynamical instabilities such as kelvin - helmholtz or rayleigh - taylor is very short compared to the acceleration time .
the flow could consist of a true two - phase medium in which cool clouds could continuously reform , but unless the densities are higher than we infer , the timescale to cool down from the hot phase is longer than the crossing time of the absorbing region .
because of the very low filling factor of the cool clouds , the mass of gas in the hot phase will be larger than in the clouds unless it has a very high temperature , @xmath18 .
if the temperature of the hot phase would be around the compton temperature of the agn radiation field , the mass loss in the hot phase would be uncomfortably large , of the order of tens of solar masses per year . however , gas at a temperature of @xmath18 is not able to confine the small clouds thought to exist in the absorbing region , because the mean free path of the hot particles is longer than the size of the clouds . finally , there is a problem with the fact that the ionization parameter of the absorbing clouds is observed to remain approximately constant as they are accelerated ( e.g.turnshek 1988 ) .
any hot gas flowing out with the clouds will cool down quickly due to adiabatic expansion losses , causing the confining pressure to drop much too rapidly to keep @xmath5 approximately constant .
thus , any hot gas confinement model has to have some way to release a lot of energy into the hot gas as it is being accelerated .
confinement by magnetic field avoids several of the problems associated with hot gas
. if the clouds travel along magnetic field lines , hydromagnetic instabilities are likely to be much less severe , and the drag forces could be very small , although these effects have not been studied in much detail yet .
a magnetic field could provide confinement in the direction perpendicular to the field , but other processes have to be invoked for confinement along the field .
one advantage is that the cloud can expand only in one dimension , so that it will take a much longer time for the density to drop than in the case of spherical expansion .
confinement along the field could possibly be provided by ram pressure or radiation pressure from the central source , since the development of the rayleigh - taylor instability normally associated with this process will be suppressed by the magnetic field .
there is in fact some indication that the clouds are being pushed , since the small size of the clouds is consistent with the scaleheight in the cool gas assuming that they experience an `` artificial gravity '' equal to their mean acceleration @xmath19 ( de kool and begelman 1995 ) .
another possibility is that the flow is self - confining in the direction along the flow , i.e. it consists of very thin flow tubes with a very low filling factor and small size perpendicular to the field , that are completely filled with cool material .
such a picture would fulfil all the constraints derived above . simple magnetic field models ( such as completely radial or azimuthal field lines ) also can not reproduce a confining pressure that is approximately constant as the clouds are accelerated , since the field strength drops too rapidly .
more complicated field geometries are clearly required .
the large drag forces that the absorbing clouds experience when they move relative to any confining gas led weymann et al .
1985 to the conclusion that the clouds are likely to be dragged along in an accelerating confining medium rather than being accelerated themselves .
the low density and high pressure in the confining gas make it a good candidate for a pressure driven wind .
pressure driven winds typically reach end speeds of a few times the sound speed in the wind , so the high velocity observed in balqso outflows requires either an extremely hot wind ( @xmath20 ) , or a wind driven by non - thermal particles such as cosmic rays .
two models have been proposed along these lines .
the first was by weymann et al .
1982 , which attempted to explain the broad emission line region as a hot outflow with embedded clouds , driven by relativistic electrons diffusing out of the central engine . the second model ( begelman et al .
1991 ) was specifically aimed at explaining the broad absorption lines . here
the primary cosmic rays were highly relativistic neutrons that are produced in the central engine , that can travel out to parsec scale before they decay to protons and couple locally to the magnetic field . because the neutrons decay over a wide range of radii , there is a steady energy injection into the wind as it is being accelerated , compensating for the adiabatic losses that would otherwise make the confining pressure drop too fast .
this model can produce bals with characteristics similar to the observed ones , but it does require significant fine tuning of the parameters . note that cosmic rays have very long mean free paths , and will penetrate the bal clouds easily , so that the cosmic ray pressure itself can not provide confinement .
thus , dissipation of cosmic ray energy into the thermal intercloud gas ( to equipartition ) seems to be required anyway .
the origin of the clouds in these models is not clear .
for stability reasons , it would be preferable if large velocity differences between the clouds and the confining medium would never develop , and from this point of view it is most attractive to inject the clouds at the base of the wind and let the ensemble accelerate together
. on the other hand , there are many bal profiles in which the absorption sets in sharply at some high velocity , which would be hard to understand in a radial outflow picture .
an alternative injection mechanism is that the wind is accelerated on smaller scales , and hits some large ( molecular ? ) cloud at high speed , ripping off material from the surface . in this picture
, we would only see a bal if we are looking through the wake of such a cloud .
it is doubtful however that the clouds of cold material could survive long enough to be accelerated to velocities as high as observed , because ( at least simple ) estimates of the hydrodynamical destruction time are much shorter than the acceleration time .
numerical models along these lines ( schiano et al . 1995 ) have not been conclusive .
observationally , we do not have much direct evidence for the existence of a very hot wind in balqsos .
stocke et al .
( 1992 ) have argued that the radio emission from radioquiet balqsos and seyferts is consistent with a hot wind model , and exhibits the proper scaling of outflow speed with luminosity .
the studies of this outflow mechanism have so far been restricted to the formation of broad emission lines ( emmering et al .
1992 , bottorff et al . 1997 ) or dusty winds at large disk radii that could form the obscuring torus in seyfert 2 galaxies ( knigl and kartje 1994 ) .
clearly it is also a good candidate for the origin of blueshifted absorption lines , but there are as yet no models exploring the formation of absorption lines in such flows in detail .
all these models are based on the basic self - similar outflow model proposed by blandford and payne ( 1982 ) .
it should be kept in mind that this model is highly simplified to make it mathematically tractable .
there are no reasons why the flow should be self - similar , since self - similarity is only obtained for a special set of boundary conditions that is not preferred over others . in this model
( see figure 1 ) cool clouds of gas emerge from the surface of the disk threaded by magnetic field lines that are anchored in disk .
if the angle between the field line and the disk is small enough , the cloud will move outward but is forced to corotate with the point in the disk where the field line is anchored , making it rotate faster than local kepler speed .
this causes a centrifugal force on the cloud , that tends to move it to even larger radii .
the speeds attained scale with the kepler speed at the field line footpoint , and with the ratio of magnetic flux to mass flux on the surface of the disk . for reasonable values , outflow speeds of the order 3 times the kepler speed are reached .
even the simple self - similar model has about ten free parameters to specify the exact flow geometry ( bottorff et al .
1997 ) . in the light of this
, it may not be surprising that this model is able to fit the bel profiles of many sources quite well . emmering et al .
1992 did however need to invoke an extra source of scattering for the line photons in order to obtain wide enough line profiles if the base of the outflow is restricted to two decades in radius on the surface of the disk .
an independent test of this model was performed by bottorff et al . who showed that for ngc 5548 , a model of this kind for which the parameters have been fixed by fitting the line profile alone can reproduce the response of the line profile to continuum variability quite accurately ( see these proceedings ) .
to what extent this type of outflow can also be responsible for bals is not clear .
there may be a problem in obtaining high enough speeds , since bal speeds are typically several times higher than bel speeds , and the model already has problems obtaining bels as broad as observed .
also the basic geometry of the self - similar model is not favorable to explain the fact that high velocity bal material is obscuring low velocity material , since in these models the high velocities occur on much smaller scale than the low velocities , typically scaling with @xmath21 .
thus the high velocity material with bal - like speeds would typically be expected to be 10 - 100 times closer in than the material with typical broad emission line speeds . for seyferts however , which show absorption lines that typically have speeds similar to the broad emission line width , this seems a more promising model .
a theoretical problem may be the neglect of the effects of radiation pressure on the clouds , which should be important unless the clouds are very optically thick .
one suggestion about the nature of balqsos is that they are qsos near the end of their forming stage , that are blowing out the remaining circumnuclear gas after the central engine has turned on ( e.g. voit et al . 1993 ) .
an important driving force for this outflow would be radiation pressure on dust .
this model is not very popular , because the argument that the upper limit on the covering factor of the bal region is about equal to the observed fraction of balqsos to non - balqsos suggests that all qsos must have bal regions .
however , the point about the likely importance of radiation pressure on dust is a very good one .
scoville and norman ( 1995 , see also these proceedings ) have worked out a model in which bal material derives from the stellar winds of cool giants , that is accelerated outward by radiation pressure acting on the dust in the wind .
the high outflow speeds of @xmath22 30,000 km sec@xmath3 are a natural consequence of this model , given that the dust opacity is relatively well known , and the starting radius of the outflow must be approximately equal to the radius ( @xmath22 1 pc ) where the equilibrium temperature of the dust is equal to the evaporation temperature of @xmath22 1800 k. a weak point in this model appears to be the confinement of the cool gas in the wind trails from the giants , which is due to the trails being compressed between the radiation pressure from the inside , and the ram pressure due to the non - comoving surrounding medium from the outside , which would appear to be very unstable . from an observational point of view , there is the problem that many high - ionization balqsos show no signs of dust absorption in their uv spectra ( e.g. weymann et al .
1991 , korista et al .
1992 ) .
there are several compelling arguments that acceleration by uv line photons plays an important role in balqsos .
it has the advantage over the other models that the source of the acceleration is directly observed . making a simple estimate of the total momentum in the outflow : @xmath23 with @xmath24 the covering fraction of the bal outflow in units of 0.1
, @xmath25 the column density of the bal region in units of @xmath26 , @xmath27 its size in units of @xmath11 cm , and @xmath28 the wind speed in units of @xmath29 @xmath30 , and comparing this to the momentum absorbed from the radiation field in the bal troughs : @xmath31 where @xmath32 is the number of important driving lines divided by 10 , and @xmath33 , we find that line radiation pressure must play an important role . furthermore , there is some direct evidence for the dynamical importance of line radiation by the occurrence of line locking in some si iv profiles ( turnshek et al .
1988 ) and extra acceleration of the flow at velocities where the ly@xmath4 broad emission line overlaps with the n v absorption , resulting in a steeper velocity gradient and a reduced optical depth in the bal profiles of other species ( arav and begelman 1994 ) .
the standard problems associated with small clouds and low filling factors ( drag , hydrodynamical destruction , confinement with constant u ) also plague this model , but recent studies of the instabilities occurring in line driven winds from o - stars may provide some solutions to these problems ( e.g. owocki 1994 , and the contribution by feldmeier and norman in these proceedings ) . because of the similarity of many bal profiles with stellar p - cygni profiles of hot stars with fast winds driven by line radiation , this model was in fact among the first to be considered for balqsos . in analogy with hot stars , drew and boksenberg ( 1984 ) constructed spherically symmetric wind models with a filling factor of 1 .
they were not very successful in reproducing the observed profiles because of the very reasons that led to the constraints on the filling factor above : unrealistically high mass loss rates are necessary to keep the density high enough , and thus the ionization parameter sufficiently low to avoid heating the gas to too high temperatures , and it was very difficult to keep the ionization parameter approximately constant as the flow is accelerated .
also the model could not account for the absence of emission at high velocities , a consequence of the spherical symmetry that was assumed .
an early radiatively driven disk wind model for the broad emission line region was proposed by shlosman et al .
( 1985 ) . in this model
the wind is not driven radially outward , but vertically up from the uv - emitting part of the accretion disk .
while it is still in the shadow of the atmosphere of the inner disk the wind is in a single , cool phase .
as it gets exposed to the full hard agn continuum above the atmosphere it becomes thermally unstable and turns into a two - phase state , with cool clouds embedded in a hot gas .
this effectively launches clouds with a velocity perpendicular to the disk , and it is assumed that after this radiation pressure is no longer important . in this model
, we would expect to see broad absorption lines under most viewing angles , with the highest velocities occurring when we are observing the disk face - on , in contrast to most other disk wind models where the wind is visible only if our line of sight is close to the disk plane .
more recently , arav et al .
( 1994 ) reinvestigated the possibility of line radiation pressure driven winds in balqsos .
they decided to neglect all the problems associated with low filling factors , and just _ assumed _ that the absorbing clouds are infinitely small , and embedded in a massless confining medium that exerts no drag forces , and has the correct run of confining pressure with radius to be consistent with observations .
once these assumptions are made , the classic cak theory for line driven winds ( castor et al .
1975 ) can be applied to derive the properties of the wind .
this theory is remarkably successful and correctly predicts the important driving lines , and the final velocity if it is assumed that the flow starts at radii comparable to the broad emission line region , and is exposed to the optical to x - ray continuum produced close to the center of an agn .
they also pointed out that if the confining pressure varies with radius in such a way that the ionization parameter is approximately constant , the gradient of this pressure should play a non - negligible role in the acceleration , and a model taking this into account was presented .
a possible candidate for the massless confining medium with just the right properties was suggested by de kool and begelman ( 1995 ) .
their model is a hybrid between the radiation pressure and magnetic disk wind models , that is based on the simple estimate that line radiation pressure will accelerate clouds much faster than the magnetic centrifugal effect ( if they are not too optically thick ) . just as in the model of emmering et al .
( 1992 ) cool matter emerges from the disk threaded by magnetic field .
as it is exposed to the central continuum , the matter is heated and expands until it is in pressure equilibrium with the magnetic field around it .
line radiation pressure then accelerates the clouds , dragging the magnetic field along with it ( see figure 2 ) .
the radiation pressure alone would accelerate the flow in a very thin wedge over the surface of the disk .
however , all of the magnetic flux emerging from the inner parts of the disk is forced into this wedge , and if it becomes too thin large magnetic pressure gradients develop that will widen the outflow wedge again .
thus an equilibrium configuration will develop in which the magnetic pressure gradients balance the radiation pressure , which is only possible if the magnetic pressure is a significant fraction of the radiation pressure .
this is just what is required to obtain an ionization parameter in the clouds that is self - regulating and approximately constant .
the most recent example of a radiatively driven disk wind model is the one by murray et al .
this differs radically from all the other ones , since it assumes that the filling factor of the absorbing material is unity . in order to avoid the earlier arguments against this
, the authors argue that the ionization parameter of the bal gas is about three orders of magnitude higher than previously assumed .
this is possible if one assumes that the ionizing spectrum irradiating the bal gas is not the the full central continuum , but rather one that has already passed through a high column density highly ionized absorber ( figure 3 ) .
this would absorb the far uv and the soft x - rays out of the spectrum , and there would not be enough hard photons left to ionize the observed bal species like c iv completely away .
a very strong point of this model is that indeed all balqsos appear to have strong absorption in their soft x - ray spectra , since rosat observations have shown them to be very weak soft x - ray emitters ( green and mathur 1996 ) . in the one case where a column density could be determined ( phl 5200 , mathur et al .
1995 ) a value consistent with the ones needed to keep the ionization low was found . in subsequent papers ( chiang and murray 1996 , murray and chiang 1997 ) it is argued that the emission lines formed at the base of the disk wind can also account for the reverberation mapping results and broad emission line profiles from seyfert galaxies ( see also these proceedings ) .
this is a very attractive picture because it avoids all the problems associated with clouds .
however , a few problems with this model have not been satisfactorily resolved .
the first one is that the model requires that the bals are formed at very small radii @xmath34 cm .
the reason for this is that the highly ionized wind is not as easily accelerated because of the low abundance of the ions that drive the wind . to obtain speeds
as high as observed in balqsos , it is necessary to start the flow very close in . since the evidence that the bel region lies inside or is cospatial with the bal region is incontrovertible
, this also implies that the bel region in qsos has a much much smaller size than commonly accepted . with these small scales ,
the crossing time of the bal region is only a month or so , and it is difficult to understand that the highly complex kinematical structures that bals often exhibit do not appear to vary on timescales of 10 years ( barlow 1994 ) . even if the absorption components we see are not actual cloud complexes being accelerated but rather standing flow patterns , it would be likely that they are anchored to certain positions on the disk , and the short rotation period of the disk would be likely to cause changes on the observed timescale .
any structure in the wind that is due to the line driving instability is not likely to be axisymmetric , and the analogy with o - star winds suggest that such structures would be seen to move in a few dynamical times ( owocki 1994 , braun and milgrom 1990 ) .
it is not clear if the full range of ions observed to show bal profiles can be explained by the very high ionization parameters required by this model . as an example , the ionization model presented by murray et al .
would have difficulties explaining the presence of broad absorption lines from ions with ionization potentials like o iii or lower .
it would also be interesting to explore the consequences of the small broad emission line region in more detail , to see if the proper line ratios can be obtained , and if the density limits obtained from semi - forbidden lines are not violated .
the origin and dynamics of the highly ionized absorber assumed to exist between the hard continuum source and the wind needs to be worked out in more detail .
murray et al .
argue that it is a `` failed wind '' which attempts to start from the disk at radii interior to the real wind , which becomes too highly ionized to achieve escape speed but which could still get far enough above the disk surface to shield the outer wind .
their derivation of the dynamics is very approximative , and a more rigorous calculation ( which would involve some fairly complicated 2-d radiation hydrodynamics ) would lend more credibility to this model .
it is clear that the problem of the origin of the blueshifted uv absorption lines in agn is far from being solved .
looking back at the the basic questions posed in the introduction , it appears that there are clear favorites for the accelerating force ( radiative acceleration by uv line photons ) and geometry ( disk - like , compare the great similarities between figures 1 , 2 and 3 ) , but the origin of the bal material and the regulation mechanism for the ionization parameter are still very uncertain .
one of the outstanding problems is the lack of variability in the kinematic structures in the balqso line profiles , which is beginning to be an embarrassment even for the models operating on parsec scale , since small velocity shifts in the components could have been easily detected .
no model has as yet attempted to explain the highly irregular structure of the absorption line profiles , with both very smooth broad and some quite narrow ( @xmath22 a few hundred km sec@xmath3 ) components .
it seems likely that most of the processes that play a role in the formation of the absorption lines have been recognized .
it now remains to look at them in much more detail to see if the primitive models that have been proposed so far can really fit all the observational constraints .
brad peterson suppose that broad emission line clouds are not confined and can freely expand ( e.g. expanding atmospheres of stars ) .
the clouds , driven outward by radiation pressure , expand and become optically thin , producing the optically thin component of the broad - line emission .
these clouds will continue to expand , and will heat up to the compton temperature ( higher in seyferts than in qsos because of the flatter @xmath35 ) ; in seyferts , you get a highly ionized warm absorber , and in qsos you get high velocities , but lower ionization levels like bal clouds .
what are the major failings of this very simple model ?
|
in this paper i will review the dynamical models that attempt to explain the outflows in qsos and seyfert galaxies that are responsible for the blue shifted absorption lines observed in some of these objects .
most models face the difficulty that the absorbing gas appears to consist of very small , low filling factor clouds , that are likely to be hydrodynamically unstable , and require an intercloud medium at high pressure to keep them confined .
two types of intercloud medium have been proposed : hot gas and magnetic field .
hot gas confinement leads to many apparent physical contradictions , and until these are resolved can not be considered a serious candidate .
magnetic confinement seems more promising , but needs more study to assess its viability .
a third possibility that should be seriously considered in the light of the confinement problem is that some of the arguments leading to the conclusion that a small filling factor is required is wrong . as an example , murray et al . showed that a model that allows the ionization parameter of the absorbing gas to be much higher than generally thought can avoid the confinement problem .
three types of acceleration mechanisms have been considered for wind - like outflows in agn : hot gas or cosmic ray pressure driven winds , winds driven by radiation pressure and centrifugally driven magnetic disk winds .
the first two have been shown to be able to produce absorption lines similar to the observed ones , with line radiation pressure requiring the smallest leaps of faith .
the formation of absorption lines in magnetic winds is still unexplored .
|
the discriminants of the chebyshev @xmath0 and @xmath1 polynomials are given by simple and elegant formulas : @xmath2 and @xmath3 is there anything comparable for linear combinations or integral transforms of chebyshev polynomials ? for a special type of linear combination , formula was generalized in @xcite to @xmath4 where @xmath5 for various formulas related to this type of linear combination , see @xcite and @xcite .
what can be said about discriminants in @xmath6 of an integral transform @xmath7 where @xmath8 is some chebyshev polynomial ?
our goal here is to show that the resulting polynomials have discriminants that factor in a remarkable way .
more precisely , when @xmath9 the discriminant factors into two polynomials , one of which has coefficients that are much larger than the coefficients of the other .
moreover , these two polynomials have `` almost '' the same roots , and their discriminants involve exactly the same prime factors .
for example , when @xmath10 , these two polynomials are @xmath11 the discriminants of @xmath12 and @xmath13 are @xmath14 and @xmath15 respectively .
the roots of @xmath12 rounded to 5 digits are @xmath16 whereas those of @xmath13 are @xmath17 we notice that in this case , the discriminants of these two polynomials have the same ( rather small ) prime factors . and
except for the root 0.98272 of @xmath12 and 0.0032902 of @xmath13 , the remaining roots are pairwise close .
in fact , for @xmath18 large we can show , after deleting one root from each of @xmath12 and @xmath13 , that the remaining roots can be paired in such a way that the distance between any two in a given pair is at most @xmath19 .
we call the small polynomial @xmath12 the `` mutt '' polynomial and the large polynomial @xmath13 the `` jeff '' polynomial after two american comic strip characters drawn by bud fisher .
they were an inseparable pair , one of whom ( `` mutt '' ) was very short compared to the other ( `` jeff '' ) .
these two names are suggested by kenneth stolarsky .
now , what will happen if we take the discriminant in @xmath20 of @xmath21 instead of taking the integral ?
it is not difficult to show that this discriminant is @xmath22 where @xmath23 is a constant depending on @xmath18 .
as we will see , one remarkable property about the polynomial @xmath24 is that its discriminant has the same factors as the discriminants of our polynomials @xmath12 and @xmath13 and its roots are almost the same as those of these two polynomials .
the main results are given by theorems 2 , 3 and 4 in sections 4 , 5 and 6 respectively . in section 2
we provide the reader with a convenient summary of notations and properties of chebyshev polynomials , discriminants , and resultants .
section 3 of this paper defines the mutt and the jeff polynomials .
section 4 and 5 of this paper analyze the discriminants of @xmath13and @xmath12 .
section 6 shows that the roots of these two polynomials are pairwise close after deleting one root from each .
the discriminant of a polynomial @xmath25 of degree @xmath18 and leading coefficient @xmath26 is @xmath27 where @xmath28 are the roots of @xmath25 .
the resultant of two polynomials @xmath25 and @xmath29 of degrees @xmath30 and leading coefficients @xmath31 respectively is @xmath32 the discriminant of @xmath25 can be computed in terms of the resultant between this polynomial and its derivative @xmath33 in this paper we compute the discriminants and resultants of polynomials related to chebyshev polynomials . here
we review some basic properties of the chebyshev polynomials .
the chebyshev polynomial of the second kind @xmath1 has the derivative @xmath34 the derivative of the chebyshev polynomial of the first kind @xmath0 is @xmath35 the connections between these polynomials are given by @xmath36 mourad ismail @xcite applied the following theorem from von j. schur @xcite to compute the generalized discriminants of the generalized orthogonal polynomials : let @xmath37 be a sequence of polynomials satisfying the recurrence relation @xmath38 and the initial conditions @xmath39 , @xmath40 .
assume that @xmath41 for @xmath42 .
then @xmath43 for @xmath44 .
ismail s idea is that if we can construct @xmath45 and @xmath46 so that @xmath47 then @xmath48 in the special case of the chebyshev polynomial satisfying the recurrence relation @xmath49 and the differental equation @xmath50 we obtain @xmath51 and @xmath52 for the discriminants of various classes of polynomials , see @xcite .
in this section we will show that the discriminant of the integral transform of chebyshev polynomial factors into the square of the product of the mutt and jeff polynomials whose formulas will be provided . while the formula for the mutt polynomial can be given explicitly in terms of the chebyshev polynomials
, we can only describe the jeff polynomial by its roots . by integration by parts
, we have @xmath53 this is a polynomial in @xmath6 with the leading coefficient @xmath54 the discriminant of this polynomial in @xmath6 is @xmath55 where @xmath56 is a rational number depending only on @xmath18 and can be different in each occurrence . also the factors of its numerator and denominator
can only be 2 or factors of @xmath57 , @xmath58 . from this
, we define the mutt polynomial @xmath59 also , we can define , within a plus or minus sign , the jeff polynomial @xmath13@xmath60 $ ] as the polynomial of degree @xmath61 whose coefficients are relatively prime and for which @xmath62 where @xmath63 is a suitable rational number .
in this section , we find the factors of the discriminant of @xmath13
. in particular we will prove the theorem : the discriminant of @xmath13 has the same prime factors as those of the discriminant of @xmath24 .
also @xmath64 since we do not have an explicit formula for @xmath13 , we will compute the discriminant as the square of the product of the distances between the roots .
we first note that @xmath65 is a polynomial in @xmath66 whose leading coefficient is @xmath67 thus from , the factors of the leading coefficient of @xmath13 can only be 2 or factors of @xmath68 , @xmath69 , @xmath18 .
we now consider the roots of @xmath13 .
according to the formula , @xmath13 has @xmath61 real roots given by the formula @xmath70 where @xmath71 and @xmath72 .
the derivative formula implies that @xmath73 thus the equation gives @xmath74 hence the roots of @xmath13 can be written as below : @xmath75 thus from the definition of discriminant as the product of differences between roots , it suffices to consider the discriminant of @xmath24 . the formulas and give @xmath76 we note that @xmath24 is a polynomial of degree @xmath61 with the leading coefficient @xmath77 from the definition of discriminant in terms of resultant and the formulas and , we have @xmath78 where @xmath79 can be computed from its trigonometric definition : @xmath80 thus @xmath64 thus @xmath81 has factors 2 , 3 , and factors of powers of @xmath18 , @xmath57 , @xmath58 .
in this section , we will present an explicit formula for the discriminant of @xmath12 and show that this discriminant also has factors 2 , 3 , and factors of powers of @xmath18 , @xmath57 , @xmath58 .
in particular , we will prove the theorem : the discriminant of @xmath12 is @xmath82 we note that @xmath12 is a polynomial in @xmath66 of degree @xmath61 whose leading coefficient is @xmath83 . to compute @xmath84 we notice the following fact : @xmath85 from this ,
we have @xmath86 thus the definition of discriminant yields @xmath87 where the value of the free coefficient of @xmath12 can be obtained from @xmath88 by substituting this value to the equation above , we obtain @xmath89 to compute the product above , we follow an idea from jemal gishe and mourad ismail @xcite by writing @xmath12 as a linear combination of @xmath90 and @xmath91 . in particular this combination can be obtained from and as below : @xmath92 thus @xmath93 where the last equality is obtained from . hence @xmath82 from this we conclude that the discriminants of @xmath12 and @xmath13 have the same factors 2 , 3 , and all the factors of @xmath18 , @xmath57 , @xmath58 .
in this section we will show that the roots of @xmath12 and @xmath13 are pairwise similar except for one pair .
in particular , we will prove the theorem : we first show that with one exception , the roots of @xmath12 and @xmath24 are pairwise close . from the definition ,
the roots of these two polynomial are positive real numbers . to simplify the computations , we consider @xmath96 and the polynomial @xmath97 whose roots ( except 0 ) are the square roots of the positive roots of @xmath13 and @xmath12 respectively . and it will suffice to consider the positive roots of @xmath98 and @xmath96 .
let @xmath99 be a positive root of @xmath96 .
we will show that for a certain small value @xmath100 , the quantities @xmath101 and @xmath102 have different signs and thus @xmath98 admits a root in the small interval @xmath103 .
first , we observe that the equation gives @xmath104 to prove @xmath102 and @xmath109 have different signs , it remains to show that @xmath110 is sufficiently large in magnitude and has the same sign as @xmath101 .
to ensure this fact , we first impose the following two conditions on @xmath111 : where the value of @xmath113 and the existence of @xmath111 with respect to these conditions will be determined later . with these conditions ,
we obtain the following lower bound for @xmath114 : @xmath115 we now need to show that @xmath116 is not too small compared to @xmath117 .
the taylor formula gives @xmath118 where @xmath119 . from
, we have , for @xmath120 , the following trivial bound for @xmath121 : @xmath122 thus the equation , with @xmath123 replaced by @xmath66 , implies that for large @xmath18 @xmath124 now the equation implies that @xmath116 and @xmath117 have the same sign and @xmath127 this inequality combined with and show that @xmath102 and @xmath101 have different signs if @xmath128 thus @xmath98 has a root in the interval @xmath129 next we will show that @xmath111 exists with respect to all the imposed conditions and that its value is small as soon as @xmath99 is not the smallest positive root of @xmath96 .
let @xmath99 be such a root . from , and
, it suffices to show @xmath99 satisfies the identity @xmath130 from this graph , it suffices to show that @xmath99 is not too small and not too close to 1 .
let @xmath132 be the second smallest root of @xmath96 .
this root has an upper bound @xmath133 since the right side is the second smallest positive root of @xmath90 .
the smallest positive root of @xmath90 is @xmath134 similarly , let @xmath142 be the largest root of @xmath96 .
the trivial upper bound @xmath143 , the largest root of @xmath90 , does not guaranty the existence of @xmath113 .
however for some @xmath144 we have @xmath145 the approximation yields @xmath146 thus @xmath147 now the condition is satisfied provided that @xmath148 .
thus we can let @xmath113 be any value in the interval @xmath149 since for large @xmath18 , @xmath113 and @xmath111 are arbitrarily bigger than @xmath150 and @xmath151 respectively , we have that for every root @xmath99 of @xmath96 except the smallest root , there is a root of @xmath98 in the small interval @xmath152 this implies that in each pair , the root @xmath153 of @xmath24 is bigger than the root of @xmath154 of @xmath12 by at most @xmath155 combining this fact with , we conclude that for every root @xmath94 of @xmath13 except the smallest root , there is a root of @xmath12 in the interval @xmath156 for large @xmath18 .
this concludes the proof .
|
the discriminants of certain polynomials related to chebyshev polynomials factor into the product of two polynomials , one of which has coefficients that are much larger than the other s .
remarkably , these polynomials of dissimilar size have `` almost '' the same roots , and their discriminants involve exactly the same prime factors .
|
the early arrival of a neutrino burst from the 1987a supernova @xcite still motivates speculations about a possible superluminal nature of neutrinos , even if it is generally assumed that the delay in the arrival of electromagnetic radiation ( light ) is caused by the time the shock wave from the core collapse needs in order to reach the surface of the exploding star .
if neutrinos are ever so slightly superluminal , then they may emit erenkov radiation in the form of light lepton pairs . in this paper
, we attempt to answer three questions : _ ( i ) .
_ how would the energy threshold for the decay channel @xmath0 ( lepton pair erenkov radiation , lpcr ) have to be calculated if we assume a strictly lorentz - covariant , space - like dispersion relation for the relevant neutrino flavor eigenstate ? _ ( ii ) .
_ how would the decay rate and the energy loss rate have to be calculated under this assumption ?
can the tachyonic dirac equation @xcite and its bispinor solutions @xcite be used in that context ? _ ( iii ) .
_ what implications could be derived for astrophysics , under the assumption that a possible cutoff seen by icecube for neutrinos with energies @xmath3 is confirmed by future experiments ?
theoretical arguments can be useful in restricting the possible degree of superluminality of neutrinos and maximum attainable neutrino velocities @xcite . in refs .
@xcite , a lorentz - noncovariant dispersion relation @xmath4 was used , where @xmath5 is a constant parameter .
this assumption leads to an energy - dependent effective `` mass '' square @xmath6 .
the effective mass @xmath7 then grows linearly with the neutrino energy .
( natural units with @xmath8 are used in this paper , yet we shall include explicit factors of @xmath9 when indicated by the context . ) indeed , at the time , a best fit to the available experimental neutrino mass data including the initial opera claim @xcite suggested the conceivable existence of an `` energy - dependent mass '' of the neutrino , as evidenced in fig . 1 of ref
. @xcite .
the choice of the relation @xmath4 made in ref .
@xcite was consistent with the need to model the initial opera claim @xcite , and is perfectly compatible with the concept of perturbative lorentz breaking terms in the neutrino sector @xcite .
a dirac - type equation leading to the lorentz - noncovariant dispersion relation used by cohen and glashow @xcite can be obtained @xcite from the current operator given in eq .
( 2 ) of ref .
@xcite upon a particular choice of the @xmath10 parameters in the generalized fermionic current operator ( in the notation adopted in ref .
@xcite ) . then , assuming a constant neutrino speed @xmath5 , one can effectively describe the apparent absence of an energy dependence of the deviation of the neutrino speed from the speed of light @xmath11 ( in the range @xmath12 ) , according to the ( falsified ) initial claim made by opera @xcite , while remaining comptabile with the framework of perturbative lorentz breaking @xcite .
however , while there are advantages to assuming a lorentz - noninvariant dispersion relation for superluminal neutrinos ( such as the preservation of the timelike positive quantity @xmath13 ) , there are also a number of disadvantages . for example , if the dispersion relation @xmath4 holds in one particular lorentz frame , then under a lorentz boost , in general , one has @xmath14 in the moving frame @xcite .
in order to illustrate the consequences of lorentz noncovariance , let us consider a boost along the positive @xmath15 axis into a frame which moves with velocity @xmath16 .
a particle moving along the positive @xmath15 axis of the lab frame with four - momentum @xmath17 is mapped onto @xmath18 and thus is `` at rest '' in the moving frame .
however , the general dispersion relation in the moving frame , @xmath19 is much more complicated .
( throughout this paper , we denote the spatial components of the four - vector @xmath20 by @xmath21 and keep @xmath22 explicitly , in order to avoid confusion between @xmath23 with @xmath24 . )
an alternative , commonly accepted dispersion relation for so - called tachyons ( these are space - like , faster - than - light particles described by a lorentz - invariant wave equation ) reads as @xmath25 , i.e. , it is the `` normal '' dispersion relation with the negative sign of the mass square term ( see refs .
@xcite ) . here
, we calculate the threshold energy and the decay rate under the assumption of a lorentz - invariant dispersion relation for the neutrino .
we find that the alternate dispersion relation imposes a tight restrictions on superluminality , and has important phenomenological implications for neutrino masses .
for tachyonic particles , starting from the pioneering work of sudarshan _ et al . _
@xcite , continuing with the works of feinberg @xcite , and including the tachyonic neutrino hypothesis @xcite , the following dispersion relation has been assumed for the tachyonic ( space - like ) solutions , @xmath26 \label{spencer2 } \qquad p^\mu \ , p_\mu = e_\nu^2 - \vec p_\nu^{\,2 } = -m_\nu^2 \,,\end{aligned}\ ] ] where we use the suggestive subscript @xmath27 for `` neutrino '' .
these relations imply that @xmath28 instead of @xmath29 . here
, the tachyonic lorentz factor appears which is @xmath30 .
tachyonic and tardyonic dispersion relations are unified upon assuming an imaginary value for @xmath31 in the tachyonic case ( starting from the tardyonic case , one has @xmath32 , where the latter equation holds for tachyons ) . with the standard definitions of @xmath21 and @xmath33
, one has @xmath34 for both tardyons and tachyons .
conventions for tachyonic neutrino decay . ] in order to obtain the threshold energy for the lpcr decay @xmath35 , we use the following conventions ( see fig . [ fig1 ] ) , inspired by chap .
10 of ref .
@xcite , and define @xmath36 and @xmath37 as the oncoming and outgoing neutrino energies , with @xmath38 being the four - momentum of the @xmath1 .
pair production threshold is reached for @xmath39 and @xmath40 . for collinear geometry , with all momenta pointing along the @xmath15 axis
, we have @xmath41 & \ ; - ( p_{1z } - p_{3z } ) ^2 = 4m_e^2 \,.\end{aligned}\ ] ] furthermore , threshold obviously requires @xmath42 .
( this is possible for tachyonic particles , when @xmath43 . in this limit ,
the tachyonic particles becomes infinitely fast , and loses all of its energy , which implies that it is impossible to detect it @xcite .
the counterintuitive loss of energy for tachyons under acceleration is a consequence of standard tachyonic kinematics @xcite . ) when the relations @xmath42 and @xmath43 are substituted into eq .
, this yields @xmath44 identifying @xmath45 with the threshold momentum , one easily finds @xmath46 the threshold energy is then easily found as @xmath47 because we are using a tachyonic dispersion relation , the threshold energy can be expressed a function of only the mass parameters .
larger tachyonic masses @xmath48 lead to lower threshold energies . in view of the tachyonic dispersion relation @xmath49 , where @xmath50 is the neutrino velocity _ at threshold _ , we may convert the threshold energy into a function of the electron mass and the neutrino threshold velocity . for given @xmath33 , the limit @xmath51 is equivalent to the limit @xmath52 because @xmath53.in this limit
, we have @xmath54 \rightarrow & \ ; e_{\mathrm{th}}\approx \frac{\sqrt{2 } \ , m_e } { ( v_{\mathrm{th}}^2 - 1 ) ^{1/4 } } \,.\end{aligned}\ ] ] substituting the exact dispersion relation into the threshold condition @xmath55 , and solving for @xmath56 , one obtains @xmath57 = & \ ; \left\ { \begin{array}{cc } \dfrac{\sqrt{2}\ , m_e}{\delta_{\mathrm{th}}^{1/4 } } \qquad & \delta_{\mathrm{th}}\ll 1 \\ 2 \ , m_e + \dfrac{m_e}{4 \ , \delta_{\mathrm{th } } } \qquad & \delta_{\mathrm{th}}\gg 1 \end{array } \right .
\,.\end{aligned}\ ] ] the exact expression confirms eq . in the limit @xmath58 , which corresponds to the phenomenologically important limit of high - energy neutrinos .
smaller values of @xmath59 ( approaching zero ) correspond to smaller tachyonic neutrino masses and therefore , to larger threshold energies . for given neutrino speed @xmath50 , neutrinos with energy @xmath56 ( or larger ) ,
under the hypothetical assumption of the tachyonic dispersion relation , have a tachyonic neutrino mass term large enough to make the decay via lpcr kinematically possible .
expressed differently , the tachyonic mass term @xmath49 in this case is large enough to lead to lpcr decay at energy @xmath56 , according to eq . .
( a ) : region of allowed outgoing momenta @xmath60 for the decay of an incoming superluminal neutrino with @xmath29 .
the neutrino is incoming along the positive @xmath15 axis ( @xmath61 ) .
the boundary of allowed @xmath60 vectors constitutes a distorted ellipsoid with a `` sharpened tip '' , obtained as a solution of setting @xmath62 in eq . .
( b ) : region of allowed @xmath60 vectors for an incoming tachyonic neutrino with @xmath63 and @xmath64 , producing an electron - positron pair of mass @xmath65 ( dispersion relation @xmath66 ) .
final wave vectors @xmath67 correspond to evanescent waves and are thus to be excluded @xcite . ]
given the complexities of calculating the decay rate due to lpcr using a tachyonic dispersion relation , it is extremely useful to first discuss the case of a lorentz noncovariant form @xmath29 , using lab frame variables . for collinear incoming and outgoing neutrinos , threshold for pair production is reached at @xmath68 , from which one derives ( setting @xmath69 ) the following threshold values ( in agreement with ref .
@xcite ) , @xmath70 here , @xmath71 is fermi s coupling constant and the @xmath72 and @xmath73 are the standard fundamental positive - energy and negative - energy bispinor solutions of the dirac equation @xcite .
the invariant matrix element is @xmath74 \ , \nonumber\\[0.1133ex ] & \ ; \times \left [ \overline u(p_4 ) \ , \left ( c_v \gamma_\lambda - c_a \ , \gamma_\lambda \ , \gamma^5 \right ) \ , v(p_2 ) \right ] \,.\end{aligned}\ ] ] here , @xmath75 , and @xmath76 [ see eq . ( 5.57 ) on p. 153 of ref .
following @xcite , we now make the additional assumption that the functional form of the projector sum over the spin orientations remains the same as for the ordinary dirac equation even if the underlying dispersion relation is lorentz - noncovariant ( for a general discussion on such models see ref .
@xcite ) . in this case , the sum over final state and the averaging over the initial spins leads to @xmath77 .
this enters the lab - frame expression for the decay rate @xcite , @xmath78 & \ ; \times \left .
( 2 \pi)^4 \ , \delta^{(4 ) } ( p_1 - p_3 - p_2 - p_4 ) \ ,
\left [ \frac12 \ , \sum_{\rm spins } | { \mathcal{m}}|^2 \right ] \right ) \nonumber\\[0.1133ex ] = & \ ; \frac{g_f^2}{12 \ , \pi^4 \ , ( 2 e_1 ) } \ , \int \frac{{\mathrm{d}}^3 p_3}{2 e_3 } \ , \left ( p_1 \cdot p_3 \ , q^2 + 2 \ , ( p_1 \cdot q ) \ , ( p_3 \cdot q ) \right ) \,,\end{aligned}\ ] ] where @xmath79 .
the azimuthal symmetry suggests the use of cylindrical coordinates .
the domain of integration contains , for given @xmath80 , all permissible @xmath81 , where @xmath82 . with @xmath29 ,
the momentum transfer is @xmath83 where we require @xmath84 . solving eq . for @xmath85 ,
one obtains the boundary of the region of permissible @xmath60 vectors .
an example is given in fig .
[ fig2](a ) , in the form of a `` sharpened ellipsoid '' with a `` sharp '' top near @xmath86 , @xmath87 , and a `` rounded '' bottom with @xmath86 , and @xmath88 \ , p_{1z}$ ] . after a somewhat tedious integration over the allowed @xmath60 vectors , one obtains @xmath89 \frac{{\mathrm{d}}e_\nu}{{\mathrm{d}}x } \approx & \ ; -\frac{g_f^2}{96 \ , \pi^4 \ , ( 2 e_\nu ) } \ , \int\limits_{q^2 > 0 } \frac{{\mathrm{d}}^3 p_3}{2 e_3 } \ , ( e_\nu - e_3 ) \nonumber\\[0.1133ex ] & \ ; \times \left [ ( p_1 \cdot p_3 ) \ , q^2 + 2 \ , ( p_1 \cdot q ) ( p_3 \cdot q ) \right ] \nonumber\\[0.1133ex ] = & \ ; -\frac{g_f^2}{86016 \ , \pi^3 } \frac{p_{1z}^6 \delta_\nu^3}{v_\nu } \approx -\frac{25}{448 } \ ,
\frac{g_f^2 \ , e_\nu^6 \ ,
\delta_\nu^3}{192 \ , \pi^3 } \,,\end{aligned}\ ] ] for the energy loss per unit length , confirming the results given in eq . ( 2 ) and ( 3 ) of ref .
@xcite , and in ref .
this confirmation of the results given in ref .
@xcite ( under the assumptions made in the cited paper , namely , the dispersion relation @xmath4 ) , but using a different method , namely , phase - space integration directly in the laboratory frame , encourages us to apply the same method to the calculation of the tachyonic neutrino decay rate , where the use of the laboratory frame is indispensable .
the confirmation also underlines the consistency of the theoretical formalism under a change of the assumptions made in the calculation .
the world line @xmath90 describes the tachyonic neutrino decay into a zero - energy , infinitely fast neutrino .
complete reversal of the time ordering of the decay process takes place in the primed frame ; the observer interprets the process as the decay of an incoming antineutrino along the trajectory @xmath91 . ]
for an incoming tachyon , the particle state ( space - like neutrino ) may transform into an antiparticle state upon lorentz transformation , and its trajectory may reverse the time ordering ( see fig .
[ fig3 ] ) .
thus , the interpretation of a tachyonic neutrino state as a particle or antiparticle may depend on the frame of reference , and we should calculate the process directly in the lab frame . the necessity to transform certain tachyonic particle field operators into antiparticle operators under lorentz boosts has been stressed in ref . @xcite .
incoming and outgoing states are required to be above - threshold positive - energy states in the lab frame ( causality and tachyonic trajectories are discussed in refs .
@xcite and appendix a.2 of ref .
@xcite ) .
we consider the matrix element @xmath92 \nonumber\\[0.1133ex ] & \ ; \times \left [ \overline u(p_4 ) \
, \left ( c_v \gamma^\lambda - c_a \ , \gamma^\lambda \ , \gamma^5 \right ) \ , v(p_2 ) \right ] \,.\end{aligned}\ ] ] here , the @xmath93 , and @xmath94 are dirac spinor solutions of the tachyonic dirac equation @xcite .
the bar denotes the dirac adjoint . in the helicity basis
( see chap .
23 of ref.@xcite and refs .
@xcite ) , these are given by @xmath95 where the @xmath96 are the fundamental helicity spinors ( see p. 87 of ref .
following @xcite , we use the tachyonic sum rule of the fundamental tachyonic bispinor solutions [ see eq .
( 34a ) of ref .
@xcite ] @xmath97 where @xmath98 is the four - momentum , and @xmath99 is a helicity quantum number .
we refer to refs .
@xcite for a thorough discussion ; roughly speaking , the factor @xmath100 in eq . restores the correct sign in the calculation of the time - ordered product of tachyonic field operators ( the propagator ) , for the contribution of all virtual degrees of freedom of the tachyonic field [ see eqs .
( 46)(57 ) and eq .
( 73)(75 ) of ref .
the @xmath101 matrix in eq . is a part of the natural dirac `` adjoint '' for the tachyonic spinor .
namely , the adjoint equation to the tachyonic dirac equation , @xmath102 , reads as @xmath103 \ , ( { \mathrm{i}}\gamma^\mu \overleftarrow{\partial}_\mu - \gamma^5 m ) \psi(x ) = 0 $ ] . as explained in eqs .
( 73)(75 ) of ref .
@xcite , right - handed particle and left - handed antiparticle states ( those with the `` wrong '' helicity ) are excluded from the physical spectrum of the tachyonic field by a gupta bleuler condition ; these can not contribute to the oncoming and outgoing neutrino states in fig .
[ fig1 ] [ while they do contribute to the virtual states , i.e. , the propagator , see eqs .
( 46)(57 ) of ref .
@xcite ] . both the incoming as well as the outgoing neutrinos in fig . [ fig1 ]
are real rather then virtual neutrinos .
hence , in order to calculate the lpcr decay rate , we use the modified sum over tachyonic spinors @xmath104 where @xmath105 is a time - like unit vector , @xmath106 is the unit vector in the @xmath21 direction , and upon promotion to a four - vector , we have @xmath107 , so that @xmath108 becomes a left - handed helicity projector .
we thus calculate with an incoming , positive - energy , left - helicity tachyonic neutrino .
one obtains the modified sum over spins @xmath109 in the matrix element , @xmath110 & \ ; \cdot \gamma^5 \ , \gamma_\lambda \ , ( 1- \gamma^5 ) \frac12 \ , \left ( 1 + \gamma^5 \ , \cancel{\tau } \ , \cancel{\hat{p}}_1 \right ) \nonumber\\[0.1133ex ] & \ ; \left .
\cdot ( \cancel{p}_1 - \gamma^5 \ , m_{\nu } ) \ , \gamma^5 \ , \gamma_\nu \ , ( 1- \gamma^5 ) \right ] \ , k^{\lambda\rho } \,.\end{aligned}\ ] ] here , @xmath111 $ ] is the familiar trace from the outgoing fermion pair . the decay rate is given by eq .
, under the replacement @xmath112 .
the integrals over the momenta of the outgoing fermion pair ( @xmath113 and @xmath114 ) are done using ( @xmath115 ) @xmath116 & = \frac{\pi}{24 } \sqrt{1 - \frac{4 m_e^2}{q^2 } } \left [ g_{\lambda \rho } \left ( q^2 - 4 m_e^2 \right ) + 2 q_\lambda q_\rho \left ( 1 + \frac{2 m_e^2}{q^2 } \right ) \right ] .\end{aligned}\ ] ] it remains to analyze the domain of allowed @xmath60 vectors [ see the `` cupola structure '' in fig . [
fig2](b ) ] , which is defined by the requirement @xmath117 , for @xmath118 .
the dispersion relation @xmath119 implies that @xmath120 here , @xmath121 is the polar angle in spherical coordinates , @xmath122 and @xmath123 , by solving eq . for @xmath124 , setting @xmath125 . after a somewhat tedious integration over the allowed @xmath60 vectors ( no masses can be neglected ) ,
one obtains [ gammadedx1 ] @xmath126 \dfrac{g_f^2 \ , m_\nu^4}{288 \pi^3 } \ , e_\nu & \qquad e_\nu \gg e_{{\mathrm{th } } } \end{array } \right .
\end{aligned}\ ] ] for the decay rate , and @xmath127 \dfrac{g_f^2 \ , m_\nu^4}{144 \pi^3 } \ , e^2_1 & \qquad e_\nu \gg e_{{\mathrm{th } } } \end{array } \right .
\end{aligned}\ ] ] for the energy loss rate . in the high - energy limit ,
one may ( somewhat trivially ) rewrite the expressions as follows ( @xmath128 ) , @xmath129 these results confirm that it is possible to use the tachyonic bispinor formalism @xcite for the calculation of decay rates of tachyonic particles .
our threshold relation eq . is based on a lorentz - covariant dispersion relation . only neutrinos with @xmath130 survive the possibility of generalized leptonic erenkov radiation over a sufficiently long path length .
the hypothetical observation of an absence of neutrinos above some energy @xmath56 could thus be interpreted as a constraint on the neutrino mass .
let us assume a neutrino mass of @xmath131 , where @xmath132 is generally assumed to be of order unity or less .
then , threshold is reached for @xmath133 , @xmath134 , and @xmath135 .
the icecube experiment @xcite has observed 37 neutrinos having energies @xmath136 during 3 years of data taking .
three of these events had energies @xmath137 , and one ( often referred to as `` big bird '' ) had @xmath138 .
according to the icecube collaboration @xcite , the spectrum of the 37 neutrinos is well fitted by a slope @xmath139 , which includes astrophysical as well as background atmospheric neutrinos , the latter being exclusively below @xmath140 .
however , their best fit to the spectrum predicts 3.1 additional events for @xmath141 pev , and yet none were seen .
preliminary data for the fourth year includes 17 additional events , with none seen for @xmath142 @xcite .
these facts suggest to the icecube authors @xcite the possibility that there may be a cutoff for the spectrum for neutrinos above @xmath143 . the hypothesis is given further support by models which show that the glashow resonance @xcite ( resonant @xmath144 ) should add between zero and three times the number of events that appear in the interval @xmath145 as part of a broad peak centered around @xmath146 @xcite . while evidence for the cutoff is disputed and alternative explanations have been proposed @xcite , the significance of such a cutoff has been analyzed in the light of superluminal neutrinos @xcite .
let us add a few clarifying remarks here .
first , we note that the plots in the paper @xcite refer to the neutrino flux as a function of neutrino energy ; the events were apparently sufficiently well reconstructed so that no excess neutrino energy in addition to the energy deposited inside the detector is expected .
our fig .
[ fig4 ] is based on fig . 4 of ref .
meanwhile , members of the icecube collaboration have presented preliminary evidence for a through - going muon of energy @xmath147 which could be interpreted as a decay product of a neutrino of even higher energy @xcite .
if the through - going muon could indeed be assigned to an ultra - high - energy neutrino of non - atmospheric origin , then it would push the conceivable cutoff seen by icecube to even higher energies , further constraining the tachyonic mass term of the relevant neutrino flavor .
so far , the authors of ref .
@xcite ( see the right column on page 4 of ref .
@xcite ) observe that `` this [ the lack of high - energy events ] may indicate , along with the slight excess in lower energy bins , either a softer spectrum or a cutoff at high energies . ''
three years of data from the icecube experiment showing the flux @xmath148 multiplied by @xmath149 plotted against the neutrino energy @xmath33 .
the solid and dashed curves show what would be expected for an @xmath150 power law for the flux , with an @xmath151 threshold , and two arbitrarily assumed values for the source distance , @xmath152 ( dashed curve ) and @xmath153 ( solid curve ) .
the drop to zero above @xmath56 only occurs for those neutrino flavors having a tachyonic mass consistent with a @xmath154 threshold . ] assuming @xmath155 we would find using eq . that @xmath156 and furthermore , that @xmath157 ( i.e. , @xmath158 ) for one or more of the three neutrino flavors ( conceivably , the one with the smallest absolute value of @xmath159 ) .
a shifted cutoff @xcite of @xmath160 , would be consistent with a tachyonic neutrino mass of @xmath161
. one might object that it is not possible to have one ( or more ) tachyonic flavor masses ( @xmath162 ) and satisfy both neutrino oscillation data and the recent findings from cosmology for the sum of the flavor masses , i.e. , @xmath163 @xcite .
however , such consistency can be achieved using 3 active - sterile @xmath164 ( tardyon - tachyon ) neutrino pairs @xcite .
the curves in fig .
[ fig4 ] were generated using an assumed pure @xmath150 power law for the flux @xmath165 beyond the assumed threshold , @xmath56 .
we then use our @xmath166 formula [ eq . ] for @xmath167 to find the modified @xmath168 spectrum .
good agreement is found with the icecube data at a threshold @xmath169 , although much more statistics will be needed to determine if the cutoff is real .
three main conclusions of the current investigation can be drawn . _
( i ) . _ as described in sec .
[ sec2 ] , the assumption of a lorentz covariant , tachyonic dispersion relation leads to tight bounds on conceivable tachyonic neutrino mass terms
. the tachyonic decay rate due to lpcr is most conveniently calculated in the laboratory frame , because the space - like kinematics involved in the process , which leads to a non - unique time ordering of the trajectories , as discussed in sec [ sec4 ] . _
_ we may apply the formalism of the tachyonic bispinor solutions of the tachyonic dirac equation @xcite recently developed in ref .
@xcite to the calculation of the tachyonic neutrino decay , as outlined in sec .
[ sec4 ] . _
( iii ) . _
a comparison of recent icecube data with the results for the calculated tachyonic decay rates reveals that a tachyonic neutrino could possibly explain a possible sharp cutoff in icecube data , but only if the neutrino flavor involved has a very specific tachyonic mass . in a more general context
, the calculation of tachyonic thresholds and decay rates based on lorentz - covariant dispersion relations could be of phenomenological significance for string theories , some of which predict the existence of tachyons @xcite .
the same is true for the precise calculation of the tail of the beta decay spectrum , which is influenced by a conceivably tachyonic neutrino mass term @xcite .
the authors declare that there is no conflict of interests regarding the publication of this paper .
this research has been supported by the national science foundation ( grant phy1403973 ) .
51ifxundefined [ 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 * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop _ _ ( , , ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop _ _ , ed .
( , , ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop ( ) @noop ( ) @noop * * , ( ) @noop * * , ( ) @noop * * , ( ) @noop _ _ ( , , ) @noop _ _ ( , , ) @noop * * , ( )
|
* abstract : * current experiments do not exclude the possibility that one or more neutrinos are very slightly superluminal or that they have a very small tachyonic mass .
important bounds on the size of a hypothetical tachyonic neutrino mass term are set by lepton pair erenkov radiation ( lpcr ) , i.e. , by the decay channel @xmath0 which proceeds via a virtual @xmath1 boson . here
, we use a lorentz - invariant dispersion relation which leads to very tight constraints on the tachyonic mass of neutrinos ; we also calculate decay and energy loss rates . a possible cutoff seen in the icecube neutrino spectrum for @xmath2 , due to the potential onset of lpcr
, is discussed .
|
the series - parallel networks are one of the simplest models of real networks .
then even if it is at last one century old ( mac mahon 1892 ) , it still occurs frequently in both theoretical or applied studies .
some questions like the reliability ( feo and johnson 1990 ) , the survival probability after failures , the routing problem and the queuing delay ( tamir 1993 ) , etc .
are now an important branch of probability theory applied to communication , computer or power networks . as usual in this kind of model , the elementary components ( resistances or switches )
are chosen all identical .
then the interest is in the behavior of the structure made of series or parallel connections .
some authors ( riordan 1958 , lomnicki 1972 ) investigated this problem with a combinatorial point of view , and , in particular , give recurrences to compute the number of different structures built with @xmath0 elements .
then , they consider the problem as formally `` solved '' . in this article , we study this model by using analogies with statistical physics models like self - avoiding walks ( des cloizeaux and jannink 1987 ) , lattice animals and percolation ( derrida and de seze 1982 ) , folding and meanders model ( di francesco et al .
1996 ) , etc .
in all these models , the number of configurations for @xmath0 elements ( steps , links , folds ) grows asymptotically like @xmath4 where @xmath5 depends on the details of the model and @xmath3 is an `` universal '' exponent , common to a broad set of models ( `` universality '' class ) . in section 2
, we define the problem . in section 3 , we explain how to enumerate the series - parallel networks . finally ,
in section 4 , we give by numerical extrapolation a value for @xmath5 and we show that @xmath6 .
the two - terminal series - parallel networks are electric network considered geometrically .
they are built with _ elements _
@xmath7 which are conductor ( @xmath8 ) or insulator ( @xmath9 ) .
when two structures ( with conductivities @xmath10 and @xmath11 ) are in parallel , their ( global ) conductivity is @xmath12 where @xmath13 is the boolean operator _
or_. when they are in series , it is @xmath14 where the product is the boolean operator _
a network with @xmath0 elements , which connects two - terminal nodes , has a boolean representation @xmath15 .
we consider that two networks are _ equivalent _ if their boolean representations differ only by a permutation of variables @xmath16 .
for example , @xmath17 , @xmath18 and @xmath19 are equivalent .
we note that this definition is less restrictive than topological definition . in other words , equivalence consists of interchange inside a group of structures in series , or in parallel .
the networks for @xmath20 are given in table [ ta ] .
.two - terminal series - parallel networks for @xmath20 . on each row
, the two networks are dual .
the @xmath21 s ( number of networks ) are 1 , 2 , 4 , 10 , the @xmath22 s ( number of essentially series or , by duality , essentially parallel networks ) are 1 , 1 , 2 , 5 , [ cols="^,^,<,^,<,^ " , ]
a brief investigation of table [ li ] shows that @xmath23 .
then , we can define a non - vanishing entropy per element @xmath24 .
more generally , it is expected that the exponential dominant behavior is corrected by sub - leading power - law for large @xmath0 @xmath25 where @xmath26 is a constant .
in fact , in eq .
( [ lomni ] ) , for large @xmath0 , @xmath27 , then @xmath28 .
this last relation is identical apart from a shift @xmath29 to the definition @xmath30 of the catalan numbers @xmath31 $ ] . using the stirling formula
, we see that @xmath32 have the asymptotic behavior of eq .
( [ ra ] ) with @xmath33 and @xmath34 .
the fact that the @xmath21 and @xmath32 have the same asymptotic relation do not imply that they have the same @xmath5 . as @xmath35 ,
in the sum , the extremal terms are dominant : @xmath36
. then the value of @xmath5 strongly depends of the value of the first @xmath32 ( @xmath0 small ) and the exponential increase of @xmath21 and @xmath32 are not described by the same @xmath5 . on the other hand , we expected the same @xmath3 as we see below .
this behavior is common in statistical physics .
the enumeration of random walks , self - avoiding walk , animals , etc . of @xmath0 step ( or links ) drawn on a regular lattice always exhibits a exponential increase @xmath37 where @xmath5 depend on the connectivity of the lattice , and a sub - leading @xmath38 , where @xmath3 is `` universal '' , i.e. is dependent on the physical nature of the problem ( dimension , phases , ) but independent on the details of the lattice .
let us take a big two - terminal series - parallel network .
at each scale , it is `` decorated '' by small parts @xmath22 ( @xmath39 ) .
then , the contribution of these small @xmath22 is a large fraction of the total entropy @xmath40 ; they are similar to the connectivity of the lattice and govern the dominant behavior @xmath37 . nevertheless , the sub - leading @xmath38 depends on more general properties and @xmath3 can be see as an `` universal '' exponent .
it described the singularity of the generating function .
we suppose that the @xmath21 have the following large @xmath0 expansion @xmath41 and we want estimate the values of @xmath5 and @xmath3 by numerical extrapolation .
it is more convenient to study @xmath42 and @xmath43 @xmath44 by plotting @xmath42 ( resp .
@xmath43 ) _ versus _
@xmath45 , the hypothesis ( [ expa ] ) is consistent if the data draw a straight line ; its intersection with the y - axis gives @xmath40 ( resp .
@xmath3 ) .
a more systematic scheme of extrapolation is the richardson polynomial extrapolation process ( brezinski and redivo zaglia 1991 ) . supposing eq .
( [ expa ] ) , @xmath46 where @xmath47 is the difference operator @xmath48 .
this method gives the limits ( @xmath5 and @xmath3 ) and show whether the hypothesis ( [ expa ] ) is consistent or not .
when @xmath49 is large , the acceleration of convergence is faster ; unfortunately , the numerical instabilities are amplified and after a given @xmath0 , extrapolated values ( eq . [ dvn ] ) become chaotic . in this work ,
better extrapolations , obtained with @xmath49 between 3 and 5 , are @xmath50 we will now prove that @xmath51 , by examining the behavior of the generating function , which is numerically plotted on fig . 1 .
clearly , the radius of convergence of the series @xmath52 ( i.e. the location of its singularity ) is @xmath53 . at @xmath54 , @xmath55 converges if @xmath56 , and its first derivative @xmath57 diverges if @xmath58 .
then , fig .
1 indicates that @xmath59 .
more generally , the behavior of @xmath60 around @xmath61 is given by @xmath3 .
let us consider the singular part @xmath62 with @xmath63 . for small @xmath64 , @xmath65 / i^\alpha$ ] . for @xmath66 ,
the sum is dominated by small @xmath67 and @xmath68 , i.e. the first derivative @xmath69 is finite . for @xmath70 ,
the sum is dominated by @xmath71 and @xmath72 .
the case @xmath51 corresponds to @xmath73 . rewriting eq .
( [ jat ] ) @xmath74 = { 1\over 2 } [ a(t ) + t ] + r(t ) \ \ \ { \rm with } \ \ \
r(t ) = \sum_{k \ge 2 } { 1 \over 2k } [ a(t^k ) + t^k],\ ] ] for @xmath75 , @xmath76 is a regular and increasing function . by defining @xmath77 , @xmath78 .
the function @xmath79 has a maximum at @xmath80 with @xmath81 and a parabolic shape ( i.e. a finite and negative second derivative ) . as @xmath82
, this gives a singularity for @xmath83 with a square root shape @xmath84 .
this proves that @xmath51 .
unfortunately , with this argument , we can not compute the value of @xmath53 , because of the complexity of @xmath76 .
then @xmath5 is sensitive to the details of the regular part of the recurrence : @xmath5 is not `` universal '' . but @xmath3 depends only of the kind of singularity : it is an `` universal '' exponent .
as we have seen above , it is not surprising that it is equal to the @xmath3 of the catalan numbers .
we have seen that some concepts of statistical physics can be transposed to a problem of graph theory .
the partition function of two - terminal series - parallel networks have an asymptotic behavior similar to those of self - avoiding walks or lattice animals .
its exponential growth depends on details on the model . on the other hand , a power - law correction , with an `` universal '' exponent
can be computed exactly and have a fractional value @xmath85 . for the self - avoiding walks ,
this exponent is sensitive to the dimension and the kind of boundaries conditions and others independent exponents are related to geometrical properties . in our problem
, it will be interesting to have a similar interpretation .
for example , the number of terminal nodes ( two in this article ) could play the role of boundary conditions . by studying quantities like the average depth of elements ( i.e. the number of decomposition needed to isolate this element ) ,
it is probable that new exponents would appear .
des cloizeaux j. et janning g. , 1987 , _ les polymres en solution : leur modlisation et leur structure _ ( ed . phys .
, les ulis ) ; 1990 , _ polymers in solution , their modeling and structure _ ( clarendon press ) .
|
this paper discusses the enumeration of two - terminal series - parallel networks , i.e. the number of electrical networks built with @xmath0 identical elements connected in series or parallel with two - terminal nodes .
they frequently occur in applied probability theory as a model for real networks .
the number of networks grows asymptotically like @xmath1 , as for some models of statistical physics like self - avoiding walks , lattice animals , meanders , etc . by using a exact recurrence relation ,
the entropy is numerically estimated at @xmath2 , and we show that the sub - leading `` universal '' exponent @xmath3 is 3/2 . #
1#2 ( c # 1 + # 2 )
|
in the last years , the literature on networks has been characterized by exponential growth .
empirical and theoretical contributions in very diverse fields such as physics , sociology , economics , etc . have increasingly highlighted the pervasiveness of networked structures .
examples range from www , the internet , airline connections , scientific collaborations and citations , trade and labor market contacts , friendship and other social relationships , business relations and r&s partnerships , all the way through cellular , ecological and neural networks @xcite .
the empirical research has thoroughly studied the ( often complex ) topological properties of such networks , whereas a large number of theoretical models has been proposed in order to investigate how networks evolve through time @xcite .
structural properties of networks have been shown to heavily impact on the dynamics of the socio - economic systems that they embed @xcite . as a result ,
their understanding has become crucial also as far as policy implications are concerned @xcite .
the simplest mathematical description of a network is in terms of a graph , that is a list of nodes @xmath0 and a set of arrows ( links ) , possibly connecting any two nodes @xcite .
alternatively , one can characterize a network through a @xmath1 real - valued matrix @xmath2 , where any out - of - diagonal entry @xmath3 is non - zero if and only if an arrow from node @xmath4 to @xmath5 exists in the network .
entries on the main diagonal are typically assumed to be all different from zero ( if self - interactions are allowed ) or all equal to zero ( if they are not ) .
networks are distinguished in binary ( dichotomous ) or weighted . in binary networks
all links carry the same intensity .
this means that in binary networks a link is either present or not , i.e. @xmath6 . in this case , @xmath7 is called an `` adjacency '' matrix .
weighted networks allow one instead to associate a weight ( i.e. a positive real number ) to each link , typically proportional to its interaction strength or the flux intensity it carries @xcite .
any non - zero entry @xmath3 thus measures the weight of the link originating from @xmath4 and ending up in @xmath5 , and the resulting matrix @xmath7 is called the `` weights '' matrix.$ ] , all @xmath8 ( the more general case @xmath9 can be reduced to the former simply by dividing all weights by their maximum level in @xmath7 ) . ]
both binary and weighted networks can be undirected or directed .
formally , a network is undirected if all links are bilateral , i.e. @xmath10 for all @xmath11 .
this means that in undirected networks all pairs of connected nodes mutually affect each other .
one can thus replace arrows with non - directed edges ( or arcs ) connecting any two nodes and forget about the implicit directions .
this greatly simplifies the analysis , as the tools for studying undirected networks are much better developed and understood .
directed networks are instead not symmetric , as there exists at least a pair of connected nodes wherein one directed link is not reciprocated , i.e. @xmath12 , but @xmath13 .
studying the topological properties of directed networks , especially in the weighted case , can become more difficult , as one has to distinguish inward from outward links in computing synthetic indices such as node and average nearest - neighbor degree and strength , clustering coefficient , etc .. therefore , it is not surprising that the properties of such indices are much less explored in the literature . from a theoretic perspective , it is easy to distinguish undirected from directed networks : the network is undirected if and only if the matrix @xmath7 is symmetric .
when it comes to the empirics , however , researchers often face the following problem . if the empirical network concerns an intrinsically mutual social or economic relationship ( e.g. friendship , marriage , business partnerships , etc . ) then @xmath7 , as estimated by the data collected , is straightforwardly symmetric and only tools designed for undirected network analysis are to be employed .
more generally , however , one deals with notionally non - mutual relationships , possibly entailing directed networks . in that case
, data usually allow to build a matrix @xmath7 that , especially in the weighted case , is hardly found to be symmetric .
strictly speaking , one should treat all such networks as directed .
this often implies a more complicated and convoluted analysis and , frequently , less clear - cut results .
the alternative , typically employed by practitioners in the field , is to compute the ratio of the number of directed ( bilateral ) links actually present in the networks to the maximum number of possible directed links ( i.e. @xmath14 ) .
if this ratio is `` reasonably '' large , then one can symmetrize the network ( i.e. making it undirected , see @xcite ) and apply the relevant undirected network toolbox . however , as shown in ref .
@xcite , this procedure has several drawbacks .
in particular , it is heavily dependent on the density of the network under analysis ( i.e. , the ratio between the total number of existing links to @xmath14 ) .
moreover , and most important here , if the network is weighted , the ratio of bilateral links does not take into account the effect of link weights .
indeed , a bilateral link exists between @xmath4 and @xmath5 if and only if @xmath10 , i.e. irrespective of the actual size of the two weights . of course , as far as symmetry of @xmath7 is concerned , the sub - case where @xmath15 will be very different from the sub - case where @xmath16 . in this paper , we present a simple procedure that tries to overcome this problem .
more specifically , we develop a simple index that can help in deciding when the empirically - observed @xmath7 is sufficiently symmetric to justify an undirected network analysis .
our index has two main properties .
first , it can be applied with minor modifications to both binary and weighted networks .
second , the standardized version of the index distributes as a standard normal ( over all possible matrices @xmath7 ) .
therefore , after having set a threshold @xmath17 , one might conclude that the network is to be treated as if it is undirected if the index computed on @xmath7 is lower than @xmath17 .
of course , the procedure that we propose in the paper is by no means a statistical test for the null hypothesis that @xmath7 involves some kind of symmetry .
indeed , one has almost always to rely on a single observation for @xmath7 ( more on that in section [ section : conclusions ] ) .
nevertheless , we believe that the index studied here could possibly provide a simple way to ground the `` directed vs. undirected '' decision on more solid bases .
the paper is organized as follows . in section [ section : definition ] we define the index and we derive its basic properties .
section [ section : statistical_properties ] discusses its statistical properties , while in section [ section : examples ] we apply the procedure to the empirical networks extensively studied in @xcite .
finally , section [ section : conclusions ] concludes .
consider a directed , weighted , graph @xmath18 , where @xmath19 is the number of nodes and @xmath20 is the @xmath1 ( real - valued ) matrix of link `` weights '' @xcite . without loss of generality
, we can assume @xmath21 , \forall i\neq j$ ] and @xmath22 .
, i.e. no self - interactions ) or all equal to one ( @xmath23 , i.e. self - interactions are allowed ) . ] in line with social network analysis , we interpret the generic out - of - diagonal entry @xmath24 , as the weight associated to the directed link originating from node @xmath4 and ending up in node @xmath5 ( i.e. , the strength of the directed link @xmath25 in the graph ) .
a directed edge from @xmath4 to @xmath5 is present if and only if @xmath26 .
the idea underlying the construction of the index is very simple . if the graph @xmath27 is undirected , then @xmath28 , where @xmath29 is the transpose of @xmath30 . denoting by @xmath31 any norm defined on a square - matrix , the extent to which directionality of links counts in the graph @xmath27 can therefore be measured by some increasing function of @xmath32 , suitably rescaled by some increasing function of @xmath33 ( and possibly of @xmath34 ) . to build the index we first define , again without loss of generality :
@xmath35 where @xmath36 is the @xmath1 identity matrix .
accordingly , we define the graph @xmath37 .
notice that @xmath38 for all @xmath11 , while now @xmath39 for all @xmath4 .
consider then the square of the frobenius ( or hilbert - schmidt ) norm : @xmath40 where all sums ( also in what follows ) span from 1 to @xmath19 .
notice that @xmath41 is invariant with respect to the transpose operator , i.e. @xmath42 .
we thus propose the following index : @xmath43 ^ 2 .
\label{eq : index_s_tilde}\ ] ] by exploiting the symmetry of @xmath44 , one easily gets : @xmath45 alternatively , by expanding the squared term at the numerator , we obtain : @xmath46 the index @xmath47 has a few interesting properties , which we summarize in the following : [ lemma : properties ] for all real - valued @xmath1 matrices @xmath48 s.t .
@xmath49 , i\neq j$ ] and @xmath50 , then : 1 . @xmath51 .
2 . @xmath52 , i.e. if and only if the graph is undirected .
3 . @xmath53 see appendix [ sec : appendix_1 ] . furthermore , when @xmath54 is binary ( i.e. , @xmath55 for all @xmath8 ) , the index in eq .
[ eq : index_s_tilde ] turns out to be closely connected to the density of the graph ( i.e. , the ratio between the total number of directed links to the maximum possible number of directed links ) and the ratio of the number of bilateral directed links in @xmath54 ( i.e. links from @xmath4 to @xmath5 s.t .
@xmath56 ) to the maximum possible number of directed links .
more precisely : [ lemma : properties_binary ] when @xmath54 is binary , i.e. @xmath55 , all @xmath8 , then : @xmath57 where @xmath58 is the density of @xmath54 and @xmath59 is the ratio between the number of bilateral directed links to the maximum number of directed links
. see appendix [ sec : appendix_2 ] .
notice that , in the case of undirected graphs , @xmath60 and @xmath61 . on the contrary , when there are no bilateral links , @xmath62 .
hence , @xmath63^{-1}$ ] , which is maximized when @xmath64 , i.e. @xmath65 , as shown in lemma [ lemma : properties ] . obviously , the larger @xmath59 , the more the graph @xmath54 is undirected .
as mentioned in section [ section : introduction ] , @xmath59 can be employed to check for the extent to which directionality counts in @xmath54 . however , such index is not very useful in weighted graphs , as it does not take into account the size effect ( i.e. the size of weights as measured by @xmath66 $ ] ) . in the case of binary graphs ,
the index in almost coincides with the one proposed in @xcite .
they suggest to employ the correlation coefficient between @xmath67 entries in the adjacency matrix ( excluding self loops ) .
it can be shown that this alternative index unlike the one in
increases with the number of reciprocated links , only if both the total number of links that are in place and the density of the network remain constant .
see also @xcite for applications .
since @xmath68 $ ] , in what follows we shall employ its rescaled version : @xmath69 which ranges in the unit interval and thus has a more straightforward interpretation .
in this section we study the distribution of the index @xmath70 as defined in eqs .
[ eq : index_s_tilde ] and [ eq : index_s ] . indeed , despite the range of @xmath70 does not depend on @xmath19 , we expect its distribution to be affected by : ( i ) the size of the matrix ( @xmath19 ) ; ( ii ) whether the underlying graph @xmath54 is binary ( @xmath55 ) or weighted ( @xmath49 $ ] ) . to do so , for each @xmath71 we generate @xmath72 random matrices @xmath73 obeying the restriction that @xmath39 , all @xmath4 . in the binary case , out - of - diagonal entries @xmath74 are drawn from i.i.d .
bernoulli random variables with @xmath75 . in the weighted case ,
entries @xmath76 are i.i.d .
random variables uniformly - distributed over @xmath77 $ ] .
we then estimate the distributions of @xmath70 in both the binary and the weighted cases , and we study their behavior as the size of the graph increases .
let us denote by @xmath78 ( respectively , @xmath79 ) the sample mean of the index @xmath70 in the binary ( respectively , weighted ) case , and by @xmath80 ( respectively , @xmath81 ) the sample standard deviation of the index @xmath70 in the binary ( respectively , weighted ) case .
simulation results are summarized in the following points . 1 . in both the binary and the weighted case
, the index @xmath70 approximately distributes as a beta random variable for all @xmath19 . as @xmath19 increases , @xmath78 decreases towards 0.50 whereas @xmath79 increases towards 0.25 . both standard deviations decrease towards 0 .
more precisely , the following approximate relations hold ( see figures [ fig : scaling_bern ] and [ fig : scaling_unif ] ) : + @xmath82 2 . given the approximate relations in eqs .
[ eq : mean_binary]-[eq : sd_weighted ] , let us standardize the index @xmath70 as follows : + @xmath83 + simulations indicate that the standardized versions of the index , i.e. @xmath84 and @xmath85 , are both well approximated by a @xmath86 , even for small @xmath19s ( @xmath87 ) .
indeed , as figures [ fig : rescaled_bern ] and [ fig : rescaled_unif ] show , the mean of the distributions of @xmath84 and @xmath85 vs. @xmath19 converges towards zero , while the standard deviation approaches one ( we actually plot standard deviation minus one to have a plot in the same scale ) . also the third ( skewness ) and the fourth moment ( excess kurtosis ) stay close to zero .
we also plot the estimated distribution of @xmath84 and @xmath85 vs. @xmath19 , see figures [ fig : distr_bern ] and [ fig : distr_unif ] .
it can be seen that all estimated densities collapse towards a @xmath86 .
notice that the y - axis is in log scale : this allows one to appreciate how close to a @xmath86 are the distributions for all @xmath19 on the tails .
+ notice finally that as @xmath19 increases , the distribution maintains a constant second moment but the range increases linearly with @xmath19 , see [ fig : bounds_bern ] and [ fig : bounds_unif ]
. the lower bound ( lb ) and the upper bound ( ub ) indeed read approximately : + @xmath88 + where @xmath89 stands for binary ( b ) and weighted ( w ) .
since the standardized index is well approximated by a @xmath86 for all @xmath19 , this means that extreme values become more and more unlikely .
this is intuitive , because as @xmath19 grows the number of matrices with highest / lowest values of the index are very rare .
the index developed above can be easily employed to assess the extent to which link directionality matters in real - world networks .
let us suppose to have estimated a @xmath90 matrix @xmath91 describing a binary ( b ) or a weighted ( w ) graph .
we then compute the index : @xmath92.\end{aligned}\ ] ] where @xmath89 and @xmath93 are as in eqs .
[ eq : mean_binary]-[eq : sd_weighted ] .
since we know that @xmath94 is approximately @xmath86 , we can fix a lower threshold in order to decide whether the network is sufficiently ( un)directed . for instance , we could set the lower threshold equal to 0 ( i.e. equal to the mean ) , and decide that if @xmath95 ( above the mean ) we shall treat the network as directed ( and undirected otherwise ) .
more generally , one might set a threshold equal to @xmath96 and conclude that the graph is undirected if @xmath97 . on the contrary
, one should expect the directional nature of the graph to be sufficiently strong , so that a digraph analysis is called for . to test the index against real - world cases , we have taken the thirteen social and economic networks analyzed in @xcite , see table [ tab : example ] .
all networks are binary and directed , apart from freeman s ones ( which are _ weighted _ and directed ) and padgett s ones ( which are binary and _ undirected _ ) .
table [ tab : example ] reports both the index @xmath70 and its standardized versions @xmath98 , for all cited examples .
.the index @xmath70 and its standardized version @xmath99 for social networks studied in @xcite , cf .
chapter 2.5 . [ cols=">,<,>,>,>",options="header " , ] suppose to fix the lower threshold equal to zero .
padgett s networks , being undirected , display a very low value ( in fact , the non standardized index is equal to zero as expected ) .
the table also suggests to treat all the binary trade networks as undirected .
the same advice applies for freeman s networks , which are instead weighted .
the only networks which have an almost clear directed nature ( according to our threshold ) are krackhardt s ones . in that case
our index indicates that a directed graph analysis would be more appropriate .
in this paper we have proposed a new procedure that might help to decide whether an empirically - observed adjacency or weights @xmath1 matrix @xmath7 , describing the graph underlying a social or economic network , is sufficiently symmetric to justify an undirected network analysis .
the index that we have developed has two main properties .
first , it can be applied to both binary or weighted graphs .
second , once suitably standardized , it distributes as a standard normal over all possible adjacency / weights matrices .
therefore , given a threshold decided by the researcher , any empirically observed adjacency / weights matrix displaying a value of the index lower ( respectively , higher ) than the threshold is to be treated as if it characterizes an undirected ( respectively , directed ) network .
it must be noticed that setting the threshold always relies on a personal choice , as also happens in statistical hypothesis tests with the choice of the significance level @xmath100 . despite this unavoidable degree of freedom ,
the procedure proposed above still allows for a sufficient comparability among results coming from different studies ( i.e. where researchers set different threshold ) if both the value of the index @xmath70 and the size of the network are documented in the analysis . in that case
, one can easily compute the probability of finding a matrix with a lower / higher degree of symmetry , simply by using the definition of bounds ( see eq .
[ eq : bounds ] ) and probability tables for the standard normal .
a final remark is in order . as mentioned
, our procedure does not configure itself as a statistical test . since the researcher often relies on a single observation of the network under study ( or a sequence of serially - correlated network snapshots through time ) , statistical hypothesis testing will be only very rarely feasible .
nevertheless , in the case where a sample of @xmath101 i.i.d .
observations of @xmath7 is available , one might consider to use the the sample average of the index ( multiplied by @xmath102 ) and employ the cental limit theorem to test the hypothesis that the observations come from a undirected ( random ) graph .
this work has enormously benefited from insightful suggestions by michel petitjean .
thanks also to javier reyes and stefano schiavo for their useful comments .
17 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , * * , ( ) . ,
* * , ( ) . ,
_ _ ( , ) . , ,
, eds . , _ _
( , ) . ,
_ _ ( , ) . ,
_ _ ( , ) . , _ _ ( , ) . , _ _ ( , ) . ,
_ _ ( , ) . , , , ,
* * , ( ) . , , , , ( ) .
, , , , * * , ( ) . , , , , , ( ) . , , , _ _ ( , ) . ,
* * , ( ) . , * * , ( ) . , * * , ( ) .
points ( 1 ) and ( 2 ) simply follow from the definition in eq .
[ eq : index_s_tilde ] . as to ( 3 ) , let us suppose that there exists a matrix @xmath73 satisfying the above restrictions and such that @xmath103 .
then , using eq .
[ eq : max ] : the best case for such an inequality to be satisfied is when the the left hand side is minimized .
this is achieved when there are @xmath105 entries equal to one and @xmath105 entries equal to zero in such a way that @xmath106 for all @xmath11 ( e.g. , when the upper diagonal matrix is made of all ones and the lower diagonal matrix is made of all zeroes ) . in that case
the left hand side is exactly equal to @xmath107 , leading to the absurd conclusion that @xmath108 .
|
this paper proposes a simple procedure to decide whether the empirically - observed adjacency or weights matrix , which characterizes the graph underlying a socio - economic network , is sufficiently symmetric ( respectively , asymmetric ) to justify an undirected ( respectively , directed ) network analysis .
we introduce a new index that satisfies two main properties .
first , it can be applied to both binary or weighted graphs .
second , once suitably standardized , it distributes as a standard normal over all possible adjacency / weights matrices . to test the index in practice , we present an application that employs a set of well - known empirically - observed social and economic networks .
|
first , liu et al . ( 2015 ) note correctly that we failed to cite a previous large - scale empirical study with similar aims .
in particular , liu ( 2008 ) compares average dependency length in attested sentences of 20 languages to dependency length in random trees . not acknowledging this important prior work was an error on our part .
the reason for this omission is that , in all honesty , we did not fully understand this paper and its relationship to ours until conversations with liu and colleagues after publication .
but these are not good reasons : we acknowledge that we should have made more of an effort to understand and acknowledge prior similar work .
consequently , we apologize and we urge anyone pursuing research relating to our paper to also study liu ( 2008 ) .
this prior work will be acknowledged in a correction to the pnas article .
nevertheless , we believe the difference between the liu ( 2008 ) baselines and ours is non - trivial , such that our work represents new large - scale evidence for the dlm hypothesis . liu ( 2008 ) uses a `` random tree '' baseline , comparing dependency length in attested dependency trees to dependency length in random ordered trees with the same numbers of nodes .
for example , the dependency length of a sentence with a tree such as in figure [ fig : examplesent ] is compared to the dependency length induced by random ordered trees as in figure [ fig : randomtrees ] .
the baseline trees do not share any syntactic structure with the attested trees they are compared to , beyond their length . in contrast , gildea & temperley ( 2010 ) and futrell et al .
( 2015 ) use `` random word order '' baselines , keeping the syntactic dependency structure of attested sentences constant and investigating random word orders given that syntactic structure , subject to a number of linguistic constraints .
for example , dependency length for a sentence such as in figure [ fig : examplesent ] is compared to dependency length in a sentence with different word order but the same ( unordered ) dependency tree structure , as in figure 3 .
attested dependency length is shorter than both the random tree and random word order baselines .
john & threw & out & the & trash + total dependency length = 6 + a & b & c & d & e + total dependency length = 8 + a & b & c & d & e + total dependency length = 11 + a & b & c & d & e + total dependency length = 7 + john & threw & the & trash & out + total dependency length = 7 + our finding that attested dependency length is shorter than random word order baselines shows that , _ given _ a syntactic structure , language users and language grammars tend to prefer the word order that minimizes dependency length .
this finding supports the dlm hypothesis and provides direct evidence for a specific mechanism ( word order preferences ) by which dependency length minimization is accomplished . on the other hand , the finding that attested
dependency length is shorter than the random _ tree _
baselines supports the dlm hypothesis in a more general form and is consistent with many possible mechanisms that shorten dependency length , including non - syntactic mechanisms .
for example , it is consistent with the idea that languages might disprefer structures which inevitably create long dependencies , such as high arity trees .
it is also consistent with the hypothesis that language users prefer sentences with structures that create long dependencies , and might structure discourse to avoid such sentences .
for example , the sentence ( 1 ) `` a man who was wearing a hat arrived '' has a long dependency between the subject `` man '' and the verb `` arrived '' because the relative clause `` who was wearing a hat '' intervenes between them .
language users might prefer to instead say ( 2 ) `` a man arrived '' , avoiding the relative clause between the subject and the verb , and perhaps mentioning the information about the hat in another sentence later in discourse , or perhaps dropping it altogether . though language users are ultimately achieving the same or similar communicative goals in saying sentence ( 1 ) and sentence ( 2 ) , they are doing so by expressing different propositional content in each sentence
. the mechanisms by which dependency length minimization is accomplished in comparison to a random tree baseline are thus highly general : in addition to word order preferences , languages might have tree structure preferences ; and language users might strategically choose _ what content to express _ , in addition to what word order to use , in order to avoid long dependencies . in summary , comparing to random tree baselines can show dlm as a result of many mechanisms , including the content that people choose to express and/or the word orders they use in sentences .
so the finding that attested dependency length is shorter than this baseline supports an influence of dlm on discourse structure or syntactic structure or both . comparing to the random word order baseline , on the other hand ,
shows specifically that the word orders that people prefer , _ given _ the content they choose to express , are those that minimize dependency length .
that is , it shows unambiguously that dlm as a pressure affects syntactic structure and word order in particular . because our findings are _ only _ compatible with dependency - length - minimizing preferences in word order , we believe they provide novel , strong evidence for the dlm hypothesis as it pertains to syntax .
our claim is that , all else being equal , language users prefer linearizations with short dependency length .
only the comparison to a random word order baseline supports this claim unambiguously .
so we see this work as a complement of liu ( 2008 ) and related work , strengthening the body of evidence for the dlm hypothesis , rather than a repetition .
the difference between random tree baselines and random word order baselines can also explain some discrepancies between our work and previous findings .
for example , we find relatively long dependency lengths for head - final languages such as japanese and turkish , whereas hiranuma ( 1999 ) finds that dependency length in japanese is highly optimized .
hiranuma ( 1999 ) s finding is specifically that japanese speakers drop verbal arguments to achieve dependency length minimization , trusting that the language comprehender will be able to infer the missing arguments from discourse context .
our finding is that , _ given _ the set of words and the dependency tree that japanese speakers want to express , they choose orders with longer dependency length than , say , english speakers .
( this finding remains unexplained . )
the second major issue raised in both liu et al . ( 2015 ) and ferrer - i - cancho & gmez - rodrguez ( 2015 ) is our choice of baselines for comparison .
we use projective linearizations , meaning that when a dependency tree is drawn over a linearized sentence , none of the arcs of the tree cross .
we also use linearizations incorporating other factors that might conceivably influence word order : a pressure for fixed word order , and a pressure for consistency in head direction .
these three factors projectivity , head direction consistency , and fixed word order all have the effect of reducing dependency length , and so it has been argued for the first two that they need not be considered separate factors , but rather the result of dlm .
ferrer - i - cancho and gmez - rodrguez ( 2015 ) argue that our use of these baselines is redundant for this reason .
we believe comparison to these baselines provides stronger evidence for dlm than comparison only to a fully nonprojective baseline , because it shows that the phenomenon of short dependencies must be explained _ even if _
independent factors affecting word order are assumed .
since dlm can explain the phenomena attributed to these other factors , the most parsimonious theory seems to be that dlm is the only factor influencing word order .
but we can only make this argument after showing that the shortness of dependencies persists as a phenomenon even after controlling for these other hypothetical factors .
for example , suppose we had found that attested dependency length was _ not _ shorter than the projective random baselines .
one would be left with the question of why , if dlm is the main factor influencing language structure , german speakers pass up opportunities to minimize dependency length
. then one could argue that dlm is not a good explanation for projectivity , since word orders are not minimized for dependency length beyond what is needed to establish projectivity , which itself might have independent motivations ( such as enabling polynomial - time parsing ) . since we found that dependency length _ is _ shorter than this baseline in many languages , this line of argumentation is no longer available . for the sake of completeness
, we provide a comparison of attested dependency lengths with dependency lengths in random nonprojective linearizations in figure [ fig : nonproj ] . for this baseline
, the dependency tree is linearized by shuffling nodes at random .
the baselines from futrell et al .
( 2015 ) are also shown .
the figure shows that dependency length is much shorter than the nonprojective baseline , and that the projective baselines are much more conservative than the nonprojective baseline . we felt that including the nonprojective baselines in the original paper would be redundant , since ferrer - i - cancho ( 2006 ) showed that projective trees on average have shorter dependency length than nonprojective trees , and kuhlmann ( 2013 ) ( among others ) showed that natural language dependency trees are overwhelmingly projective .
we also want to stress that , contra ferrer - i - cancho & gmez - rodrguez ( 2015 ) , controlling for these possible alternative factors affecting word order does not imply that we are accepting traditional nativist or universal grammar - based hypotheses .
these factors have possible functional explanations , just as dlm does .
fixed word order can be motivated by efficient communication of relation types ; consistent head direcion can be motivated by compression of grammars ; and projectivity can be motivated by the time complexity of parsing , where parsing to projective trees is cubic - time but parsing to fully nonprojective trees is np - hard .
in general , we aimed to include the most conservative reasonable baselines .
liu et al . ( 2015 ) also raise a number of more specific criticisms .
they claim that the uniformity of genres of the text in our corpora could be a confounding factor .
the criticism is valid : it is true that our corpora were primarily ( but not entirely ) written text from newspapers and novels .
nevertheless , we would find it surprising if dlm universally influenced novels and newspapers but not language use in general .
we welcome any work which controls for this possible issue .
finally , liu et al . (
2015 ) also note that in our original paper we state that head - final languages appear to have longer dependencies than more head - initial or head - medial languages , but we do not provide statistical tests of this claim .
we intended this remark not as a main claim of the paper , but as a conjecture intended to draw attention to the wide variation between languages in their dependency length , and possible typological implications of that variation .
working out the correct statistical methodology and gathering the right data to make this a strong empirical claim would require another whole paper .
the question of variation in dependency length has also been a major focus of liu s research .
we feel that explaining this variation is the most interesting direction for future dependency length research , and we hope to join our present critics in future investigations of this phenomenon .
|
we address recent criticisms ( liu et al . , 2015 ;
ferrer - i - cancho and gmez - rodrguez , 2015 ) of our work on empirical evidence of dependency length minimization across languages ( futrell et al . , 2015 ) .
first , we acknowledge error in failing to acknowledge liu ( 2008 ) s previous work on corpora of 20 languages with similar aims
. a correction will appear in pnas .
nevertheless , we argue that our work provides novel , strong evidence for dependency length minimization as a universal quantitative property of languages , beyond this previous work , because it provides baselines which focus on word order preferences .
second , we argue that our choices of baselines were appropriate because they control for alternative theories . in recent work
, we addressed the question of whether dependency length
the distance between syntactically related words in natural language sentences is shorter than one would expect under random baselines ( futrell et al . , 2015 )
this idea has linguistic relevance because if one hypothesizes a universal pressure to minimize dependency length , one can explain a variety of universal properties of languages , including many of the word - order universals noted by greenberg ( 1963 ) .
evidence that language users perfer word orders with shorter dependency length than chance supports this hypothesis , known as the dependency length minimization ( dlm ) hypothesis .
the dlm hypothesis is theoretically attractive because it is motivated by general human information processing constraints : minimizing dependency length minimizes the online memory load for human sentence parsing and generation .
two recent articles have raised important criticisms of our work ( liu et al . , 2015 ;
ferrer - i - cancho & gmez - rodrguez , 2015 ) .
|
two closed riemannian manifolds @xmath1 and @xmath2 are said to be _ isospectral _ if the laplace beltrami operators acting on functions on @xmath1 and on @xmath2 have the same spectrum , equivalently if the spectral zeta functions @xmath9 and @xmath10 are equal .
the manifolds are said to be _ strongly isospectral _ if the spectra of every natural self - adjoint elliptic differential operator on @xmath1 and @xmath2 agree .
in particular , if @xmath1 and @xmath2 are strongly isospectral , then the laplace de rham operators acting on differential @xmath11-forms on the respective manifold have the same spectrum , equivalently the zeta functions @xmath12 and @xmath13 encoding those spectra are equal . following kac s famous question `` can one hear the shape of a drum ? ''
@xcite , the following broad questions have received a lot of attention @xcite .
[ q:1 ] which isometry invariants of closed riemannian manifolds are isospectral invariants ? which ones are strongly isospectral invariants ?
for example , two riemannian manifolds that are strongly isospectral have the same dimension , volume and betti numbers . by analogy with the number theoretic
notion of so - called arithmetically equivalent number fields , sunada @xcite proposed a general construction showing that there exist riemannian manifolds that are strongly isospectral , but are not even homeomorphic .
let us briefly recall sunada s construction .
let @xmath14 be a @xmath0-covering of closed manifolds , where @xmath0 is a finite group .
let @xmath15 , @xmath16 be two subgroups of @xmath0 with the property that there is an isomorphism @xmath17\cong \c[g / u_2]$ ] of linear permutation representations of @xmath0 , or equivalently that for every conjugacy class @xmath18 of @xmath0 we have @xmath19 .
in this case we say that the formal linear combination @xmath20 is a @xmath0-relation .
sunada proves that then the intermediate coverings @xmath21 and @xmath22 are strongly isospectral . in this paper , we want to take the analogy with number fields further .
the cheeger - mller theorem @xcite , proving a conjecture of ray and singer @xcite , gives a special value formula for the spectral zeta functions of a riemannian manifold , somewhat analogous to the analytic class number formula for the dedekind zeta function of a number field @xcite .
it implies that if @xmath1 and @xmath2 are strongly isospectral riemannian manifolds , then @xmath23 where , for a riemannian manifold @xmath24 , @xmath25 is the covolume of the lattice @xmath26 in the vector space @xmath27 with respect to a certain canonical inner product ( see notation [ not : pairings ] [ not : reg ] ) .
the study of torsion homology and regulators of manifolds has attracted a lot of attention recently , see for instance @xcite . in light of equation ( [ eq : cheegermuller ] ) ,
a natural instance of question [ q:1 ] is whether , for @xmath1 and @xmath2 as above , an integer @xmath28 , and a prime number @xmath8 , we must necessarily have @xmath29=\#h_i(m_2,\z)[p^\infty],\end{aligned}\ ] ] where @xmath30 $ ] denotes the @xmath8-primary torsion subgroup . here
, we answer this and similar questions in the context of sunada s construction . more broadly
, we study homological properties of covering manifolds in @xmath0-relations , analogous to an old and fruitful line of research in number theory .
for example , the following basic result is analogous to results of @xcite on class groups of number fields .
[ prop : intro1 ] let @xmath14 be a @xmath0-covering of closed riemannian manifolds , let @xmath31 be a @xmath0-relation , and let @xmath32 be an integer .
then for all primes @xmath8 that do not divide @xmath33 , we have @xmath34 \cong h_i(x / u_2,\z)[p^\infty].\ ] ] we actually prove a more general result , see theorem [ thm : nomackeyfunctor ] .
in providing counter examples , we have concentrated on closed hyperbolic @xmath7-manifolds , a class of manifolds that plays an important rle in other branches of geometry , as well as in number theory .
the following is a result in the opposite direction to that of proposition [ prop : intro1 ] .
[ prop : intro2 ] let @xmath6 be a prime number .
then there exist strongly isospectral closed hyperbolic @xmath7-manifolds @xmath1 and @xmath2 such that @xmath35\neq \#h_1(m_2,\z)[p^\infty].\ ] ] moreover , if @xmath36 , then there exist @xmath7-manifolds @xmath1 and @xmath2 as above with @xmath35=1,\;\;\text { and } \;\;\#h_1(m_2,\z)[p^\infty]=p.\ ] ] note that hyperbolic @xmath7-manifolds have torsion - free homology in degrees @xmath37 , @xmath38 , and @xmath7 , so in the situation of proposition [ prop : intro2 ] , equation ( [ eq : chmtors ] ) is trivially satisfied in those degrees .
it is already known that for every prime @xmath8 , there exist strongly isospectral closed riemannian @xmath39-manifolds @xmath1 and @xmath2 such that @xmath40\neq \#h_1(m_2,\z)[p^\infty]$ ] .
this is shown in @xcite , using the following two results .
firstly , every finite group can be realised as the fundamental group of a closed smooth 4-manifold @xcite .
secondly , for every prime @xmath8 , there exists a finite @xmath8-group @xmath0 with a @xmath0-relation @xmath31 , where the abelianisations of @xmath15 and @xmath16 have different orders ( * ? ? ?
* , example 3 ) . at the other extreme , if @xmath24 is a closed oriented 2-manifold , then @xmath41 is torsion - free .
riemannian @xmath7-manifolds present an in - between case .
it follows from the elliptisation theorem , as proven by perelman @xcite , and from a theorem of ikeda @xcite , that any two isospectral @xmath7-manifolds with finite fundamental groups are homeomorphic .
in particular , sunada s construction for 4-manifolds can not possibly generalise to @xmath7-manifolds .
instead , we prove proposition [ prop : intro2 ] by a judicious choice of @xmath0-relation @xmath20 , a different one for each prime @xmath8 , and then by performing a computer search among @xmath0-coverings of hyperbolic @xmath7-manifolds and computing the resulting torsion homology of the intermediate coverings corresponding to @xmath15 and @xmath16 . nevertheless , we believe that the evidence of proposition [ prop : intro2 ] justifies the following conjecture in @xcite using techniques introduced in this paper , see in particular proposition [ prop : impliesconj ] ] .
[ conj : intro ] for every prime number @xmath8 , there exist strongly isospectral closed hyperbolic @xmath7-manifolds @xmath1 and @xmath2 such that @xmath40\neq \#h_1(m_2,\z)[p^\infty]$ ] .
it follows from equation ( [ eq : cheegermuller ] ) that if @xmath1 and @xmath2 are strongly isospectral , then @xmath42 can be expressed in terms of torsion homology of the two manifolds , and in particular is a rational number .
moreover , if @xmath1 and @xmath2 arise from sunada s construction , then it follows from proposition [ prop : intro1 ] that the numerator and denominator of this rational number are only divisible by prime divisors of @xmath33 .
like in the case of torsion , it would be interesting to understand the quotient of regulators in each degree separately . in this direction
we have the following result , which we now state for arbitrary @xmath0-relations ( see definition [ defn : relation ] ) .
below , for a group @xmath43 we write @xmath44 $ ] where @xmath45 $ ] denotes the derived subgroup of @xmath43 .
[ thm : intro3 ] let @xmath14 be a @xmath0-covering of closed oriented riemannian @xmath46-manifolds , and let @xmath47 , @xmath48 be a @xmath0-relation .
then : 1 . for every integer @xmath32 ,
the product @xmath49 is a rational number that is a product of powers of prime divisors of @xmath33 ; 2 .
[ item : regcompq ] for every @xmath50 , the class of @xmath51 in @xmath52 only depends on the isomorphism class of the @xmath0-module @xmath53 ; 3 .
[ item : regcompz ] if @xmath8 is a prime number that does not divide @xmath54 for any @xmath55 , then the @xmath8-part of @xmath56 only depends on the isomorphism class of the @xmath0-module @xmath57 . the representation theoretic invariant that appears in parts ( [ item : regcompq ] ) and ( [ item : regcompz ] ) of the theorem was first introduced by dokchitser dokchitser in the context of regulators of elliptic curves ( see e.g. @xcite ) , and is called the regulator constant of the representation with respect to the given @xmath0-relation .
we actually relate the quotients of regulators to regulator constants in greater generality , namely without the hypothesis on @xmath8 in part ( [ item : regcompz ] ) of theorem [ thm : intro3 ] .
but in general , there is an `` error term '' , which measures the failure of @xmath0-descent for homology in the covering
see theorem [ thm : regcomparison ] for the precise statement , but see also remark [ rmrk : noclue ] .
vignras @xcite introduced another well - known construction of isospectral manifolds , of a more arithmetic flavour . in @xcite , calegari and venkatesh study torsion homology in pairs of manifolds that are of a similar nature to the examples of vignras , called `` jacquet langlands pairs '' , that are not isospectral but
whose spectra are closely related .
they raise , in ( * ? ? ?
* section 7.10 ) , the question of giving algebraic interpretations to the quotients of torsion homology and regulators separately , analogously to theorem [ thm : intro3 ]
. it would also be interesting to investigate the same type of questions in the context of the vignras examples . in section [ sec
: examples ] , we give examples of how theorem [ thm : intro3 ] allows one to deduce concrete information about the @xmath58$]-module structure of @xmath59 from the torsion homology of quotients of @xmath60 .
it also allows one to formulate a possible representation theoretic line of attack on conjecture [ conj : intro ] , such as the following result , which will be proved as corollary [ cor : impliesconj ] .
[ prop : impliesconj ] let @xmath8 be an odd prime , let @xmath61 , and define the subgroups @xmath62 of @xmath0 . suppose that there exists a @xmath0-covering @xmath14 of closed hyperbolic @xmath7-manifolds such that the difference of the betti numbers @xmath63 is odd .
then conjecture [ conj : intro ] holds for @xmath8 . to fully appreciate the representation theoretic nature of proposition [ prop : impliesconj ] , see the full statement of corollary [ cor : impliesconj ] .
the structure of the paper is as follows . in section [ sec :
brauer ] we recall the formalism of burnside groups , representation groups , and @xmath0-relations , and the definition of regulator constants . in section [ sec : manifolds ]
we investigate the behaviour of homological invariants of riemannian manifolds in @xmath0-relations , and prove proposition [ prop : intro1 ] and theorem [ thm : intro3 ] . in section [ sec : regconst ]
we compute the regulator constants of the rational representations of @xmath64 for odd primes @xmath8 , and of @xmath65 , with respect to certain @xmath0-relations
. these computations will allow us to deduce concrete information about the @xmath58$]-module structure of the rational homology of @xmath0-coverings from the torsion homology of intermediate coverings .
together with theorem [ thm : regcomparison ] , these calculations will also imply proposition [ prop : impliesconj ] .
we prove proposition [ prop : intro2 ] by a direct computer search , and section [ sec : calc ] is devoted to the methods and algorithms used to that end . in section
[ sec : examples ] we have collected some interesting examples , illustrating the various phenomena that we investigate here . all our riemannian manifolds will be assumed to be finite - dimensional , oriented , and closed . by an automorphism of a riemannian manifold we mean an orientation preserving diffeomorphism from the manifold to itself that is a local isometry . by a hyperbolic @xmath7-manifold , we mean a quotient of the hyperbolic @xmath7-space by a discrete subgroup of orientation preserving isometries .
if @xmath8 is a prime number , we will write @xmath66 for the localisation of @xmath67 at @xmath8 , i.e. the subring @xmath68 of @xmath69 . for a rational number @xmath70 , with @xmath71 and @xmath72 , the @xmath8-adic valuation of @xmath73
is defined to be @xmath74 .
when @xmath24 is a @xmath67-module , we write @xmath75 for the torsion submodule of @xmath24 and @xmath76 for the torsion - free quotient of @xmath24 .
we would like to thank nicolas bergeron , nathan dunfield , derek holt , emilio lauret , and karen vogtmann for helpful discussions , and the anonymous referee for suggestions that improved the exposition .
we recall some standard definitions , for which we refer to ( * ? ? ?
11 ) . by `` module
'' we will always mean a finitely generated left module .
let @xmath0 be a finite group .
the _ burnside group _ of @xmath0 is the free abelian group on isomorphism classes of transitive @xmath0-sets .
the set of transitive @xmath0-sets is in bijection with the set of conjugacy classes of subgroups of @xmath0 via the map that assigns to the subgroup @xmath43 the set of cosets @xmath77 . using this identification
, we will represent elements of the burnside group as formal sums @xmath78 , where @xmath79 and @xmath80 .
let @xmath81 be a domain .
representation group _ of @xmath0 over @xmath81 is the free abelian group on isomorphism classes of @xmath81-free indecomposable @xmath82$]-modules , where @xmath82 $ ] is the group ring of @xmath0 over @xmath81 .
let @xmath0 be a finite group and @xmath81 be a domain .
we write @xmath83 for the free @xmath81-module of rank @xmath84 with trivial @xmath0-action .
we have a natural group homomorphism @xmath85}$ ] from the burnside group of @xmath0 to the representation group of @xmath0 over @xmath81 , which sends a @xmath0-set @xmath60 to the permutation module @xmath86 $ ] with an @xmath81-basis indexed by the elements of @xmath60 , and with the @xmath81-linear @xmath0-action given by permutations of the basis .
[ defn : relation ] let @xmath81 be a domain .
@xmath82$]-relation _ is an element of the kernel of @xmath85}$ ] .
a @xmath58$]-relation will be referred to simply as a @xmath0-relation .
[ ex : dihedral ] let @xmath8 be a prime , and let @xmath87 be a dihedral group of order @xmath88 . then @xmath89 is a @xmath0-relation .
concretely , this means that there is an isomorphism of linear permutation representations of @xmath0 over @xmath69 , @xmath90\oplus \q[g / g]^{\oplus 2}\cong \q[g / c_2]^{\oplus 2 } \oplus
\q[g / c_p].\ ] ] moreover , for every prime @xmath91 , the relation @xmath92 is in fact a @xmath93$]-relation , but it is not a @xmath94$]-relation see ( * ? ? ?
* proof of proposition 3.9 ) .
[ ex:2-gassman ] let @xmath0 be the affine linear group over @xmath95 , i.e. the group of linear transformations @xmath96 of @xmath95 , where @xmath97 and @xmath98 .
consider the subgroups @xmath99 and @xmath100 .
the group @xmath0 is isomorphic to the semi - direct product @xmath101 , and the subgroups @xmath15 and @xmath16 are both isomorphic to @xmath102 .
then @xmath20 is a @xmath94$]-relation for every odd prime @xmath8 , but is not a @xmath103$]-relation , as can be deduced from a direct character computation and lemma [ lem : grelations ] below .
[ ex : p - gassman ] let @xmath8 be an odd prime , and let @xmath104 . consider the two subgroups @xmath105 then for every prime @xmath106 , @xmath20 is a @xmath93$]-relation if and only if @xmath91 . observe that @xmath107 contains the subgroup @xmath108 , which is central in @xmath0 .
set @xmath109 , and @xmath110 for @xmath111 and @xmath38 .
we then get the @xmath112-relation @xmath113
. moreover , the triple @xmath114 minimises the index @xmath115 among all triples @xmath116 , where @xmath117 is a finite group , and @xmath118 is a @xmath117-relation that is not a @xmath119$]-relation ( see @xcite ) .
[ lem : grelations ] let @xmath0 be a finite group , let @xmath8 be a prime number not dividing @xmath33 , and let @xmath92 be an element of the burnside group of @xmath0
. then the following are equivalent : 1 .
@xmath92 is a @xmath120$]-relation ; 2 .
@xmath92 is a @xmath58$]-relation ; 3 .
@xmath92 is a @xmath94$]-relation ; 4 .
@xmath92 is an @xmath121$]-relation .
the equivalence of ( 1 ) and ( 2 ) follows from the fact that if @xmath122 and @xmath123 are @xmath58$]-modules , then @xmath124 ( see @xcite ) .
the remaining equivalences are proven in ( * ? ? ?
* chapter 5 ) .
[ not : burnsidefunctions ] let @xmath125 be any function on the set of conjugacy classes of subgroups of a finite group @xmath0 with values in an abelian group @xmath126
. then @xmath125 extends to a unique group homomorphism from the burnside group of @xmath0 to @xmath126 , defined by @xmath127 , where @xmath128 are subgroups of @xmath0 and @xmath48 .
let @xmath129 be a subring of @xmath130 that is a pid .
the main examples we are thinking of are @xmath67 , the localisation @xmath66 where @xmath8 is a prime , and @xmath69 .
given an @xmath131$]-module @xmath126 , let @xmath132 denote the @xmath131$]-submodule consisting of @xmath129-torsion elements of @xmath126 , and let @xmath133 . since @xmath129 is a pid , for every @xmath131$]-module @xmath126 , the quotient @xmath134 is an @xmath129-free @xmath131$]-module .
moreover , @xmath135 is _ self - dual _ , i.e. it is isomorphic to the @xmath136$]-module @xmath137 .
thus there exists an @xmath130-valued @xmath129-bilinear @xmath0-invariant non - degenerate pairing on @xmath134 . from now on ,
we will just say _ pairing on @xmath126 _ when we mean an @xmath130-valued @xmath129-bilinear @xmath0-invariant pairing on @xmath126 that is non - degenerate on @xmath134 .
let @xmath0 be a finite group , let @xmath129 a subring of @xmath130 that is a pid , let @xmath81 be a subring of @xmath69 , and let @xmath126 be an @xmath131$]-module .
let @xmath138 be a pairing on @xmath126 , and let @xmath139 be an @xmath82$]-relation .
the _ regulator constant _ of @xmath126 with respect to @xmath92 is defined as @xmath140 where the @xmath55-th determinant is computed with respect to any @xmath129-basis on @xmath141 .
the class in @xmath142 of each determinant is independent of the choice of basis .
[ thm : regconst ] the value of @xmath143 is independent of the pairing @xmath138 on @xmath126 . see ( * ? ? ?
* theorem 2.17 ) .
it follows from theorem [ thm : regconst ] that if @xmath144 , then @xmath145 is well - defined .
indeed , @xmath146 , and the pairing can always be chosen to take values in @xmath69 .
similarly , if @xmath147 , then @xmath148 is well - defined , while if @xmath149 , then @xmath143 defines a class in @xmath52 .
[ cor : multiplicative ] let @xmath150 , @xmath151 be @xmath131$]-modules , and @xmath92 a @xmath0-relation .
then @xmath152 .
we can choose the pairing on @xmath153 so that the direct summands are orthogonal to each other , making all the matrices block diagonal .
[ ex : regconsttriv ] let @xmath0 be a finite group , and let @xmath154 be a @xmath0-relation .
then @xmath155 .
[ lem : degree0 ] let @xmath0 be a finite group and let @xmath139 be a @xmath0-relation .
then we have @xmath156 . by definition ,
the virtual representation @xmath157^{\oplus n_j}$ ] is zero .
the result follows by taking the inner product with the trivial character of @xmath0 .
[ prop : orthoreg ] let @xmath0 be a finite group , and let @xmath158 be a @xmath0-relation .
let @xmath126 be a @xmath58$]-module that has no simple summand in common with any @xmath159 $ ] . then @xmath160 .
see ( * ? ? ?
* lemma 2.26 ) .
let @xmath0 be a finite group , and @xmath161 a subgroup .
the operation on @xmath0-sets of restricting the action of @xmath0 to @xmath161 extends linearly to a restriction map on the burnside groups , and induces a restriction map on relations . with respect to the bases of the burnside groups given by transitive @xmath0-sets , respectively transitive @xmath161-sets ,
the restriction map is given by mackey s formula : @xmath162 we have the following form of frobenius reciprocity for regulator constants . [
prop : frobenius ] let @xmath0 be a finite group , let @xmath161 be a subgroup , let @xmath126 be an @xmath163$]-module , and let @xmath92 be a @xmath0-relation .
then we have @xmath164 see ( * ? ? ? * proposition 2.45 ) .
in this section , we investigate the quotients of torsion homology and of regulators of riemannian manifolds arising from @xmath0-relations .
we begin by recalling some basic definitions and fixing the notation .
the reader is referred to @xcite for the details .
if @xmath165 , @xmath166 are two abelian groups equipped with @xmath130-valued bilinear forms @xmath167 , respectively @xmath168 , a _ similitude from @xmath165 to @xmath166 with factor @xmath169 _ is a homomorphism @xmath170 such that @xmath171 for all @xmath172 , @xmath173 .
an _ isometry _ is a similitude with factor 1 . [ not : pairings ] let @xmath60 be a @xmath46-dimensional riemannian manifold , and let @xmath32 be an integer . 1 . _ harmonic forms .
_ the laplace de rham operator acts on the differential @xmath11-forms on @xmath60 , and we let @xmath174 denote the space of differential @xmath11-forms that lie in the kernel of that operator , the space of harmonic @xmath11-forms .
the riemannian metric on @xmath60 induces a canonical inner product on @xmath174 , which induces a canonical isomorphism between @xmath174 and the @xmath130-linear dual @xmath175 .
explicitly , the inner product on @xmath174 is given by @xmath176 where @xmath177 , @xmath178 , and where @xmath179 denotes the hodge star operator @xmath180 ( see @xcite ) .
we equip @xmath181 with the inner product induced by the one on @xmath174 , i.e. the unique inner product that makes the map @xmath182 an isometry .
[ not : reg ] _ regulators .
_ we have a homomorphism @xmath183 given by @xmath184 .
it follows from the hodge theorem ( ( * ? ? ?
* theorem 1.45 ) ) and de rham s theorem that @xmath185 and that the image of @xmath186 spans @xmath181 over @xmath130 .
@xmath11-th regulator _
@xmath187 of @xmath60 is defined as the covolume of @xmath188 with respect to the inner product on @xmath181 .
+ let @xmath189 denote the pullback via @xmath186 of the pairing on @xmath181 , i.e. the unique pairing on @xmath190 that makes @xmath186 an isometry .
poincar duality .
_ by poincar duality , the map @xmath191 given by cap product with the fundamental class of @xmath60 is an isomorphism .
it follows that @xmath192 .
[ rmrk : equivariant ] _ actions of automorphism groups .
_ let @xmath0 be a group of automorphisms of @xmath60 .
then @xmath0 acts on the left by linear transformations on @xmath190 via pushforward , @xmath193 for @xmath194 and @xmath195 .
also , @xmath0 acts linearly on the right on @xmath174 via pullback , @xmath196 for @xmath194 and @xmath197 , which induces the dual action on the left on @xmath198 .
it follows from the adjointness of pullbacks and pushforwards that the map @xmath186 is a homomorphism of left @xmath0-modules .
moreover , recall that our automorphisms are assumed to be orientation preserving , so the poincar duality map @xmath199 is an isomorphism of @xmath0-modules .
[ lem : regdet ] let @xmath60 be a riemannian manifold
. then we have @xmath200 , where the left hand side is computed with respect to any @xmath67-basis of @xmath201 .
in particular , the pairing @xmath202 is non - degenerate on @xmath201 .
the claimed equality is the expression of the covolume of a lattice in terms of the determinant of its gram matrix .
[ rmrk : poincarereg ] explicitly , let @xmath203 be an orthonormal basis of @xmath174 , and let @xmath204 be a @xmath67-basis of @xmath201 .
then we have @xmath205 and for all @xmath206 , @xmath207 , we have @xmath208 .
we now prove proposition [ prop : intro1 ] as a consequence of the following result .
[ thm : nomackeyfunctor ] let @xmath14 be a finite @xmath0-covering of riemannian manifolds , and let @xmath209 be a @xmath94$]-relation , where @xmath210 and @xmath211 are positive integers . then for every integer @xmath32 , we have @xmath212^{n_j}\cong \bigoplus_k h_i(x / u_k',\z)[p^\infty]^{n_k'}.\ ] ] in particular we have @xmath213 ( see notation [ not : burnsidefunctions ] ) .
let @xmath43 be any subgroup of @xmath0 .
then the @xmath214-torsion of @xmath215 is isomorphic to the torsion subgroup of @xmath216 . by definition of homology with local coefficients
( see ( * ? ? ?
* h ) ) , we have @xmath217)$ ] . since homology with local coefficients is additive in direct sums of modules , the result follows .
let @xmath14 and @xmath0 be as in theorem [ thm : nomackeyfunctor ] , let @xmath8 be a prime number that does not divide @xmath33 , and let @xmath92 be a @xmath0-relation .
then @xmath212^{n_j}\cong \bigoplus_k h_i(x / u_k',\z)[p^\infty]^{n_k'}.\ ] ] this is a direct consequence of theorem [ thm : nomackeyfunctor ] and lemma [ lem : grelations ] .
the rest of the section is devoted to the behaviour of regulators in @xmath0-relations .
in particular , we will prove theorem [ thm : intro3 ] .
[ lem : h0 ] let @xmath218 be a finite @xmath0-covering of riemannian manifolds of dimension @xmath46 , and let @xmath139 be a @xmath0-relation . then @xmath219 for all @xmath220 , we have @xmath221 .
it follows that @xmath222 where the last equality follows from example [ ex : regconsttriv ] and lemma [ lem : degree0 ] .
[ lem : simil ] let @xmath223 be a finite @xmath0-covering of riemannian manifolds , and let @xmath32 be an integer
. then the following maps are similitudes with factor @xmath33 : 1 .
[ item : prepnotisom ] the pullback map @xmath224 ; 2 .
[ item : notisom ] its dual @xmath225 ; 3 .
[ item : descentpairing ] the pushforward map @xmath226 4 .
[ item : ascentpairing ] any map @xmath227 satisfying @xmath228 . 1 .
let @xmath229 $ ] and @xmath230 $ ] denote the fundamental class on @xmath60 , respectively on @xmath231 . then @xmath232 = \#g\cdot [ y]$ ] , so for all @xmath233 , we have @xmath234 2 .
the assertion immediately follows from part ( [ item : prepnotisom ] ) , from the fact that @xmath235 is an isomorphism , and from the definition of the inner product on the dual space , see notation [ not : pairings ] .
3 . by notation
[ not : pairings ] [ rmrk : equivariant ] , the map @xmath186 sends @xmath236 to @xmath237 , and it follows from the adjointness of pullbacks and pushforwards that the diagram @xmath238^{h_x^i } \ar[r]^(.56){f _ * } & h_i(y,\z ) \ar[d]^{h_y^i } \\ ( \ch^i(x)^g)^{\lor } \ar[r]^(.6){(f^*)^{\lor } } & \ch^i(y)^{\lor } } \ ] ] is commutative . note that @xmath237 is canonically isomorphic , as an inner product space , to the @xmath0-coinvariants @xmath239 , which in turn is canonically isomorphic to @xmath240 .
this shows that @xmath241 is an isometry . by definition , @xmath242 is also an isometry , and by part ( [ item : notisom ] ) , @xmath243 is a similitude with factor @xmath244 .
it follows that @xmath245 is also a similitude with factor @xmath244 .
4 . by part ( [ item : descentpairing ] ) , the map @xmath245 is a similitude with factor @xmath244 , and the claim immediately follows . [ lem : exconvenient ] let @xmath218 be a finite @xmath0-covering of riemannian manifolds , and let @xmath32 be an integer
. then the composition @xmath246 satisfies @xmath247 .
the pushforward of the fundamental class of @xmath60 is equal to @xmath33 times the fundamental class of @xmath231 .
it therefore follows from the naturality of the cap product that @xmath248 , which proves the result .
[ lem : index ] let @xmath218 be a finite @xmath0-covering of riemannian manifolds , and let @xmath249 be an integer .
let @xmath227 be any map satisfying @xmath228 , and let @xmath250 then @xmath251 is finite and is independent of @xmath252 .
the quantity @xmath251 is finite since @xmath253 is an isomorphism .
if @xmath254 is another map satisfying the conditions of the lemma , then @xmath255 .
it follows from this , and from the fact that @xmath256 is an isomorphism , that @xmath257 .
therefore @xmath258 which proves the lemma .
given a finite @xmath0-covering @xmath14 and @xmath11 as in lemma [ lem : index ] , we define @xmath259 for any @xmath252 satisfying the hypotheses of the lemma .
[ thm : regcomparison ] let @xmath218 be a @xmath0-covering of riemannian manifolds , let @xmath260 be a @xmath0-relation ( see definition [ defn : relation ] ) , and let @xmath32 be an integer .
then we have @xmath261 by lemma [ lem : regdet ] , the pairing @xmath262 on @xmath190 is non - degenerate on the quotient modulo torsion .
moreover , by lemma [ lem : simil ] ( [ item : descentpairing ] ) applied to @xmath263 for each @xmath194 , this pairing is @xmath0-invariant .
for any @xmath80 , let @xmath252 be any map satisfying the conditions of lemma [ lem : index ] for the covering @xmath264 .
then we have @xmath265 where the second equality follows from lemma [ lem : simil]([item : ascentpairing ] ) , and the last one from lemma [ lem : regdet ] .
this proves the theorem , by taking the product over @xmath55 corresponding to @xmath92 .
[ cor : modsquares ] let @xmath218 be a @xmath0-covering of riemannian manifolds , let @xmath260 be a @xmath0-relation , and let @xmath32 be an integer .
then we have @xmath266 .
let @xmath43 be a group .
the abelianisation of @xmath43 is the quotient @xmath267 $ ] , where @xmath45=\langle uvu^{-1}v^{-1}\colon u , v\in u\rangle$ ] , i.e. the maximal abelian quotient of @xmath43 .
[ cor : regcomparison ] let @xmath223 be a finite @xmath0-covering of riemannian @xmath46-manifolds , let @xmath139 be a @xmath0-relation , let @xmath8 be a prime , and let @xmath32 be an integer .
assume that one of the following holds : 1 .
@xmath268 and for each subgroup @xmath80 , the order of @xmath269 is not divisible by @xmath8 ; 2 . for each subgroup @xmath80 ,
the order of @xmath128 is not divisible by @xmath8 . then @xmath270 . in particular , for all @xmath32 the rational number
@xmath271 is a product of the primes dividing the order of @xmath0 .
let @xmath43 be any subgroup of @xmath0 .
let @xmath272 be the map of lemma [ lem : exconvenient ] . in case
( 1 ) , the 5-term exact sequence coming from the cartan
leray spectral sequence in cohomology ( see ( * ? ? ?
vii , 7 ) ) is @xmath273 by the universal coefficient theorem , we have @xmath274 , which has order coprime to @xmath8 by assumption . in case ( 2 ) , the group @xmath43 has cohomological dimension @xmath37 over @xmath66 , since @xmath8 does not divide its order , so by the cartan
leray spectral sequence we have an isomorphism @xmath275 in both cases , it follows that @xmath276 has cokernel of order coprime to @xmath8 , and since the poincar duality maps @xmath199 and @xmath277 are isomorphisms , the cokernel of @xmath278 is also of order coprime to @xmath8 .
the first equality therefore follows from theorem [ thm : regcomparison ] . for the second equality , note that firstly , @xmath279 , and secondly that for an arbitrary @xmath280$]-module @xmath126 , we have @xmath281 . [ rmrk : noclue ]
we do not know whether the regulator quotient is a purely representation theoretic invariant in full generality without the restrictive assumptions of corollary [ cor : regcomparison ] , that is whether it only depends on the @xmath280$]-module structure of @xmath190 .
this seems to us to be an interesting question .
in this section , we compute the regulator constants of rational representations with respect to the relations of examples [ ex:2-gassman ] and [ ex : p - gassman ] .
first , let @xmath61 .
denote by @xmath282 the subgroup of upper - triangular matrices in @xmath0 , let @xmath283 , and let @xmath15 , respectively @xmath16 be the subgroup of matrices in @xmath282 with square lower right , respectively upper left entry .
then @xmath284 is the @xmath0-relation of example [ ex : p - gassman ] . in this section ,
we compute the regulator constants of the rational irreducible representations of @xmath0 with respect to @xmath92 .
we begin by recalling the classification of complex representations of @xmath0 . for
any 1-dimensional representations @xmath285 , @xmath286 of @xmath287 , let @xmath288 be the @xmath84-dimensional representation of @xmath282 defined by @xmath289 .
[ prop : classif ] the following is a classification of the complex irreducible representations of @xmath0 : * @xmath84-dimensional representations : they are all of the form @xmath290 , where @xmath291 is an irreducible character of @xmath287 .
* irreducible principal series , of dimension @xmath292 : they are the inductions @xmath293 , where @xmath294 ; we have @xmath295 if and only if @xmath296 .
* special representations , of dimension @xmath8 : they are the @xmath8-dimensional summands @xmath297 of @xmath298 . * cuspidal representations , of dimension @xmath299 : they are the irreducible representations of @xmath0 that are not summands of any @xmath300 .
see @xcite .
let @xmath301 denote the complex irreducible representation @xmath302 , the steinberg representation of @xmath0 , and let @xmath303 denote the complex irreducible representation @xmath304 , where @xmath305 is the unique character of @xmath287 of order 2 .
note that both @xmath301 and @xmath303 are realisable over @xmath69 , so in fact give rise to irreducible @xmath58$]-representations .
[ prop : regconstpgassman ] let @xmath306 be an irreducible @xmath58$]-representation .
then@xmath307 if @xmath308 , and @xmath309 otherwise .
we have @xmath17\cong \c[g / u_2 ] \cong \triv\oplus \st\oplus i$ ] .
so if @xmath306 is an irreducible @xmath58$]-representation that is not isomorphic to any of these three direct summands , then @xmath310 by proposition [ prop : orthoreg ] .
moreover , we have @xmath311 by example [ ex : regconsttriv ] . to compute @xmath312 ,
observe that by corollary [ cor : multiplicative ] and by proposition [ prop : frobenius ] , we have @xmath313 let @xmath314 for @xmath111 , @xmath38 .
an elementary calculation ( or the bruhat decomposition ) shows that for @xmath111 and @xmath38 , @xmath315 has cardinality 2 , with double coset representatives @xmath84 and @xmath316 .
moreover , we have @xmath317 , and @xmath318 .
it follows from equation ( [ eq : res ] ) that @xmath319 we conclude from equation ( [ eq : steinberg ] ) and example [ ex : regconsttriv ] that @xmath320 similarly , we have @xmath321 since @xmath322 is 1-dimensional , the contribution from each subgroup @xmath43 in @xmath323 is 1 if @xmath324 is zero , and is @xmath325 otherwise .
it follows that @xmath326 .
[ cor : impliesconj ] let @xmath8 be an odd prime , let @xmath0 , @xmath282 , @xmath15 , and @xmath16 be as above , and let @xmath14 be a @xmath0-covering of closed hyperbolic @xmath7-manifolds
. then the following are congruent modulo 2 : 1 . the @xmath8-adic valuation @xmath327 ; 2 . the multiplicity in @xmath328 of the irreducible @xmath120$]-representation @xmath303 , as defined after proposition [ prop : classif ] ; 3 . the difference of the betti numbers @xmath329 .
first , we show the equality between ( 1 ) and ( 2 ) . for closed hyperbolic @xmath7-manifolds ,
all homology groups with integral coefficients except possibly the first are torsion - free .
it therefore follows from equation ( [ eq : cheegermuller ] ) , remark [ rmrk : poincarereg ] , lemma [ lem : h0 ] , and example [ ex : regconsttriv ] , that @xmath330 by corollary [ cor : modsquares ] , the @xmath8-adic valuation of the right hand side is congruent to @xmath331 , where @xmath332 .
recall that the right hand side of that last congruence is only well - defined modulo 2 .
the equality of ( 1 ) and ( 2 ) therefore follows from corollary [ cor : multiplicative ] and proposition [ prop : regconstpgassman ] .
we now prove the equality between ( 2 ) and ( 3 ) .
below , we will identify representations of @xmath0 with their characters .
let @xmath305 be an arbitrary rational representation of @xmath0 .
we have @xmath333 , so the multiplicity of @xmath303 in @xmath305 is equal to @xmath334 where @xmath335 denotes inner products of characters of @xmath43 . the equality between ( 2 ) and ( 3 ) follows from this calculation , with @xmath336 . as in example
[ ex : p - gassman ] , let @xmath337 , let @xmath109 , @xmath110 for @xmath111 , @xmath38 , and let @xmath338 , which is a @xmath112-relation . there is a canonical bijection between irreducible @xmath58$]-representations with kernel containing @xmath339 and irreducible @xmath340$]-representations , given by projecting to the quotient . for any irreducible @xmath58$]-representation @xmath306 with kernel containing @xmath339 , let @xmath341 denote its image under this bijection .
let @xmath341 be an irreducible @xmath340$]-representation .
then@xmath342 if @xmath343 , and @xmath344 otherwise . by (
* proposition 2.45 ) , we have @xmath345 . the result therefore immediately follows from proposition [ prop : regconstpgassman ] .
now let @xmath0 be the group of affine linear transformations @xmath96 of @xmath95 , where @xmath97 and @xmath98 , let @xmath99 and @xmath100 , as in example [ ex:2-gassman ] .
let @xmath305 be an irreducible faithful representation of @xmath346 .
then @xmath347 is a 4-dimensional irreducible @xmath120$]-representation , and is realisable over @xmath69 , so it gives rise to an irreducible @xmath58$]-representation .
its isomorphism class does not depend on the choice of the representation @xmath305 .
all the other complex irreducible representations of @xmath0 are also realisable over @xmath69 , and are 1- and 2-dimensional .
an explicit calculation , which we omit , gives the regulator constants of all the irreducible @xmath58$]-representations with respect to @xmath332 as follows .
[ prop:2regcst ] let @xmath306 be an irreducible @xmath58$]-representation .
then@xmath348 if @xmath308 , and @xmath349 otherwise .
in this section , we present the computational methods used to prove proposition [ prop : intro2 ] .
the computations were run on the warwick number theory cluster .
we have made the resulting data available at https://arxiv.org/abs/1601.06821 .
for each prime @xmath8 , we look for explicit pairs of manifolds @xmath1 and @xmath2 as in proposition [ prop : intro2 ] .
the pairs we construct all arise as intermediate coverings of @xmath0-coverings @xmath14 of closed hyperbolic @xmath7-manifolds , for suitable finite groups @xmath0 .
we compute many such coverings , and explicitly compute the homology groups of suitable intermediate covers .
there are several choices involved , which we now explain in detail .
we only compute with arithmetic hyperbolic @xmath7-manifolds .
we will recall the definitions , assuming basic facts and terminology on quaternion algebras ; a standard reference is @xcite .
let @xmath350 denote the hyperbolic 3-space .
we identify the group of orientation preserving isometries of @xmath350 with @xmath351 via the poincar extension ( @xcite ) . for any order @xmath352 in a quaternion algebra , let @xmath353 denote the group of elements in @xmath352 of reduced norm @xmath84 .
a number field is called _ almost totally real ( atr ) _ if it has exactly one complex place .
a quaternion algebra over an atr number field @xmath354 is called _ kleinian _ if it ramifies at every real place of @xmath354 . for any ring @xmath81 ,
let @xmath355 denote the algebra of @xmath356 matrices over @xmath81 .
an _ arithmetic hyperbolic @xmath7-manifold _ is a manifold commensurable with a quotient @xmath357 , where @xmath358 arises as follows .
let @xmath359 be a kleinian quaternion algebra over an atr number field , and let @xmath360 be an isomorphism induced by a complex embedding of @xmath354 . then define @xmath361 an arithmetic hyperbolic @xmath7-manifold has finite volume and is compact if and only if @xmath282 is a division algebra see ( * ? ? ?
* theorem 8.2.2 ) .
let @xmath359 be a quaternion algebra over a number field , let @xmath362 denote the ring of integers in @xmath354 , and let @xmath363 be a maximal order in @xmath282 .
let @xmath364 be an ideal of @xmath362 coprime to the discriminant of @xmath282 , so that @xmath365 , and let @xmath366 be the preimage of the subring of upper - triangular matrices under such an isomorphism .
then @xmath367 is an order in @xmath282 .
[ lem : torfree ] let @xmath354 be a number field , and let @xmath282 be a quaternion algebra over @xmath354 that is a division algebra .
let @xmath368 be the set of primes @xmath106 such that @xmath369 is isomorphic to a quadratic extension of @xmath354 contained in @xmath282 , where @xmath370 denotes a primitive @xmath371-th root of unity in an algebraic closure of @xmath354 . for each @xmath372 , and for each primitive @xmath371-th root of unity @xmath373
, choose a prime ideal @xmath374 of @xmath362 such that the minimal polynomial of @xmath373 is irreducible modulo @xmath374 .
let @xmath375 be the set of such @xmath374 .
then : 1 .
the set @xmath368 is finite ; 2 . for every ideal @xmath364 of @xmath362
divisible by all @xmath376 , the group @xmath377 is torsion - free
. if @xmath372 is odd , then @xmath378\le [ f(\zeta_{2q})\colon\q ] = 2[f\colon\q]$ ] , so that @xmath379 + 1 $ ] .
this proves ( 1 ) . to prove ( 2 ) , assume for a contradiction that there exists an element @xmath380 such that @xmath73 has prime order @xmath106 modulo @xmath381 . replacing @xmath73 by @xmath382 if necessary
, we may assume that @xmath73 has order @xmath383 : if @xmath106 is odd then either @xmath73 or @xmath382 has order @xmath383 ; if @xmath384 , then the only elements of order @xmath38 are @xmath385 , since @xmath282 is a division algebra , so that @xmath73 has order @xmath39 . in particular , @xmath386 is a subfield of @xmath282 isomorphic to @xmath369 .
we claim that this is a quadratic extension of @xmath354 , i.e. that @xmath106 belongs to @xmath368 : if @xmath369 is not quadratic , then @xmath387 , but @xmath388 , and @xmath73 does not have order @xmath383 .
let @xmath389 be the minimal polynomial of @xmath73 , and let @xmath376 be the corresponding prime ideal . by definition , the image of @xmath73 in @xmath390 upper - triangular . since @xmath391
, this implies that @xmath389 is not irreducible modulo @xmath392 : a contradiction . in our search
, we extract atr number fields of degree up to 6 and discriminants with absolute value up to @xmath393 from the pari number fields database @xcite .
then , using the algorithms from @xcite implemented in the computer algebra system magma @xcite , we compute 30,000 kleinian quaternion algebras @xmath282 whose maximal orders @xmath394 are such that @xmath395 has covolume at most @xmath396 . using the sufficient condition of lemma [ lem : torfree ]
, we find many orders @xmath367 such that @xmath397 is torsion - free , and therefore acts freely on @xmath350 , and hence we produce many arithmetic hyperbolic @xmath7-manifolds @xmath398 .
the usual way of proving that arithmetic groups have a torsion - free subgroup of finite index is to consider principal congruence subgroups , which have large index .
lemma [ lem : torfree ] allows us to use groups of the form @xmath399 , which have smaller covolume , reducing the cost of the computation .
theorem [ thm : nomackeyfunctor ] implies that in order to produce examples as in proposition [ prop : intro2 ] using sunada s method , we need to look for @xmath0-coverings @xmath14 , where @xmath0 is a finite group that admits a @xmath0-relation @xmath332 that is not a @xmath94$]-relation . for @xmath400 ,
we take @xmath0 and @xmath92 as in example [ ex:2-gassman ] , and for @xmath8 odd , we use the relation of example [ ex : p - gassman ] . in order to find many such @xmath0-coverings of every manifold @xmath401 obtained in section [ sec : y ] , we first compute a finite presentation of @xmath117 , using the algorithms of @xcite . since @xmath402 , @xmath0-coverings of @xmath231 correspond to surjective homomorphisms @xmath403 . using the presentation of @xmath117 , we enumerate surjective homomorphisms @xmath404 up to conjugacy , using the methods described in @xcite .
the complexity of the enumeration depends heavily on @xmath244 , so we use the following improvement .
let @xmath405 be the free group on @xmath406 generators .
recall that for any group @xmath407 there is a canonical bijection @xmath408 .
moreover , if @xmath407 is an abelian group , then this bijection is an isomorphism of abelian groups .
we will tacitly use this identification in the next result .
let @xmath117 be a group with a finite presentation @xmath409 where @xmath81 is the normal closure of the subgroup of @xmath405 generated by @xmath410 elements @xmath411 .
let @xmath0 be a group , let @xmath407 be a subgroup of the center of @xmath0 , and set @xmath412 . let @xmath413 be a homomorphism , and let @xmath414 be an arbitrary lift of @xmath415 .
let @xmath416 , and let @xmath417 denote the linear map induced by evaluation on the @xmath418 . then : 1 .
we have @xmath419 ; and 2 .
the lifts @xmath420 of @xmath421 are exactly the homomorphisms @xmath422 , where @xmath125 ranges over @xmath423 . 1 . since @xmath424 is a lift of @xmath413 , we have @xmath425 , and @xmath419 .
2 . since @xmath407 is central , every lift of @xmath415 is of the form @xmath422 , where @xmath426 .
such a lift descends to a homomorphism @xmath404 if and only if it vanishes on the generators of @xmath81 . for every such generator @xmath316 , we have @xmath427 if and only if @xmath428 , so that the homomorphism descends if and only if @xmath429 . applying this proposition to @xmath61 , @xmath430 and @xmath431 , we reduce the enumeration of @xmath432 to : * enumerating @xmath433 , * computing inverse images under linear maps @xmath434 with @xmath435 , which can be done efficiently by linear algebra . since the enumeration algorithm is expensive , another simple improvement was useful : if there exists a surjection @xmath403 , then @xmath436 surjects onto @xmath437 .
checking this condition before the enumeration is fast and rules out many groups . given a surjective homomorphism @xmath403 , we can compute the homology of the isospectral manifolds as follows .
for every @xmath220 , we have @xmath438)\cong h_1(\gamma,\z[g / u])\text{,}\ ] ] and we can compute the first homology group by linear algebra from a presentation of the group @xmath117 . recall that we want to find examples of @xmath0-coverings @xmath218 such that @xmath439 \neq 1 $ ] . since linear algebra over @xmath67 is costly , we used the following strategy : we first compute @xmath440 for @xmath111 , @xmath38 , which only uses linear algebra over @xmath441 , and we only compute the homology over @xmath67 if @xmath442 .
that condition by itself already implies that @xmath443\not\cong h_1(x / u_2,\z)[p^\infty]$ ] . in practice , when @xmath444 , this condition often yields examples in which @xmath445 $ ] is actually trivial for one @xmath55 , and isomorphic to @xmath446 for the other one . by computing explicit examples ,
we obtain proposition [ prop : intro2 ] , which we restate here for convenience .
[ intro2bis ] let @xmath6 be a prime number .
then there exist strongly isospectral closed hyperbolic @xmath7-manifolds @xmath1 and @xmath2 such that @xmath35\neq \#h_1(m_2,\z)[p^\infty].\ ] ] moreover , if @xmath36 , then there exist @xmath7-manifolds @xmath1 and @xmath2 as above with @xmath35=1,\;\;\text { and } \;\;\#h_1(m_2,\z)[p^\infty]=p.\ ] ]
in this section , we give several numerical examples , illustrating some features of the connection between the torsion on the homology of isospectral manifolds , and the @xmath0-module structure of the integral homology of their covering manifold .
[ ex:1 ] let @xmath447 , let @xmath15 , @xmath16 be as in example [ ex : p - gassman ] . using the notation of section [ sec : y ] ,
let @xmath354 be the atr quartic number field generated by a root @xmath448 of @xmath449 .
consider the kleinian quaternion algebra @xmath450 over @xmath354 .
let @xmath364 be the principal ideal of @xmath362 generated by @xmath451 , which has norm @xmath452 , and let @xmath453 , which is well - defined up to conjugacy in @xmath351 .
a fundamental domain for @xmath454 acting on the ball model of @xmath350 is displayed in figure [ pic37 ] .
the group @xmath454 has a presentation by generators and relations as follows : @xmath455 let @xmath456 .
there is a surjective group homomorphism @xmath457 , given by @xmath458 which gives rise to a @xmath0-covering @xmath14 of arithmetic hyperbolic @xmath7-manifolds . a direct computation , as described in section [ sec : homcomp ] , yields that @xmath459 where @xmath460 denotes the cyclic group of order @xmath406 . here
, we have @xmath461 .
it follows from corollary [ cor : impliesconj ] , that the multiplicity of @xmath462 ( see section [ sec : regconst ] for the definition ) in the @xmath0-module @xmath59 is odd . in this particular case
, we computed directly that this multiplicity is 3 .
let @xmath463 .
let @xmath354 be the atr number field generated by a root @xmath448 of @xmath464 , and let @xmath282 be the kleinian quaternion algebra @xmath465 let @xmath364 be the principal ideal of @xmath362 generated by @xmath466 , of norm @xmath467 , and let @xmath468 , as in section [ sec : y ] .
as in the previous example , @xmath469 is well - defined up to conjugacy in @xmath351 .
a fundamental domain for @xmath469 is displayed in figure [ pic19 ] .
the hyperbolic manifold @xmath470 has a @xmath0-covering @xmath14 ( which we do not give explicitly this time ) , such that @xmath471\cong c_{19 } \times c_{19},\;\;\;h_1(x / u_2,\z)[19^\infty]\cong c_{19 ^ 2}.\ ] ] here , @xmath472 .
corollary [ cor : impliesconj ] implies that the multiplicity of @xmath303 in the @xmath0-module @xmath59 is even , and in fact , in this example it is 0 .
moreover , unlike in the previous example , the non - isomorphic torsion can not be detected by the regulator constant of the @xmath0-module @xmath473 .
let @xmath474 , and let @xmath332 be as in example [ ex : p - gassman ] .
let @xmath354 be the atr number field generated by a root @xmath448 of @xmath475 , and let @xmath282 be the kleinian quaternion algebra @xmath465 let @xmath364 be the principal ideal of @xmath362 generated by @xmath476 , of norm @xmath477 , and let @xmath478 , as in section [ sec : y ] .
as in the previous examples , @xmath479 is well - defined up to conjugacy in @xmath351 .
a fundamental domain for @xmath479 is displayed in figure [ pic5 ] .
the hyperbolic manifold @xmath480 has a @xmath0-covering @xmath14 such that @xmath481\cong c_{5}^3,\;\;\;h_1(x / u_2,\z)[5^\infty]\cong c_{5}.\ ] ] here , @xmath482 . since the quotient @xmath483 is a square , corollary [ cor : modsquares ] together with equation ( [ eq : cheegermuller ] ) imply that the regulator constants @xmath484 are trivial modulo squares , which we confirm by computing that the multiplicity of @xmath303 in @xmath59 is @xmath38 .
in particular , the non - triviality of @xmath485 can not be detected in the rational homology of @xmath60 .
however , by corollary [ cor : regcomparison ] , the non - triviality of @xmath486 is detected in the @xmath0-module structure of the integral homology via the equality @xmath487 .
c. gordon , p. perry and d. schueth , isospectral and isoscattering manifolds : a survey of techniques and examples , geometry , spectral theory , groups , and dynamics , contemp . math . * 387 * , amer . math .
( 2005 ) , 157179 .
|
given a finite group @xmath0 , a @xmath0-covering of closed riemannian manifolds , and a so - called @xmath0-relation , a construction of sunada produces a pair of manifolds @xmath1 and @xmath2 that are strongly isospectral .
such manifolds have the same dimension and the same volume , and their rational homology groups are isomorphic . here
, we investigate the relationship between their integral homology .
the cheeger
mller theorem implies that a certain product of orders of torsion homology and of regulators for @xmath1 agrees with that for @xmath2 .
we exhibit a connection between the torsion in the integral homology of @xmath1 and @xmath2 on the one hand , and the @xmath0-module structure of integral homology of the covering manifold on the other , by interpreting the quotients @xmath3 representation theoretically .
further , we prove that the @xmath4-torsion in the homology of @xmath1 is isomorphic to that of @xmath2 for all primes @xmath5 . for @xmath6 ,
we give examples of pairs of strongly isospectral hyperbolic @xmath7-manifolds for which the @xmath8-torsion homology differs , and we conjecture such examples to exist for all primes @xmath8 .
-0.8 cm
|
recent years have seen progress in our understanding of both the type ii supernova mechanism and heavy element nucleosynthesis .
an intriguing suggestion has been that the origin of the @xmath0-process is in the neutrino - driven wind of a young neutron star ( woosley & hoffman 1992 , see also meyer et al .
1992 ; howard et al . 1993 ; witti , janka , & takahashi 1994 ; and takahashi , witti , & janka 1994 ) .
the winds leave the neutron star about 1 to 20 s after its creation by the stellar core collapse .
the mass loss is sustained by neutrino heating during the kelvin - helmholtz cooling phase of the neutron star .
recently , two groups ( woosley et al . 1994 ; takahashi et al . 1994 ) have published calculations of @xmath0-process in these neutrino - heated ejecta based upon different supernova models .
these calculations yield tantalizing results compared with the observed solar system abundance distribution of @xmath0-process elements , but the calculations share some drawbacks and exhibit differences .
one failure common to all calculations so far is the overproduction of nuclei in the vicinity of the magic neutron shell , @xmath1 , specifically @xmath2sr , @xmath3y , and @xmath4zr .
interestingly , this difficulty can be circumvented and even used to provide some attractive nucleosynthesis of the light @xmath5-process nuclei if the electron fraction @xmath6 in the ejecta is increased slightly from the value found by wilson in woosley et al .
hoffman et al . ( 1996a ) have shown that for @xmath7 , the problematic nucleosynthesis disappears .
this result highlights the need for a detailed and accurate analysis of @xmath6 obtained in the wind , and also hints at some critical uncertainties in the supernova models .
a second difficulty is that various calculations do not agree on the entropy in the wind at the time the @xmath0-process is alleged to occur .
woosley et al . ( 1994 ) based their calculations on a supernova model computed by jim wilson , where high entropies , @xmath8 per baryon ( hereafter , we frequently refer to entropy without its units in boltzmann constant @xmath9 per baryon ) , were obtained at times later than about 10 seconds .
these high entropies were very beneficial , perhaps even necessary for the production of nuclei with mass number @xmath10 and especially the @xmath0-process abundance peak at @xmath11 .
however , in the supernova model evolved by janka and employed in the calculations of witti et al .
( 1994 ) and takahashi et al .
( 1994 ) , entropies of only @xmath12 were obtained .
in fact , takahashi et al . (
1994 ) had to artificially impose an increase by a factor of @xmath13 in the entropy to obtain a successful @xmath0-process .
similar small values of the entropy were also obtained in numerical calculations by woosley ( 1993 and unpublished ) .
this difference further demonstrates the necessity of an unambiguous understanding of the physical conditions obtained in the neutrino - driven wind . finally ,
meyer ( 1995 ) has recently shown that the @xmath0-process in the neutrino - heated supernova ejecta might be seriously affected by neutrino spallation reactions on @xmath14-particles under conditions obtained in wilson s supernova model .
these reactions were neglected in the previous @xmath0-process calculations .
inclusion of these reactions reduces the production of nuclei with @xmath11 in wilson s supernova model . to counter the damaging effects of neutrino spallation reactions on @xmath14-particles requires an entropy even higher than @xmath15 , a lower electron fraction , a shorter dynamic time scale , a larger distance of the @xmath0-process site from the neutrino source , or some combination of the above .
it is clear then that a careful analysis of the physical conditions in the neutrino - driven wind is needed .
these conditions include the entropy per baryon , the electron fraction , the dynamic time scale , and the mass outflow rate .
while simple analytic calculations can not give the full detail of a multidimensional numerical study with non - lte treatments of neutrino transport , they can be very useful for understanding the output from the numerical codes and delimiting , from first principles , likely ranges for the conditions . that is the chief goal of this paper .
because the conditions relevant for heavy element nucleosynthesis exist , in terms of neutron star time scales , long after the shock has been launched in a successful supernova explosion , the neutrino - driven wind can be reasonably approximated as a quasi - steady spherical outflow .
this assumption is borne out by the recent two - dimensional numerical calculations by janka and mller ( 1995 ) .
this fact facilitates analytic studies of the neutrino - driven wind , and greatly simplifies the numerical treatment ( though we shall also briefly examine non - steady outflow ) .
it also has the consequence that our key results are independent of the details of how the shock gets launched . in @xmath16
, we set up the basic equations for analytic studies of the neutrino - driven wind .
our work here is similar in ways to that of duncan , shapiro , and wasserman ( 1986 ) , but more extensive and directed towards nucleosynthetic issues .
the relevant initial and boundary conditions and the input neutrino physics to solve these equations are discussed , and we present the analytic results for the entropy per baryon , the dynamic time scale , and the mass outflow rate in the ejecta in @xmath17 . in @xmath18 , we compare these analytic results with numerical calculations using an implicit hydrodynamic code kepler while varying the neutron star mass , radius , and neutrino luminosity . some dependencies upon the newtonian approximation , the steady - state assumption , boundary pressure , and additional energy sources are also briefly explored . in @xmath19 , we discuss how the electron fraction @xmath6 is determined in the ejecta , and present a novel neutrino two - color plot " that aids in understanding , qualitatively , the nature of heavy element nucleosynthesis in supernovae . in @xmath20 , we discuss the dependence of a successful @xmath0-process on the set of three physical parameters : the entropy per baryon , the electron fraction , and the dynamic time scale , and speculate further on how various factors can affect the properties of the neutrino - driven wind . in a separate paper ( hoffman et al .
1996b ) , we will give a detailed discussion of the implications of our results for heavy element nucleosynthesis in supernovae .
the time of interest here is later than about 1 s after the shock has been launched in a successful type ii or ib supernova . by this time
the shock wave responsible for the supernova explosion has reached a radius of about 10 thousand kilometers and the temperature at that shock has declined to about a billion degrees .
a hot neutron star with a radius of @xmath21 km lies near the center of this supernova .
its kelvin - helmholtz phase of cooling by neutrino emission lasts about 10 s , during which time the gravitational binding energy of the final neutron star , a few times @xmath22 erg , is radiated approximately equally in @xmath23 , @xmath24 , @xmath25 , @xmath26 , @xmath27 , and @xmath28 ( janka 1995 ) . therefore , the average luminosity for a specific neutrino species at this late time is of order 10@xmath29@xmath30 erg s@xmath31 .
as this neutrino flux streams through the region above the neutron star , the following weak interactions can take place : absorption of @xmath23 and @xmath24 , neutrino - electron scattering , and neutrino - antineutrino annihilation .
the last two are experienced by all flavors of neutrinos .
these interactions transfer energy from neutrinos to the matter above the neutron star and heat it .
consequently , this material expands away from the neutron star and develops into a mass outflow we call the neutrino - driven wind . " intuitively ,
once the convection associated with the launching of the shock has ceased , one expects this mass outflow to be relatively smooth , and adequately described by a quasi - steady - state approximation .
in fact , this expectation is verified by recent two - dimensional numerical simulations ( janka and mller 1995 ) .
this is because during the kelvin - helmholtz cooling phase of the neutron star , ( i ) the global characteristics of the neutrino flux , e.g. , luminosity and energy distribution , change rather slowly ; ( ii ) the properties of the neutron star , e.g. , mass , radius , and surface temperature , also evolve very slowly ; and ( iii ) the supernova shock wave is at very large radii , and has little influence on the conditions close to the neutron star .
such is the ideal world .
in fact , the real local situation in the wind may be more complex .
woosley & weaver ( 1995 ) find considerable mass continues to accrete onto the neutron star for several hours following its formation .
convection may also continue at some level even 10 s after the shock is launched .
rotation may lead to a breaking of spherical symmetry .
however , in order to proceed , we shall adopt the steady - state assumption , in which case the dynamic equations can be written as ( duncan et al .
1986 ) : @xmath32 where @xmath33 is the rest mass density , @xmath34 is the outflow velocity , @xmath35 is the temperature , @xmath36 and @xmath37 are the total pressure and specific internal energy corresponding to non - relativistic matter and relativistic particles in the outflow , respectively , @xmath38 is the net specific heating rate due to neutrino interactions above the neutron star , @xmath39 is the mass of the neutron star , and @xmath40 is the constant mass outflow rate in the ejecta . in writing the above equations ,
we have made the following assumptions : ( i ) the outflow velocity is non - relativistic ; ( ii ) the general relativistic effects above the neutron star are small ; and ( iii ) the neutron star is the predominant source of gravity .
we will discuss the validity of these assumptions later .
as we will see in @xmath17 and @xmath18 , the characteristics of the neutrino - driven wind are mostly determined at @xmath41 mev . at these temperatures ,
the material is composed of non - relativistic free neutrons and protons , relativistic electrons and positrons , and photon radiation .
the expression for pressure , @xmath36 , and internal energy , @xmath37 , can then be approximately written as @xmath42 where @xmath43 is the nucleon rest mass , and @xmath44 is the degeneracy parameter with @xmath45 the electron chemical potential .
the degeneracy parameter , @xmath46 , is related to the electron fraction @xmath6 through @xmath47 in equations ( 4 ) to ( 6 ) , we have used the units in which the planck constant @xmath48 , the speed of light @xmath49 , and the boltzmann constant @xmath9 are taken to be unity .
it is easy to obtain the correct dimension of a physical quantity if one knows that @xmath50 mev fm .
for example , equation ( 6 ) in normal units would be @xmath51 .
the electron fraction @xmath6 is , in turn , determined by @xmath52 where @xmath53 , @xmath54 , @xmath55 , and @xmath56 are the rates for the forward and reverse reactions in the following equations : @xmath57 given the input neutrino physics ( i.e. , the expressions of @xmath38 , @xmath53 , @xmath54 , @xmath55 , and @xmath56 ) , the initial conditions at the neutron star surface , and the boundary conditions at the shock wave , we can think of equations ( 1 ) through ( 7 ) as a complete set of equations for an `` eigenvalue '' problem of @xmath40 in the neutrino - driven wind .
sometimes it is convenient to think of these equations as describing a lagrangian mass element moving away from the neutron star with velocity @xmath58 . in this case
, we can introduce a time variable @xmath59 , with @xmath60 being some initial reference radius .
then the derivative with respect to @xmath61 is equivalent to @xmath62 .
it is helpful to estimate the general conditions before a detailed description of the neutrino - driven wind .
the ejecta leave the neutron star surface with a small , very subsonic initial velocity . in order to escape to large radii
, a nucleon has to gain enough energy from the neutrino flux to overcome its gravitational potential at the neutron star surface . for a typical neutron star ,
the mass is @xmath63 , and the radius is @xmath64 km .
the amount of energy provided by neutrino heating for a nucleon has to be at least @xmath65 mev . because the neutrino flux decreases as @xmath66 away from the neutron star
, we expect that most of the heating takes place close to the neutron star .
the surface temperature of a nascent neutron star is several mev , and the thermal kinetic energy of a nucleon close to the neutron star is of the same order .
since the initial velocity of the nucleon is also small , the nucleon is incapable of carrying the amount of energy obtained from the neutrino flux .
almost all of this energy has to go into photon radiation and relativistic electron - positron pairs .
this is consistent with the neutrino heating processes . in absorption of @xmath23 and @xmath24 ,
essentially all the neutrino energy goes into the produced electron or positron .
neutrino - antineutrino annihilation produces electron - positron pairs , and neutrino - electron scattering also transfers neutrino energy directly to electrons and positrons .
therefore , the ejecta become radiation - dominated a short distance above the neutron star .
the energy initially stored in photon radiation and electron - positron pairs is converted into the mechanical energy of the nucleons at much larger radii where temperatures are low . in terms of the local thermodynamic conditions
, the dominance of radiation means that @xmath67 and @xmath68 .
this is clear from equations ( 4 ) to ( 6 ) . under these conditions , it is convenient to introduce a thermodynamic quantity @xmath69 where @xmath70 is the temperature in mev , and @xmath71 is the density in @xmath72 g @xmath73 .
it is easy to see from equations ( 4 ) to ( 6 ) that @xmath74 is the entropy per baryon in relativistic particles for @xmath75 .
the ejecta become radiation - dominated when @xmath76 .
we now calculate the heating and cooling rates resulting from interactions between the neutrino flux and material in the ejecta .
we assume that neutrinos are emitted from a neutrinosphere with radius @xmath77 . at radius @xmath78 ,
one only sees neutrinos within the solid angle subtended by the neutrinosphere at this radius . because the neutrino interaction cross sections have a power law dependence on neutrino energy , the heating rates can be cast in terms of the neutrino luminosity and various neutrino energy moments , without specifying a particular neutrino energy distribution .
our approach here thus parallels pioneering analytic calculations of the supernova mechanism by bethe ( e.g. , 1993 ) .
the most important heating and cooling processes are those given in equations ( 8a ) and ( 8b ) , i.e. , neutrino absorption and electron capture on free nucleons . the specific heating rate due to neutrino absorption is @xmath79 { 1-x\over r_{\nu6}^2 } \ { \rm mev}\ { \rm s}^{-1}\ { \rm g}^{-1},\end{aligned}\ ] ] where @xmath80 is a function of radius , @xmath81 is the avogadro s number , @xmath82 is the individual neutrino luminosity in @xmath83 erg s@xmath31 , @xmath84 is the neutrinosphere radius in @xmath85 cm , and @xmath86 is an appropriate neutrino energy @xmath87 in mev , defined through @xmath88 , where @xmath89 denotes the @xmath90th neutrino energy moment of the neutrino energy distribution .
the specific cooling rate due to capture of relativistic electrons and positrons is @xmath91 in deriving the above rates , we have neglected the neutron - proton mass difference and the pauli blocking effects for leptons in the final state .
we have also used the fact that there are almost equal numbers of relativistic electrons and positrons in the radiation - dominated ejecta .
neutrinos of all flavors can scatter on the electrons and positrons in the ejecta .
these scattering processes also contribute to the heating .
the corresponding specific heating rate is @xmath92 where @xmath93 is @xmath94 in mev , and where we have assumed the same characteristics for the @xmath25 , @xmath26 , @xmath27 , and @xmath28 fluxes . in deriving equation ( 12 ) ,
we have also made the approximation that the average energy transfer from neutrino to electron is @xmath95 for each scattering , instead of the more accurate value @xmath96 found in tubbs and schramm ( 1975 ) .
this is a reasonable approximation at @xmath97 a few mev where the neutrino - electron scattering processes are expected to be most effective .
cooling can also occur through annihilation of relativistic electron - positron pairs into @xmath98 , @xmath99 , and @xmath100 pairs .
the corresponding specific cooling rate is @xmath101 of course , there are other cooling processes operating through emission of neutrino - antineutrino pairs .
however , electron - positron pair annihilation into neutrino - antineutrino pairs is the most important one for the radiation - dominated conditions in the ejecta .
finally , we give the specific heating rate for neutrino - antineutrino annihilation into electron - positron pairs : @xmath102{\phi(x)\over\rho_8r_{\nu6}^4},\end{aligned}\ ] ] where @xmath103 .
it is also useful to give the total heating rate due to neutrino - antineutrino annihilation above the neutrinosphere : @xmath104r_{\nu6}^{-1}\ { \rm erg } \ { \rm s}^{-1}.\ ] ] the heating and cooling rates presented above correspond to the major heating and cooling mechanisms for the ejecta . from these rates
, we can make the following observations : \(1 ) the specific heating rate due to neutrino - antineutrino annihilation decreases rapidly away from the neutrinosphere , as is evident from the expression for @xmath105 . at @xmath106 , @xmath107 .
physically , this is because ( i ) the neutrino and antineutrino fluxes decrease as @xmath66 ; ( ii ) the chance for a neutrino to collide with an antineutrino , especially the occurrence of almost head - on collisions decreases quickly away from the neutrinosphere ; and ( iii ) the center - of - mass energy of the neutrino - antineutrino pair , and hence , the available phase space for the produced electron - positron pair , also decrease very fast away from the neutrinosphere .
unfortunately , our prescription for the neutrino flux does not work well close to the neutrinosphere .
therefore , the rates given for this process can only be taken as a rough estimate .
extensive studies of the heating rate due to neutrino - antineutrino annihilation were carried out by janka ( 1991a , b ) using monte carlo methods . in the subsequent ( sub)sections
, we will discuss to what extent the simple approximations made in deriving equations ( 14 ) and ( 15 ) will affect our understanding of the neutrino - driven wind .
\(2 ) the relative importance of neutrino - scattering and neutrino absorption can be seen from the ratio @xmath108 , where @xmath109 mev is the temperature at the neutrinosphere .
therefore , neutrino - electron scattering become important only at high entropies .
\(3 ) similarly , the relative importance of the cooling processes can be gauged by the ratio @xmath110 . however , both cooling processes have sensitive temperature dependence , and the absolute cooling rates decrease rapidly away from the neutron star surface .
we refer to the conditions at radius @xmath111 , where we start our analytic treatment of the neutrino - driven wind , as the initial conditions at the neutron star surface .
these conditions are closely related to the heating and cooling processes taking place in the region between the neutrinosphere at radius @xmath77 and the neutron star surface at radius @xmath111 . in this region ,
the heating and cooling processes between the intense neutrino flux and material at high temperature and density proceed at very high rates .
this results in a kinetic equilibrium between the neutrino flux and the material ( burrows & mazurek 1982 ) , and the temperature in this region almost stays constant .
the material in this region is also in close hydrostatic equilibrium , and we have @xmath112 for slowly varying temperature , @xmath113 because previous deleptonization has led to @xmath114 at the neutrinosphere , we can neglect the first term in the above equation , and obtain @xmath115 for @xmath116 .
taking @xmath117 , @xmath118 km , and @xmath109 mev , we see that the density decreases by one @xmath119-fold over a distance scale @xmath120 km . as a result ,
the entropy in relativistic particles @xmath121 quickly rises .
the pressure and specific internal energy of the material become dominated by contributions from the relativistic particles when @xmath122 . at this point
, the temperature has to decrease sufficiently fast to break the kinetic equilibrium in order to maintain the approximate hydrostatic equilibrium .
the cooling rates decrease with the falling temperature , and net heating takes place to drive a mass outflow , and to further increase the entropy .
this picture is then consistent with our expectation of the general radiation - dominated condition in the ejecta .
our initial conditions then correspond to the last radius where kinetic equilibrium can still be maintained . at this radius @xmath111 , @xmath122 , so the dominant cooling process is electron capture on free nucleons .
since the rate for this process has a sharp temperature dependence , the temperature @xmath123 at radius @xmath111 is relatively insensitive to the uncertainties in the rates for the counteracting heating processes .
we can then take neutrino absorption on free nucleons as the dominant heating process , and further approximate the corresponding rate in equation ( 10 ) as @xmath124 in deriving the above rate , we have made the following approximations : ( i ) @xmath125 ; ( ii ) @xmath126 ; and ( iii ) @xmath127 .
the last approximation follows from the observation that @xmath128 ranges from 25/24 to 5/4 for neutrino energy distributions of the form @xmath129 $ ] with @xmath130 and @xmath131 .
the temperature @xmath123 at radius @xmath111 is then @xmath132^{1/6}{l_{\bar\nu_e,51}^{1/6 } \over r_6^{1/3}}\epsilon_{\bar\nu_e,\rm mev}^{1/3 } \
{ \rm mev}.\ ] ] for @xmath133 erg s@xmath31 , @xmath134 mev , @xmath135 , @xmath136 0.51 , and @xmath137 cm , we have @xmath138 mev
. we can now compare the rates @xmath139 and @xmath140 with @xmath141 . as we have discussed in @xmath142 , @xmath143 . for @xmath122 and @xmath144 , @xmath145 .
taking @xmath146 erg s@xmath31 and @xmath147 cm , we obtain @xmath148 .
clearly , taking into account the contributions from neutrino - electron scattering and neutrino - antineutrino annihilation only increases @xmath123 by a small amount .
however , to guard against miscalculation of the heating rates , especially @xmath140 , we can multiply @xmath141 by a correction factor @xmath149 .
this then gives the temperature @xmath123 at radius @xmath111 as @xmath150 where we have assumed an average value for @xmath151^{1/6}$ ] . in the unlikely case where we have underestimated @xmath140 by more than one order of magnitude
, we note that a correction factor @xmath152 only increases @xmath123 by @xmath153% .
equation ( 21 ) then constitutes our basic initial condition at the neutron star surface .
before we start our analytic treatment of the neutrino - driven wind , we note that equations ( 1 ) to ( 3 ) describe the hydrodynamic and thermodynamic evolution of the ejecta , while equation ( 7 ) determines the chemical evolution of the ejecta .
the former equations are coupled with the latter only through the dependence on @xmath6 in the equations of state ( 4 ) to ( 6 ) and the expression of @xmath38 . from the previous discussions in this section , we know that the ejecta become radiation - dominated immediately above the neutron star surface .
equations ( 4 ) and ( 5 ) are approximately reduced to @xmath154 which are independent of @xmath6 .
the expression of @xmath38 has a weak dependence on @xmath6 , mainly through equation ( 10 ) .
we have eliminated this dependence by taking @xmath125 in @xmath155 .
this is justified because @xmath156 for any @xmath6 as long as @xmath157 and @xmath158 . at any rate ,
we have introduced a correction factor @xmath149 in @xmath155 to account for the uncertainties in deriving the expression of @xmath38 .
therefore , we can separate the discussion of the ejecta into two almost independent parts . in this subsection
, we derive the analytic expressions for the entropy , the dynamic time scale , and the mass outflow rate in the ejecta .
these physical parameters are determined by equations ( 1 ) to ( 3 ) .
we devote @xmath19 to a detailed analysis of the electron fraction in the ejecta .
taking advantage of equations ( 22 ) and ( 23 ) , we can rearrange equations ( 1 ) to ( 3 ) into @xmath159 where @xmath160 is the adiabatic sound speed in the ejecta , given by @xmath161 for a particular lagrangian mass element , the above equations describe the evolution of its velocity , total `` flow '' energy , and entropy .
we note that the total flow energy is given by the sum of enthalpy and mechanical energy , i.e. , @xmath162
. given the initial conditions at the neutron star surface , i.e. , @xmath35 and @xmath74 at radius @xmath111 , we can choose any three independent equations from equations ( 1 ) to ( 3 ) and ( 24 ) to ( 26 ) , and integrate these equations for various values of the mass outflow rate @xmath40 .
the eigenvalue of @xmath40 is then selected to meet the boundary conditions at the shock wave . however , the qualitative features of the solutions to these equations can be readily obtained from equations ( 24 ) to ( 26 ) . from equations ( 25 ) and ( 26 ) , we see that @xmath163 and @xmath74 always increase as long as @xmath164 .
neutrino heating becomes negligible for @xmath165 mev because in the region at these temperatures , ( i ) free nucleons are bound into @xmath14-particles and heavier nuclei ; ( ii ) electron - positron pairs annihilate into photons ; and ( iii ) the neutrino flux has decreased by a factor @xmath166 . therefore , @xmath163 and @xmath74 reach their final values at @xmath167 mev .
we note that equations ( 22 ) and ( 23 ) are invalid for @xmath168 mev due to the disappearance of electron - positron pairs .
however , because entropy is conserved during the annihilation of electron - positron pairs into photons , we can show that after all the electron - positron pairs are gone , the correct equations to describe the ejecta are @xmath169 where @xmath170 and @xmath171 are the final values of @xmath163 and @xmath74 at @xmath167 mev , respectively , and @xmath172 .
the behavior of @xmath34 is more complicated . at the neutron star surface
, we have @xmath173 and @xmath174 . according to equation ( 24 )
, @xmath34 always increases initially .
the subsequent evolution of @xmath34 depends on the mass outflow rate @xmath40 .
given the initial conditions at the neutron star surface , the initial velocity is proportional to @xmath40 . for small values of @xmath40 , the ejecta move slowly and
there is more time for the ejecta to gain energy from the neutrino flux .
the right - hand side of equation ( 24 ) can become zero due to the increase in @xmath163 before @xmath34 reaches @xmath160 .
when this occurs , @xmath175 , and @xmath34 reaches its maximum value .
the velocity decreases afterwards as the right - hand side of equation ( 24 ) becomes positive . in this case , the ejecta are always subsonic .
when @xmath40 increases , it can happen that the right - hand side of equation ( 24 ) becomes zero at the same time as @xmath34 reaches @xmath160 .
this case corresponds to a critical value @xmath176 of @xmath40 . in this critical case
, the velocity increases through the sound speed to supersonic values , and eventually becomes a constant when all the flow energy is converted into the mechanical kinetic energy .
mass outflow rates larger than @xmath176 are unphysical , because for these values of @xmath40 , @xmath34 will reach @xmath160 when the right - hand side of equation ( 24 ) is still negative , resulting in an unphysical infinite acceleration .
therefore , we only need to focus our attention on cases of @xmath177 . from equations ( 1 ) and ( 30 )
, we see that @xmath178 and @xmath179 as @xmath34 approaches its final value in the critical case .
therefore , @xmath176 can only correspond to vanishing boundary conditions at large radii . on the other hand , in the cases of @xmath180
, one can impose a boundary pressure , or equivalently a boundary temperature , in the radiation - dominated ejecta .
the supernova shock wave at large radii can be regarded as such a boundary condition . as @xmath35 in the ejecta
approaches the boundary temperature @xmath181 , @xmath33 also approaches a constant value , and @xmath34 decreases as @xmath66 , according to equations ( 1 ) and ( 30 ) .
however , for @xmath182 mev , there is little difference between the subsonic structures of the ejecta for the critical and subcritical mass outflow rates . in the subcritical case , @xmath183 everywhere , and we can approximate equation ( 29 ) as @xmath184 in particular , this equation is also satisfied at @xmath167 mev because @xmath170 corresponds to the flow energy at @xmath167 mev . if @xmath182 mev , we can take @xmath185 as an effective boundary condition . in the critical case
, we have @xmath186 when @xmath187 . equation ( 29 ) then gives @xmath188 where @xmath189 is the temperature when @xmath34 reaches @xmath160 . if @xmath190 mev , i.e. , @xmath167 mev belongs to the subsonic region , we obtain the same effective boundary condition as in equation ( 32 )
. therefore , the mass outflow rates are close to @xmath176 for small boundary temperatures , resulting in similar subsonic structures to that in the critical case . typical boundary temperatures for the supernova shock wave are @xmath191 mev .
therefore , the mass outflow rate in the neutrino - driven wind corresponds to the effective boundary condition in equation ( 32 ) . as we have stated before
, our discussion of the physical conditions in the ejecta can be limited to the region at @xmath41 mev .
the above discussion of the neutrino - driven wind is mainly based on equations ( 24 ) to ( 26 ) .
strictly speaking , these equations are valid only when @xmath192 , where @xmath193 is the entropy per baryon in non - relativistic nucleons . the number 11 in the above expression for @xmath194 could be 12 if one distinguishes between neutrons and protons , and takes account of their spin states .
however , the qualitative features given in the above discussion apply to the neutrino - driven wind as long as the pressure @xmath36 and the specific energy @xmath37 are dominated by the contributions from electron - positron pairs and photon radiation , i.e. , @xmath195 . in this case
, the analytic treatments of the neutrino - driven wind are very similar for both cases of @xmath192 and @xmath196 , if we use the general forms of equations ( 25 ) and ( 26 ) : @xmath197 equations ( 35 ) and ( 36 ) can be easily derived from equations ( 2 ) to ( 5 ) with the approximation @xmath68 .
we first derive the final entropy @xmath171 in the ejecta . from equation ( 35 ) , we have @xmath198 where we have used the effective boundary condition in equation ( 32 ) and have taken the initial flow energy to be @xmath199 . from equation ( 36 ) , we have @xmath200 we can define an effective temperature @xmath201 through @xmath202 or equivalently , @xmath203 we can regard equation ( 39 ) as defining @xmath204 by weighing @xmath205 over @xmath38 , and expect @xmath201 approximately corresponds to the temperature at which @xmath38 reaches the maximum value . taking @xmath206 , we find @xmath207 at @xmath201 , where we have used the same correction factor introduced in equation ( 21 ) to crudely account for the heating processes other than neutrino absorption on free nucleons .
we note that in general this correction factor @xmath149 should vary with radius , for example , due to the increasing importance of @xmath139 relative to @xmath141 at higher entropies .
however , for the purpose of simple analytic treatments , we will take @xmath149 to be constant and subsequently derive its suitable value to be used in the analytic estimates for the entropy per baryon , the dynamic time scale , and the mass outflow rate in the wind . because hydrostatic equilibrium approximately holds in the subsonic region , we have @xmath208 where we have introduced a numerical factor @xmath209 to account for the contribution to @xmath36 from non - relativistic free nucleons .
for @xmath210 , @xmath211 , and for @xmath212 , @xmath213 .
equations ( 41 ) and ( 42 ) then give @xmath214 where @xmath215 and @xmath216 are the radius and entropy , respectively , corresponding to @xmath201 .
hereafter , we use the subscript `` eff '' to denote parameters at the radius where @xmath38 reaches the maximum value . to proceed further
, we make the following approximation : @xmath217 and @xmath218 the numerical factors in equations ( 44 ) and ( 45 ) are conveniently chosen to reflect that the concerned quantities at @xmath215 are in the `` middle '' of some characteristic evolution .
using equations ( 40a ) , ( 40b ) , ( 44 ) , and ( 45 ) , we find @xmath219 substituting equations ( 45 ) , ( 46a ) , and ( 46b ) into equation ( 43 ) , we have @xmath220 combining equations ( 21 ) , ( 40a ) , ( 40b ) , ( 47a ) , and ( 47b ) , we obtain @xmath221 the total entropy per baryon @xmath222 in the ejecta is @xmath223 where we have evaluated @xmath194 at @xmath167 mev using @xmath171 and taking advantage of the logarithmic dependence of @xmath194 on @xmath35 .
we can also derive an approximate value for the correction factor @xmath149 .
assuming that the main heating process other than neutrino absorption on free nucleons is neutrino - electron scattering , we can take the correction factor to be @xmath224 from equations ( 12 ) and ( 19 ) , we obtain @xmath225 where we have used equations ( 45 ) , ( 46a ) , and ( 46b ) to evaluate @xmath226 in equation ( 12 ) , and where we have assumed @xmath227 . typically , we have @xmath228 , @xmath229 ( 0.51)@xmath230 , and @xmath231 mev .
so numerically , @xmath232/ ( \epsilon_{\nu_e,\rm mev}^2+\epsilon_{\bar\nu_e,\rm mev}^2)\approx 0.1.$ ] next , we derive the mass outflow rate @xmath40 in the ejecta .
the total net neutrino heating rate above @xmath215 is given by @xmath233 where @xmath234 to estimate the integral in equation ( 52 ) , we need the density scale height @xmath235 at @xmath215
. we can relate @xmath235 to the corresponding temperature scale height @xmath236 by @xmath237 for the radiation - dominated ejecta ( @xmath122 ) . in turn
, @xmath236 can be estimated from equation ( 42 ) as @xmath238 now the integral in equation ( 52 ) can be approximated as @xmath239 where we have used equation ( 9 ) to express @xmath240 in terms of @xmath201 and @xmath216 . from equation ( 35 )
, we have @xmath241 where we have used equation ( 45 ) to evaluate @xmath242 . combining equations ( 52 ) , ( 53 ) , ( 56 ) , and ( 57 ) , we obtain @xmath243 where we have also used equations ( 21 ) , ( 46a ) , ( 46b ) , ( 47a ) , and ( 47b ) .
the total power of net neutrino heating @xmath244 is approximately given by @xmath245 finally , we derive the dynamic time scale in the ejecta .
we define this time scale to be @xmath246 from equations ( 1 ) , ( 9 ) , ( 32 ) , ( 48a ) , ( 48b ) , ( 58a ) , and ( 58b ) , we find @xmath247 for both cases of @xmath210 and @xmath212 . the time scale derived in equation ( 61 ) will prove important for heavy element nucleosynthesis in the neutrino - driven wind .
we illustrate its usefulness as follows .
we expect @xmath248 , or equivalently @xmath249 , to approximately hold between @xmath167 mev and @xmath250 mev .
if we ignore the effect of annihilation of electron - positron pairs into photons , we roughly have @xmath251 and @xmath252 , which leads to @xmath253 for a constant mass outflow rate @xmath254 .
therefore , the time for the temperature to decrease from @xmath255 mev to , say 0.2 mev , is @xmath256 . between the radius where the temperature reaches @xmath257 mev for the first time and the position of the shock wave ,
both the temperature and density of the ejecta remain approximately constant . according to @xmath258 , we have @xmath259 , from which the time dependence of the radius for a particular lagrangian mass element can be obtained with @xmath260 at the initial radius for @xmath261 . to summarize , we have derived the total entropy per baryon , the mass outflow rate , and the relevant dynamic time scale for heavy element nucleosynthesis in the neutrino - driven wind . in particular , we have shown that these physical conditions for heavy element nucleosynthesis only depend on the neutrino luminosity and energy distributions , and the mass and radius of the neutron star .
in order to test some of the approximations derived in the previous section , a series of calculations were carried out using the one - dimensional implicit hydrodynamic code , kepler ( weaver , zimmerman , & woosley 1978 ) .
this lagrangian - coordinate - based code has an appropriate equation of state for the ions , radiation , electrons , and pairs under all conditions of interest here .
for example , this equation of state adequately treats electrons in any arbitrarily relativistic and degenerate regime .
it is most important that the code is implicitly differenced and thus capable of accurately following the evolution in a situation where , as here , tight hydrostatic equilibrium exists .
though the code has provision for radiation transport , energy transport other than by neutrinos is negligible here .
only the outer layers of the neutron star were carried , approximately 0.01 @xmath262 .
the remainder of the neutron star was represented by a boundary condition of given radius , gravitational potential , and neutrino luminosity .
typical densities and temperatures at the basis of the problem were 10@xmath263 g @xmath73 and 810 mev .
the wind of course originated at much lower densities and temperatures in a small amount of mass .
also important was the code s ability to smoothly and repeatedly rezone the material in the wind as its density decreased .
typical zoning in the wind was about @xmath264 .
the velocity structure and , in the asymptotic region , the mass outflow rate , @xmath265 , were , at all times following a start - up transient , well behaved and smooth . on the other hand , the neutrino physics employed was primitive by the standard of modern supernova codes .
the inner boundary acted as a light bulb " and the overlying layers were assumed to subtract nothing from the flux , i.e. , the optically thin limit was assumed .
the neutrino luminosity was shared equally among the six varieties and each neutrino species was characterized by its mean energy and flux . for a neutrinosphere radius of @xmath266 km ( also the approximate neutron star radius , @xmath111 ) ,
the assumed mean energies were @xmath267 = 12 , 22 , 34 mev for @xmath23 , @xmath268 , and @xmath269 , respectively .
the corresponding values for @xmath270 km were 16 , 20 , 32 mev .
these neutrinos could interact with nucleons and pairs ( and with one another ) in the neutron star crust and wind according to equations ( 10 ) , ( 11 ) , ( 12 ) , and ( 14 ) .
neutrino losses due to pair annihilation , plus plasma and photoneutrino processes were calculated separately assuming @xmath271 and the fitting functions of munakata , koyama , & itoh ( 1985 ) .
this is more accurate than equation ( 13 ) . nuclear physics , other than the neutrino interactions , was very simple .
since in the region of interest , nuclear statistical equilibrium ( nse ) exists and the composition is solely neutrons , protons , and @xmath14-particles , only these three constituents were carried .
since the neutrino interactions with the two nucleon species are similar only one species was carried .
the transition from nucleons to @xmath14-particles , which effectively shuts off the neutrino - nucleon interaction ( neutrino - electron scattering and neutrino - antineutrino annihilation still contribute ) was approximated using the following prescription for the free nucleon mass fraction : @xmath272 or unity , whichever was the smaller ( woosley & baron 1992 ) .
the free nucleon mass fraction @xmath273 given by this prescription was used to multiply the rates in equations ( 10 ) and ( 11 ) to give the capture rates of neutrinos and pairs on nucleons .
nine models were then calculated . in each model
the neutron star was represented by an inner boundary condition of given interior mass , radius , and neutrino luminosity .
the parameters employed as well as sample results are given in table 1 and figs . 1 and 2 .
each model was constructed by allowing a coarsely - zoned version of the neutron star crust and atmosphere to come into approximate hydrostatic equilibrium with the inner boundary conditions .
this typically took a few milliseconds .
then the zoner was turned on .
both adzoning and dezoning were employed .
this acted to give smooth temperature and density profiles .
typically in the steady - state solution there were about 200 zones and about a dozen zones per decade of density in the region of interest .
typical zone masses were @xmath264 except near the inner boundary where the zoning smoothly increased on a logarithmic scale to @xmath274 .
zones were removed from the surface of the problem , typically when the density had declined to about 1 g @xmath73 so as to inhibit a runaway accumulation of zones .
each problem was run long enough that a steady state had clearly been attained .
for total neutrino luminosities of 10@xmath275 erg s@xmath31 this took about one second . for higher luminosities ,
the corresponding time was about 0.2 s. here we point out that our numerical approach to model the neutrino - driven wind differs from that of duncan et al .
( 1986 ) . their numerical model of the neutrino - driven wind was based on the wind equations ( 1 ) to ( 3 ) , and mathematically was an eigenvalue problem looking for the mass outflow rate @xmath40 . in our approach , we started from some initial configuration of the material above the neutron star surface , and evolved the mass outflow into a steady - state neutrino - driven wind . therefore , not only did we obtain the mass outflow rate @xmath40 and other physical parameters in the steady state , but also showed that a steady state could actually be reached over some relaxation time scale of the problem .
the relaxation time scale could then be compared with the evolution time scales of the inner boundary ( especially the neutrino luminosity ) to establish the quasi - steady - state nature of the neutrino - driven wind . for typical decay time of @xmath276 several seconds for the neutrino luminosity ,
the steady - state approximation in our analytic treatments of the neutrino - driven wind is valid . because the code did not include general relativistic corrections to gravity , two different baryon masses , 1.4 and 2.0 @xmath262 , were employed for the neutron star .
the smaller value is more typical of the gravitational mass of neutron stars and the baryon mass of many presupernova model calculations ( e.g. , timmes , woosley , & weaver 1996 ) , but the latter gives a more realistic gravitational potential , at least near the neutron star surface , when the neutron star has approximately reached its final radius of about 10 km .
table 1 compares numerical and analytic results for the total power of net neutrino heating , the mass outflow rate , the final total entropy per baryon , and the dynamic time scale in the neutrino - driven wind . as we can see
, the analytic results for these parameters agree with the numerical output from kepler within a factor of 2 in all nine models .
the agreement between the numerical and analytic results is the best for the final entropy per baryon in the wind , as can be expected from the shallow dependence on the correction factor @xmath149 in equations ( 48a ) and ( 48b ) .
this correction factor indicates the uncertainties in the analytic treatment of various neutrino heating processes . in our numerical models ,
neutrino absorption on free nucleons is the main heating process .
neutrino - electron scattering can make significant contribution to the heating for the cases with high entropies , i.e. , the models with @xmath277 km .
for all nine models , neutrino - antineutrino annihilation contributes at most @xmath278 to the total heating .
while the general agreement between numerical and analytic results is important to show , we also highlight the predicted power - law dependences on the neutrino luminosity and mean energy , and the mass and radius of the neutron star for the physical parameters listed in table 1 . as can be easily verified from the numerical results in table 1 , the power - law dependences on the neutrino luminosity for these physical parameters follow the analytic predictions with good accuracy .
the models of the previous subsection were all calculated using a simple newtonian theory of gravity .
a neutron star is of course a relativistic system . to zeroth order the gravity of the neutron star
can simply be increased , e.g. , by turning up its mass , as was done in models 10d , 10e , and 10f .
however , the actual radial dependence of gravity differs from @xmath66 in general relativity . to check that this does not have an appreciable effect on the mass outflow rate , entropy , and dynamic time scale ,
two additional models were calculated using a version of kepler modified to include first - order post - newtonian corrections to the gravitational force ( e.g. , fuller , woosley , & weaver 1986 ; shapiro & teukolsky 1983 ) .
the gravitational term in the force equation ( weaver , zimmerman , & woosley 1978 ) , @xmath279 , was replaced by @xmath280 where @xmath281 is the mass included within a radius @xmath0 .
the term corresponding to @xmath38 in the energy equation was unchanged .
this includes the leading post - newtonian corrections to the wind dynamics ( with the wind velocity always much less than @xmath49 ) .
unfortunately it does not include the gravitational effects on the neutrino energy spectra and angular distributions .
models 10b and 10e were then recalculated using the modified gravitational field .
the former had a neutron star mass at infinity of 1.4 @xmath262 and a total neutrino luminosity of @xmath282 erg s@xmath31 .
in steady state , the post - newtonian model had a mass loss rate of @xmath283 @xmath262 s@xmath31 , an asymptotic entropy of 116 , and a dynamic time scale of 0.044 s. compared to quantities in table 1 , we see that the effect of the modified gravitational potential is to increase the entropy by about 1/3 ( from 87 to 116 ) and decrease the dynamic time scale by about 1/3 .
the post - newtonian version of model 10b in fact resembles model 10e , though perhaps an effective mass of 1.8 @xmath262 instead of 2 @xmath262 might have been more appropriate .
we also considered the extreme case of model 10e ( neutron star mass at infinity of 2.0 @xmath262 ) _ and _ post - newtonian corrections .
neglecting general relativistic corrections to the neutrino transport is clearly wrong here . while allowed by causality , this is also a heavier neutron star than commonly derived in binary systems .
we consider this as an example of the state a neutron star might pass through while evolving to a black hole on a kelvin - helmholtz cooling time scale .
the asymptotic entropy for this model is 205 , the mass loss rate is @xmath284 @xmath262 s@xmath31 , and the dynamic time scale is 0.035 s ; i.e. , a 60% increase in the entropy and about a factor of 2 reduction in the dynamic time scale . in summary ,
the effect of post - newtonian corrections to gravity is to increase the entropy and shorten the dynamic time scale by substantial amounts . in this regard ,
a stronger gravitational potential is more favorable to the @xmath0-process .
except in the case of the accretion - induced collapse of a neutron star the wind will not be able to flow unencumbered to infinity .
it will be impeded by external matter and radiation in that portion of the supernova through which the shock has already passed .
woosley & weaver ( 1995 ) show that the pressure behind the shock will be largely due to radiation in a nearly isothermal sphere . at a time of several seconds
this temperature is very approximately two billion k. it declines to under one billion k by the time the neutron star is 10 seconds old . to illustrate the effect ,
we have recalculated model 10e with an external boundary pressure appropriate to a temperature of @xmath285 k. this is an extreme case ; a realistic model would have less boundary pressure at 10 s. the converged model was run for one second with this boundary pressure and reached an approximate steady state though the outer boundary was being gradually pushed out at 200 km s@xmath31 ( corresponding to @xmath286 work of @xmath287 erg s@xmath31 ) . figure 3 shows that there is little change in the structure of the wind , except near its outer boundary .
the asymptotic entropy was increased from 129 to 140 and the mass loss rate reduced from @xmath288 to about @xmath289 @xmath262 s@xmath31 .
the dynamic time scale ( at 0.5 mev ) was lengthened from 0.066 s to 0.11 s. in summary , the effect of a moderate external boundary pressure is to increase the entropy slightly , and decrease the expansion rate appreciably .
overall there is little modification of the parameters of the wind in the region where the @xmath0-process occurs except that the material has a longer time to capture neutrons .
the slower time scale at temperatures of 3 to 5 billion k will also give a smaller neutron - to - seed ratio in the ejecta .
this is an adverse effect , but not a large one .
1 and 2 show , the mass outflow rate is determined closer to the neutron star than the entropy .
one consequence of this is that any attempt to increase the entropy of the wind simply by raising the overall neutrino heating rate is thwarted by an increased outflow rate .
the extra energy goes into doing additional work against the gravity of the neutron star and not into raising the entropy of the wind .
this is also to be expected from the dependences of @xmath74 and @xmath40 on the neutrino luminosity and the correction factor @xmath149 in equations ( 48a ) , ( 48b ) , ( 58a ) and ( 58b ) : @xmath290 vs. @xmath291 .
conversely , energy that is added after the mass loss rate has already been determined can have a very beneficial effect , both in raising the entropy and increasing the velocity , both good for the @xmath0-process .
energy that is added too late however , after the temperature has already gone below two billion k , has little nucleosynthetic effect .
the neutron - to - seed ratio for the @xmath0-process has already been set .
this suggests that an additional energy source between about 20 and 50 km might have considerable leverage on the nucleosynthesis in the wind .
possible sources of this energy are discussed later in the paper ( @xmath20 ) .
we thus explored the effect of adding an additional 10@xmath292 erg s@xmath31 to model 10e at a radius between 20 and 30 km .
the total energy deposited thus became @xmath293 erg s@xmath31 instead of @xmath294 erg s@xmath31 .
of this 10@xmath292 erg s@xmath31 , roughly 10% has a physical origin in the recombination of nucleons to @xmath14-particles ( @xmath295 erg g@xmath31 times @xmath40 ) .
this occurs at @xmath296 mev or a radius of about 1520 km .
the mass outflow rate in the modified model increased slightly from @xmath288 @xmath262 s@xmath31 to @xmath297 @xmath262 s@xmath31 , and the asymptotic entropy rose substantially , from 129 to 190 .
most importantly the expansion time scale at 0.5 mev decreased from 0.066 to 0.010 s. the velocity at 0.5 mev was 3500 km s@xmath31 .
these are conditions quite favorable to the @xmath0-process ( hoffman et al .
1996b ) . model 10f in fig .
4 shows what a relatively small amount of energy will do if it can be generated in the right place and time .
an additional volume energy term ( @xmath298 erg @xmath73 s@xmath31 ) was added for radii between 15 and 25 km giving a total added energy of @xmath299 erg s@xmath31 , a moderate perturbation on the @xmath300 erg s@xmath31 the neutrinos were already adding . as a result ,
the mass loss rate increased a little to @xmath301 @xmath262 s@xmath31 , the entropy climbed to 192 and the dynamic time scale at 0.5 mev shrank to 0.022 s. these conditions should be quite favorable for the @xmath0-process for @xmath302 ( hoffman et al .
1996b ) .
larger energy deposition was not explored as it would no longer be a perturbation on the neutrino - driven wind , but the driving term for the mass loss .
of the three interactions that deposit energy in the wind neutrino capture on nucleons , neutrino scattering on electrons , and neutrino - antineutrino annihilation the last is probably worst approximated in our calculations .
the radial dependence of equation ( 14 ) is very steep .
detailed and better treatment of this dependence was given by janka ( 1991a , b ) .
multidimensional and multi - group neutrino transport calculations by jim wilson ( 1996 ) show that within two neutron star radii , the approximation in equation ( 14 ) is probably a gross underestimate of the actual heating rate due to neutrino - antineutrino annihilation .
even larger deviations may be expected when a fully general relativistic transport calculation is ultimately done .
the effects of general relativity on neutrino transport calculations were also discussed by janka ( 1991a ) . for someone seeking high entropy and rapid expansion neutrino - antineutrino annihilation has the attractive feature that it deposits energy at a rate independent of the local matter density .
it can thus boost the specific energy and entropy of a wind after its density has already declined .
the problem is the steep fall - off of the neutrino - antineutrino annihilation rate with radius and its low efficiency . for model 10e , of the @xmath294 erg s@xmath31 being deposited in the steady - state model ,
about 10% is due to neutrino - antineutrino annihilation .
we thus experimented with varying the radial dependence and efficiency of neutrino - antineutrino annihilation . in one calculation model 10e
was run with an additional 10@xmath292 erg s@xmath31 deposited as a constant volume term ( @xmath303 erg @xmath73 s@xmath31 ) between 10 and 20 km ( the surface of the neutron star and twice its radius ) . as a result ,
the asymptotic entropy increased from 129 to 137 , the mass outflow rate climbed from @xmath304 to @xmath305 @xmath262 s@xmath31 , and the dynamic time scale at 0.5 mev declined from 0.066 s to 0.040 s. a second calculation maintained the functional form of equation ( 14 ) , but multiplied the neutrino - antineutrino annihilation efficiency by a factor of six inside 20 km .
this resulted in increases of the energy deposition to @xmath306 erg s@xmath31 and mass outflow rate to @xmath307 , and decreases in entropy to 127 and dynamic time scale to 0.061 s. finally , we explored , again in model 10e , changing the radial dependence of equation ( 14 ) .
the term @xmath308 in @xmath105 was raised to the third power instead of the fourth and the overall efficiency multiplied by 3 ( at all radii ) .
this raised the energy deposition to @xmath309 erg s@xmath31 , the entropy to 134 , the mass outflow rate to @xmath305 @xmath262 s@xmath31 , and decreased the dynamic time scale to 0.044 s. we conclude that reasonable variations in the efficiency of neutrino - antineutrino annihilation can give a modest increase in the entropy and decrease in the dynamic time scale .
however , very large increases in the entropy , like a doubling , would require more radical alterations than we think are justified at the present time .
unfortunately any energy source that is peaked near the neutron star surface tends to increase the mass loss rate more than the entropy .
our analytic solutions assume a steady state in which the neutrino luminosity does not vary . in the real situation
the local neutrino flux in the outflowing wind may vary considerably because of rotation , accretion , and convective flows .
we thus explored the effect on model 10e of varying the total neutrino luminosity on a time scale comparable to the flow time across the wind zone .
in particular , we take the total neutrino luminosity to be @xmath310 $ ] erg s@xmath31 with @xmath311 = 0.1 s. much shorter time scales would have had no effect .
much longer time scales would have given the steady - state solutions previously described .
the chief effect of this variation in the neutrino luminosity was to cause a periodic variation in the outflow velocity and dynamic time scale .
the entropy , e.g. , at 0.5 mev , did not change greatly from 129 in the steady - state model , being 127 @xmath312 3 during three oscillations .
but the outflow velocity varied almost linearly with the neutrino luminosity so that the dynamic time scale at 0.5 mev oscillated between 0.044 and 0.15 s ( the steady - state value for model 10e was 0.066 s )
. a similar calculation in which @xmath311 = 0.2 s gave the same range of dynamic time scales , but with an entropy oscillating between 115 and 136 . in both cases
the larger entropy was associated with the faster outflow .
note that an inverse dependence of the expansion time scale on the neutrino luminosity is predicted by equation ( 61 ) for the steady - state wind .
we conclude that reasonable variations in the neutrino luminosity on intervals of order 0.1 s can give a large range of expansion time scales at nearly constant entropy .
thus it is possible to have , in a fraction of the ejecta , material which has experienced the high entropy appropriate to a low time - averaged neutrino luminosity , but with the rapid expansion time scale characteristic of the temporary peaks .
in this section , we describe how the electron fraction @xmath6 is determined in the neutrino - driven wind .
as we will see , the final @xmath6 in the ejecta is principally set by the characteristics of the @xmath23 and @xmath24 fluxes . in this regard
, we also discuss general aspects of neutrino emission in supernovae , and present a novel neutrino two - color plot " to illustrate the time evolution of the @xmath23 and @xmath24 energy distributions .
to conclude the section , we discuss the implications of this evolution plot for heavy element nucleosynthesis in supernovae .
we begin our discussion by calculating the rates for the forward and reverse reactions in equations ( 8a ) and ( 8b ) .
these reactions are the most important processes which set the electron fraction @xmath6 in the neutrino - driven wind .
the cross sections for the forward reactions in equations ( 8a ) and ( 8b ) are given by @xmath313 where @xmath314 , @xmath315 and @xmath316 are the momentum and total energy , respectively , of the electron or positron in the final state , and where we have made the approximation @xmath317 .
this approximation is very good because ( i ) the energy of the final state electron in equation ( 8a ) is at least the neutron - proton mass difference ; and ( ii ) in any case , the typical neutrino energy is about 10 mev or more , which makes the final state electrons and positrons extremely relativistic . following the prescription for the neutrino flux in @xmath142 , we find the rates for the forward reactions in equations ( 8a ) and ( 8b ) to be @xmath318 where @xmath319 mev is the neutron - proton mass difference . at @xmath320 , we can write the above rates as @xmath321 where @xmath322 is @xmath323 in mev , and where we have taken an average value of @xmath324 .
in fact , @xmath325 ranges from 16/15 to 4/3 for neutrino energy distributions of the form @xmath326 $ ] with @xmath130 and @xmath131 .
we have taken into account the neutron - proton mass difference to give accurate values of the above reaction rates .
as we will see in @xmath327 , the final @xmath6 in the ejecta is sensitive to the ratio of these reaction rates .
the cross sections for the reverse reactions in equations ( 8a ) and ( 8b ) are given by @xmath328 where @xmath329 is the neutrino energy in the final state .
the rates for these reactions are @xmath330 where we have neglected the neutron - proton mass difference , and have assumed that the initial state electrons and positrons are extremely relativistic .
these approximations are reasonable at @xmath331 mev , and become invalid when @xmath35 approaches 0.5 mev .
however , these reaction rates are negligible compared with those in equations ( 65a ) and ( 65b ) at @xmath296 mev .
this is because at these low temperatures , ( i ) the number density of electron - positron pairs decreases significantly ; and ( ii ) the cross sections decrease rapidly , especially so for electron capture on proton , which has to overcome the neutron - proton mass difference .
therefore , the breaking - down of the above approximations has no serious consequences for setting the final value of @xmath6 in the ejecta . as noted in @xmath16 , the electron fraction @xmath6 in the ejecta
is governed by equation ( 7 ) .
we can rewrite this equation as @xmath332 where @xmath333 , @xmath334 , and @xmath335 .
regardless of the particular forms of @xmath336 and @xmath337 , the general solution to the above equation is given by @xmath338i(0,t)+ { \lambda_1(t)\over\lambda_2(t)}-\int_0^ti(t^\prime , t){d\over dt^\prime } \left[{\lambda_1(t^\prime)\over\lambda_2(t^\prime)}\right]dt^\prime,\ ] ] where @xmath339 is taken as the time when the ejecta leave the neutron star surface at radius @xmath111 , and where @xmath340 is a `` memory '' function of the interaction history , with @xmath341 the total number of interactions on a pair of neutron and proton between @xmath342 and @xmath61 ( qian 1993 ) .
it is easy to show that the first term on the right - hand side of equation ( 69 ) quickly vanishes at @xmath343 .
the gravitational binding energy of a nucleon at the neutron star surface is @xmath344 mev .
a nucleon has to obtain at least this amount of energy from the neutrino fluxes in order to escape to large radii .
the main heating reactions , @xmath23 and @xmath24 absorption on free nucleons , are also responsible for determining @xmath6 . from each interaction with the @xmath23 or @xmath24 flux , an amount of energy @xmath2120 mev per nucleon is absorbed .
therefore , a nucleon in the ejecta has to have at least 10 interactions above the neutron star surface , and the total number of interactions on a pair of neutron and proton in the ejecta satisfies @xmath345dt>20,\ ] ] where @xmath346 is a symbolic time when @xmath337 becomes negligibly small . from the discussions in @xmath347 , we know that most of the heating interactions take place at temperatures around @xmath348 mev .
so for @xmath349 mev , we can safely assume @xmath350 and neglect the first term on the right - hand side of equation ( 69 ) .
making use of the relation @xmath351 we can perform integration by parts in the third term on the right - hand side of equation ( 69 ) , and obtain @xmath352\ ] ] when @xmath350 . in deriving the above equation ,
we have assumed @xmath353\right| \ll{\lambda_1(t)\over\lambda_2(t)},\ ] ] and therefore , have neglected terms of higher orders in @xmath354 , i.e. , the remaining integral @xmath355 \right\}dt^\prime.\ ] ] the condition in equation ( 74 ) is called the quasi - equilibrium ( qse ) condition .
this is because under this condition , @xmath356 where @xmath357 is the instantaneous equilibrium value at time @xmath61 . from the rates and discussions in @xmath358
, we see that for @xmath296 mev , @xmath336 and @xmath337 are dominated by the rate(s ) for neutrino absorption on free nucleons .
consequently , @xmath359 , and @xmath6 assumes a constant equilibrium value @xmath360 for @xmath296 mev ( qian et al .
. the @xmath361 given by equation ( 77 ) would be the final value of @xmath6 in the ejecta if equations ( 7 ) and ( 68 ) were valid for all temperatures .
however , as we have discussed before , free nucleons begin to be bound into @xmath14-particles and heavier nuclei at @xmath362 mev ( cf .
therefore , equations ( 7 ) and ( 68 ) are to be replaced by @xmath363 at @xmath362 mev , where @xmath364 and @xmath365 . in deriving equation ( 78 ) ,
we have neglected the rates for neutrino absorption on @xmath14-particles and heavier nuclei and all the electron capture rates , compared with the neutrino absorption rates on free nucleons .
we have also assumed @xmath366 , where @xmath367 is the mass fraction of free nucleons , with @xmath368 and @xmath369 the mass fractions of protons and neutrons , respectively .
this approximation for @xmath6 is good for material mainly composed of free nucleons and @xmath14-particles
. equation ( 78 ) is mathematically identical to equation ( 68 ) . at the transition from equation ( 68 ) to ( 78 ) , we have @xmath370 and @xmath371 . from the general solution given in equation ( 69 )
, we can easily show that @xmath372 provided the qse condition @xmath373\right|\ll{\lambda_1^\prime(t ) \over\lambda_2^\prime(t)}\ ] ] is satisfied . from the expression for @xmath374
, we see that the qse condition in equation ( 80 ) depends on how fast @xmath273 is changing . in turn , the rate at which @xmath273 changes is determined by the thermodynamic and hydrodynamic conditions in the ejecta ( cf .
[ 62 ] ) .
typically , the qse condition in equation ( 80 ) is no longer met when @xmath375 at this point , the neutrino absorption reactions are not frequent enough to change @xmath6 any more , and @xmath6 `` freezes '' out at the last achievable instantaneous equilibrium value @xmath376 therefore , if the qse condition for equation ( 78 ) is still satisfied when a significant fraction of the material is in @xmath14-particles and heavier nuclei , the final @xmath6 in the ejecta will deviate from @xmath361 given in equation ( 77 ) .
the influence of @xmath14-particles and heavier nuclei on @xmath6 in the ejecta was first pointed out by fuller & meyer ( 1995 ) . because the evolution of @xmath6 in the ejecta at @xmath362 mev is coupled with the change in the nuclear composition of the ejecta , and depends sensitively on the thermodynamic and hydrodynamic conditions in the ejecta ( cf .
[ 62 ] and [ 81 ] ) , we will follow @xmath6 numerically in the nucleosynthesis calculations to be presented in a separate paper ( hoffman et al . 1996b )
. however , whether the final @xmath6 in the ejecta is given by equation ( 77 ) or ( 82 ) , mathematically , @xmath377 corresponds to @xmath378 barring the unlikely case where the qse condition for equation ( 78 ) holds until all the free nucleons disappear .
the qualitative feature of heavy element nucleosynthesis depends crucially on whether @xmath379 or @xmath380 . from equations ( 65a ) , ( 65b ) , and ( 83 ) , we see that neutrino luminosities and energy distributions have direct bearings on the nature of heavy element nucleosynthesis in supernovae . as the neutrinos diffuse towards the edge of the neutron star , their interactions with matter decrease with the decreasing temperature and density .
gradually , they decouple from the thermodynamic equilibrium with the neutron star matter , keeping energy distributions corresponding to the conditions ( e.g. , temperature , density , chemical composition ) in the decoupling region . because only energy - exchange processes are directly responsible for achieving thermodynamic equilibrium , and different neutrino species have different ability to exchange energy with matter
, the thermodynamic decoupling processes of different neutrino species occur in different regions of the neutron star . in turn , each neutrino species emerges from the neutron star with its own characteristic energy distribution .
all neutrino species exchange energy with electrons .
in fact , the neutral - current scattering on electrons is the main energy - exchange process for @xmath25 , @xmath26 , @xmath27 , and @xmath28 . on the other hand , although @xmath23 and @xmath24 have the extra charged - current scattering on electrons which @xmath25 , @xmath26 , @xmath27 , and @xmath28 lack , they exchange energy with matter even more efficiently through the charged - current absorption reactions on nucleons given in equations ( 8a ) and ( 8b ) .
therefore , @xmath25 , @xmath26 , @xmath27 , and @xmath28 decouple at higher temperature and density inside the neutron star than @xmath23 and @xmath24 , and correspondingly , their energy distributions are harder than those of @xmath23 and @xmath24 .
as we have seen in @xmath327 , the luminosities and energy distributions for @xmath23 and @xmath24 have significant effects on @xmath6 , or equivalently , the neutron excess in the wind .
in addition , emission of @xmath23 and @xmath24 is coupled with the deleptonization process of the neutron star , which makes the problem of @xmath23 and @xmath24 transport in the neutron star even more complicated . in what follows ,
we give a more detailed discussion of the time evolution of @xmath23 and @xmath24 emission in supernovae . shortly after the creation of the neutron star by the core collapse and subsequent supernova explosion , its neutron concentration is larger than , but still comparable to its proton concentration . because @xmath23 and @xmath24 exchange energy with matter mainly through absorption on neutrons and protons , respectively , @xmath23 can stay in thermodynamic equilibrium with the neutron star matter down to somewhat lower temperatures and densities than @xmath24 .
in other words , @xmath24 decouple at slightly higher temperatures than @xmath23 , and have a slightly larger average energy , i.e. , @xmath381 . the luminosities for @xmath23 and @xmath24
are about the same , so the number flux of @xmath23 is larger than that of @xmath24 .
this results in a net electron lepton number flux leaking out of the neutron star , which reflects the net deleptonization process of converting protons into neutrons through electron capture inside the neutron star .
meanwhile , the neutron star is shrinking to its final equilibrium configuration .
the heating caused by the compression raises the temperature of the neutron star interior . as a result , both @xmath382 and @xmath383 increase initially while keeping @xmath384 . because of the diffusive nature of the neutrino transport processes
, we expect to see the strongest deleptonization in the surface layers of the neutron star .
as these surface layers become more and more neutron rich due to the deleptonization , the decoupling regions for @xmath23 and @xmath24 are progressively separated from each other .
the deleptonization effects on @xmath383 and @xmath382 are most pronounced when the neutron star approximately reaches its final radius and the compressional heating becomes negligible . at this point
, @xmath385 still increases because @xmath24 decouple deeper with the deleptonization . on the contrary , @xmath383 begins to decrease as @xmath23 decouple further out in the neutron star due to the increasing neutron concentration in the surface layers .
schematically , we can illustrate the time evolution of the @xmath23 and @xmath24 energy distributions on a two - dimensional plot .
we choose the abscissa and ordinate to be @xmath386 and @xmath230 , respectively , because these energies are more directly related to our discussion of @xmath6 in the ejecta ( cf .
[ 65a ] and [ 65b ] ) . borrowing a term from optical photometry , we refer to this plot as the `` neutrino two - color plot . '' for a particular supernova model , the time evolution of @xmath386 and @xmath230 is represented by a contour line on this plot . taking @xmath386 and @xmath230 as functions of time from fig . 3 in woosley et al .
( 1994 ) , we present an example plot in fig .
5 . the solid contour line in fig . 5 shows the evolution track of @xmath386 and @xmath230 .
the open circles on this track indicate the time after the onset of the collapse in intervals of approximately 1/3 s for @xmath3874 s and approximately 1 s for @xmath38818 s. time @xmath387 corresponds to the open circle at the lower left corner of the plot , and time increases thereafter along the track to @xmath389 s at the last open circle .
the evolution of @xmath386 and @xmath230 slows down considerably at @xmath390 s. as a result , there are two overlapping open circles at @xmath391 , 14 , and 17 s ( i.e. , the open circles for @xmath392 , 15 , and 18 s can not be distinguished ) . from this plot , we can see that @xmath386 and @xmath393 both increase for @xmath3871 s , and @xmath386 starts to decrease while @xmath230 increases for @xmath394 s , in agreement with our previous discussion .
the evolution of @xmath386 and @xmath230 during @xmath3952 s is more complicated and may only be understood by more involved argument than our previous qualitative discussion .
numerical supernova calculations show that shortly after the explosion , the ratio @xmath396 always stays close to 1 ( see , e.g. , fig . 2 in woosley et al .
therefore , without specific detailed neutrino transport calculations , we can divide the neutrino two - color plot into three regions corresponding to @xmath380 , @xmath125 , and @xmath379 in the ejecta .
the region corresponding to @xmath125 is bounded by the range of values for @xmath396 , and is determined by equation ( 83 ) . in fig .
5 , we give an example situation where this region is bounded by @xmath397 from below and by @xmath398 from above .
the parameter space to the left of this region corresponds to @xmath380 , and the parameter space for @xmath379 lies to the right of this region . in general
, our neutrino two - color plot can be presented similarly for any supernova model .
the implications of this plot for heavy element nucleosynthesis in supernovae are discussed in the next subsection . as discussed in @xmath399 ,
a severe failure of @xmath0-process calculations in the neutrino - driven wind is the overproduction of nuclei in the vicinity of @xmath1 at @xmath400 s after the supernova explosion .
hoffman et al . ( 1996a ) have shown that this failure can be turned into an attractive feature if @xmath6 in the ejecta is slightly increased from the value in wilson s supernova model to @xmath7 , for which some light @xmath5-process nuclei can be produced . from fig . 5
, we see that the evolution trajectory of @xmath230 and @xmath386 for @xmath400 s lies entirely in the region of @xmath125 on the neutrino two - color plot .
this implies that some small uncertainties in the supernova model predictions for @xmath230 , @xmath386 , and/or @xmath401 , @xmath402 may explain the overproduction of the nuclei in the vicinity of @xmath1 .
it also hints that future supernova models with more accurate neutrino transport calculations may not encounter the overproduction problem , but would facilitate the nucleosynthesis of some light @xmath5-process nuclei instead . on the other hand
, we see that @xmath230 and @xmath386 predicted by wilson s supernova model will clearly give @xmath380 in the neutrino - heated ejecta for @xmath403 s. for any @xmath0-process nucleosynthesis to take place in the ejecta , we must have @xmath380 .
therefore , the neutrino two - color plot in fig . 5 at least signifies the possibility of an @xmath0-process for @xmath403 s in wilson s supernova model , even in the presence of some small uncertainties in the predicted neutrino characteristics .
while the exact values of @xmath6 in the ejecta depend on the accuracy of supernova neutrino transport calculations on the one hand , and the thermodynamic and hydrodynamic conditions in the ejecta on the other hand , we feel that the neutrino two - color plot has some interesting implications for the nature of supernova nucleosynthesis .
perhaps the nucleosynthesis of some light @xmath5-process nuclei can occur at @xmath400 s , and the @xmath0-process nucleosynthesis is produced later when the neutrino characteristics evolve to the neutron - rich region of the neutrino two - color plot .
the principal conclusions of our paper are simple analytic expressions , verified by comparison to numerical simulation , for the entropy ( eqs .
[ 48a ] and [ 48b ] ) , mass loss rate ( eqs . [ 58a ] and [ 58b ] ) , and dynamic time scale ( eq . [ 61 ] ) in the neutrino - driven wind .
we have also given a detailed analysis of how @xmath6 is determined ( @xmath19 ) in the wind .
we began this work in the hope that a simple physical model would yield the conditions required by woosley et al .
( 1994 ) for the @xmath0-process , in particular rapid expansion time scales and entropies of order 300 or more .
while pursuing this quest , a great deal was learned about the properties of neutrino - driven winds , which eventually came to appear quite simple .
mutually confirming analytic and numerical calculations give the solutions in table 1 .
these models span a range of gravitational potential and neutrino luminosity that should typify the kelvin - helmholtz evolution of most neutron stars . unfortunately
, none of the models in table 1 will give a good @xmath0-process unless @xmath6 is very low ( hoffman et al
. 1996b ) .
nor can one continually increase the entropy by turning the neutrino luminosity down .
the entropy only increases as the sixth root of @xmath404 while the mass loss rate declines as @xmath405 ( assuming that the neutrino energy distributions stay more or less the same over the period of interest , cf . fig .
moreover , the dynamic time scale becomes unacceptably long .
we conclude that one of the following must be true : ( 1 ) the ( heavy ) @xmath0-process isotopes are not made in the neutrino - powered wind of young neutron stars ( of the type given in table 1 ) ; ( 2 ) the nuclear physics of the @xmath0-process is at fault ( presently unlikely ) ; or ( 3 ) important physics has been left out of the models in table 1 .
we shall spend the remainder of the paper discussing the third possibility .
what has been left out ?
first , there are effects that we know exist and can estimate general relativity and nucleon recombination . in @xmath4064.2
we provided numerical calculations in a post - newtonian approximation to show the effects of general relativity .
any effect that strengthens the gravitational potential leads to higher entropy in the wind ( @xmath4063.3.1 ) .
the essential consequences of general relativity on our models can be obtained simply by increasing the effective mass of the neutron star e.g. , models 10def vs. models 10abc .
we also considered , however , in a post - newtonian calculation , the extreme case of a neutron star having a radius of 10 km and a gravitational mass ( at infinity ) of 2 @xmath262 .
this was the only calculation without an artificial energy source that gave an entropy over 200 .
perhaps the heavy @xmath0-process is made in neutron stars that are on their way to becoming black holes .
however , one can not extend this approach indefinitely . at some point ,
not much beyond the extreme case considered , as the gravitational potential deepens , the neutron star becomes unstable for any equation of state and turns into a black hole on a dynamic ( collapse ) time scale .
whether sufficient matter is ejected from such objects to make the solar @xmath0-process abundances remains to be determined .
the recombination of nucleons into @xmath14-particles releases an amount of energy of about @xmath407 mev per nucleon .
assuming most of the nucleons recombine at @xmath408 mev , we get an additional entropy of about @xmath409 .
this was not included in either the numerical or analytic models and should be added to all the entropies calculated in the paper .
our treatment of neutrino - antineutrino annihilation is very approximate ( cf .
janka 1991a , b ) and there are expectations that a more realistic treatment might lead to a moderate increase in the entropy ( @xmath4064.5 ) and reduction in the dynamic time scale .
our opinion is that this is too small a change to restore the @xmath0-process .
it is rather unfortunate that we did not get or understand the high entropies obtained in wilson s supernova model used by woosley et al .
( 1994 ) for the @xmath0-process calculations .
we note that wilson s supernova model derives from a fully general relativistic hydrodynamic code with a consistent and detailed treatment of neutrino transport .
however , the origin of the discrepancies in entropy between our study and wilson s numerical calculations remains unknown and is still under investigation . in @xmath4064.6 we considered the effect of a time variable luminosity and found that , while the entropy did not change much , considerable leverage could be exercised on the expansion time scale .
perhaps the @xmath0-process is made in those regions of the wind that begin their ejection at low neutrino luminosity , but absorb additional energy and expand faster as the local neutrino luminosity increases . indeed , the dynamic time scale is an important third parameter not sufficiently emphasized in previous studies of the @xmath0-process in the wind . for a given @xmath6 , increasing the neutron - to - seed ratio can be achieved either by reducing the reaction rates that assemble @xmath14-particles into heavy nuclei , i.e. , by reducing the density ( raising the entropy ) or by decreasing the time during which these reactions can operate .
since the rate of @xmath14-burning goes approximately as @xmath410 ( woosley & hoffman 1992 ) , one can obtain approximately the same @xmath0-process at one - half the requisite entropy if the time scale is about 8 times shorter .
one general result of our study is that it is easier to achieve a more vigorous outflow than it is to increase the entropy , thus to some extent the quest for @xmath411 300 may be misguided .
entropies of @xmath412 , which are readily achievable may function equally well if the expansion time scale can be cut by an order of magnitude to @xmath2760.01 s. the most effective way of accelerating the expansion , and , as it turns out , of increasing the entropy , is to provide an additional source of energy of order 10@xmath292 erg s@xmath31 at between 1.5 and 3 times the neutron star radius .
this is a region where the mass loss rate has , for the most part , already been determined , but the freeze - out of @xmath14-particles from nse ( @xmath413 k ) has not yet occurred .
the effect of a purely arbitrary energy source was explored in @xmath4064.4 .
the origin of such energies brings us to effects that might be there , but which are very difficult to estimate .
first , we know that fall - back will occur as the supernova explosion develops .
convection becomes less important to the explosion mechanism as time passes , but material is still decelerated to below the escape speed as the shock interacts with the stellar mantle .
woosley & weaver ( 1995 ) find that typically @xmath2760.1 @xmath262 falls back in about 10@xmath414 s ( more earlier than later ) , so one expects this accreted material to be adding mass faster than the wind is removing it .
however , the hydrodynamic interaction is complex .
it is uncertain whether _ any _ material finds its way back to the neutron star during the first 20 s of interest .
but the entropy of the infalling material is much lower and it may descend in plumes while the wind rises in large bubbles .
it may even be that the wind is derived from the accreted material rather than the neutron star itself .
it is difficult to guess what the actual interaction will be of the plumes with the neutron star , but it may be that the wind starts with an initial entropy higher ( from the accretion shock ) than the steady - state value derived here .
electron capture near the neutron star would also influence the starting value of @xmath6 .
a ten - second - old neutron star will also be the site of continuing energetic activity other than its neutrino emission .
in particular , the star may be violently vibrating , rotating rapidly , or have a strong magnetic field .
small oscillations at the base of the developing neutron star crust will steepen into shocks in the steep density gradient at the neutron star surface ( woosley 1996 ) .
this must certainly occur at some level , but whether it leads to @xmath415 erg s@xmath31 being deposited at greater than 1.5 neutron star radii for a ten - second - old neutron star is unclear .
the magnetic field configuration near the surface of a newly born neutron star is very uncertain both its spatial structure and strength but there are speculations ( duncan & thompson 1992 ) that the magnetic field may be very large , @xmath416 gauss . since the outer layers of the neutron star have recently been convective , the field structure near the surface may be highly tangled .
a 10@xmath417 gauss field could confine , or at least impede the escape of plasma having a temperature of 4 mev , i.e. , greater than the neutrinosphere temperature at late times .
such a field could have several effects .
to the extent that it impedes the outflow of the wind , its effect resembles that of a stronger gravitational potential .
higher entropy is a likely outcome . to the extent the field reconnects at higher altitudes ,
because of rotational shearing or stretching in the wind , it provides the desired extra heating . flow in the field
may also develop clumps , the analogue of photon bubbles " studied in x - ray pulsars ( klein & arons 1989 , 1991 ) .
if the dynamic time scale and entropy are inadequate , one may have to turn to a significant lowering of @xmath6 .
as discussed in @xmath19 , @xmath6 in the ejecta is determined by neutrino emission characteristics of the neutron star . in turn , the neutrino luminosity and energy spectrum are determined by neutrino interactions in the hot dense neutron star ( @xmath418 ) .
any uncertainty in our understanding of neutrino interactions in hot dense matter causes an inaccurate prediction of the emission properties . for example
, some recent studies ( keil , janka , & raffelt 1995 ; sawyer 1995 ) suggest that the strength of neutrino interactions is reduced by the medium response of nuclear matter , with respect to the case of neutrino interactions with a single nucleon . although it is not clear whether the effects found by these studies on neutrino characteristics can change @xmath6 in the ejecta significantly , a complete picture of supernova nucleosynthesis may await better knowledge of neutrino interactions in hot dense matter .
finally , the neutrino characteristics above the neutron star surface can be altered by the process of neutrino flavor transformation .
for example , if @xmath25 or @xmath27 has a cosmologically significant mass in the range of 1100 ev and the mass of @xmath23 is much smaller , then @xmath25 or @xmath27 will be transformed into @xmath23 above the neutron star surface through the mikheyev - smirnov - wolfenstein ( msw ) mechanism ( fuller et al .
1992 ; qian et al .
because @xmath25 and @xmath27 have a higher average energy than @xmath24 , conversion of @xmath25 or @xmath27 into @xmath23 will increase @xmath6 in the ejecta ( cf .
such flavor transformation is undesirable ( qian et al . 1993 ) . on the other hand ,
if @xmath23 has a mass of a few ev and @xmath25 or @xmath27 is much lighter , then @xmath26 or @xmath28 will be transformed into @xmath24 above the neutron star surface ( qian 1993 ; qian & fuller 1995 ) . in this case ,
conversion of @xmath26 or @xmath28 into @xmath24 can significantly reduce @xmath6 in the ejecta ( fuller , qian , & wilson 1996 ) . in view of the recent claim of possible evidence for @xmath419 oscillations by the lsnd experiment ( athanassopoulos et al .
1995 ) and other astrophysical evidence for massive neutrinos ( see , e.g. , fuller , primack , & qian 1995 ) , this latter process of flavor transformation is worth further exploring .
we are grateful to rob hoffman for making the first version of the neutrino two - color plot for us , and for providing us with the results of nuclear reaction network calculations for various conditions .
we thank the referee ( h .- thomas janka ) and adam burrows for helpful comments on improving the paper .
we also want to thank wick haxton , craig hogan , jim wilson , and especially george fuller for interesting and informative conversations on the subject of the paper .
qian thanks sterl phinney and kip thorne for reading the draft of the paper and caltech for a fellowship during the final stage of this work .
this work was supported by the department of energy under grant no .
de - fg06 - 90er40561 at the institute for nuclear theory , where y .- z .
qian was a postdoctoral research associate .
woosley was supported by nsf grants no .
ast 91 - 15367 and no .
ast 94 - 17161 at ucsc , and by an alexander von humboldt stiftung in germany .
athanassopoulos , c. et al .
1995 , , 75 , 2650 bethe , h. a. 1993 , , 412 , 192 burrows , a. & mazurek , t. j. 1982 , , 259 , 330 duncan , r. c. , shapiro , s. l. , & wasserman , i. 1986 , , 309 , 141 duncan , r. c. & thompson , c. 1992 , , 392 , l9 fuller , g. m. , mayle , r. , meyer , b. s. , & wilson , j. r. 1992 , , 389 , 517 fuller , g. m. & meyer , b. s. 1995 , , 453 , 792 fuller , g. m. , primack , j. r. , & qian , y .- z .
1995 , , 52 , 1288 fuller , g. m. , qian , y .- z . , &
wilson , j. r. 1996 , in preparation fuller , g. m. , woosley , s. e. , & weaver , t. a. 1986 , , 307 , 675 hoffman , r. d. , woosley , s. e. , fuller , g. m. , & meyer , b. s. 1996a , , 460 , 478 hoffman , r. d. , woosley , s. e. , qian , y .- z .
, & fuller , g. m. 1996b , in preparation howard , w. m. , goriely , s. , rayet , m. , & arnould , m. 1993 , , 417 , 713 janka , h .- th .
1991a , ph.d .
thesis , mpa - report 587 janka , h .- th .
1991b , , 244 , 378 janka , h .- th .
1995 , astroparticle physics , 3 , 377 janka , h .- th . & mller , e. 1995 , , 448 , l109 keil , w. , janka , h .- th .
, & raffelt , g. 1995 , , 51 , 6635 klein , r. i. & arons , j. 1989 , in proc . of the 23rd eslab symposium in x - ray astronomy , vol . 1 , eas sp-296 , ed . n. white & t. guyenne ( paris : european space agency ) , 89 klein , r. i. & arons , j. 1991 , in stellar atmospheres : beyond classical models , ed . l. crivillari ( boston : kluwer ) , 205 meyer , b. s. 1995 , , 449 , l55 meyer , b. s. , howard , w. m. , mathews , g. j. , woosley , s. e. , & hoffman , r. d. 1992 , , 399 , 656 munakata , h. , kohyama , y. , & itoh , n. 1985 , , 296 , 197 , see also ( errata ) , 304 , 580 qian , y .- z . 1993 , ph.d .
thesis , university of california , san diego qian , y .- z . , fuller , g. m. , mathews , g. j. , mayle , r. w. , wilson , j. r. , & woosley , s. e. 1993 , , 71 , 1965 qian , y .- z . &
fuller , g. m. 1995 , , 52 , 656 sawyer , r. f. 1995 , , 75 , 2260 shapiro , s. l. , & teukolsky , s. a. 1983 , _ black holes , white dwarfs and neutron stars _ , ( wiley & sons : new york ) , p. 125
takahashi , k. , witti , j. , & janka , h .- th .
1994 , , 286 , 857 timmes , f. x. , woosley , s. e. & weaver , t. a. 1996 , , 457 , 834 tubbs , d. l. & schramm , d. n. 1975 , , 201 , 467 weaver , t. a. , zimmerman , g. b. , & woosley , s. e. 1978 , , 225 , 1021 wilson , j. r. 1996 , private communication witti , j. , janka , h .- th . , & takahashi , k. 1994 , , 286 , 841 woosley , s. e. 1993 , , 97 , 205 woosley , s. e. 1996 , in preparation woosley , s. e. & baron , e. 1992 , , 391 , 228 woosley , s. e. & hoffman , r. d. 1992 , , 395 , 202 woosley , s. e. , wilson , j. r. , mathews , g. j. , hoffman , r. d. , & meyer , b. s. 1994 , , 433 , 229 * fig . 1 * conditions in the numerical model for a neutrino - driven wind assuming newtonian physics , a total neutrino luminosity in all flavors of @xmath420 erg s@xmath31 , and a neutron star mass and radius of 2.0 @xmath262 and 10 km , respectively ( model 10e ) .
the top frame gives the dimensionless entropy per baryon @xmath222 ( long dashed line ) , rate for net energy deposition by neutrinos , @xmath38 , in 10@xmath421 erg g@xmath31 s@xmath31 ( short dashed line ) , and mass outflow rate , @xmath422 , in units of 10@xmath423 @xmath262 s@xmath31 ( solid line ) , as functions of radius in km .
note the abrupt decline in @xmath38 at 35 km as the nucleons recombine .
the bottom frame gives the density , @xmath33 , in 10@xmath424 g @xmath73 ( solid line ) , temperature , @xmath35 , in 10@xmath425 k ( long dashed line ) , and outflow velocity , @xmath34 , in 10@xmath426 cm s@xmath31 ( short dashed line ) , all on the same radial scale .
2 * conditions in the numerical model for a neutrino - driven wind from a neutron star of 1.4 @xmath262 with a radius of 30 km and a total neutrino luminosity of @xmath427 erg s@xmath31 ( model 30b ) . notation and quantities edited are the same as in fig .
1 , except that the density is in 10@xmath426 g @xmath73 and the mass outflow rate is in units of 10@xmath428 @xmath262 s@xmath31 for display purposes . *
3 * effect of a boundary pressure equivalent to a boundary temperature of @xmath429 k in model 10e .
the total entropy , temperature , and outflow velocity are given as functions of radius .
the solid lines are for model 10e without the boundary pressure , and the corresponding conditions for the same model with the boundary pressure are shown as dashed lines . * fig .
4 * conditions in the numerical model for a neutrino - driven wind assuming newtonian physics , a total neutrino luminosity of @xmath430 erg s@xmath31 , and a neutron star mass and radius of 2.0 @xmath262 and 10 km , respectively ( model 10f ) , when an additional ( artificial ) energy source of @xmath299 erg s@xmath31 is evenly spread ( by volume ) between 15 and 25 km ( 1.5 and 2.5 neutron star radii ) . notation and quantities edited are the same as in fig
5 * neutrino two - color plot .
the time evolution of the @xmath24 and @xmath23 mean energies , @xmath230 and @xmath386 , respectively , predicted by wilson s supernova model , is shown as a solid contour on the plot .
the open circles on this contour indicate the time after the onset of the collapse from @xmath431 at the lower end to @xmath389 s at the upper end .
time increases along the contour in intervals of approximately 1/3 s for @xmath3874 s and approximately 1 s for @xmath38818 s. note that due to the slow evolution of @xmath230 and @xmath386 at @xmath390 s , there are two overlapping open circles at @xmath391 , 14 , and 17 s ( i.e. , the open circles for @xmath392 , 15 , and 18 s can not be distinguished ) .
three regions , separated by the two dashed lines , correspond to values of @xmath230 and @xmath386 which would give @xmath379 , @xmath432 , and @xmath380 , respectively , in the neutrino - driven wind . see text for detailed explanation .
|
during the first 20 seconds of its life , the enormous neutrino luminosity of a neutron star drives appreciable mass loss from its surface .
previous investigations have shown that this neutrino - driven wind could be the site where the @xmath0-process occurs .
the nucleosynthesis is sensitive to four physical parameters characterizing the wind : its mass loss rate , the entropy per baryon , the electron fraction , and the dynamic time scale .
different authors , using numerical models for supernovae , have arrived at qualitatively different values for these key parameters . here
we derive their values analytically and test our analytic results by numerical calculations using an implicit hydrodynamic code . employing our analytic and numerical methods , we also investigate how various factors can affect our results .
the derived entropy typically falls short , by a factor of two to three , of the value required to produce a strong @xmath0-process .
various factors that might give a higher entropy or a more rapid expansion in the wind are discussed .
|
lepton flavor violating ( lfv ) tau decays provide one of the most efficient indirect tests of supersymmetric ( susy ) models with extended neutrino sector , if the seesaw mechanism for neutrino mass generation is implemented . here
we assume susy - seesaw models with the mssm particle content plus three right handed neutrinos , @xmath4 @xmath5 , and their corresponding susy partners , @xmath6 , and use the parameterisation for the yukawa couplings given by @xmath7 , with @xmath8 defined by three complex angles @xmath9 ; @xmath10 , @xmath11 gev ; @xmath12 denotes the three light neutrino masses , and @xmath13 the three heavy ones . with this parameterisation
it is easy to accommodate the @xmath14 data and also get large @xmath15 , by choosing large @xmath16 and/or @xmath9 .
the sensitivity to the higgs sector of these susy - seesaw models can appear only via the lfv processes that are mediated by higgs particles .
this is the case of the tau decay channels considered here , whose present experimental bounds are respectively at br@xmath17 , br@xmath18 , br@xmath19 and br@xmath20 ( assuming br@xmath21 ) .
the interest of these channels is that for scenarios with heavy susy soft masses of the order of 1 tev , where the predicted rates for the @xmath22 channel lay below the present experimental bound , br@xmath23 , still some of the higgs - mediated processes can indeed be at the present experimental reach if the relevant higgs mass is light enough , say of the order of 100 - 250 gev .
we will focus here in the type of constrained susy - seesaw scenarios called nuhm - seesaw ( standing for non universal higgs mass ) where this kind of spectrum with light higgs and heavy susy particles is possible .
the input parameters are @xmath24 , @xmath25 , @xmath26 @xmath27 , sign(@xmath28 ) , @xmath29 and @xmath30 . in refs .
@xcite and @xcite the proper choices of @xmath31 and @xmath32 leading to the wanted light higgs sector can be found .
notice that @xmath33 corresponds to the usual constrained model ( cmssm - seesaw ) with all scalar masses being universal , but this model does not lead to the scenario with heavy susy and light higgs particles that we are interested here , so we will not considered it next .
most of the results reported here are extracted from the works @xcite and @xcite to which we refer the reader for more details .
the numerical results for the branching ratios of the studied lfv tau decays are summarized in fig .
[ fig : allbr ] .
they are full one loop results and do not make use of any approximation like the mass insertion , large @xmath34 , nor the leading logarithmic approximations .
the mass spectra for all the involved particles in the loops that contribute to these processes are computed within the nuhm - seesaw model , by solving the rges also to one loop level . in the case of the semileptonic
decays we have used the standard techniques in chiral theory to describe the final hadrons in terms of quark bilinears . in particular , the channels with pseudo goldstone bosons ( pgb ) , @xmath35 , like @xmath36 , @xmath37 and @xmath38 , are treated within chiral perturbation theory ( @xmath39pt ) to leading order , @xmath40 , where the results are given in terms of @xmath41 mev and @xmath42 .
the additional contributions from resonances , @xmath8 , in channels of the type @xmath43 are taken into account within resonance chiral theory ( r@xmath39 t ) , where the results are given in terms of @xmath44 , @xmath42 , @xmath45 and well established form factors .
in particular for the @xmath1 channel the contributions from the @xmath46 , @xmath47 and @xmath48 are considered via the electromagnetic vector form factor , @xmath49 . on the other hand ,
the @xmath50 is defined via mixing between the octet , @xmath51 , and singlet , @xmath52 , components of the @xmath53 nonet of pgb in @xmath39pt .
concretely we assume here a mixing angle of @xmath54 .
the @xmath55 is defined via mixing between the octet , @xmath56 , and singlet , @xmath57 , components of the @xmath58 nonet of resonances in r@xmath39 t .
concretely we assume here two choices for this mixing angle , @xmath59 and @xmath60 .
notice that we have selected the semileptonic channels where the final hadrons have a relevant strange quark content , and consequently the sensitivity to the higgs particles is greater than in those with just up and/or down quarks . besides the total rates
, we also show separately in fig .
[ fig : allbr ] the various contributions to these processes : 1 ) @xmath0 can be mediated by a @xmath61 , a @xmath62 boson , boxes and @xmath63 , @xmath64 and @xmath65 @xcite , 2 ) @xmath1 by a @xmath61 ( also @xmath62 , but it is negligible ) and @xmath63 , @xmath64 , 3 ) @xmath66 by a z boson and @xmath65 , and 4 ) @xmath3 by @xmath63 and @xmath64 . we conclude that although the higgs contributions in all these processes grow very fast with @xmath27 , still at large @xmath27 values these are fairly dominated by the @xmath61 contribution in the cases of @xmath67 and @xmath68
. therefore , these are not sensitive to the higgs sector .
we have checked that this is true even for a very heavy susy spectra where the @xmath61 contribution gets reduced considerably . in @xmath68
the higgs contribution is relevant for @xmath69 gev , but there the rates are too small compared to the present bound . in contrast , the @xmath66 and @xmath3 channels are clearly sensitive to the higgs sector . in fig . [
fig : allbr ] we see that the @xmath65 contribution dominates br(@xmath66 ) for @xmath70 and the @xmath64 contribution dominates br(@xmath3 ) at all @xmath27 values .
we also conclude from this figure that the approximate formulas found in refs .
@xcite and @xcite for large @xmath27 , whose simplest forms are given by , @xmath71 { \rm br}(\tau \to \mu f_0(980))_{{\rm approx } } & = & \ ! \ ! \ !
\left ( \ ! \begin{array}{c } 7.3 \times 10^{-8}\,\ , ( \theta_s=7^\circ ) \\ 4.2 \times 10^{-9}\,\ , ( \theta_s=30^\circ ) \end{array } \ ! \ !
\right ) \left| \delta_{32 } \right|^2 \left ( \frac{100 } { m_{h^0}({\rm gev } ) } \right)^4 \left ( \frac{\tan \beta}{60 } \right)^6 \ ! \ ! , \nonumber \end{aligned}\ ] ] provide a very good approximation to the full result .
this is also shown in fig.[fig : breta - brf0-mh ] where br(@xmath66 ) and br(@xmath72 ) are displayed as a function of the relevant higgs mass .
we see clearly in this figure that these two channels are sensitive to masses within the range 100 - 250 gev , for large @xmath27 , @xmath73 and @xmath74 .
finally , fig.[fig : final ] illustrates several examples for the relevant parameter @xmath75 that meassures approximately the size of the lfv in the tau - mu sector in seesaw scenarios with a ) hierarchical and b ) degenerate heavy neutrinos @xmath76 .
we see that in both scenarios values as large as @xmath77 can be obtained .
therefore , with such large values and the present experimental upper limits one can extract lower bounds for the relevant higgs mass and upper bounds for @xmath27 .
this is the main conclusion of this work .
the case of @xmath3 is illustrated in the last plot of fig.[fig : final ] , where one can see the excluded regions in the @xmath78 plane .
9 e. arganda , m. j. herrero and j. portoles , jhep * 0806 * ( 2008 ) 079 [ arxiv:0803.2039 [ hep - ph ] ] .
m. j. herrero , j. portoles and a. m. rodriguez - sanchez , phys .
d * 80 * , 015023 ( 2009 ) [ arxiv:0903.5151 [ hep - ph ] ] .
e. arganda and m. j. herrero , phys .
d * 73 * ( 2006 ) 055003 [ arxiv : hep - ph/0510405 ] .
|
here we study and compare the sensitivity to the higgs sector of susy - seesaw models via the lfv tau decays : @xmath0 , @xmath1 , @xmath2 and @xmath3 .
we emphasize that , at present , the two latter channels are the most efficient ones to test indirectly the higgs particles .. address = departamento de fsica terica and instituto de fsica terica , ift - uam / csic + universidad autnoma de madrid , cantoblanco , e-28049 madrid , spain address = ific , universitat de valncia - csic , apt .
correus 22085 , e-46071 valncia , spain address = departamento de fsica terica and instituto de fsica terica , ift - uam / csic + universidad autnoma de madrid , cantoblanco , e-28049 madrid , spain
|
superclusters ( scs ) are the ultimate manifestation of hierarchical large - scale structure ( lss ) in the universe . with a median major axis of @xmath3mpc ( jaaniste et al .
1998 ) , they are the most massive and the largest possibly bound structures known
. `` great walls '' of galaxies have been discovered that are larger than what one normally calls a sc , but they are almost certainly not gravitationally bound ( geller & huchra 1989 ) .
observations and numerical simulations suggest that a rich hierarchy of constituents trace a sc .
massive galaxy clusters are linked by smaller galaxy groups in the high - density regions , while filamentary distributions of galaxies fill out the expansive , low - density domains .
scs are not isolated entities but the top - level components of a `` cosmic web '' of lss ( tully 1987 ; bond , kofman , & pogosyan 1996 ; einasto et al.1997 ) . despite their ubiquity in our overall impression of the universe ,
scs are nonetheless exceedingly rare relative to other types of objects .
the extreme properties of scs are the results of the initial conditions of the nascent universe developed by structure formation mechanisms .
scs are so large that the crossing time of the member clusters is greater than the hubble time
. hence the imprint of the initial conditions has not been erased by internal motions . therefore understanding the nature of scs is of great cosmological importance .
much effort has been expended to locate scs , primarily in the optical regime .
galaxy surveys provide nearly all - sky sampling of the local universe out to @xmath4 ( e.g. cfa , huchra , vogeley , & geller 1999 ) , while more restricted slice surveys ( e.g. lcrs , shectman et al .
1996 ) reach to @xmath5 .
galaxy clusters are more sparsely distributed but efficiently probe scs to intermediate redshifts ( @xmath6 ) across essentially the entire sky ( e.g. bahcall & soneira 1984 ; batuski & burns 1985 ; einasto et al .
1997 ; ebeling et al.1998 ) . here
we report on the successful application of x - ray data , featuring high sensitivity over a large , contiguous solid angle , to characterize scs and lss .
we have used the rosat all - sky survey to detect a substantial sc in the north ecliptic pole ( nep ) region a structure discovered in the optical by batuski & burns ( 1985 ) with subsequent x - ray detections by burg et al .
we have for the first time detected a significant number of x - ray emitting galaxy groups in a sc yielding a greatly enhanced understanding of its extent and content .
these x - ray data combined with optical , iras , and hst / uv observations make the nep supercluster ( nepsc ) one of the most broadly detected superstructures known . throughout this letter
we assume a hubble constant of @xmath7 km s@xmath8mpc@xmath9 and a deceleration parameter of @xmath10 , however the major conclusions do not change for other reasonable cosmologies .
co - moving distances are reported unless otherwise indicated .
the rosat all - sky survey ( rass ) is the first all - sky survey with an imaging x - ray detector ( trmper 1983 ; voges et al .
the nep region of the rass possesses the deepest exposure ( @xmath11 ks ) and consequently the greatest sensitivity ( median flux limit , @xmath12 = 7.8 @xmath13 10@xmath14 erg s@xmath9 @xmath15 [ 0.52.0 kev ] ) of the entire survey ( voges et al .
we have identified the physical nature of 99.6% of the 445 x - ray sources with fluxes measured at greater than @xmath16 significance in the 9@xmath17 @xmath13 9@xmath17 survey region centered on the nep ( henry et al .
2001 ) . of these , 64 ( 14.4% ) are galaxy clusters and groups .
x - ray luminous clusters are detected out to @xmath18 , while low - luminosity groups of galaxies are well - sampled to @xmath19 .
our primary motivation for pursuing the rosat nep survey was to construct a complete sample of x - ray selected galaxy clusters , and use it to examine x - ray cluster luminosity evolution and to characterize lss .
the initial evolution results are presented by gioia et al.(2001 ) , while the first findings relevant to lss are presented in this letter .
we emphasize the cluster sample is unique in that it is both deep in flux sensitivity and contiguous on the sky . in the top panel of fig . 1 we plot the observed rosat nep cluster redshift distribution along with the expected distribution for a spatially homogeneous , non - evolving cluster population .
the latter is calculated by folding the local cluster x - ray luminosity function ( xlf , ebeling et al .
1997 ; de grandi et al . 1999
; h. bhringer 2000 , private communication ) through the rosat nep survey selection function .
the observed redshift distribution shows a striking feature at @xmath0 .
sixteen clusters lie in the redshift interval @xmath20 , 25% of the entire cluster sample . depending on which xlf determination
is used , this interval is 3.3 to 4.3 times more populated than expected which is significant at the @xmath21 to @xmath22 level .
active galactic nuclei ( agn ) are the dominant class of x - ray emitters in the rosat nep survey ( 49.0% of all sources ) and they are another means for examining lss . though they are more sporadic markers than clusters , they reach to greater depths ( @xmath23 ) . a @xmath24 density enhancement around @xmath0 is also seen in the redshift distribution of rosat nep agn . the middle panel of fig .
1 shows the observed distribution compared to a spatially homogeneous , evolving agn population .
twelve agn are observed in the redshift interval @xmath25 where 5.6 are predicted from the agn xlf ( miyaji , hasinger , & schmidt 2000 ) .
note , an additional peak ( @xmath26 ) in the agn redshift distribution at @xmath27 is comprised of two separate groups of agn .
using _ only _ the rosat x - ray data and supporting optical observations from within the nep survey boundaries , we have detected a significant lss consisting of _ at least _
16 galaxy clusters and groups , plus 12 agn , one bl lac , and an isolated elliptical galaxy with an x - ray luminosity commensurate with that of a poor cluster .
note there are additional scs and clusters of agn in the rosat nep data .
these structures along with spatial correlation analyses for the cluster and agn samples will be discussed in forthcoming papers . _ _ _ _ the presence of lss in the vicinity of the nep was first noted in a pioneering study by batuski & burns ( 1985 ) .
they produced a finding list of sc candidates ( scc ) based on a percolation analysis of abell clusters , and found an association of six clusters ( a2301 , a2304 , a2308 , a2311 , a2312 , a2315 ) approximately 5@xmath17 from the nep ( scc # 47 ) .
only two of these clusters had spectroscopic redshifts at the time , the remainder having photometric estimates .
more recent percolation studies , working with improved redshift data , recovered the sc candidate ( sc#34/2 , zucca et al .
1993 ; # 97 , # 170 , einasto et al . 1994 , 1997 ) , with the latter two analyses noting an additional cluster a2295 .
four of the abell clusters ( a2295 , a2301 , a2304 , a2308 ) lie within the rosat nep survey boundaries and are easily detected in x - rays .
the other three ( a2311 , a2312 , a2315 ) , though barely outside the nep region , are detected in the rosat bright source catalog ( bsc , voges et al .
1999 ) . during the early phase of the rosat mission burg et al .
( 1992 ) combined a 50 ks pointed observation with all - sky scans to identify five x - ray clusters and groups between redshifts 0.08 and 0.09 within 1.5@xmath17 of the nep . except for one very low luminosity group discovered in the pointed data ( hasinger , schimdt , & trmper 1991 ) which is below our flux limit
, our survey recovers these detections .
we examined all x - ray sources in the rosat bsc within 10@xmath17 of the known sc members .
there are no additional candidate galaxy clusters or groups in the redshift interval @xmath20 .
the nearest candidate ( rxj1827.6 + 6135 , @xmath28 , bhringer et al .
2000 ) is just outside both the nep survey region and the nominal sc redshift boundary , and is not included in our analysis .
the iras satellite mapped the infrared sky in a mode similar to rosat , and hence also has enhanced sensitivity at the nep .
ashby et al.(1996 ) tested models of starburst galaxy evolution using a sample of iras galaxies within 1.5@xmath17 of the nep and noted the strong effect of lss .
fifteen galaxies , 20% of their ir - selected sample , lie between redshifts 0.085 and 0.090 with @xmath29 .
rinehart et al .
( 2000 ) have expanded the iras survey to 56 deg@xmath30 .
this extended iras sample ( fig . 1 , lower panel ) has 61 galaxies in the redshift interval @xmath25 where 30.9 are expected representing a @xmath31 excess .
these iras sources are predominantly spirals found outside the dense cluster environment .
the galaxies form a sheet estimated to be three times as dense as the great wall ( geller & huchra 1989 ) .
additional evidence of structure in the nep region is a ly@xmath32 absorber detected at @xmath33 in the direction of a well - known qso h1821 + 643 ( @xmath34 ) .
tripp et al .
( 1998 ) identified this absorbing system during a study of low - redshift weak ly@xmath32 absorbers using hst uv spectroscopy .
this qso was originally discovered via its high x - ray luminosity and is the brightest source in the rosat nep survey .
a three - dimensional view of the conical nep survey volume , showing the distribution of clusters , groups , and agn out to @xmath35 , is presented in fig . 2 ( right ) .
the nepsc is clearly visible as an over - density of x - ray emitters at @xmath0 ( @xmath36mpc ) .
this complex is relatively disconnected , with a candidate void lying on the high redshift side ( @xmath37 ) .
the nearest non - sc clusters are rxj1827.6 + 6135 , a2293 , and a2275 at respectively 64.7 , 87.2 , and @xmath38mpc .
we use the non - weighted inertia tensor for the nepsc groups and clusters to determine the principal axes and estimate the size , shape , and orientation of the structure .
the maximal extents in three orthogonal directions are @xmath39mpc)@xmath2 with a 12@xmath17 extent on the sky .
the major and minor axes are oriented 42@xmath17 and 85@xmath17 relative to our line - of - sight .
the nepsc is basically an oblate ellipsoid seen nearly edge - on .
consequently , the effect of flattening the sc in the line - of - sight direction in going from real space to redshift space is less important than would be in a face - on scenario such as with the great wall .
all eight abell clusters associated with the nepsc are detected in x - rays with luminosities in the range of 0.23.6 @xmath40 erg s@xmath41 [ 0.52.0 kev ] .
note that three clusters discovered by the rosat nep survey in this luminosity range are missing from the abell catalog , indicative of the incompleteness inherent to optical selection .
furthermore , there are an additional 10 groups of galaxies with x - ray luminosities in the range 29 @xmath42 erg s@xmath9 [ 0.52.0 kev ] .
this is the first time a large sample of x - ray emitting galaxy groups has been identified in a sc .
. 2 ( middle ) shows a close - up view of the nepsc where the groups and clusters have been linked together by a minimum spanning tree .
minimum , median , and maximum link distances are respectively 5.6 , 24.2 , and @xmath43mpc .
the massive abell clusters that initially delineated the sc are linked together by newly discovered , less massive x - ray clusters and groups which bridge the region between the previously known sc and a very rich cluster of galaxies ( a2255 ) at the same distance , but not previously associated with this structure .
the new realization of the nepsc has doubled in physical size and tripled in cluster content .
if this sc is characteristic of others , then scs can be substantially larger than what is shown by the abell clusters alone .
filaments are the fundamental element of cosmic structure present in all modern numerical simulations of lss .
the nepsc is oriented in fig . 2 ( middle ) to display a possible filament containing two groups and two clusters .
the length of the filament , the distance between a2304 and nep 4990 , is @xmath44mpc .
we also show in fig.2 ( left ) a filament from a virgo consortium @xmath45+cdm ( @xmath46 ) numerical simulation ( jenkins et al .
1998 ) of lss selected to resemble the nep structure .
the nepsc filament is potentially the best match yet observed in x - rays to what is seen in the simulations ( see briel & henry 1995 and scharf et al .
2000 for previous observations ) .
the fortunes of experimental design and serendipity have converged at the nep to reveal a prominent example of lss in the universe .
the nepsc ( @xmath0 ) is one of the most robustly sampled superstructures known 21 x - ray emitting clusters of galaxies demark the highest density regions while 12 x - ray selected agn and over 60 iras galaxies trace out the lower density domains .
the @xmath47mpc)@xmath2 complex is strongly detected in x - ray , optical , ir , and uv wavebands .
detailed follow - up observations with chandra , xmm - newton and ground - based facilities will permit us to examine the dynamics of the structure , the evolutionary state of its constituents , and the degree of alignments between the member clusters .
the sc s edge - on orientation makes it an attractive candidate in which to look for filaments of diffuse x - ray emitting gas .
we have combined contiguous coverage with high sensitivity in x - rays to efficiently detect objects in a sc ranging from agn and poor galaxy groups to rich galaxy clusters .
this is the first time a sizable number of x - ray emitting galaxy groups has been identified in a sc , thus yielding a better reckoning of a sc s girth and content .
the rosat nep survey demonstrates the effectiveness of x - rays for studying lss and advocates a next generation , large format x - ray survey mission to propel this field of study to the next level .
it is a pleasure to thank harald ebeling , brent tully , and gnther hasinger for fruitful discussions , and karen teramura for graphics assistance . the kind support of the uh tac and the expertise of the mauna kea observatory personnel are gratefully acknowledged .
we also thank stephen rinehart for providing access to the extended iras nep sample prior to publication .
c.r.m . acknowledges partial financial support from the nasa graduate student researchers program ( ngt5 - 50175 ) and the arcs foundation .
support has also come from the nsf ( ast91 - 19216 , ast95 - 00515 ) , nasa ( go-5402.01 - 93a , go-05987.02 - 94a ) , the smithsonian astrophysical observatory , nato ( crg91 - 0415 ) , and the italian asi - cnr .
|
we have used the rosat all - sky survey to detect a known supercluster at @xmath0 in the north ecliptic pole region .
the x - ray data greatly improve our understanding of this supercluster s characteristics , approximately doubling our knowledge of the structure s spatial extent and tripling the cluster / group membership compared to the optical discovery data .
the supercluster is a rich structure consisting of at least 21 galaxy clusters and groups , 12 agn , 61 iras galaxies , and various other objects .
a majority of these components were discovered with the x - ray data , but the supercluster is also robustly detected in optical , ir , and uv wavebands . extending @xmath1mpc)@xmath2 , the north ecliptic pole supercluster has a flattened shape oriented nearly edge - on to our line - of - sight . owing to the softness of the rosat x - ray passband and the deep exposure over a large solid angle , we have detected for the first time a significant population of x - ray emitting galaxy groups in a supercluster .
these results demonstrate the effectiveness of x - ray observations with contiguous coverage for studying structure in the universe .
|
a @xmath1-manifold is called _ virtually fibred _ if it has a finite cover which is a surface bundle over the circle . a link in a connected @xmath1-manifold is said to be _ virtually fibred _ if its exterior is a virtually fibred @xmath1-manifold .
thurston conjectured that all closed hyperbolic @xmath1-manifolds and all hyperbolic links in closed @xmath1-manifolds are virtually fibred .
this conjecture , which has been named as _ virtually fibred conjecture _
, is one of the most fundamental and difficult problems in @xmath1-manifold topology .
recall that a link @xmath2 in @xmath3 is called a _ generalized montesinos link _ if the double branched cover @xmath4 of @xmath5 is a seifert fibred @xmath1-manifold .
such a link @xmath2 is further said to be _ of type @xmath0 _ if the canonical geometric structure on @xmath4 is from the @xmath0-geometry .
when every component of the branched set in @xmath4 is not a fiber of the seifert fibration of @xmath4 , @xmath2 is called a _ classic montesinos link_. recent work of walsh [ wa ] and work of agol - boyer - zhang [ abz ] combined together solved the virtually fibred conjecture for all generalized montesinos links in @xmath3 which are not classic montesinos links of type @xmath0 .
agol - boyer - zhang [ abz ] also gave an infinite family of virtually fibred classic montesinos links of type @xmath0 .
in this paper , we extend the latter result of [ abz ] to a larger family of classic montesinos links of type @xmath0 .
note that every classic montesinos link has a cyclic rational tangle decomposition of the form @xmath6 with all @xmath7 as shown in figure [ cml ] .
we prove classic montesinos link ] [ main ] if a classic montesinos link @xmath2 has a cyclic rational tangle decomposition of the form @xmath8 with @xmath9 odd , then @xmath2 is virtually fibred .
this theorem is proved in [ abz ] when @xmath10 is a multiple of @xmath11 .
our approach follows closely to that of [ abz ] . at a number of places we need to deal with some new issues that arise .
we shall describe our ways of dealing with these issues when we reach these places .
let @xmath2 be a link as given in theorem 1.1 .
the base orbifold @xmath12 of the seifert fibred space @xmath4 is a @xmath13-sphere with @xmath10 cone points each has order @xmath11 .
let @xmath14 be the seifert fibration which is invariant under the covering involution @xmath15 .
the euler number @xmath16 of @xmath4 and the orbifold euler characteristic @xmath17 of @xmath18 are given by the following formulas : @xmath19 note that @xmath4 has the @xmath0-geometry precisely when @xmath20 and @xmath21 .
thus @xmath2 is of type @xmath0 precisely when * case ( 1)*. @xmath22 , @xmath23 , and @xmath24 , or * case ( 2)*. @xmath25 and @xmath26 .
we shall split our proof of theorem [ main ] into these two cases , given in section 2 and section 3 respectively .
all main ingredients of the proof will occur already in case ( 1 ) ; the proof of case ( 2 ) will be a quick generalization . in the remainder of this section , we shall give a few additional notes and notations which will be used throughout this paper .
from now on we assume that @xmath27 is a link satisfying the conditions of case ( 1 ) or ( 2 ) listed above . first note that @xmath2 is a single component knot if @xmath28 is odd , and a two component link otherwise .
the double branched cover @xmath4 has the @xmath0-geometry and its base orbifold @xmath18 is hyperbolic .
let @xmath29 be the corresponding branched set in @xmath4 and let @xmath30 , where @xmath31 is the seifert quotient map . by the orbifold theorem we may assume that the covering involution @xmath32 on @xmath4 is an isometry and the branch set
@xmath33 is a geodesic which is orthogonal to the seifert fibres by lemma 2.1 of [ abz ] .
also @xmath34 is a geodesic in @xmath18 , and is an equator of @xmath18 containing all the @xmath10 cone points which we denote by @xmath35 ( indexed so that they appear consecutively along @xmath34 ) .
the order of each cone point @xmath36 is @xmath11 .
since @xmath11 is odd , the restriction map of @xmath37 to each component of @xmath29 is a double covering onto @xmath34 .
in this section we prove theorem [ main ] in the case @xmath22 , @xmath23 and @xmath24 . throughout this section , without further notice , @xmath38 as well as numbers expressed by them will be considered as non - negative integers mod @xmath11 , and @xmath39 will also be assumed .
here is an outline of the proof .
* step 1 * take a specific @xmath41-fold orbifold cover @xmath42 of @xmath43 such that @xmath42 is a smooth surface .
there is a corresponding free cover @xmath44 of the same degree such that the base orbifold of the seifert space @xmath45 is @xmath42 and such that the following diagram commutes @xmath46 where @xmath47 is the seifert quotient map .
note that @xmath45 is a locally - trivial circle bundle over @xmath42 since @xmath42 is a smooth surface .
let @xmath48 .
then @xmath49 has exactly @xmath41 components , which we denote by @xmath50 .
let @xmath51 .
then @xmath52 are @xmath11 mutually disjoint simple closed geodesics in @xmath42 . * step 2 * construct a surface semi - bundle structure on @xmath53 , where @xmath54 is a small regular neighborhood of @xmath55 in @xmath45 , @xmath56 .
note that @xmath57 is a graph manifold with non - empty boundary , with vertices @xmath58 and @xmath59 , where @xmath60-\overset{\circ}{n}(l_{j , j})$ ] , @xmath56 .
we construct a surface semi - bundle structure on @xmath57 by using @xcite and following @xcite .
* step 3 * isotope all @xmath62 , @xmath63 , such that they are transverse to the surface bundle in @xmath64 , denoted by @xmath65 .
this is one of the key parts of the proof .
let @xmath66 , the vertical torus over @xmath67 . by the construction in step 2
, there are exactly two singular points in the induced foliation by @xmath68 on every component of @xmath69 .
we show that @xmath49 can be oriented such that the arcs @xmath70 whose images intersect in @xmath42 , all travel from one of @xmath71 and @xmath72 to the other .
this property allows us to arrange these arcs in @xmath69 such that they always travel from the @xmath73 " side of the surface bundle in @xmath64 to the @xmath74 " side .
* step 4 * construct a double cover of @xmath57 , denoted by @xmath75 , so that @xmath75 has a surface bundle structure .
denote the corresponding double cover of @xmath45 as @xmath76,the lift of @xmath59 as @xmath77 , and the lift of @xmath62 as @xmath78
. perform certain dehn twist operations on the surface bundles of @xmath77 so that the @xmath79 are transverse to the new surface bundles , @xmath80 .
this is another key part of the proof .
we perform these dehn twist operations along a union of vertical tori , @xmath81 .
@xmath82 contains some boundary parallel tori of @xmath83 as in @xcite , different form @xcite , these boundary parallel tori intersect some arcs of @xmath84 two times with different direction or do not intersect at all , @xmath80 .
we call such arcs of @xmath85 `` bad '' arcs .
we construct four specific extra tori as members of @xmath82 to deal with the `` bad '' arcs , @xmath80 . by steps 1 - 4
, the exterior of the inverse image of @xmath29 in @xmath86 has a surface bundle structure .
it is a free cover of the exterior of @xmath29 in @xmath4 , which in turn is a free double cover of the exterior of @xmath2 in @xmath3 .
thus @xmath2 is virtually fibred in @xmath3 .
now we fill in the details . * step 1 * instead of taking the @xmath11-fold cyclic cover of @xmath87 as in @xcite , we construct the @xmath41-fold cover @xmath42 of @xmath88 , by a composition of two @xmath11-fold cyclic covers .
let @xmath89 be the orbifold fundametal group of @xmath90 .
it has a presentation @xmath91 where @xmath92 is represented by a small circular loop in @xmath93 centered at @xmath94 .
let @xmath95 be the @xmath11-fold cyclic orbifold cover of @xmath93 corresponding to the homomorphism : @xmath96 where @xmath97 , and @xmath98 .
@xmath99 is as shown in figure [ csf ] ( while @xmath100 ) . covering space @xmath99 ]
the order of @xmath102 and @xmath103 are both @xmath11 , so @xmath104 and @xmath105 are two points in @xmath99 ( not cone points ) .
the order of @xmath106 is @xmath107 , so @xmath108 are the only cone points in @xmath99 , each has order @xmath11 .
we have @xmath109 .
@xmath110 is a set of @xmath11 geodesics @xmath111 , such that @xmath112 goes through the cone point @xmath113 ( cf .
figure [ csf ] ) .
let @xmath114 be the deck transformation of @xmath115 corresponding to @xmath116 .
we may assume that @xmath118 .
orient @xmath119 and give @xmath112 the induced orientation , @xmath120 .
@xmath99 admits an orientation such that @xmath121 is a counterclockwise rotation by @xmath122 near @xmath123 and a clockwise rotation by @xmath122 near @xmath124 on @xmath99 , since @xmath125 and @xmath126 .
note that @xmath127 intersects @xmath128 at @xmath123 in an angle of @xmath129 , also intersects @xmath128 at @xmath124 in an angle of @xmath130 .
( we always suppose that the counterclockwise direction is the positive direction of angles . )
let @xmath131 be the orbifold fundamental group of @xmath99 .
it has the presentation @xmath132 where @xmath133 is represented by a small circular loop in @xmath134 centered at @xmath135 .
let @xmath136 be the @xmath11-fold cyclic orbifold cover of @xmath99 corresponding to the homomorphism : @xmath137 where @xmath138 , @xmath56 .
the order of @xmath139 is @xmath11 , so @xmath140 is a point , @xmath141 .
hence @xmath42 is a smooth closed orientable surface without cone points . for each @xmath142
, @xmath143 is a set of @xmath11 points , which we denote by @xmath144 . for each @xmath145
, @xmath146 is a set of @xmath11 simple closed geodesics , which we denote by @xmath147 .
denote the deck transformation of @xmath148 corresponding to @xmath149 by @xmath150 .
we may assume that @xmath152 , @xmath153 , and @xmath154 .
then @xmath155 .
now we fix an orientation for each of @xmath156 such that @xmath157 goes through @xmath158 , @xmath159 and @xmath160 in order , and give @xmath161 the induced orientation .
@xmath42 admits an orientation such that @xmath150 is a counterclockwise rotation by @xmath162 near the fixed points @xmath163 , so @xmath164 intersect at @xmath165 , and @xmath166 intersects @xmath167 at @xmath165 in an angle of @xmath168 , @xmath169 . since @xmath170 is not a cone point on @xmath99 , a small regular neighborhood of @xmath171 on @xmath42
is a copy of a small regular neighborhood of @xmath172 on @xmath99 , @xmath173 . recall that @xmath112 intersects @xmath128 at @xmath123 in an angle of @xmath174 , and at @xmath124 in an angle of @xmath168 .
then @xmath175 intersects @xmath157 at @xmath160 in an angle of @xmath174 , and at @xmath159 in an angle of @xmath168 .
see the schematic picture , figure [ lij * ] , for example @xmath100 .
here we did nt depict the genus of the surface , so some curves meet in the picture actually do not meet . for convenience ,
we only draw a part of @xmath157 in figure [ lij * ] , @xmath56 .
@xmath176 with @xmath100,width=576 ] summarizing the above discussion , we have the following remark .
[ lij ] @xmath177 only intersects at the points @xmath178 on @xmath42 , and @xmath175 goes through @xmath165 , @xmath159 and @xmath160 in order following the given orientation of @xmath175 , @xmath179 .
in particular @xmath157 goes through @xmath180 , and @xmath160 , so @xmath181 are mutually disjoint , and @xmath182 note that @xmath183 are considered as integers mod @xmath11 .
let @xmath184 , which is a @xmath41-fold orbifold covering .
we have @xmath185 , @xmath186 , and @xmath187 .
let @xmath188 be the @xmath11-fold cover of @xmath4 corresponding to @xmath115 , @xmath189 the @xmath11-fold cover of @xmath190 corresponding to @xmath148 , and @xmath191 .
then @xmath192 is a @xmath41-fold cover of @xmath4 corresponding to @xmath193 .
@xmath45 has locally - trivial circle bundle seifert structure with base surface @xmath42 .
set @xmath194 , which is a geodesic link in @xmath45 .
we have the following commutative diagram : note that @xmath49 has exactly @xmath41 components .
let @xmath195 , where @xmath196 .
@xmath197 is a 2-fold cover .
* step 2 * by remark [ lij ] , @xmath181 are mutually disjoint .
let @xmath198 $ ] , for some small positive number @xmath199 , be a neighborhood of @xmath157 in @xmath42 , such that @xmath200 s are mutually disjoint .
let @xmath201 , @xmath202 , and @xmath203 .
we may suppose that @xmath204 is on the left side of @xmath157 . by remark [ lij ]
, @xmath205 goes through @xmath206 , @xmath207 and @xmath165 , so @xmath205 is separated into @xmath208 arcs by @xmath209 , and @xmath210 .
we denote the six arcs of @xmath205 by @xmath211 , @xmath212 , @xmath213 , @xmath214 , so that @xmath215 , @xmath216 , @xmath217 , and @xmath218 , @xmath219 , and @xmath220 .
we also require that as we travel along @xmath205 in its orientation , we will pass @xmath214 , @xmath221 , @xmath213 , @xmath211 consecutively . from the construction , we have remark [ arcs ] and table [ l * ] .
[ arcs ] @xmath222 .[l*]the induced orientation on @xmath223 . [
cols="^,^,^,^,^,^,^,^,^ " , ] note : @xmath224 is considered as a nonnegative integer mod @xmath11 , @xmath225 or @xmath226 .
fix a transverse orientation on the surface bundle @xmath65 in @xmath64 . by the construction of @xmath227 in step 2 ,
there are two singular points in the foliation of @xmath228 given by @xmath229 if @xmath230 is small enough , @xmath231 ( cf .
figure [ foliation ] ) .
we can isotope @xmath232 in @xmath228 such that @xmath233 is transverse to @xmath234 , and travels from the @xmath73 " side of @xmath65 to the @xmath74 " side ( @xmath235 ) .
we only discuss the isotopy in @xmath236 , and others are similar .
@xmath237 and @xmath238.,width=480 ] by the definition of @xmath239 , ( @xmath240 , @xmath241 .
suppose that @xmath230 is small enough so that the singular foliation on @xmath237 , @xmath238 or @xmath242 induced by @xmath243 has two singular points .
the singular points on @xmath244 @xmath245 , and the singular points on @xmath246 , @xmath241 . recall that @xmath247 , @xmath248 , so @xmath249 .
let @xmath250 , or @xmath251 .
fix @xmath252 . by lemma 2.1 in @xcite ,
@xmath232 is always perpendicular to the original fibers in @xmath228 .
if @xmath232 passes from the `` @xmath74 '' to the `` @xmath73 '' side of @xmath243 s leaves along its orientation , we isotope @xmath232 passing the singular point smoothly along the original fibres in a small neighborhood of @xmath253 , and place them as in figure [ foliation ] . by
( [ tail ] ) , @xmath254 travels from @xmath255 to @xmath256 ; @xmath257 travel from @xmath256 to @xmath255 .
then @xmath258 and @xmath259 are in the different sides of their own singular points , so are @xmath260 and @xmath261 . also by ( [ tail ] ) , @xmath257 and @xmath262 are both travel from @xmath256 to @xmath255 , so @xmath263 is similar to @xmath264 .
@xmath265 is as shown in figure [ foliation]-1 , and @xmath264 is as shown in figure [ foliation]-2 , in other words , they alternate with the singular points in the original circle fiber direction .
the isotopies are similar to that in @xcite .
next , we want to show that the above isotopies of @xmath232 wo nt block each other . since @xmath266 , @xmath267 , and @xmath268 , @xmath269
, potential obstructions happen in the set @xmath270 , @xmath271 , or @xmath272 .
@xmath233 and @xmath232 are in the different halves of @xmath228 ( cf .
figure [ foliation ] ) , so they wo nt block each other since the smooth @xmath273-vertical isotopy can not cross the punctures on @xmath228 .
next we check @xmath274 , and the discussions for @xmath275 and @xmath276 are similar . by ( [ tail ] ) , @xmath277 travels from @xmath255 to @xmath256 for any @xmath278 .
we may assume that we need isotope @xmath279 smoothly along @xmath273 above its singular point , as shown in figure [ isotope]-1 , @xmath278 .
( there is only half of @xmath280)$ ] in figure [ isotope]-1 ) . although the singular points of @xmath281 s in @xmath282 are at different positions , and the distribution of the intersection points of @xmath282 and @xmath283 may be different for different @xmath284 , we can isotope @xmath279 in the positive direction of @xmath282 above all singular points , so they wo nt block each other .
figure [ isotope]-2 illustrates the situation near @xmath285 $ ] . in figure
[ isotope ] , @xmath286 denotes the singular point of @xmath287 , for @xmath288 and @xmath289 .
the positions of @xmath258 and @xmath259.,width=566 ] by generalizing the above discussion to @xmath64 , @xmath290 , we have the following proposition which is analogous to proposition 6.2 in @xcite . [ m2j ]
suppose @xmath290 .
fix a transverse orientation on @xmath65 in @xmath64 .
there is a smooth @xmath273-vertical isotope of @xmath291 and @xmath292 @xmath293 in @xmath294 , fixed outside a small neighborhood of @xmath295 , which reposition @xmath296 and @xmath297 to be transverse to @xmath65 and to pass from the negative to the positive side of @xmath65 s leaves while traveling along the orientation on @xmath296 and @xmath298 , where @xmath299 .
proposition [ m2j ] finishes step 3 .
* step 4 * at first we construct a double cover of @xmath57 similarly to @xcite .
set @xmath300 to be the 3-manifold obtained by cutting @xmath57 open along @xmath301 , @xmath302 .
take two copies of @xmath300 , denoted @xmath303 and @xmath304 .
we have @xmath305 , where @xmath306 is a copy of @xmath59 , and @xmath307 is a copy of @xmath64 , @xmath56 and @xmath80 .
@xmath75 is a 3-manifold obtained by gluing @xmath308 to @xmath309 along @xmath310 , and @xmath311 to @xmath312 along @xmath313 , where @xmath314 is a copy of @xmath315 , @xmath56 and @xmath80 .
the gluing map is the same as the one used to glue @xmath64 back to @xmath59 to get @xmath57 for every @xmath56 .
this construction is shown in figure [ double ] .
it s clear that @xmath75 is a free double cover of @xmath57 .
we denote @xmath316 to be the covering map .
then @xmath317 @xmath318 @xmath319 @xmath75 is also a graph manifold and figure [ double ] illustrates the graph of the jsj - decomposition of @xmath75
. the graph decomposition of @xmath75.,width=384 ] by construction , @xmath75 is fibred . in step 2 , we construct @xmath320 as a leaf of the semi - bundle structure of @xmath57 .
let @xmath321 .
it s not hard to see that @xmath322 is a connected orientable surface in @xmath75 .
let @xmath323 where @xmath324 .
then @xmath325 is a surface bundle in @xmath75 , and @xmath322 is one leaf of @xmath325 . fix a transverse orientation for @xmath326 and let @xmath327 and @xmath328 have the induced transverse orientation , @xmath329 , @xmath56 .
@xmath316 can be extended to a double cover of @xmath45 .
denote the covering space by @xmath76 .
@xmath76 inherits the seifert fibred structure from @xmath45 with fibre @xmath330 .
let @xmath330 have the inherited orientation from @xmath273 .
@xmath331 where @xmath332 and @xmath333 , @xmath329 , @xmath56 . by construction , @xmath334
are two copies of @xmath55 , denoted @xmath335 , and @xmath336 .
@xmath337 , @xmath80
. then @xmath338 equip @xmath339 with the inherited orientation from @xmath55 , @xmath340 . by the construction of @xmath75
, @xmath341 is a @xmath330-vertical torus , @xmath342 .
let @xmath343 , and @xmath344 .
denote @xmath345 , @xmath346 , where @xmath347 , @xmath348 , and @xmath349 .
denote @xmath350 .
@xmath351 has two components , and two - fold trivially covers its seifert quotient image , since @xmath62 is a two - fold cyclic cover of @xmath352 .
then @xmath353 has @xmath354 components .
suppose @xmath355 to be the lift of @xmath356 .
equip @xmath357 with the inherited orientation , @xmath358 .
now we fix @xmath329 in the following discussion . by construction ,
proposition [ m2j ] holds for every @xmath359 , @xmath56 .
in particular , we can isotope @xmath360 smoothly along @xmath330 such that @xmath361 pass from the negative side to the positive side of @xmath328 s leaves when traveling along their orientation in @xmath359 .
this isotopy is fixed outside a small neighborhood of @xmath362
. then @xmath363 are transverse to @xmath364 .
next , we perform dehn twist operations a sufficiently large number of times on @xmath365 along a set of tori @xmath82 , such that the new surface fiber is transverse to @xmath366 everywhere . unlike in @xcite , here we need more vertical tori .
@xmath82 consists @xmath367 @xmath330-vertical tori,@xmath368 .
@xmath369 are @xmath370 vertical tori , where @xmath371 parallel to @xmath372 for @xmath373 .
in addition to these boundary parallel tori , we need @xmath374 extra tori , which we will give the construction later .
@xmath375 is a @xmath330-vertical annulus , @xmath346 .
@xmath376 , and transverse to @xmath330 by lemma 2.1 in @xcite , @xmath377 . by the construction of @xmath378 in step 2
, we may assume that the oriented arcs @xmath379 are always transverse from the negative to the positive side of @xmath380 near the boundary of @xmath375 , @xmath377 . by the construction of @xmath381 in step 2
, @xmath380 is transverse to the original fibre @xmath330 .
@xmath382 gives another foliation of @xmath375 , @xmath377 .
one leaf of @xmath382 is shown in figure [ dehntwistu ] .
dehn twist of @xmath327.,width=576 ] let @xmath371 be a @xmath330-vertical torus in a small neighborhood of @xmath372 in @xmath383 and parallel to @xmath372 , @xmath340 . by the above discussion we may assume that @xmath371 intersect @xmath379 transversely , @xmath377 .
now we introduce a sign for every intersection point of an arc @xmath384 and a @xmath330-vertical torus @xmath385 in @xmath75 .
suppose @xmath386 and @xmath387 are the images of @xmath384 and @xmath385 under the seifert quotient map respectively .
we say @xmath384 intersects t _ positively _ if the algebraic intersection number @xmath388(cf .
figure [ sign ] ) , otherwise we say @xmath384 intersects t _ negatively_. equip the image of @xmath371 under the seifert quotient map the same orientation as @xmath339 , @xmath56 .
it s easy to see that if the tail of @xmath389 is in @xmath390 , @xmath389 intersects @xmath371 positively,(cf .
figure [ dehntwistu]-1),@xmath391 .
algebraic intersection number ] the following remark is obtained from table [ orientl ] .
[ intersection ] any of @xmath392 , intersects @xmath393 twice and in different signs , and any of @xmath394 intersects @xmath393 once and positively .
note that @xmath393 only intersects the above arcs . by the above remark , when we perform dehn twist along @xmath369 , the relation between @xmath392 and the new fibres are the same as the relation of those with the old fibres . to make @xmath392 transverse to the new fibres , we need extra dehn twists .
we now construct @xmath395 .
@xmath396 is the lift of @xmath273-vertical torus @xmath397 , where @xmath398 .
@xmath399 and @xmath400 are two simple closed curves in @xmath401 and constructed as follows .
recall that @xmath201 , @xmath202 , @xmath56 .
let @xmath402 be the intersection of @xmath403 and a small disk centered at @xmath160 with radius bigger than @xmath199 , @xmath56 .
push @xmath402 into @xmath401 a little bit , as shown in figure [ l1l2 ] . by remark [ lij ] , @xmath404 intersect @xmath157 at @xmath160 .
then @xmath405 , @xmath169 , @xmath406 , @xmath142 .
since @xmath407 intersects @xmath157 only at @xmath160 , @xmath159 and @xmath158 , we can take an arc @xmath408 in @xmath401 which is parallel to @xmath403 and between @xmath159 and @xmath158 , @xmath56 , @xmath142 ( cf .
figure [ l1l2 ] ) .
@xmath399 and @xmath400,width=566 ] by remark [ lij ] and remark [ arcs ] , we can connect @xmath409 @xmath410 .
note that the indices are in @xmath411 .
then we get two simple closed curve @xmath399 , and @xmath400 in @xmath401 , ( cf .
figure [ l1l2 ] ) . @xmath399 and @xmath400 can be oriented as in figure [ l1l2 ] , such that the orientations on @xmath412 and @xmath413 are the same as the orientation on @xmath157 . by the construction of @xmath399 and @xmath400 , they only intersect @xmath205 at the arcs @xmath412 and @xmath413 , @xmath56 .
@xmath414 only intersects the arcs of @xmath205 between @xmath165 and @xmath415 , and between @xmath415 and @xmath207 , i.e. @xmath211 and @xmath416 by remark [ arcs ] . @xmath417 intersects @xmath211 or @xmath416 exactly once and does not intersect @xmath221 , @xmath418 , @xmath39 .
moreover @xmath419 , since the angle between @xmath157 and @xmath175 at @xmath420 after the reorientation in step 3(cf .
figure [ l1l2 ] ) .
since @xmath421 and @xmath422 are the lifts of @xmath423 and @xmath424 respectively , and their images under the seifert quotient map are @xmath425 and @xmath426 , respectively , @xmath427 , we have the following remark .
[ intersection2 ] @xmath428 intersects @xmath429 and @xmath430 exactly once and negatively , and @xmath431 , @xmath418 , @xmath39 , @xmath432 note that @xmath433 , so we may suppose that the tori in @xmath82 are mutually disjoint .
let @xmath434 be a small regular neighborhood of @xmath371 in @xmath383 consisting of @xmath330-circle fibres , @xmath435 .
we may suppose that @xmath436 if we take the regular neighborhoods small enough .
define @xmath437 .
@xmath438 and @xmath439 are two @xmath330-vertical tori , @xmath435 . by remark [ intersection ]
@xmath440 intersects each of @xmath441 and @xmath442 in a single arc , @xmath443 , @xmath39 , @xmath80 .
we may assume that the tail of this arc ( with the induced orientation ) is contained in @xmath438 , since @xmath444 intersects @xmath440 in the same sign .
on the other hand , @xmath445 intersects each of @xmath446 in a single arc , by remark [ intersection2 ] .
we also assume that the tail of this arc ( with the induced orientation ) contained in @xmath447 , @xmath142 .
since @xmath327 is transverse to the old fibres , @xmath448 is a foliation of @xmath434 by annuli , @xmath435 , @xmath80 .
the dehn twist operation on @xmath327 , denoted by @xmath449 wraps these annuli sufficiently large number of times around the @xmath330-fibres in the direction opposite to the transverse orientation of @xmath327 as we pass from @xmath438 to @xmath439 , @xmath435 .
figure [ dt ] shows one dehn twist . dehn twist in @xmath434 in the direction opposite to the transverse orientation of @xmath327.,width=412 ] we perform dehn twist operations on @xmath380 , by two steps . at first , perform @xmath450 , then perform @xmath451 and @xmath452 . let @xmath453 be the new surface bundle obtained by dehn twist operations @xmath450 on @xmath327 . by remark [ intersection ] , there are two cases for the foliation of @xmath375 by @xmath454 , @xmath80 .
case a. @xmath455 and @xmath456 , or @xmath457 and @xmath458 .
@xmath342 in this case @xmath459 and @xmath460 are both _ two _ arcs .
in particular @xmath461 and @xmath462 intersect @xmath463 and @xmath464 , and @xmath465 and @xmath466 intersect @xmath467 twice by table [ orientl ] .
then @xmath468 intersect @xmath440 two times , and in different signs by remark [ intersection ] .
if the tail of one arc of @xmath469 lies in @xmath470 , the tail of the other one lies in @xmath471 .
the same thing happens for @xmath472 .
@xmath473 is obtained from @xmath474 by @xmath449 s ( @xmath56 , @xmath80 ) in a regular neighborhood of _ two _ @xmath330-circle fibres in @xmath475 in the direction opposite to the transverse orientation of @xmath327 as we pass from @xmath476 to @xmath477 , @xmath377 .
figure [ dehntwistu]-2 shows @xmath478 , and @xmath479 is similar to @xmath478 .
now we perform an isotope of @xmath453 in a small regular neighborhood of @xmath475 in @xmath480 , say @xmath481 , such that @xmath473 changes back to @xmath474 as shown in figure [ dehntwistu]-1 , and @xmath482 always keeps to be the same .
the isotope is described as following .
push the whole arc @xmath483 ( like a finger move ) along @xmath330 in @xmath475 in the direction opposite to the dehn twist and the same times as the dehn twist , meanwhile keep @xmath484 all the time ( cf . figure [ dehntwistu]-2 ) .
we still call the surface bundle @xmath453 after this isotope . in case
a , @xmath473 is same as @xmath474 . case b. @xmath485 . in this case ,
@xmath459 and @xmath460 are both _ one _ arc .
the foliation of @xmath486 determined by @xmath487 is obtained from the foliation @xmath488 by @xmath449 s @xmath489 in a regular neighborhood of one @xmath330-circle fibre in @xmath486 in the direction opposite to the transverse orientation of @xmath327 as we pass from @xmath490 to@xmath491 ( cf .
figure [ dehntwistu]-3 ) .
now we perform the dehn twist operations @xmath451 , @xmath452 on @xmath492 .
suppose @xmath493 is the new surface bundle of @xmath480 obtained by @xmath451 and @xmath452 . by remark [ intersection2 ] , @xmath494 and @xmath495 are both one arc , @xmath496 .
then the situation of @xmath497 is the same as case b in the first step ( cf .
figure [ dehntwistu]-3 ) , @xmath289 .
the foliation of @xmath486 determined by @xmath498 is the same as @xmath487 because @xmath499 . finally like @xcite
, we adjust @xmath493 by isotope which is the identity in a small regular neighborhood of @xmath500 and outside a small regular neighborhood of @xmath501 such that the interval foliation in each @xmath475 becomes transverse to @xmath502 , @xmath377 ( cf .
figure [ dehntwistu]-4 ) .
we denote the resulting surface bundle @xmath503 . since @xmath504 is same as @xmath505 , the resulting surface bundle @xmath506 in @xmath75 is transverse to the link @xmath507 every where . in other words ,
the exterior of @xmath507 in @xmath75 has a surface bundle structure , i.e. , @xmath507 is a fibred link , i.e. @xmath2 is a virtually fibred montesinos knot .
now we finish the proof of theorem [ main ] when @xmath2 is a classic montesinos knot , and @xmath22 .
the proof in this case is mostly same as in section 2.1 , except for the following changes . in this case
, @xmath509 is a trivial 2-fold cover .
then @xmath194 is a geodesic link with @xmath354 components .
let @xmath510 .
@xmath511 , @xmath142 .
now we take @xmath57 to be the complement of @xmath512 in @xmath45 . as in the proof of proposition [ semibundle ]
, we have @xmath513 , where @xmath514)-\overset{\circ}{n}(l_{j , j}^1)$ ] , @xmath56 .
note that @xmath515 is a fibre of the seifert fibre structure on @xmath516 .
since @xmath2 has two components , and @xmath11 is odd , we have @xmath517 is even .
@xmath518 so @xmath519 is even and non - zero .
we use the definitions of @xmath520 in the proof of proposition [ semibundle ] , and in this case @xmath521 . we can take @xmath522 and @xmath523 .
i.e. @xmath524 in the following discussion , we will always assume that @xmath56 except for special indication . by a similar analysis in proposition [ semibundle ]
, we can see that there are orientable horizontal surfaces @xmath381 in @xmath59 , and @xmath525 in @xmath64 which piece together to form the semifibre @xmath227 where the projection of @xmath381 to @xmath401 has degree @xmath526 and that of @xmath525 to the base of @xmath64 has degree @xmath527 .
further , if we suppose that the slope of @xmath381 on @xmath528 is given by @xmath529 , and that of @xmath525 on @xmath528 is @xmath530 . then we get the following result from ( [ slope12 ] ) and ( [ slope3 ] ) @xmath531 since @xmath532 is even , the values of the coefficients @xmath533 are given as following @xmath534 ( -1 , \displaystyle{\frac{\tilde{e}}{4 } } ) & \text{if $ \tilde{e}=4k$ } , \end{cases } k\in\mathbb{z}.\end{aligned}\ ] ] as in 2.1 @xmath525 is a surface which interpolates between the slope @xmath535 on @xmath536 and @xmath537 on @xmath315 .
@xmath538 has @xmath539 component @xmath173 , and @xmath540 has @xmath541 component @xmath186 .
still @xmath542 implies that @xmath57 is a surface semi - bundle .
denote the associated fibring in @xmath64 by @xmath65 .
@xmath65 is transverse to all the new fibres , and in particular to @xmath515 .
now @xmath543 only intersects the original fibre once , so @xmath544 is a once - punctured annulus , and there is only one singular point on @xmath545 .
@xmath237 and @xmath238.,width=480 ] the intersection of @xmath237 with @xmath243 is depicted in figure [ foliation2]-1 , @xmath264 is depicted in figure [ foliation2]-2 , and @xmath263 is similar to @xmath546 .
the previous argument can be proceed without significant change to produce the desired surface bundle structure on @xmath359 which is transverse to @xmath547 . by similar dehn twist operation on @xmath327
, we can get a surface bundle in @xmath480 , say @xmath453 such that @xmath453 is transverse to @xmath548 .
the proof of theorem [ main ] is finished in case ( 1 ) .
in this section , we consider classic montesinos link @xmath549 , where @xmath25 , @xmath11 odd and @xmath9 .
we do not consider the case @xmath550 , otherwise @xmath2 is virtually fibred by theorem 1.2 @xcite .
the seifert structure is @xmath551 , and @xmath552 has a presentation @xmath553 define @xmath554 similarly as in sec .
@xmath96 where @xmath97 , and @xmath555 , for @xmath556 .
let @xmath557 be the @xmath11-fold cyclic orbifold cover of @xmath558 corresponding to @xmath554 .
then @xmath559 and @xmath560 define @xmath561 where @xmath562 for @xmath563 .
suppose that @xmath136 is the @xmath11-fold cyclic orbifold cover of @xmath99 corresponding to @xmath564 , and @xmath565 .
now let @xmath192 be the associated @xmath41-fold cover .
@xmath194 is still a geodesic link in @xmath45 with exactly @xmath41 components .
we use the same notations as in section 2 .
@xmath566 , @xmath567 , @xmath568 , and @xmath569 , @xmath570 .
let @xmath571 . by the same process from step 2
to step 4 as described in section 2.1 , we can find a double cover of @xmath57 , say @xmath75 and a surface bundle @xmath572 in @xmath75 such that it is transverse to the lift of @xmath291 , @xmath573 .
the only thing different here is the construction of @xmath82 . in step 4 ,
we perform dehn twist operations on @xmath327 along a set of @xmath330-vertical tori @xmath82 such that the new surface fibre is transverse to @xmath468 everywhere , @xmath574 , @xmath80 , @xmath418 , @xmath39 .
@xmath575 contains four tori not @xmath367 as in case ( 1 ) .
they are the lifts of two @xmath273-vertical tori , @xmath576 , @xmath173 .
@xmath399 and @xmath400 are two simple closed curves on @xmath42 , which intersect all @xmath577 s once or twice when @xmath578 in the same signs , @xmath579 , and @xmath580 .
@xmath402 is still the intersection of @xmath403 and a small disk centered at @xmath160 with radius bigger than @xmath199 , @xmath56 , @xmath142 .
push @xmath402 into @xmath401 a little bit .
@xmath408 is different .
we take @xmath408 in @xmath401 which is parallel to @xmath403 and between @xmath159 and @xmath596 , @xmath56 , @xmath142 .
then @xmath597 intersects @xmath587 once , where @xmath496 , and @xmath598 , intersects @xmath599 and @xmath600 both once and intersect @xmath587 twice where @xmath601 , @xmath418 , @xmath39 .
we still can connect @xmath402 s and @xmath408 s by the same arcs as in ( [ connect ] ) , @xmath56 ,
@xmath142 . see figure [ n=4 ] in the case @xmath602 , and figure [ n=5 ] in the case @xmath603 .
when @xmath604 , @xmath399 and @xmath400 are similar to the ones in the case @xmath603 .
now we perform a dehn twist operation @xmath614 which wraps these annuli @xmath615 times around the @xmath330-fibres in the direction opposite to the transverse orientation of @xmath327 as we pass from @xmath612 to @xmath616 , @xmath142 , @xmath80 . by remark [ intersection3 ] @xmath468 intersects
@xmath617 once , @xmath473 is similar as in figure [ dehntwistu]-3 , when @xmath609 .
when @xmath601 , @xmath468 intersects @xmath617 twice and in the same sign .
@xmath473 is shown in figure [ foliationu2]-1 . in this case
, @xmath468 goes from @xmath618 to @xmath619 .
now @xmath473 is wrapped @xmath620 times around the @xmath330-fibres in the direction opposite to the transverse orientation of @xmath327 as we pass from @xmath621 to @xmath622 , @xmath418 , @xmath39 , @xmath601 .
we adjust @xmath453 by isotope , and denote the resulting surface bundle @xmath572 in @xmath75 such that it is transverse to @xmath468 as shown in figure [ foliationu2]-2 .
99 ian agol , steven boyer , and xingru zhang , _ virtually fibred montesinos links _ , journal of topology , to appear .
g. walsh , _ great circle links and virtually fibred knots _ , topology , 44 ( 2005 ) no . 5 , 947 - 958
. s. wang and f. yu , _ graph manifolds with non - empty boundary are covered by surface bundles _ , math .
( 1997 ) 447 - 455 .
|
we find a larger class of virtually fibred classic montesinos links of type @xmath0 , extending a result of agol , boyer and zhang .
|
neutron stars are associated with the most diverse and energetic phenomena in the universe , from gamma - ray bursts to the emission of gravitational waves and from periodic millisecond radio signals to month - long x - ray outbursts .
the strength and even the occurrence of some of these phenomena depend on the neutron star mass . for decades
, precise dynamical measurements of the masses of neutron stars resulted in highly clustered values around 1.251.4 solar masses @xcite .
this paradigm has recently changed with the measurement of the neutron star mass for psr j1614@xmath02230 .
psr j1614@xmath02230 is a 3.1 ms radio pulsar in a 8.7 day orbit around a massive white dwarf @xcite .
its very high inclination , at 89.17@xmath4 , allowed the detection of a strong shapiro delay signature in highly accurate timing observations with the green bank telescope and a precise measurement of the neutron star mass .
the inferred value of 1.97@xmath5 is by far the highest observed from any neutron star to date @xcite .
the high mass of psr j1614@xmath02230 provides a lower limit on the maximum mass of neutron stars and sets the dividing line between neutron stars and black holes to at least this value .
this has significant implications for the fraction of neutron stars that will collapse into black holes as a result of mass accretion in x - ray binaries . for a maximum neutron star mass that is @xmath6 , the relative frequency of black hole binaries that are formed through
this evolutionary channel is at most @xmath7 ( pfahl , rappaport , & podsiadlowski 2003 ) .
such a small fraction would help explain the lack of low - mass black holes in binary systems with dynamical mass measurements @xcite .
the highest neutron star mass that can be supported against collapse depends very sensitively on the underlying equation of state .
in particular , if quark , hyperon , or boson degrees of freedom are excited at high densities , the equation of state softens and can not support massive neutron stars ( see , e.g. , lattimer & prakash 2001 ) .
the discovery of even a single very high mass neutron star can , therefore , provide strong constraints on the fundamental properties of ultradense matter . in this _ letter _ , we explore in detail the constraints imposed by the recent mass measurement on the radii of neutron stars and on the properties of quark matter in their interiors .
the maximum neutron star mass also plays a crucial role in determining the outcome of the coalescence of two neutron stars .
these astrophysical events are thought to be the central engines of short - duration gamma - ray bursts ( see reviews by nakar 2007 and lee & ramirez - ruiz 2007 ) .
furthermore , they are expected to be the primary sources of gravitational waves that will be detected by ground - based observatories such as ligo , virgo , and geo600 . in 3 , we investigate the viability of the coalescing neutron - star scenario for short - duration gamma - ray bursts in view of the above lower limit on the maximum mass of neutron stars .
finally , in 4 , we discuss the prospect of measuring the equation of state of neutron - star matter using ground - based gravitational wave observatories .
the high mass of the psr j1614@xmath02230 , in conjuction with the requirement of causality for the neutron star matter equation of state , places a strong lower limit on its radius , independent of the composition of its interior . assuming an equation of state that arises from normal interactions in the outermost neutron star layers ( where the density @xmath8 is less than a characteristic value @xmath9 ) and requiring that the sound speed @xmath10 everywhere in the neutron star core is less than the speed of light @xmath11 , such that @xmath12 , leads to a firm lower limit on the radius @xmath13 of a neutron star as a function of its mass @xmath14 @xcite
this is , of course , true only for neutron stars , which are gravitationally bound and have densities that drop smoothly to zero in their atmospheres ; the causality limit does not apply to strange stars .
the bps equation of state ( baym , pethick , & sutherland 1971 ) is a standard choice for low - density matter .
assuming it is accurate up to a density @xmath15 g @xmath16 , which is comparable to the nuclear saturation density of atomic nuclei @xmath17 g @xmath16 , the bound imposed by causality is @xmath18 all realistic equations of state for gravitationally bound stars predict radii that are either very weakly dependent on or decreasing with increasing neutron - star mass @xcite . the causality bound , in conjunction with the recent spectroscopic measurements of neutron - star radii in x - ray binaries that place them at values @xmath19 km ( zel , baym , & gver 2010 , steiner , lattimer , & brown 2010 ) , lead to the very narrow range @xmath20 km for all neutron stars . from shapiro delay observations
indicates that , if the transition to quark matter occurs at densities that are relevant to neutron star interiors , such a massive star can be supported against collapse only if the quarks are strongly interacting ( @xmath21 ) . ]
[ fig : fig1 ] stronger constraints on the composition of the neutron star interior can be placed by requiring that the equation of state support a maximum neutron star mass that is at least as high as the measured value . in general
, the appearance of new degrees of freedom at and above the nuclear saturation density , such as quarks , hyperons , or bosons , softens the equation of state and lowers the maximum mass that can be supported against collapse . in figure 1
, we show how the mass measurement for psr j1614@xmath02230 constrains the quark matter equation of state .
we use a specific model of quark matter , the phenomenological equation of state proposed by alford et al .
( 2005 ) , which has three parameters : @xmath22 , @xmath23 , and the bag coefficient . the quartic coefficient @xmath22 measures the degree of interaction between quarks , with @xmath24 corresponding to free quark plasma .
the quadratic coefficient @xmath23 depends on the mass of the strange quark and the color superconducting energy gap ; it is large in unpaired quark matter with a heavy strange quark , and small in color superconducting ( for example color - flavor locked ) quark matter with a light strange quark .
we calculate the maximum neutron star mass @xmath25 for a range of values of @xmath22 and @xmath23 , while fixing the bag coefficient so that the transition from nuclear matter to quark matter occurs when the baryon density @xmath26 of the nuclear matter is @xmath27 , where nuclear density is @xmath28 . for the nuclear matter equation of state , we use bps ( baym , pethick , & sutherland 1971 ) at low densities and apr ( akmal , pandharipande , & ravenhall 1998 ) , which we assume is valid up to a density @xmath29 .
we chose the apr equation of state because it describes well the nucleon - nucleon scattering data and is believed to be accurate up to the nuclear saturation density .
the excluded region in the lower right part of the figure is the region where , after the transition from nuclear to quark matter at @xmath30 , there is another transition back to nuclear matter at @xmath31 .
such a transition seems unlikely to be physical , so we do not calculate masses in this region .
we see from figure 1 that values of @xmath25 compatible with the measurement for psr j1614@xmath02230 can only be achieved if @xmath32 and @xmath33 . in alford et al .
( 2005 ) , such values were associated with strong interactions between the quarks ( pushing @xmath22 well below unity ) and color superconductivity ( which allows @xmath23 to remain small by cancelling the contribution to @xmath23 from the strange quark mass ) .
short gamma - ray bursts have characteristic timescales @xmath34 s , which are much longer than the dynamical timescale for the coalescence of two neutron stars . in the coalescing neutron star scenario ,
the required longer timescales can be achieved in two ways @xcite .
the initial remnant may be a differentially rotating massive neutron star that is temporarily supported by centrifugal forces .
the dissipation of the differential rotation on timescales @xmath35 s leads to the delayed collapse of the remnant into a black hole and produces the gamma - ray burst . in order for this channel to occur , the sum of the masses in the initial system has to be larger than the maximum neutron star mass for solid body rotation but less than the maximum mass that can be supported by differential rotation .
numerical simulations of spinning neutron stars place these limits at @xmath36 and @xmath37 , respectively , where @xmath25 is the maximum mass of a non - spinning neutron star @xcite ; the exact value of this limit depends on the radial profile of the differential rotation . as a result , the delayed collapse channel requires that @xmath38 the second channel involves the prompt collapse of the remnant of the coalescence into a black hole and the formation of a @xmath39 torus around it that is accreted onto the black hole in the required timescale .
all numerical simulations of coalescing neutron stars show that equal mass mergers hamper the formation of these massive thick accretion disks . if , on the other hand , the ratio of the merging neutron star masses is larger than @xmath40 , the lower mass neutron star fills its roche lobe before contact and loses the required amount of mass into a torus before the remnant collapses into a black hole @xcite .
( the value of this ratio depends on the mass of the torus required to achieve the timescale of the grbs by the subsequent accretion of this torus ; see rezzolla et al.2010 ) .
this prompt collapse channel , therefore , requires that the total mass of the coalescing neutron stars is @xmath41 and their mass ratio is @xmath42 . in figure 2 , we calculate the outcome of mergers in the parameter space defined by the masses of the two coalescing neutron stars and delineate the areas in which short duration gamma - ray bursts can be generated via either channel .
the two panels demonstrate the very strong sensitivity of these outcomes on the maximum mass that can be supported against collapse in a non - rotating neutron star . for a maximum neutron star mass of @xmath43 ,
the allowed area in this parameter space is marginal ( left panel ) but increases rapidly as @xmath25 grows to values larger than @xmath44 ( right panel ) .
therefore , the unequivocal detection of a 1.97 @xmath2 neutron star makes the idea of coalescing neutron stars as the central engines of short duration gamma - ray bursts viable .
figure 2 shows that the largest part of the parameter space that leads to short duration gamma - ray bursts is populated by double neutron star systems in which one of the neutron stars is significantly more massive than the other .
this is in contrast to the double neutron star systems observed in the galactic disk , where both neutron stars have surprisingly similar masses @xcite .
indeed , a significant amount of mass transfer onto one of the two neutron stars is likely to take place only in a low - mass x - ray binary system , which does not lead to the formation of a double neutron star .
double neutron stars with high mass ratios are perhaps most easily formed in globular clusters via exchange interactions .
globular clusters have large populations of low - mass x - ray binaries and , therefore , may harbor the systems responsible for the short duration gamma - ray bursts ( see also grindlay et al . 2007 ) .
this would explain the recent observations of optical counterparts of short duration gamma - ray bursts , which appear to lie outside the host galaxies ( e.g. , berger 2010 ) .
the coalescence of two neutron stars is also expected to be the primary source of gravitational waves detected with current and future ground - based detectors such as ligo , virgo , and geo 600 @xcite .
detailed modeling of the waveforms during the final stages of the coalescence can place strong constraints on the equation of state of neutron - star matter @xcite
. this will be possible with a generation of detectors beyond advanced ligo , as it requires following the gravitational wave signal to frequencies @xmath45 khz .
however , even the low - frequency waveforms may encode the characteristics of the equation of state as they are affected by the tidal deformation experienced by the two neutron stars at larger separations @xcite .
the degree of tidal deformation , as measured by the love number , is relatively insensitive to the compactness of the neutron star but depends strongly on how centrally condensed the density profile inside the stars is @xcite , because stars that are centrally condensed are weakly deformed in the presence of an external tidal gravitational field .
equations of state that allow pulsar masses @xmath46 typically yield neutron stars that are not centrally condensed ( as can be inferred from the polytropic models of hinderer 2008 )
. therefore , the discovery of the 1.97 @xmath2 pulsar indicates that coalescing neutron stars will undergo significant tidal deformation , and information about the equation of state of the stellar interior will be encoded in low - frequency ( @xmath47 hz ) gravitational waves emitted during coalescence .
detection of these waves will allow accurate measurements of the equation of state of neutron - star matter in the near future .
we thank sanjay reddy , luciano rezzola , and chryssa kouveliotou for useful discussions and comments .
f acknowledges support from the nasa adap grant nnx10ae89 g , nsf grant ast 07 - 08640 , and chandra theory grant tmo-11003x .
dp was supported by the nsf career award nsf 0746549 and chandra theory grant tmo-11003x .
pd is a jansky fellow of the national radio astronomy observatory .
the national radio astronomy observatory is a facility of the national science foundation operated under cooperative agreement by associated universities , inc .
|
the recent measurement of the shapiro delay in the radio pulsar psr j1614@xmath02230yielded a mass of [email protected] @xmath2 , making it the most massive pulsar known to date .
its mass is high enough that , even without an accompanying measurement of the stellar radius , it has a strong impact on our understanding of nuclear matter , gamma - ray bursts , and the generation of gravitational waves from coalescing neutron stars .
this single high mass value indicates that a transition to quark matter in neutron - star cores can occur at densities comparable to the nuclear saturation density only if the quarks are strongly interacting and are color superconducting .
we further show that a high maximum neutron - star mass is required if short duration gamma - ray bursts are powered by coalescing neutron stars and , therefore , this mechanism becomes viable in the light of the recent measurement . finally , we argue that the low - frequency ( @xmath3 hz ) gravitational waves emitted during the final stages of neutron - star coalescence encode the properties of the equation of state because neutron stars consistent with this measurement can not be centrally condensed .
this will facilitate the measurement of the neutron star equation of state with advanced ligo / virgo .
|
although the determination of capacity region of the interference channel ( ic ) has been a long standing open problem , several interesting recent results shed light on the problem from various perspectives . among these
we may cite the capacity region obtained for special cases @xcite , or obtained for general channel classes in both scalar and mimo settings up to approximations with bounded gaps @xcite .
when specializing to the large snr regime , it is known that the characterization of the full capacity region can be conveniently replaced with the determination of the so - called degree - of - freedom ( dof ) region .
progress on that particular front was reported in @xcite with the derivation of the dof region for the two - user mimo interference channel with @xmath0 , @xmath1 transmit antennas and @xmath2 , @xmath3 receive antennas , where the sum dof @xmath4 is shown to be optimal .
most of these advances suggest achievable schemes which require the full knowledge of channel state information ( csi ) at both the transmitter and receiver sides .
in fact , the cruciality of csi at the transmitter side in particular is demonstrated in such works as @xcite where the dof region is shown to shrink dramatically when zero csit is available .
the intermediate scenario of _ limited _ or _ incomplete _ csit was also considered in @xcite . in @xcite , the rate of limited feedback needed to preserve the dof optimality in interference alignment - enabled ic is provided .
more recently , the impact of feedback delays providing the transmitter with outdated csi over mimo channels was considered in @xcite for the broadcast channel ( bc ) and later extended to the ic @xcite .
the key contribution in @xcite was to establish the usefulness of even completely outdated channel state information in designing precoders achieving significantly better dof than what is obtained without any csit .
considering the worst case scenarios , including those where the feedback delay extends beyond the coherence period of the time varying fading channels , the authors in @xcite propose a space - time interference alignment - inspired strategy achieving an optimal sum dof of 4/3 for the two - user miso bc , in a setting when the no csit case yields no more than 1 dof .
the essential ingredient for the proposed scheme in @xcite lies in the use of multi - slot protocol initiating with the transmission of unprecoded information symbols to the user terminals , followed by the analog forwarding of the interference overheard in the first time slot .
recently , this strategy was generalized under similar principle to the interference channel setting @xcite , again establishing dof strictly beyond the ones obtained without csit in scenarios where the delayed csit bears no correlation with the current channel realization .
albeit inspiring and fascinating in nature , such results nonetheless rely on the somewhat over - pessimistic assumption that no estimate for the _ current _ channel realization is available to the transmitter . owing to the finite doppler spread behavior of fading channels , it is however the case in many real life situations that the past channel realizations can provide information about the current one
. therefore a scenario where the transmitter is endowed with delayed csi in addition to some ( albeit imperfect ) estimate of the current channel is practical relevance .
this form of combined delayed and imperfect current csit was recently introduced in @xcite for the multiple - antenna broadcast channel whereby a novel transmission scheme is proposed which extends beyond the mat algorithm in allowing the exploitation of precoders designed based on the current csit estimate .
the full characterization of the optimal dof for the hybrid csit was reported in @xcite and independently in @xcite .
the key idea behind the schemes in @xcite lies in the modification of the mat protocol where i ) the initial time slot involves transmission of _ precoded _ symbols , followed by the forwarding of _ residual _ interference overheard in the first time slot , and ii ) the taking advantage of the reduced power for the residual interference ( compared with full power interference in mat ) based on a suitable quantization method and digital transmission . in this paper
, we extend the results in @xcite and consider the two - user time - correlated multiple - antenna interference channel .
a similar hybrid csit scenario is considered whereby each transmitter has access to delayed channel samples for the links it is connected to , as well as possessing an imperfect estimate of the current channel .
the current csit estimate could be obtained from , e.g. , a linear prediction applied to past samples @xcite , although the prediction aspects are not specified in this paper .
instead , the quality level for the current csit estimate is simply modeled in terms of an exponent of the transmit power level , allowing dof characterization for various ranges of current csit quality .
thus our model bridges between previously reported csit scenarios such as the pure delayed csit of @xcite and the pure instantaneous csit scenarios .
we assume each receiver has access to its own perfect instantaneous csi and the perfect delayed csi of other receivers ( as in e.g. @xcite ) , in addition to the imperfect current csi .
in what follows we obtain the following key results : * we establish an outer bound on the dof region for the two - user temporally - correlated miso interference channel with perfect delayed and imperfect current csit , as a function of the current csit quality exponent .
this result is initially derived for the two - antenna transmitters and then generalized .
* we propose two schemes which achieve the key vertices of the outer bound with perfect delayed and imperfect current csit .
the schemes build on the principles of time - slotted protocol , starting with the zf precoded transmission of information symbols from the two interfering transmitters simultaneously and followed by forwarding of the residual interferences . as in the bc case ,
the residual interference reflects on the quality of the initial precoder and can be shown to be quantized and power scaled in a suitable way to achieve the optimal dof .
* our results coincide with previously reported dof results for the perfect csit setting ( current csit of perfect quality ) and pure delayed csit setting ( current csit of zero quality ) . * the dof region of certain mimo cases
is also provided as a function of the current csit quality exponent and the number of receive antennas .
* notation * : matrices and vectors are represented as uppercase and lowercase letters , and matrix transport , hermitian transport , inverse and determinant are denoted by @xmath5 , @xmath6 , @xmath7 and @xmath8 , respectively .
@xmath9 is the normalized orthogonal component of any nonzero vector @xmath10 .
the approximation @xmath11 is in the sense of @xmath12 , where @xmath13 is a constant that does not scale as @xmath14 .
@xmath15 means @xmath16 is symmetric positive semidefinite if @xmath16 is square and @xmath17 means @xmath18 is symmetric positive semidefinite if both @xmath16 and @xmath19 are squared matrices .
we consider a two - user miso interference channel , where two transmitters each equipped with @xmath20 antennas wish to send two private messages to their respective receivers each with a single antenna , as shown in fig . 1 .
the discrete time baseband signal model is given by @xmath21 for any time instant @xmath22 , where @xmath23 is the channel vector from tx-@xmath24 to rx-@xmath25 ; @xmath26 are normalized additive white gaussian noise ( awgn ) at the respective receivers ; the coded input signal @xmath27 is subject to the power constraint @xmath28 , @xmath29 .
[ mutually independent fading ] at any given time instant @xmath22 , the channel vectors @xmath30 are mutually independent and identically distributed ( i.i.d . ) with zero mean and covariance matrix @xmath31 .
[ perfect delayed local csit and imperfect current local csit ] at each time instant @xmath22 , tx-@xmath24 knows perfectly the delayed local csit @xmath32 ( with which link it is respectively connected ) , and somehow predict / estimate imperfectly the current local csit @xmath33 , which can be modeled by @xmath34 where the estimate @xmath35 and estimation error @xmath36 are independent and assumed to be zero - mean and with variance @xmath37 , @xmath38 , respectively ( @xmath39 ) .
[ perfect delayed csir , imperfect current csir and perfect current local csir ] at each time instant @xmath22 , rx-@xmath24 knows perfectly the delayed csir up to instant @xmath40 for all links , i.e. , @xmath41 , where @xmath42 and the imperfect current csir ( similarly modeled as at the transmitters ) up to instant @xmath22 for all links , i.e. , @xmath43 , where @xmath44 as well as the perfect current local csir , i.e. , @xmath45 .
we assume that the estimation error @xmath46 can be parameterized as an exponential function of the power @xmath14 , so that we hope to characterize the dof of the miso ic with respect to this exponent . to this end
, we introduce a parameter @xmath47 , such that @xmath48 this @xmath49 indicates the quality of current csit at high snr .
while @xmath50 reflects the case with no current csit , @xmath51 corresponds to that with perfect instantaneous csit . as a matter of fact , when @xmath52 , the quality of the imperfect current csit is sufficient to avoid the dof loss , and zf precoding with this imperfect csit is able to achieve the maximum dof @xcite .
therefore , we focus on the case @xmath53 $ ] hereafter . the connections between the above model and the linear prediction over existing time - correlated channel models with prescribed user mobility are highlighted in @xcite .
according to the definition of the estimated current csit , we have @xmath54 where @xmath55 , and ( [ eq : p - to - alpha ] ) is obtained from ( [ eq : alpha - def ] ) .
a rate pair @xmath56 is said to be achievable for the two - user interference channel with perfect delayed csit and imperfect current csit if there exists a @xmath57 code scheme consists of : * two message sets @xmath58 $ ] at the tx-1 and @xmath59 $ ] at the tx-2 , from which two independent messages @xmath60 and @xmath61 intended respectively to the rx-1 and rx-2 are uniformly chosen ; * one encoding function at the tx-@xmath24 : @xmath62 * and one decoding function at its corresponding receiver , e.g. , @xmath63 for the rx-1 when @xmath64 , and it is similarly defined for the rx-2 by replacing @xmath65 with @xmath66 , such that the average decoding error probability @xmath67 , defined as @xmath68 vanishes as the code length @xmath69 tends to infinity .
the capacity region @xmath70 is defined as the set of all achievable rate pairs .
accordingly , the dof region can be defined as follows : [ the degree - of - freedom region ] the degree - of - freedom ( dof ) region for two - user miso interference channel is defined as @xmath71 consequently , the dof region for the two - user time - correlated miso interference channel is stated in the following theorem . in the two - user miso interference channel with perfect delayed csit and imperfect current csit ( as stated in assumption 2 ) , the optimal dof region can be characterized by @xmath72 * remark * : interestingly , the above dof region is identical to that of the two - user miso broadcast channel with perfect delayed csit and imperfect current csit @xcite .
in fact , this result is consistent with previous results on the pure delayed csit case ( @xmath50 ) where it was shown that the dof region for the two - user bc and the two - user ic coincides , and also on the special case of perfect instantaneous csit ( @xmath73 ) . for illustration , the dof region for the two - user miso ic
is provided in fig . 2 .
the dof regions with no csit , pure perfect delayed csit , and perfect instantaneous csit are also plotted for comparison .
it shows that the dof region with perfect delayed csit and imperfect current csit is strictly larger than that with pure delayed csit and quickly approaches the region with perfect csit as the quality of current csit increases . ) . ]
given an @xmath49 , the dof region is a polygon whose vertices are : @xmath74 , @xmath75 , @xmath76 , @xmath77 and @xmath78 . in the following ,
we first characterize the outer bound , and then propose two schemes to show they are achievable , and in turn the entire region can be achieved by time sharing .
we adopt a strategy reminisced in @xcite to obtain the genie - aided outer bound , by assuming that ( i ) both receivers know the csi @xmath79 perfectly and instantaneously as well as the imperfect current csi @xmath80 at time @xmath22 , and ( ii ) the rx-2 has the instantaneous knowledge of the tx-1 s received signal @xmath65 .
define @xmath81 where @xmath82 denotes the channel state information and its estimated version available at the receivers from the beginning up to time instant @xmath69 . to ease our presentation ,
we denote : @xmath83 where @xmath84 tends to zero as @xmath85 by the assumption that @xmath86 .
then , we can upper - bound the achievable rate of rx-1 by applying fano s inequality : @xmath87 where ( [ eq : r-1-bound ] ) follows the fact that the rate of the point - to - point miso channel ( i.e. , tx-1 together with tx-2 are treated as the transmitter by cooperation while rx-1 as the receiver ) is bounded by @xmath88 ; ( [ eq : epsilon ] ) is due to the fact that ( a ) transmitted signals @xmath89 are determined given messages , channel matrices up to @xmath69 and the encoding functions defined in ( [ eq : enc - fun ] ) , ( b ) translation does change differential entropy , and ( c ) noise is independent of the channel matrices , the transmitted signals and the messages ; ( [ eq : r-1-remove - m1 ] ) is obtained because ( a ) the transmitted signals @xmath90 are determined provided the channel matrices , @xmath60 and the encoding functions according to ( [ eq : enc - fun ] ) , and ( b ) translation preserves differential entropy ; ( [ eq : r-1-remove - m2 ] ) follows the chain rule of differential entropy and the fact that conditioning reduces differential entropy ; the last equality is obtained due to @xmath91 is independent of @xmath92 and @xmath93 . by applying fano s inequality ,
we then also upper - bound the achievable rate of rx-2 as @xmath94 where ( [ eq : r-2-chain - rule ] ) is obtained because of the chain rule of mutual information and the independence between @xmath60 and @xmath61 ; ( [ eq : r2-remove - m1 ] ) is due to ( a ) the transmitted signals @xmath90 are determined given message @xmath60 , channel matrices and encoding functions , and ( b ) @xmath61 and @xmath95 are independent of @xmath60 ; ( [ eq : r2-markov - chain ] ) is obtained by markov chain @xmath96 and the data processing inequality ; ( [ eq : last - eq-2 ] ) is because ( a ) translation does not change differential entropy , ( b ) gaussian noise terms are independent from instant to instant , and are also independent of the channel matrices and the transmitted signals , and ( c ) the differential entropy of gaussian noise is nonnegative ; and the last equality is obtained due to the independence @xmath95 of @xmath92 and @xmath93 . according to the markov chain @xmath97 , we upper - bound the weighted sum rate as @xmath98 before preceding further , we introduce the following lemma stated in @xcite .
[ @xcite ] for an @xmath99 random vector @xmath100 where @xmath101 is independent of @xmath102 , given any @xmath103 with eigenvalues @xmath104 , we have the following upper and lower bounds : @xmath105 the difference of the upper and lower bounds can be further bounded by @xmath106 where @xmath107 is euler s constant . with the definitions @xmath108^{{\scriptscriptstyle\mathsf{t}}}\\ { { { \uppercase{{\bm{k}}}}}}(t ) & { \triangleq}{\ifthenelse{\isempty{}}{\mathbb{e}}{\mathbb{e}\left(\right)}}\{{{{\lowercase{{\bm{x}}}}}}_2(t){{{\lowercase{{\bm{x}}}}}}_2^{{{\scriptscriptstyle\mathsf{h}}}}(t)|{{\mathcal u}}(t)\}\end{aligned}\ ] ]
we further upper - bound the weighted difference of two conditional differential entropies that derived above , i.e. , @xmath109 where in ( [ eq : gaussian - input-1 ] ) @xmath110 is replaced by @xmath111 because of the independence of @xmath95 ; ( [ eq : gaussian - input-2 ] ) is obtained because gaussian distributed vector @xmath112 maximizes the weighted difference of two differential entropies over all conditional distribution of @xmath113 with the same covariance matrix constraint , where @xmath114 @xcite ; ( [ eq : gaussian - input-3 ] ) is because @xmath112 , @xmath111 and @xmath115 are independent of @xmath116 except @xmath117 ; ( [ eq : expectation - bound1 ] ) is obtained because @xmath112 is gaussian distributed and independent of @xmath118 , @xmath119 as well as the noise terms ; ( [ eq : expectation - bound3 ] ) follows the fact that putting the expectation out of the maximization increases the value ; ( [ eq : expectation - bound4 ] ) follows from the inequality @xmath120 where @xmath16 is an @xmath121 positive semidefinite matrix with entry @xmath122 ; the last two inequalities are according to lemma 1 and the quality of current csit ( @xmath123 ) . accordingly , we obtain an upper bound of @xmath124 from ( [ eq : r1 - 2r2 ] ) and ( [ eq : diff - bound ] ) , i.e. , @xmath125 as @xmath126 , from which ( [ eq : dof - bound-3 ] ) is obtained according to the definition of dof . by exchanging the roles of tx-1/rx-1 and tx-2/rx-2 , another inequality ( [ eq : dof - bound-4 ] ) can be similarly obtained by assuming rx-1 has the instantaneous knowledge of @xmath66 , where the weighted rate is bounded as @xmath127 together with the first two bounds ( [ eq : dof - bound-1 ] ) and ( [ eq : dof - bound-2 ] ) which are obtained by the constraint of antenna configuration , the dof region is completely characterized .
with perfect delayed csit , the authors in @xcite and @xcite characterize the dof region for two - user mimo interference channel , bridging between the case with no csit @xcite and that with perfect instantaneous csit @xcite . particularly , for two - user miso case , the dof pair @xmath128 is achievable with delayed csit , strictly larger than @xmath129 achieved with no cist and dominated by @xmath130 with perfect csit .
the technique exploits the advantage of interference alignment in the time domain by utilizing the delayed csit together with the space domain , which is referred to as mat alignment @xcite .
we first briefly review its application in the interference channel .
the mat alignment in the interference channel is an extension from the broadcast channel , taking into account the distributive and uncooperative nature of the transmitters @xcite .
the two - phase protocol which consumes three time slots is described as follows : in this phase , each tx transmits two independent encoded symbols to its intended receiver without precoding during a single time slot , i.e. , @xmath131 and the received signals at both receivers are @xmath132 where @xmath133 and @xmath134 are interference terms overheard at the tx-1 and tx-2 , respectively . at the end of phase - i , the delayed csit @xmath135 is available at the tx-1 , while @xmath136 is accessible at the tx-2 . together with the transmitted symbols , the overheard interference terms are reconstructible at both txs . by retransmitting the overheard interference terms @xmath137 at the tx-1 and @xmath138 at the tx-2 with time division , i.e. , @xmath139 and @xmath140 where two entire time slots are consumed ,
we cancel the interference terms @xmath133 and @xmath134 at the rx-1 and rx-2 , and importantly provide another linear combination of @xmath141 ( from @xmath134 ) and @xmath142 ( from @xmath133 ) to the rx-1 and rx-2 , respectively . by the end of phase - ii , both receivers are able to recover their own symbols with high probability .
the key idea behind is interference repetition and alignment in both space and time domain . at each receiver ,
the mutual interference aligns in one dimension , while the desired signal spans in a two - dimensional space .
this enables each receiver to retrieve the desired signal from a three - dimensional space .
the mat alignment takes into account the completely outdated csit , regardless of the correlation between current and previous channel states . as a matter of fact ,
such an assumption on the delayed csit is over - pessimistic , since the current csi can be predicted from the past states if the underlying channel exhibits some temporal correlation .
recent results demonstrate that the dof region can be enlarged in broadcast channel by using estimated current csit , even it is imperfect @xcite . in the following ,
two schemes are proposed , demonstrating the larger dof region can be achieved by utilizing estimated current csit exploited from time correlation model . instead of forwarding the interference terms in an analog fashion @xcite , we first quantize the interference and then retransmit the quantized version . by utilizing the imperfect current csit for precoding ,
the interference terms are efficiently compressed with quantization . in the following two schemes ,
we demonstrate the vertices @xmath76 and @xmath77 are all achievable .
note that we simply use @xmath35 for the range space while @xmath143 for the null space of @xmath35 .
the precoder design to improve the achievable rate is out of scope of this paper . inspired by the enhanced scheme for the two - user miso broadcast channel @xcite , a 3-time - slotted protocol , which achieves the vertex @xmath76 of the dof region for the two - user miso interference channel , is detailed as follows . in the first time slot
, the symbol vectors @xmath144 and @xmath145 are respectively sent from the two transmitters with precoding , heading to their corresponding receivers : @xmath146 { { { \lowercase{{\bm{u}}}}}}(1)\\ { { { \lowercase{{\bm{x}}}}}}_2(1 ) & = [ \hat{{{{\lowercase{{\bm{h}}}}}}}_{12}(1 ) \
\hat{{{{\lowercase{{\bm{h}}}}}}}_{12}^{\perp}(1 ) ] { { { \lowercase{{\bm{v}}}}}}(1 ) \end{aligned}\ ] ] where @xmath147^{{{\scriptscriptstyle\mathsf{t}}}}$ ] , @xmath148^{{{\scriptscriptstyle\mathsf{t}}}}$ ] satisfy @xmath149 .
the received signal at both receivers are then given as : @xmath150 where @xmath151 and @xmath152 are interferences overheard at the rx-1 and rx-2 respectively , i.e. , @xmath153 according to ( [ eq : p - to - alpha ] ) , i.e. , @xmath154 , we can make @xmath155 by allocating @xmath156 whereas @xmath157 . at the end of slot-1 , tx-1 can reconstruct @xmath158 while tx-2 can reconstruct @xmath159 . instead of forwarding the interferences in an analog fashion , we first quantize the interference term @xmath160 into @xmath161 with @xmath162 bits each , then encode the index of @xmath161 to codeword @xmath163 using a gaussian channel codebook , and forward @xmath163 as a common message to both receivers in the ensuing two time slots . to ease our presentation ,
the process of the encoding and decoding of @xmath163 is omitted hereafter , making it look as if the codeword @xmath161 itself is conveyed to the receivers .
the source codebook @xmath164 ( resp .
@xmath165 ) is generated for @xmath152 ( resp .
@xmath151 ) and maintained at the tx-1 ( resp .
the entry @xmath161 in codebook @xmath166 satisfies @xmath167 where @xmath168 is the quantization error with distortion @xmath169 and independent of @xmath161 . according to the rate distortion theory@xcite
, we let the normalized distortion @xmath170 decay as @xmath171 ( in turn @xmath172 ) so that each receiver can decode it successfully and the quantization error is drowned in the noise . during the second time slot
, the index corresponding to @xmath173 is encoded to @xmath174 and sent from tx-1 as a common message together with a new symbol @xmath175 with zf precoding , while a new symbol @xmath176 intended to rx-2 is instantaneously sent from tx-2 with zf precoding as well . by omitting the encoding and decoding process of @xmath174 , the equivalent transmitted signals can be written as @xmath177 where the cordword @xmath173 is power scaled with @xmath178 to ensure it can be recovered from noisy observation . to avoid interference from the other transmitters
, we assume the new symbols @xmath175 and @xmath176 satisfy the power constraint @xmath179 .
the received signals at both receivers are given as : @xmath180 note that in the received signal @xmath181 , @xmath182 , @xmath183 , while @xmath184 is at noise level .
with distortion @xmath185 , both receivers can retrieve @xmath173 with high probability by treating @xmath186 and @xmath187 as noise @xcite . by removing @xmath173 from the received signals ,
@xmath175 and @xmath176 can be recovered with high probability as long as their power constraints are satisfied .
the transmission in the third time slot is similar to that in the slot-2 , where the index corresponding to @xmath188 chosen from @xmath165 is encoded to @xmath189 and transmitted as a common message together with another new symbol @xmath190 from tx-2 , while only one new symbol @xmath191 intended to rx-1 is sent from tx-1 . by omitting the encoding and decoding process of @xmath189 , the equivalent transmitted signals are @xmath192 where the new symbols @xmath191 and
@xmath190 satisfy the power constraint @xmath193 .
the received signals at both receivers are given as @xmath194 similarly to the slot-2 , @xmath188 is retrievable at both receivers by treating other terms as noise , and @xmath191 and @xmath190 can be also recovered respectively by subtracting @xmath188 from the received signals at both receivers . at the end of the third slot , @xmath175 , @xmath191 , @xmath188 and @xmath173 can be successfully recovered at the rx-1 . as was modeled in @xcite
, an equivalent mimo can be formulated to find the symbols @xmath144 : @xmath195 for the rx-1 , and it is similar for the rx-2 .
the vertex @xmath76 of dof region is achievable by the above scheme .
we outline the main idea of the proof here and please refer to appendix for details . at the rx-1 for instance
, we transform the original signal model into an equivalent @xmath196 point - to - point mimo system model for @xmath144 ( resp .
@xmath145 at the rx-2 ) , together with two parallel siso signal models respectively for @xmath175 and @xmath191 ( resp .
@xmath176 and @xmath190 at the rx-2 ) . for the mimo model
, we obtain the dof of @xmath197 , while get @xmath49 dof for each parallel siso model , and finally @xmath198 dof is achieved per user . in the following ,
we extend the han - kobayashi scheme @xcite here to achieve the vertex @xmath77 . the symbols sent from the tx-1 consists of two parts @xmath199 and @xmath200 , where only @xmath200 is precoded by using imperfect current csit .
simultaneously , one symbol @xmath201 intended to rx-2 is sent from the tx-2 with zf precoding .
the transmission can be given as @xmath202 where the transmitted symbols are assumed to satisfy the power constraints @xmath203 , @xmath204 .
although the symbol @xmath199 is decodable by both receivers and hence referred to as a common message , it is only desirable by rx-1 . on the other hand , we refer to @xmath200 , @xmath201 as the private messages which can only be seen and decoded by their corresponding receivers . at the receiver side
, we have @xmath205 where the terms carrying common message are with approximated power @xmath14 , while those carrying private messages are @xmath206 , and the interference terms @xmath207 are at noise level according to ( [ eq : p - to - alpha ] ) . by firstly treating @xmath208 as noise
, rx-@xmath24 can recover the common message @xmath199 with high probability .
then , the private messages @xmath200 and @xmath201 can be retrieved from the received signals after @xmath199 being subtracted at the rx-1 and rx-2 , respectively .
the vertices @xmath77 and @xmath75 of dof region are achievable .
define @xmath209 for both receivers , the achievable rate can be given by @xmath210 for the rx-1 , and @xmath211 for the rx-2 , where ( [ eq : scheme-2-pr ] ) holds because @xmath199 and @xmath201 are independent .
the dof for both receivers can be simply obtained by definition .
the other vertex @xmath75 can be achieved by swapping the roles of tx-1 and tx-2 .
this completes the proof .
note that the vertices @xmath78 and @xmath74 are achievable by letting one pair communicate while keeping the other one silent . in conclusion ,
all vertices of the dof region for two - user miso interference channel are achievable , and in turn the entire region can be achieved by time sharing .
here , we extend the aforementioned miso case to a class of mimo settings with antenna configuration @xmath212 , where @xmath213 antennas at each transmitter and @xmath214 antennas at each receiver , satisfying @xmath215 .
this includes a generalized miso setting with more than 2 antennas at each transmitter .
the discrete time baseband signal model is given by @xmath216 for any time instant @xmath22 , where @xmath217 is the channel matrix from tx-@xmath24 to rx-@xmath25 ; @xmath218 are normalized awgn vectors at the respective receivers ; the coded input signal @xmath219 is subject to the power constraint @xmath28 , @xmath29 . in analogy to the miso case , we have the optimal dof region of two - user time correlated @xmath212 mimo interference channel . in the two - user @xmath212 mimo interference channel ( @xmath215 ) with perfect delayed csit and imperfect current csit
, the optimal dof region can be characterized by @xmath220 * remark * : the dof region is irrelevant to the number of transmit antennas as long as @xmath215 . for the @xmath221 miso case ( @xmath222 ) ,
the dof region is identical to that when @xmath223 , coinciding with the region for two - user miso broadcast channel .
following the same strategy as the miso case , we first provide the outer bound and then show the region confined by the outer bound is achievable .
the outer bound can be simply extended from the miso case . to avoid redundancy
, we outline the main difference but omit the similar parts . by defining similarly @xmath224 , @xmath225 , @xmath116 , i.e. , @xmath226 we have @xmath227 define @xmath228 and we upper - bound the weighted difference of two conditional differential entropies by @xmath229 where @xmath230 possesses the same definition as that in the miso case and the last inequality is obtained according to the following lemma . for an @xmath231 random matrix ( @xmath232 ) @xmath233 where @xmath234 is independent of @xmath235 and whose entries satisfy @xmath236 , given any @xmath103 with eigenvalues @xmath237 , we have the following upper and lower bounds : @xmath238 the difference of the upper and lower bounds can be further bounded by @xmath239 where @xmath240 and @xmath241 is the digamma function that given by @xcite @xmath242 for integer @xmath243 , where @xmath107 is euler s constant
. please refer to appendix for details .
hence , we can outer - bound the weighted sum rate as @xmath244 and similarly obtain another outer bound by exchanging the roles of rx-1 and rx-2 , i.e. , @xmath245 and therefore the outer bound of the dof is obtained by the definition . for the achievability , the vertices @xmath246 , @xmath247 , and @xmath248
are all achievable and the achievable schemes can be simply extended from the miso case . here , we only show the achievable scheme for the vertex @xmath248 for instance . for the achievability of @xmath246 and @xmath247 ,
the extension is similar and straightforward . in the extended scheme ,
three time slots are consumed , where the transmitted signals are detailed as follows : the transmitted signals from both transmitters are given by @xmath249 where @xmath250 , @xmath251 are assumed to satisfy @xmath252 and @xmath253 , @xmath254 which satisfy @xmath255 where @xmath256 and @xmath257 represent range and null spaces , respectively .
note that the range space is with dimension @xmath214 whereas the null space is with dimension @xmath258 .
similarly to the miso case , the estimation error satisfies @xmath259 . at both receivers
, we have @xmath260 where the interference vectors overheard at both receivers are @xmath261 where @xmath262 , @xmath263 , @xmath264 and @xmath265 are all @xmath266 vectors . by balancing the allocated power among those vectors , i.e. , @xmath267 we approximate the total power of interference vectors as @xmath268 and @xmath269 .
a set of source codebooks @xmath270 with size @xmath162 bits each are generated to represent the quantized elements of the interference vectors @xmath271 and @xmath272 at the tx-1 and tx-2 , respectively regardless of their mutual correlation . ] .
the codewords representing the elements of @xmath273 and @xmath274 are chosen uniformly from @xmath275 and @xmath276 and concatenated as @xmath277 and @xmath278 , respectively .
as stated in miso case , the indices of @xmath279 are encoded to @xmath280 using a gaussian channel codebook and then forwarded as common messages to both receivers in the following two slots .
for the sake of simplicity , we omit the channel encoding and decoding process of @xmath280 , and therefore , it looks as if @xmath279 itself is conveyed . the objective of the slot-2 is to convey the codeword vector @xmath277 whose information is carried on a coded common message @xmath281 together with a new symbol vector at the tx-1 , while only a new symbol vector is sent at the tx-2 . by omitting the encoding and decoding process of @xmath281 , the equivalent transmitted signals are @xmath282 where @xmath283 and @xmath284 .
we assume @xmath285 and @xmath286 to ensure they are recoverable . by treating @xmath287 and @xmath288 as noise , @xmath266 vector @xmath277 is retrievable with high probability provided @xmath214 linearly independent equations at both receivers .
after that , @xmath287 and @xmath288 are also recoverable from @xmath214 linear equations at the rx-1 and rx-2 by subtracting @xmath277 from the received signals .
the objective of the slot-3 is the same as the slot-2 but with the exchanged roles between tx-1 and tx-2 .
the equivalent transmitted signals are given as @xmath289 where @xmath290 , @xmath291 are assume to satisfy power constraint @xmath292 and @xmath293 . by firstly treating @xmath294 and @xmath295 as noise , @xmath266 vector @xmath278 can be recovered with high probability given @xmath214 linearly independent equations at both receivers . similarly to the slot-2 , @xmath294 and @xmath295 can also be recovered from the subtracted received signals . at the end of slot-3 ,
@xmath266 vectors @xmath278 and @xmath277 can be all recovered at both receivers , serving to cancel the overheard interference as well as to provide additional linearly independent equations for @xmath145 and @xmath144 , respectively . with @xmath296
linearly independent equations , the @xmath297 vectors @xmath144 and @xmath145 are both recoverable with high probability at its respective receiver .
the proof of the achievable dof pair @xmath248 is similar to that in the miso case .
take tx-1/rx-1 pair for example .
the original channel model can be transformed to an equivalent @xmath298 point - to - point mimo channel which conveys symbol vector @xmath144 yielding @xmath299 dof , and two parallel @xmath300 mimo channels which carry @xmath287 and @xmath294 respectively yielding @xmath301 dof each .
hence , the total @xmath302 dof is achieved within three time slots , and in turn the dof pair @xmath248 is achievable by symmetry .
the detailed proof can be analogically derived according to the miso case and hence omitted here .
we characterize the dof region of the two - user miso and certain mimo interference channels where the transmitter has access to both delayed csi as well as an estimate of the current csi .
in particular , these results are suited to time - correlated fading channels for which a latency - prone feedback channel provided the transmitter with the delayed samples , based on which a prediction mechanism can be applied to obtain the current imperfect csi .
our dof region covers a family of csit settings , coinciding with previously reported results for extreme situations such as pure delayed csit and pure current csit . for intermediate regimes , the dof achieving scheme relies on the forwarding to users of a suitably quantized version of prior interference obtained under imperfect linear zf precoding at the two transmitters .
we consider rx-1 for example to illustrate @xmath303 is achievable . by symmetry , @xmath304 at the rx-2 can be similarly obtained .
given the received signal model @xmath305 we have the achievable rate of tx-1 and rx-1 pair , i.e. , @xmath306 for the first term , by defining @xmath307 we have @xmath308 where @xmath309 is independent of @xmath144 . by formulating an equivalent mimo channel with gaussian input @xmath144 and output @xmath310 , i.e. , @xmath311}_{{{{\uppercase{{\bm{h } } } } } } } { { { \lowercase{{\bm{u}}}}}}(1 ) \\ & + \underbrace{{\begin{bmatrix}h_{12,1}^{*}(3 ) p^{\alpha/2}(\delta_1+e(1 ) ) - s_{12 } - s_{13 } - e(3)\\- h_{11,1}^{*}(2 ) p^{\alpha/2 } \delta_2 + t_{12 } + t_{13 } + e(2)\end{bmatrix}}}_{{{{\lowercase{{\bm{n}}}}}}}\end{aligned}\ ] ] where @xmath312 , @xmath313 , @xmath172 , and @xmath314 , we have @xmath315 where ( [ eq : translation - app ] ) holds because translation does not change differential entropy ; ( [ eq : conditioning - app ] ) follows the fact that conditioning reduces entropy ; ( [ eq : term1 - 1 ] ) is according to the definition of differential entropy with gaussian distributed input , where @xmath316 and @xmath317 ; and ( [ eq : term1 - 2 ] ) is obtained because @xmath318 , @xmath319 and @xmath320 does not scale as @xmath14 .
for the second term , we have @xmath321 where ( [ eq : term2-mi-1 ] ) holds because @xmath322 is independent of @xmath323 ; ( [ eq : term2-mi-2 ] ) is obtained by the chain rule of mutual information ; ( [ eq : term2-mi-3 ] ) is because @xmath324 are independent of @xmath325 ; ( [ eq : term2-mi-4 ] ) is due to @xmath326 are independent of @xmath325 and @xmath144 . by formulating an equivalent parallel channel with gaussian input @xmath175 ( resp .
@xmath191 ) and output @xmath327 ( resp .
@xmath328 ) , we have @xmath329 where ( [ eq : term2 - 1 ] ) is due to the chain rule of mutual information and the independence between @xmath175 ( resp .
@xmath330 ) and @xmath191 ( resp .
@xmath328 ) ; ( [ eq : term2 - 4 ] ) is according to the definition of differential entropy with gaussian distributed input ; the last equality is due to @xmath312 , @xmath313 , and @xmath331 . substituting ( [ eq : term1 - 2 ] ) and ( [ eq : term2 - 5 ] ) into ( [ eq : mutual - information ] ) , by the definition of dof @xmath332
we conclude that @xmath303 is achievable . by symmetry
, @xmath304 can be simultaneously obtained in a similar way .
this completes the proof .
first for the upper bound , we have @xmath334 where in ( [ eq : theorem-2 - 1 ] ) , @xmath335 with @xmath336 and @xmath337 ; ( [ eq : theorem-2 - 2 ] ) comes from the equality @xmath338 ; ( [ eq : theorem-2 - 3 ] ) is due to poincare separation theorem @xcite that @xmath339 for @xmath340 ; ( [ eq : theorem-2 - 4 ] ) is from the fact that @xmath341 ; ( [ eq : theorem-2 - 5 ] ) is obtained by applying jensen s inequality to a concave function . for the lower bound , we have @xmath342 where ( [ eq : theorem-2 - 10 ] ) comes from minkowski s inequality and ( [ eq : theorem-2 - 11 ] ) is from jensen s inequality by noticing that @xmath343 is a convex function in @xmath243 .
hence , the expectation part can be simplified further as @xmath344 where @xmath345 with @xmath346 being the unitary matrix containing the eigenvectors of @xmath347 , i.e. , @xmath348 with @xmath349 ; in ( [ eq : theorem-2 - 12 ] ) , @xmath350 with @xmath351 being the first @xmath214 columns of @xmath346 where @xmath352 and @xmath353 ; ( [ eq : theorem-2 - 13 ] ) is due to the fact that the capacity of a ricean fading channel is no less than that of a rayleigh fading channel in all snr region @xcite , by treating @xmath235 as the deterministic line - of - sight component ; ( [ eq : theorem-2 - 14 ] ) is because @xmath234 is independent of @xmath351 and @xmath354 is an @xmath300 i.i.d .
complex gaussian matrix with each entry being zero - mean and variance-@xmath46 ; the last equation is according to the isotropic assumption and obtained from @xcite . by the definition of @xmath355
, we have @xmath356^{+ } \\ & \ge \sum_{i=1}^n \log ( 1+\lambda_i \sigma^2 e^{\zeta } ) - n\\ & = \sum_{i=1}^n \log ( 1+\lambda_i \sigma^2 e^{\zeta } ) + o(1 ) \end{aligned}\ ] ] where @xmath357 and last inequality due to the fact @xmath358 .
the difference between upper and lower bounds can be further bounded as @xmath359 where the second inequality is due to the fact @xmath360 and the last equality is because both @xmath361 and @xmath362 are bounded .
this completes the proof . c.
vaze and m. varanasi,``the degree of freedom region and interference alignment for the mimo interference channel with delayed csi , '' _ submitted to ieee trans .
theory _ , arxiv preprint arxiv:1101.3809 , jan .
2011 .
m. kobayashi , s. yang , d. gesbert , x. yi , `` on the degrees of freedom of time correlated miso broadcast channel with delayed csit , '' _ submitted to isit 2012 _ , arxiv preprint arxiv:1202.1909 , feb .
s. yang , m. kobayashi , d. gesbert , x. yi , `` degrees of freedom of time correlated miso broadcast channel with delayed csit , '' _ submitted to ieee trans . inf .
theory _ , arxiv preprint arxiv:1203.2550 , march 2012 .
g. caire , n. jindal , m. kobayashi , and n. ravindran , `` multiuser mimo achievable rates with downlink training and channel state feedback , '' _ ieee trans .
theory _ , vol .
56 , no . 6 , pp .
28452866 , june 2010 .
x. yi , d. gesbert , s. yang , m. kobayashi , `` on the dof of the multiple - antenna time correlated interference channel with delayed csit , '' _ to appear in proc .
asilomar conference on signals and systems _ , ( invited paper ) 2012 .
|
we consider the time - correlated multiple - antenna interference channel where the transmitters have ( i ) _ delayed _ channel state information ( csi ) obtained from a latency - prone feedback channel as well as ( ii ) imperfect _ current _ csit , obtained e.g. from prediction on the basis of these past channel samples .
we derive the degrees of freedom ( dof ) region for the two - user multiple - antenna interference channel under such conditions . the proposed dof achieving scheme exploits a particular combination of the space - time alignment protocol designed for fully outdated csit feedback channels ( initially developed for the broadcast channel by maddah - ali et al , later extended to the interference channel by vaze et al . and ghasemi et al . )
together with the use of simple zero - forcing ( zf ) precoders .
the essential ingredient lies in the quantization and feedback of the residual interference left after the application of the initial imperfect zf precoder .
our focus is on the miso setting albeit extensions to certain mimo cases are also considered .
|
recently , it has been growing that the concern with elastic surface models of helfrich and polyakov - kleinert @xcite .
a considerable number of studies have been conducted on the phase transition between the smooth phase and the crumpled one over the past two decades @xcite .
curvature energies play a crucial role in smoothing the surface . according to curvature energies
, surface models can be divided into two classes ; one with an extrinsic curvature and the other with an intrinsic curvature .
it is also possible that both extrinsic and intrinsic curvatures are included in a model hamiltonian .
intrinsic curvature is the one that is defined only by using the metric tensor ( first fundamental form ) of the surface , and extrinsic curvature is defined by using the extrinsic curvature tensor ( second fundamental form ) of the surface @xcite . both of the mean curvature @xmath0 and the gaussian curvature @xmath1 defined by using the extrinsic curvature tensor , are considered as an extrinsic curvature .
however , the gaussian curvature can also be considered as an intrinsic curvature , because of the relation @xmath2 , where @xmath3 is the scalar curvature defined only by using the metric tensor .
intrinsic curvature models were first studied by baillie et .
al . in @xcite .
the shape of surfaces can be strongly influenced by intrinsic curvatures .
recently several numerical studies have been made on the phase diagram of the model with intrinsic curvature @xcite .
it was reported that the model undergoes a first - order phase transition between the smooth phase and the crumpled phase on a sphere @xcite and on a disk @xcite . as a consequence
the phase structure of the model has been partly clarified : the transition can be seen independent of whether the surface is compact or non - compact .
however , the model is not yet sufficiently understood .
remaining subject to be confirmed is whether the phase transition and the topology - change are _ compatible _ in the surface model on compact surfaces . here _
compatible _ means that both of two phenomena lead to the same result without depending on which phenomenon firstly occurs .
if we define the surface model on compact surfaces , it is reasonable to consider the topology - change of surfaces .
in fact , the partition function for the closed string model of polyakov includes the summation over topology .
moreover , a toroidal vesicle and a genus two vesicle with two holes can be observed in biological membranes @xcite .
therefore , it is interesting , even in the context of the tethered surfaces , to study whether the first - order transition can be observed on a torus , as one of the higher genus surfaces . the phase transition might be seen only on spherical surfaces , if the phase transition and the topology - change are incompatible . in this letter , in order to confirm the compatibility we will show that a first - order transition can be seen in the model on a torus with intrinsic curvature and that the phase transition is identical to that observed in the model on a sphere reported in @xcite .
( a ) , and ( b ) the real torus obtained by connecting the opposite sides of the surface in ( a).,title="fig:",width=264 ] ( b ) , and ( b ) the real torus obtained by connecting the opposite sides of the surface in ( a).,title="fig:",width=264 ] a triangulated real torus is obtained by modifying a triangulated rectangular surface of size @xmath4 as shown in fig . [ fig-1](a ) . bending the surface and connecting the sides of length @xmath5
, we have a cylinder of length @xmath5 . then the remaining two - sides of the cylinder are connected as in the first - step .
thus , we have a real torus as shown in fig . [
fig-1](b ) , which is topologically identical to the surface in fig .
[ fig-1](a ) under the periodic boundary condition .
the real torus mentioned above will be used in the simulations .
the real torus is , therefore , characterized by the ratio @xmath6 .
two kinds of tori are used in the simulations : @xmath7 and @xmath8 .
figure [ fig-1](b ) is the torus of size @xmath9 of the first type @xmath7 , where @xmath10 and @xmath11 .
every vertex has a coordination number @xmath12 on the torus .
the gaussian tethering potential @xmath13 and the intrinsic curvature @xmath14 are defined by @xmath15 where @xmath16 is the sum over bond @xmath17 connecting the vertices @xmath18 and @xmath19 , and @xmath20 in @xmath14 is the sum of angles of the triangles meeting at the vertex @xmath21 , and @xmath22 is the sum over vertices @xmath21 .
the partition function is defined by @xmath23,\qquad \\
s(x)=s_1 + \alpha s_3 , \qquad\quad \nonumber\end{aligned}\ ] ] where @xmath24 is the total number of vertices , which is equal to @xmath25 as described previously .
the expression @xmath26 shows that @xmath27 explicitly depends on the variable @xmath28 .
the coefficient @xmath29 is a modulus of the intrinsic curvature .
the surfaces are allowed to self - intersect , and the center of each surface is fixed in the partition function @xmath30 to remove the translational zero - mode .
we note , as described in @xcite , that the intrinsic curvature term @xmath31 comes from the integration measure @xmath32 @xcite in the partition function for the model on a sphere , where @xmath33 is the co - ordination number of the vertex @xmath21 .
the term @xmath34 becomes minimum when @xmath35 for all @xmath21 , and hence it becomes smaller on a smooth torus than on a crumpled one .
it is also exact that the term @xmath36 can be minimized on smooth configurations of the torus .
the reason for using the term @xmath31 , as an intrinsic curvature on the torus as in the model on a sphere @xcite , is that @xmath14 is closely related to the previously mentioned integration measure .
we use two groups of surfaces classified by the ratio @xmath6 in the simulations as mentioned above . the first is characterized by @xmath7 and is composed of surfaces of size @xmath37 , @xmath38 , @xmath39 , and @xmath40 .
the second is characterized by @xmath8 and is composed of surfaces of size @xmath41 , @xmath42 , @xmath43 , and @xmath44 . the variables @xmath28 are updated by using the canonical monte carlo technique so that @xmath45 , where the small change @xmath46 is made at random in a small sphere in @xmath47 .
the radius @xmath48 of the small sphere is chosen at the start of the simulations to maintain the rate of acceptance @xmath49 for the @xmath28-update as @xmath50 .
the total number of mcs ( monte carlo sweeps ) after the thermalization mcs is about @xmath51 in the smooth phase at the transition point of surfaces of @xmath52 , and about @xmath53 for the smaller surfaces . relatively smaller number of mcs ( @xmath54 ) is iterated in the crumpled phase , because if the surfaces become once crumpled they hardly return smooth .
this irreversibility was also seen in the model on a sphere @xcite and in the model on a disk @xcite .
a random number called mersenne twister @xcite is used in the mc simulations .
we use two sequences of random numbers ; one for 3-dimensional move of vertices @xmath28 and the other for the metropolis accept / reject in the update of @xmath28 .
figures [ fig-2](a ) and [ fig-2](b ) show @xmath55 against @xmath29 obtained on the type @xmath7 surfaces and on the type @xmath8 surfaces , respectively . we find from these figures that the expected relation @xmath56 is satisfied .
scale invariance of the partition function predicts that @xmath56 .
we expect that this relation should not be influenced by whether the phase transition is of first order or not . vs. @xmath29 obtained from the tori of types ( a ) @xmath7 and ( b ) @xmath8 .
, width=453 ] the size of surfaces can be reflected in the mean square size @xmath57 defined by @xmath58 in fact , it is expected that the surfaces become smooth in the limit @xmath59 and crumpled in the limit @xmath60 . in order to see
how large the size of surfaces of the type @xmath7 is , we plot @xmath57 against @xmath29 in fig .
[ fig-3](a ) .
figure [ fig-3](b ) shows those for surfaces of the type @xmath8 .
we find a phase transition separating the smooth phase from the crumpled one both in figs . [ fig-3](a ) and [ fig-3](b ) .
the transition seems as a first - order one because @xmath57 discontinuously changes at finite @xmath29 . vs. @xmath29 obtained on surfaces of the type ( a ) @xmath7 and the type ( b ) @xmath8 .
the surfaces in ( a ) are of size from @xmath37 to @xmath40 , and those in ( b ) are of size from @xmath61 to @xmath44.,width=453 ] vs. @xmath24 obtained just above and below the transition point @xmath62 on surfaces of the types ( a ) @xmath8 and ( b ) @xmath8 . the straight lines in ( a ) and ( b ) are drawn by fitting the data to eq .
( [ hausdorff]).,width=453 ] figures [ fig-4](a ) and [ fig-4](b ) show log - log plots of @xmath57 against @xmath24 , obtained just below and above the transition point @xmath62 where @xmath57 discontinuously changes as shown in figs .
[ fig-3](a ) and [ fig-3](b ) .
the straight line denoted by smooth ( crumpled ) is obtained by fitting @xmath57 in the smooth ( crumpled ) phase just above ( below ) @xmath62 by @xmath63 where @xmath0 is the hausdorff dimension .
thus we have @xmath64 @xmath65 the results @xmath66 in ( [ h1 ] ) and @xmath67 in ( [ h2 ] ) imply that the surfaces are almost smooth in the smooth phase . on the contrary , @xmath68 in ( [ h1 ] ) and @xmath69 in ( [ h2 ] ) imply that the surfaces are highly crumpled in the crumpled phase close to the transition point . in the smooth and the crumpled phases on the tori of the types ( a ) @xmath7 and ( b ) @xmath8 .
the auto - correlation coefficient @xmath70 obtained on the tori of the types ( c ) @xmath7 and ( d ) @xmath8 .
, width=453 ] the phase transition can also be reflected in the convergence speed of the mc simulations . in order to see the convergence speed
, we plot the autocorrelation coefficient @xmath71 in figs . [ fig-5](a ) and [ fig-5](b ) .
the autocorrelation coefficient is defined by @xmath72 ^ 2 } , \\
\nonumber \tau_{i+1 } = \tau_i + n \times 500 , \quad n=1,2,\cdots , \end{aligned}\ ] ] where @xmath73 is the bending energy defined by @xmath74 .
the horizontal axes in the figures represent @xmath75-mcs , which is a sampling - sweep between the samples @xmath76 and @xmath77 .
the coefficients @xmath71 in figs .
[ fig-5](a ) and [ fig-5](b ) are obtained on surfaces of @xmath7 and @xmath8 respectively . from these figures
, we can find that the convergence speed for @xmath73 in the smooth phase is 10 times or more larger than that in the crumpled phase .
figures [ fig-5](c ) and [ fig-5](d ) show the autocorrelation coefficient @xmath70 for the mean square size @xmath57 .
contrary to the behavior of @xmath71 , there is no difference in @xmath70 between the crumpled phase and the smooth phase .
the reason why the convergence speed for @xmath57 in the smooth phase is almost identical to the one in the crumpled phase is that the phase space volume ( @xmath78 ) , where the vertices @xmath18 take their values , is almost the same both in the smooth phase and in the crumpled phase .
the shape of surfaces in the smooth phase is considered to be almost fixed , although they extend to relatively large region in @xmath47 .
this behavior of @xmath70 is very similar to the one in a model on a sphere with extrinsic curvature @xcite but very different from the one in the model on a disk with intrinsic curvature @xcite , where @xmath70 in the smooth phase is quite larger than that in the crumpled phase close to the transition point .
therefore , we understand from @xmath71 and @xmath70 in figs . [ fig-5](a ) [ fig-5](d ) that the toroidal surface can fluctuate only locally , and there is no large - deformation of the surface in the smooth phase . it is quite likely that the shape of surface in the smooth phase is almost unchanged in the model on compact surfaces , and it largely changes in the model on non - compact surfaces so that the phase space volume may be relatively large .
the phenomenon of critical slowing down can not be seen both in @xmath71 and in @xmath70 at the transition point .
the reason of this is because of irreversibility of the transition .
the transition from the crumpled phase to the smooth one appears to be irreversible .
as mentioned in the chapter [ mc - techniques ] , if the vertices once localize to a small region in @xmath47 at the transition point , they hardly expand to be a smooth surface . vs. @xmath29 obtained on the tori of ( a ) @xmath8 and ( b ) @xmath8 .
, width=453 ] vs. @xmath29 on the tori of ( a ) @xmath8 and ( b ) @xmath8 , where @xmath79 is the total number of bonds.,width=453 ] the intrinsic curvature energy @xmath80 is shown in figs . [ fig-6](a ) and [ fig-6](b ) , and it is clear that @xmath80 changes discontinuously against @xmath29 .
this indicates the existence of a first - order phase transition .
the variation of @xmath81 against @xmath29 is shown in figs . [ fig-7](a ) and [ fig-7](b ) , where @xmath82 is the total number of bonds , which is given by @xmath83 .
the discontinuous change of @xmath81 indicates that the phase transition separates the smooth phase from the crumpled phase . and of size @xmath44 at ( a ) @xmath84 , and at ( b ) @xmath85 , and ( c ) a section of the surface in ( a ) , ( d ) sections of the surface in ( b ) .
figures are shown in the same scale .
, width=453 ] snapshots of surfaces of the type @xmath8 are shown in figs . [ fig-8](a ) and [ fig-8](b ) , which are obtained at @xmath84 and @xmath85 .
the size of surfaces in figs . [ fig-8](a ) and [ fig-8](b ) is of @xmath44 .
sections of the surfaces in figs . [ fig-8](a ) and [ fig-8](b ) are shown in figs . [ fig-8](c ) and [ fig-8](d ) respectively .
it is easy to see that the surface in fig .
[ fig-8](a ) is crumpled and that the surface in fig .
[ fig-8](b ) is smooth .
although the surface in fig .
[ fig-8](b ) seems to be a cylinder or a tube , it is almost smooth as can be seen in the section shown in fig .
[ fig-8](d ) .
we note that the cylindrical surface in fig .
[ fig-8](b ) is rotationally symmetric only around the longitudinal axis of the surface .
the appearance of an axis along such cylindrical and smooth surface in fig .
[ fig-8](b ) does not imply a spontaneous symmetry breaking , which can be seen in a surface model in @xcite .
in fact , a real torus such as the one shown in fig . [
fig-1](b ) is also rotationally symmetric only along the axis perpendicular to the hole of the torus . in this sense ,
a symmetry of the model in the smooth phase is equivalent to the one of the real torus shown in fig . [
fig-1](b ) . on the contrary , the surface in the crumpled phase
is expected to have a higher symmetry just like in the crumpled phase of the model on a sphere and on a disk .
in summary , our attentions were focused on the compatibility of the topology change and the phase transition in a tethered surface model on compact surfaces with intrinsic curvature . in order to see this compatibility ,
we have studied the model on two kinds of tori in @xmath47 ; one is characterized by @xmath7 and the other by @xmath8 , where @xmath5 and @xmath86 are the length of two sides of rectangles which are deformed to be the tori by connecting the sides .
monte carlo simulations were performed on surfaces of size up to @xmath87 in both kinds of tori .
we found that the model undergoes a first - order phase transition separating the smooth phase from the crumpled one on both kinds of tori .
the smooth phase is characterized by hausdorff dimension @xmath88 , which is identical with the topological dimension of surfaces .
the surface in the crumpled phase is completely collapsed , and characterized by large values of hausdorff dimension .
these results indicate that the phase transition on a torus is identical with that on a sphere @xcite and that on a disk @xcite .
thus we can conclude that the topology change and the phase transition is compatible in the tethered model of surfaces with the intrinsic curvature .
moreover , the autocorrelation coefficients @xmath70 obtained on the tori in this paper and on the disk in @xcite enable us to consider that compact surfaces deform very little in the smooth phase , and on the contrary , non - compact surfaces can largely deform in the smooth phase .
an interesting problem to be studied is whether the compatibility is present in the model with extrinsic curvature , which exhibits a first - order transition on a sphere @xcite .
it is also interesting to see how the phase transition of the intrinsic curvature model is influenced by the fluidity of lateral diffusion , which is realized on dynamically triangulated surfaces .
m. matsumoto and t. nishimura , `` mersenne twister : a 623-dimensionally equidistributed uniform pseudorandom number generator '' , acm trans .
on modeling and computer simulation vol .
1 , january pp.3 - 30 ( 1998 ) .
|
a tethered surface model is investigated by using the canonical monte carlo simulation technique on a torus with an intrinsic curvature .
we find that the model undergoes a first - order phase transition between the smooth phase and the crumpled one . and phase transition , intrinsic curvature , elastic membranes 64.60.-i , 68.60.-p , 87.16.dg
|
an important unsolved problem in the study of active galactic nuclei ( agn ) is how the central massive black hole is fueled . in particular ,
if the primary fuel source is interstellar gas and dust in the host galaxy , how is that material transported from kiloparsec scales into the central parsecs of the galaxy and onto the supermassive black hole ? in order to fuel the agn , this material must somehow lose nearly all of its angular momentum .
estimates of the mass accretion rate to power agn range from more than @xmath0 per year for the most luminous quasars to @xmath1 per year for the seyfert galaxies discussed here .
thus , over an agn s lifetime ( 10@xmath2 years ? ) a significant amount of material must be transported inwards to feed the central black hole .
a number of dynamical mechanisms have been proposed that could remove angular momentum from the host galaxy s gas and thus supply the fueling rates required to match the observed luminosities of agn .
these include tidal forces caused by galaxy - galaxy interactions ( toomre & toomre 1972 ) and stellar bars within galaxies ( schwartz 1981 ) .
numerical simulations of galaxy interactions ( hernquist 1989 ; barnes & hernquist 1991 ; mihos , richstone , & bothun 1992 ; bekki & noguchi 1994 ; bekki 1995 ) have shown that strong perturbations on a galaxy disk due to a close encounter with another galaxy can trigger the infall of large amounts of gas into the central 1 kpc to 10 pc , depending on the simulation resolution , the type of encounter , and the properties of the simulation .
further , relatively minor encounters ( `` minor mergers '' , hernquist 1989 ; mihos & hernquist 1994 ; hernquist & mihos 1995 ) can lead to significant gas infall , so one need not invoke disruptive encounters .
observationally , there is ample evidence of past or on - going tidal encounters and companions among nearby luminous qsos ( hutchings & neff 1988 , 1992 , 1997 ; bahcall et al .
1995a , b , 1997 ) , but at most only a mild statistical excess of interactions or close companions is seen among the lower - luminosity seyfert galaxies ( adams 1977 ; petrosian 1982 ; dahari 1984 ; keel et al .
1985 ; keel 1996 ; derobertis , hayhoe , & yee 1998 ) . in barred galaxies ,
the triaxial stellar potential leads to a family of orbits for the stars and gas that can transport material into the inner 100@xmath31000 pc of the galaxy .
models of orbits using 2-d and 3-d hydrodynamic simulations of gas and stars in barred potentials ( athanassoula 1992 ; friedli & benz 1993 ; piner , stone , & teuben 1995 ) have shown the formation of a shock front at the leading edge of a bar .
material builds up in this shock and falls into the nuclear region , generally forming a nuclear ring with a diameter approximately equal to the bar s minor axis ( @xmath4100@xmath31000 pc ) .
observations of apparent shock fronts at the leading edges of bars and velocity distributions in barred galaxies are in general agreement with inflow models ( quillen et al .
1995 ; regan , vogel , & teuben 1997 ) , and provide evidence that the dynamical behavior predicted by the models is realistic .
however , the typical semiminor axis of a bar is still @xmath5100@xmath31000 pc , and thus while bars might plausibly transport material reasonably close to the center , an additional mechanism is needed to move the material into the nucleus proper .
a mechanism proposed to complete the transport of material into the nucleus is the ` bars within bars ' scenario ( shlosman , frank , & begelman 1989 ; maciejewski & sparke 1997 ; erwin & sparke 1998 ) . in this model ,
a large - scale bar transports material into a kiloparsec - scale disk .
this gaseous disk then becomes unstable to bar formation , creating a miniature ` nuclear ' bar that transports the gas to within approximately 10 pc of the galactic nucleus , which is approximately where the central supermassive black hole s potential can take over . while large - scale bars in galaxies are readily observable due to their large angular size , small - scale , nuclear bars ( shaw et al .
1995 ) are a factor of 5 to 10 smaller and thus are more difficult to resolve in distant galaxies . because of their smaller physical size , nuclear bars were initially discovered in relatively nearby galaxies ( de vaucouleurs 1975 ; kormendy 1979 ; buta 1986a , b ) .
observationally , investigations of the relative fraction of bars in agn as compared to non - active galaxies have failed to turn up a significant excess of bars among spirals harboring agn ( kotilainen et al .
1992 ; mcleod & rieke 1995 ; alonso - herrero , ward , & kotilainen 1996 ; mulchaey & regan 1997 ) .
similarly , searches for nuclear bars using the _ hubble space telescope _
( _ hst _ ) have failed to find them in numbers sufficient to salvage the bar fueling picture ( regan & mulchaey 1999 ) .
we shall reach a similar conclusion in this paper . at most
, these observations have shown that nuclear bars are present in only a minority of seyferts .
this is not to say that bars and interactions are not viable fueling mechanisms as clearly they _ are _ operating in a number of seyferts .
the point is that neither is the dominant mechanism , nor necessarily responsible for fueling the agn in all cases .
what is responsible for fueling an agn in the majority of cases where there is no evidence for either interactions or bars ?
one way of addressing this question is to assay the amount of potential fuel in the form of interstellar dust and gas in the inner kiloparsec of nearby seyfert galaxies . indeed , while regan & mulchaey ( 1999 ) failed to find a preponderance of nuclear bars in their _ hst _ imaging sample of 12 seyferts , they _ did _ find one common morphological structure : spiral arms of dust , of greater or lesser degrees of organization , in the inner few hundred parsecs .
in this paper we present new , high angular resolution _
hst _ near - infrared nicmos images combined with archival wfpc2 images of the nuclear regions of a representative sample of 24 seyfert 2s to search for dynamical and morphological signatures of agn fueling . in section 2
we describe our sample selection , and our observations in section 3 . in section 4
we describe the nuclear morphology derived from direct flux and color maps to search for nuclear bars and circumnuclear dust structures .
section 5 describes the possible role of these structures in fueling the agn , and we summarize our results and conclusions in section 6 .
the history of the study of interactions among seyfert galaxies serves to illustrate the main difficulty in drawing conclusions on the nature of agn from studies of statistical samples , namely the problem of defining an unbiased sample and identifying a suitable control group . as discussed by keel ( 1996 ) , the reexamination of interacting galaxy samples illustrated the importance of matching host galaxy types and luminosities ( fuentes - williams & stocke 1988 ) .
in the last decade , better defined samples of agn , for example the center for astrophysics ( cfa ) sample ( huchra & burg 1992 ) and the revised shapley - ames ( sandage & tammann 1987 ) sample of seyfert galaxies ( maiolino & rieke 1995 ) , have greatly assisted statistical studies of the agn population .
the cfa redshift survey ( huchra et al . 1982 ) obtained spectra of a complete sample of 2399 galaxies down to a limiting photographic magnitude of @xmath6 .
this relatively unbiased survey avoids many of the problems of traditional surveys for agn , particularly those based on ultraviolet excesses which are biased against reddened agn and thus preferentially detecting seyfert 1s and seyferts in host galaxies with active star formation .
this sample also has the often overlooked advantage of a uniform set of spectral classifications obtained with high signal - to - noise ratio that provides comparable detection limits for weak broad - line components ( osterbrock & martel 1993 ) .
this is essential for making meaningful classifications of seyfert types 1.8 , 1.9 , and 2 ( osterbrock 1981 ) for distinguishing objects with and without obvious broad - line components .
studies of the relative frequency of close companions among the cfa seyferts have shown that this sample has only a marginal excess of companions , although a larger fraction of the seyferts do appear to be currently undergoing the final stages of a past interaction ( derobertis , hayhoe , & yee 1998 ) .
in contrast , this same study emphasizes that a small but significant number of the cfa seyferts show no morphological evidence for recent interactions .
infrared wavelength imaging also shows no signature of a violent interaction history in this sample ( mcleod & rieke 1995 ) , nor does it detect an excess of nuclear bars . for this study
, we have chosen 24 of the 25 seyfert 2s ( including types 1.8 and 1.9 as classified by osterbrock & martel 1993 ) in the cfa sample .
we are concentrating on just the seyfert 2 galaxies because they in general have fainter nuclear point sources than seyfert 1s ( nelson et al .
1996 ; malkan , gorjian , & tam 1998 ) , and thus the circumnuclear environment is relatively unaffected by the contamination from the complex _
hst _ point - spread function ( psf ) .
this is especially true with nicmos where the psf is very complex ( mackenty et al .
our observations are comprised of archival visible - band and new near - infrared _ hst _ images .
the visible wavelength wfpc2 images include all of the seyfert 2 galaxies in the cfa sample with the exception of ngc4388 and mrk461 ( neither have been observed to date with broad - band filters ) .
most of the visible images were from a snapshot survey of agn ( go-5479 , see malkan , gorjian , & tam 1998 ) and are single 500s exposures taken through the f606w filter .
the remaining archival images were taken through the f547 m filter from a variety of programs .
table 1 lists the properties of the sample galaxies and which of these two visible band filters ( hereafter referred to collectively as the @xmath7 filter ) was used to obtain the @xmath8band galaxy image used in this investigation .
the images were all obtained with the nucleus roughly centered on the pc1 detector of wfpc2 , which has a plate scale of 004553 pixel@xmath9 and a total effective field of view of 356 ( biretta et al .
1996 ) .
we obtained near - infrared @xmath10 and @xmath11band images of 23 of our seyfert sample with the _ hst _ nicmos camera 1 during cycle 7 ( project go-7867 ) .
the remaining galaxy in our sample , ngc1068 , was observed by the nicmos team as part of their guaranteed time observations .
we chose camera 1 as its plate scale ( 0043 pixel@xmath9 ) is closest to that of the pc1 camera ( @xmath5 0046 pixel@xmath9 ) , although with a substantially smaller field of view ( @xmath511 ) .
the near match in plate scale between the visible and infrared images is helpful for matching images from the two cameras and achieving the best sampling of the resolutions of the two cameras .
further , the surface brightness of the underlying host galaxies falls off sufficiently fast that we would have obtained less signal - to - noise in the outer regions of nicmos camera 2 at a cost of coarser pixel sampling ( 0075 pixel@xmath9 ) .
our images were taken through both the f110w and f160w filters ( hereafter @xmath12 and @xmath13 , respectively ) , which are each approximately 2000 wide and are centered at 1.1 @xmath14 m and 1.6 @xmath14 m . in each filter
we acquired images at four dither positions separated by 1(spiral - dith pattern ) so that the effects of bad pixels and other detector artifacts could be eliminated . at each dither position we used an exposure ramp ( step128 ) for a total of 256 seconds per position .
this allows us to correct for any saturated pixels that might occur if the nucleus is brighter at near - ir wavelengths than suggested by the archival wfpc2 imaging ( as would be expected if the nuclei are dusty ) .
the final shifted and combined frames have a cumulative exposure time of 1024 seconds per filter .
this was done for each galaxy , resulting in a homogeneous set of relatively deep near - infrared images of the central 11of these galaxies . to calibrate this dataset we had to perform several additional processing steps beyond the standard _
hst _ data - reduction pipeline . for the wfpc2 images ,
this consisted of cosmic@xmath3ray removal and absolute flux calibration .
the flux calibration included transforming our f547 m and f606w to the johnson / cousins system by using the stsdas synphot package to convolve a series of galaxy templates with both johnson @xmath7 and the _ hst _ filters . for the infrared images we shifted and added the four individual exposures from the calnica part of the data reduction pipeline ( mackenty et al .
1997 ) after masking out individual bad pixels and detector artifacts .
we chose not to use the calnicb part of the standard data reduction pipeline as this task attempts to subtract the background from the final mosaic image .
as our observations were not through filters contaminated by thermal emission , this step was not necessary . furthermore , the background subtraction attempted by the calnicb pipeline was in fact subtracting the flux level due to extended emission from these galaxies , a problem that was particularly acute for our brighter targets that covered the entire array .
we finally flux - calibrated our data using the transformations derived by stephens et al .
( 2000 ) to place the near - infrared images on the cit system .
we will present the full atlas of our observations in a future publication .
if there is cold gas flowing into the nuclear ( @xmath16 pc ) regions from the host galaxy , we should see evidence of this material in the form of dust lanes extending from large scales into the nuclear regions .
dust is a useful tracer of this material as it is generally well - mixed with gas and can be detected by its attenuation of light from background stars .
our infrared images primarily trace the stellar distribution in these galaxies , while the visible wavelength images trace both the stellar distribution and the dust .
we combined these images to form @xmath15 color maps of all of the galaxies for which we had visible band _ hst _ data to map the dust distribution .
figure 1 shows the @xmath15 color maps for the seyfert 1.8s and 1.9s in our sample , while figure 2 shows the @xmath15 color maps for the seyfert 2s . in figure 3
we show the @xmath17 color maps for mrk461 and ngc4388 , the two galaxies for which we do not have visible - band _ hst _ imaging .
these images are 5on a side , corresponding to projected spatial sizes between 2.7 kpc and 125 pc in the galaxies for @xmath18 km s@xmath9 mpc@xmath9 .
the 2@xmath3pixel resolution element is 0091 , corresponding to projected spatial scales between 50 pc and 2.3 pc . in table 1 ,
column 8 we list the physical scale corresponding to 1at the distance of each galaxy . to provide a visual reference , a bar showing 100 pc projected distance at the galaxy
is drawn in the bottom left corner of each frame . to exactly match the plate scales of the two images , we rebinned our nicmos images to the plate scale of the pc1 chip using a geometric transformation .
though the resolutions of the two cameras are different , we did not smooth our @xmath7 and @xmath13 images to a common resolution . as a result our @xmath15 color maps , particularly those galaxies with a bright nucleus at @xmath13 , show artifacts of the resolution mismatch as small , `` red '' rings at the position of the nucleus ( e.g. mrk334 , mrk744 ) as well as `` red '' spikes from the nicmos diffraction pattern ( e.g. mrk334 ) .
we did not smooth the @xmath7 and @xmath13 images to a common resolution as we found that this substantially decreased the contrast of the dust features in these color maps .
this resolution mismatch , however , introduces small uncertainties into our estimate of the amount of reddening present in the color maps .
we discuss this further in section
4.3 below .
the images shown in figures 1 , 2 & 3 reveal a wide range of morphologies , including nuclear spiral patterns , stellar bars , and irregular , clumpy dust distributions . in table 2
we list our morphological classification for each galactic nucleus .
the most common morphology we find in these images are nuclear spiral dust lanes , generally extending into the inner @xmath19 pc .
these nuclear spiral arms are wound in the same direction as the main disk spiral arms , but they do not appear to be continuations of the main disk spiral arms . in many of the galaxies with large - scale bars ,
these nuclear dust spirals are clearly connected to dust lanes along the leading edge of the bar , similar to what is seen in the hydrodynamic simulations of athanassoula ( 1992 ) .
the most striking examples are mrk471 , ugc12138 , ngc5674 , ngc3362 , ngc5347 , and ngc7674
. the nuclear spiral dust lanes in figures 1 , 2 , & 3 exhibit a variety of spiral morphologies , ranging from relatively tightly wound arms ( e.g. , ngc1144 and ugc6100 ) to loosely wound arms ( e.g. , ugc12138 and mrk573 ) .
we find examples of one - sided spiral arms ( e.g. , ngc3982 and mrk270 ) , two - arm , `` grand - design '' spirals ( e.g. , ugc12138 and ngc7682 ) , and multi - armed , `` flocculent '' spirals ( e.g. , mrk334 and ngc5674 ) .
often it is easier to see spiral arms on one half of the image than the other , indicating that the gas disks containing the spiral structure are inclined to the line of sight and we are more clearly seeing the dust lanes on the near side of the galaxy ( e.g. mrk744 and ngc1144 ) . in some of our color maps
we also see regions with very blue colors .
these include diffuse `` blue '' regions that are due to extended emission - line gas and bright , unresolved knots of blue light that are likely regions of ob star formation .
in particular , these latter regions are found in mrk334 , mrk744 , mrk266sw , and ngc7682 ( figures 1 & 2 ) . while we do not have uv imaging to confirm that there is significant star formation in these sources , some authors ( heckman et al .
1995 , 1997 ; gonzlez delgado et al .
1998 ) have proposed that massive star formation on 100 pc scales may be a significant contributor to the continuum emission in some seyfert 2s .
these galaxies may therefore be candidates for such seyfert 2starburst galaxies .
nuclear stellar bars have been implicated as a possible mechanism for removing angular momentum from gas on 100 s of parsec scales and transporting it to the inner 10 s of parsecs , where the central supermassive black hole begins to dominate the gravitational potential .
recent ground - based ( mulchaey & regan 1997 ) and _ hst _ observations ( regan & mulchaey 1999 ) have shown that most seyfert galaxies do not have nuclear bars , implying that some other mechanism is responsible for transporting interstellar gas to the central engine . our data confirm and extend their conclusions with a larger , representative sample of seyfert 2s .
we searched for nuclear bars by qualitatively examining the isophotal contour maps of our @xmath7 , @xmath12 and @xmath13 images .
in several galaxies we also found straight dust lanes in the color maps , but in all cases the bar signature was clearly visible in the isophotal contour maps ( see figure 4 ) . in a future paper
we will present a more quantitative analysis of bar selection in these galaxies .
we find at most 5 nuclear bars among the 24 galaxies in our sample ( see figure 4 ) .
mrk573 is the best example of a double - barred galaxy in our sample .
it has both a host - galaxy bar and a nuclear bar as seen in previous ground - based and _ hst _ visible - wavelength imaging ( pogge & derobertis 1993 ; capetti et al .
1996 ) , and there is a clear straight dust lane going into the nucleus in visible and visible / near - ir color maps ( pogge 1996 ; capetti et al . 1996 ; quillen et al .
the appearance of this dust lane in the @xmath15 color map is consistent with dust lanes at the leading edge of the bar ( cf .
quillen et al . 1999 ) .
the nuclear bar in mrk270 is apparent at visible and infrared wavelengths , and there is a straight dust lane in the @xmath15 color map oriented nearly perpendicular to the bar major axis .
ngc5929 has a distinct nuclear bar in the near - infrared images , but it is hidden by dust in the visible - light images .
the @xmath15 color map of this galaxy shows the nuclear region is very dusty with an irregular dust morphology .
mrk471 shows evidence for a nuclear bar in both the near - infrared surface brightness image and through straight dust lanes in the @xmath15 color map .
the lowest surface brightness contours in figure 4 also show the large - scale bar in mrk471 .
regan & mulchaey ( 1999 ) report a nuclear bar in ngc5347 .
while this galaxy has a clear large - scale bar seen at visible and ir wavelengths ( mcleod & rieke 1995 ) , the nuclear bar is less obvious in the @xmath11band image ( also shown in figure 4 ) .
while we will retain the classification of this galaxy as nuclear barred in the interests of setting a strong upper limit on the nuclear bar fraction in seyferts , we consider this the weakest case for a nuclear bar of the 5 we have found .
peletier et al .
( 1999 ) have suggested that there is a nuclear bar perpendicular to the main galaxy bar in ngc5674 at visible wavelengths , but we do not see it in our near - infrared images ( figure 4 ) . instead , we find a great deal of dust in this galaxy at the position angle of both semiminor axes and thus the `` bar '' apparent at visible wavelengths is most likely an artifact of the circumnuclear dust distribution .
this underscores the advantage of using near - infrared images to search for and confirm suspected nuclear bars , as dust can greatly distort the nuclear surface brightness profiles at visible wavelengths .
finally , nuclear bars that are heavily enshrouded by dust may not reveal their presence with straight dust lanes in @xmath15 color maps .
if nuclear bars are generally hidden by large quantities of dust , however , straight dust lanes may be visible in @xmath17 color maps .
we have created @xmath17 color maps for all of the galaxies in our sample , but find no additional evidence for nuclear bars in the @xmath17 color maps that were not detected in our @xmath7 , @xmath13 , or @xmath15 images .
seyferts , like more quiescent spirals , have a central stellar component that consists of a disk and a central bulge . in the nuclear regions
there is additional light from the agn and the extended narrow - line region .
this latter region is partially resolved in many _
hst _ images of seyferts ( bower et al .
1994 ; simpson et al .
1997 ; malkan et al .
the near - infrared continuum emission from the unresolved active nucleus proper may also include a contribution from hot dust ( e.g. , glass & moorwood 1985 ; alonso - herrero et al . 1998 ; glass 1998 ) .
the presence of circumnuclear dust manifests itself in our images in two ways .
first , we expect there to be a relatively uniform distribution of diffuse dust spread throughout the volume of the nuclear bulge and disk of the host galaxy
. this component will be difficult to measure as it is expected to be well mixed with the stars .
second , we expect discrete clouds of gas and dust , most likely organized into a disk confined to a plane ( or warped slightly ) , which could give rise to the spiral dust lanes seen in the color maps ( figures 1 , 2 , & 3 ) .
diffuse gas and dust uniformly distributed in the central regions of the galaxies would have a relatively low integrated mass , and so is unlikely to be a significant contributor to the fuel reserves of the active nuclei .
the discrete , dense dust structures we see in the nuclear disks , however , particularly in the spiral dust lanes , are a more probable fuel source .
as these spiral dust lanes are likely to be confined to a disk , their net extinction can be estimated by treating them as a thin , obscuring slab of material and using the change in @xmath15 color between the arm and interarm regions as an estimate of the total extinction , @xmath20 .
we define the color excess due to dust in the nuclear disk as : @xmath21 this excess color is used to estimate the total visual extinction , @xmath20 , via a standard interstellar extinction curve ( mathis 1990 ) .
the @xmath17 color excess could also be used for this purpose , but due to the close proximity of these filters in wavelength , they do not provide as precise a measure of @xmath20 . by relating this @xmath20 to a column density of hydrogen atoms , @xmath22 @xmath23 for an @xmath24 reddening law ( bohlin , savage , & drake 1978 ) and converting this into a mass column density , we can obtain a rough estimate of the surface density of these disks : @xmath25 the disk surface density derived by this technique actually estimates the maximum surface density rather than the average .
however , as discussed in detail below , this measure of @xmath26 will tend to underestimate the true value of @xmath20 due to uncertainties in the dust geometry and instrumental effects .
we therefore chose this conservative approach to estimating the average disk surface density so as to minimize the likelihood that we have significantly underestimated the surface density . while this estimate is computationally straightforward , there are several systematic effects at work that combine to make these estimates lower limits on the true gas surface density .
first , we have used a very simple foreground screen as our dustlane model .
a more physically realistic model would be a dust screen sandwiched between two layers of stars , corresponding to the stars behind and in front of the dust relative to the line of sight .
for an equal amount of starlight on either side of the dust lane , the arm - interarm @xmath26 defined above reaches a maximum of @xmath27 at @xmath28 for a smooth dust layer .
a patchy or clumpy distribution , as discussed below , would increase these numbers .
scattering is another vital component to a realistic dust model .
dust particles are strongly forward scattering and more likely to scatter blue light than red light , causing a `` bluing '' of the starlight that could counteract , in part , the reddening due to absorption .
significant scattering would reduce the observed @xmath26 , thus leading us to underestimate the total column density of material in these nuclear spiral arms .
we estimated the magnitude of scattering by incorporating the `` dusty galactic nuclei '' model of witt , thronson , & capuano ( 1992 ) into our dust sandwich model .
we find that scattering can decrease the observed @xmath26 by a factor of two , corresponding to an increase in the disk surface density by the same factor .
the simple foreground screen model we have used to estimate the disk surface density therefore results in an underestimate of the true disk surface density by a factor which depends on the true dust geometry , but is useful as it sets a lower limit on the surface density .
another complication to our estimate of the disk surface density is that our images do not resolve small - scale dust structures in these dust lanes .
studies of the arm - interarm extinction in overlapping galaxies ( white & keel 1992 ; white , keel , & conselice 1996 ; berlind et al .
1997 ) show that while the large - scale dust lanes are optically thin , the extinction law in the spiral arms tends to be grayer than a standard galactic extinction law , primarily due to unresolved , optically thick clumps ( berlind et al .
unresolved clumps of dust in these nuclear spirals , particularly in the more distant galaxies , would lead to a grayer extinction law than assumed here , resulting in an underestimate of the total dust mass .
a second resolution effect is due to the finite resolution of the wfpc2 and nicmos cameras .
many of the dust lanes in our @xmath15 color maps are unresolved or only marginally resolved .
we have therefore constructed a simple model of a dust lane to estimate the magnitude of this effect .
we constructed models of dust lanes by creating a series of artificial images at @xmath7 and @xmath13 of `` ideal '' dust lanes in which the dust lane was a step function in flux at a range of widths and @xmath26 ( note that @xmath29 for @xmath24 ; mathis 1990 ) .
we then convolved these images with the _ hst _ psf for the pc1 camera and nicmos camera 1 from tinytim ( krist & hook 1999 ) and measured the difference @xmath30 as a function of the width of the unconvolved dust lane in units of pc1 pixels . in figure 5 we plot the difference in @xmath26 to illustrate the effect of unresolved and marginally resolved dust lanes on measurements of the extinction ( _ solid lines _ ) .
the lines correspond to input @xmath26 values of 1.1 , 0.8 , and 0.4 with the top line corresponding to the largest @xmath26 we measured .
this figure shows that unresolved dust lanes may cause us to underestimate @xmath20 by up to a factor of three for the most heavily reddened dust arms in this sample .
a further complication is the resolution mismatch between the pc1 and nicmos camera 1 , which corresponds different widths for the @xmath7 and @xmath13 psfs . creating a @xmath15 color map corresponds to dividing a @xmath8band image of the galaxy that has been convolved with the narrower wfpc2 psf by an @xmath11band image that has been convolved with the broader nicmos psf .
the result is a weak unsharp masking effect that causes the @xmath15 color to be slightly larger in unresolved or marginally resolved dust lanes . to measure the size of this effect
, we repeated the exercise discussed above , but convolved both the @xmath7 and @xmath13 band image with the same psf ( in this case the psf for the f606w filter and pc1 detector ) .
we have plotted the results of this test ( _ dotted lines _ ) on figure 5 for the same @xmath26 values and , as expected , the model dust lanes convolved with different psfs are systematically redder than the model dust lanes convolved with identical psfs .
the magnitude of this effect , however , is much less than that due to convolving the model dust lanes with a psf in the first place .
a final factor that introduces systematic uncertainties into the measurement of @xmath26 is the possible presence of diffuse line emission .
as mentioned above , we are resolving the narrow - line region in several of these galaxies .
in particular , mrk270 , mrk573 , ngc3362 , ngc7682 , and ugc6100 all show bright , extended line emission in the f606w images that result in `` blue '' , and often cone - shaped , structures in their @xmath15 color maps . the bow - shaped emission line region in mrk573 and other seyfert 2s have been previously studied with _
hst _ by falcke , wilson & simpson ( 1998 ) .
capetti et al .
( 1996 ) have seen similar structures in _ hst _ images of the narrow line regions in seyfert 2s .
in addition to these discrete structures , there may also be faint , diffuse line - emitting gas spread throughout the circumnuclear regions . since the f606w filter contains the bright [ o iii ]
@xmath31 and h@xmath32 lines , a significant emission - line contamination is possible .
we estimated the contribution of emission lines to a variety of seyfert 2 spectra with the stsdas synphot package .
this exercise showed that if these lines have a total equivalent width of 100 , they would still only increase the @xmath8band surface brightness by @xmath33 magnitudes .
the contribution to the total flux in the f606w filter is small because the filter has an effective width of approximately 1500 .
the brightest contributors to the @xmath11band surface brightness , the hydrogen brackett lines , are significantly lower equivalent width than their visible - wavelength counterparts and therefore the emission - line contribution to the @xmath11band is expected to be unimportant .
in addition to requiring a very high equivalent width to affect the @xmath8band flux , this emission would have to be correlated or anticorrelated with the dust lanes to lower or raise , respectively , the measured @xmath26 . to summarize our results , our ( @xmath15 ) color maps of these galaxies exhibit dust lanes with a range in @xmath26 corresponding to inferred mass surface densities of @xmath34 pc@xmath35 .
while we have overestimated the average surface density by measuring the maximum value of @xmath26 , the dust model and resolution effects could offset this as they lead to an underestimate of the total column density .
these surface densities imply that these nuclear disks contain at least @xmath36 of interstellar material in the inner @xmath37 pc .
as seyfert galaxies require mass accretion rates of @xmath38 yr@xmath9 ( e.g. peterson 1997 ) to fuel their nuclear luminosities , these circumnuclear disks are massive enough to act as potential reservoirs for fueling the agn .
nuclear spirals like those we have shown here have been observed in quiescent spirals ( phillips et al .
1996 ; carollo , stiavelli , & mack 1998 ; elmegreen et al .
1998 ; laine et al . 1999 ) , as well as in seyfert galaxies ( quillen et al . 1999 ; regan & mulchaey 1999 ) .
these nuclear spirals are dynamically distinct from the large - scale spiral arms seen in spiral galaxies as they may lie inside the inner lindblad resonance of the outer spiral arms ( elmegreen , elmegreen , & montenegro 1992 ) and they may also be shielded by the dynamical influence of the galaxy s nuclear bulge ( bertin et al .
the nuclear spiral dust lanes appear to differ from the large - scale spiral structure of their hosts in two important ways : they are unlikely to be the result of gravitational instabilities in self - gravitating disks , and their sound speed is likely to be comparable to the orbital speed .
elmegreen et al .
( 1998 ) estimated the value of toomre s ( 1981 ) @xmath39 parameter for the nuclear disk in ngc2207 and determined that the nuclear disk was not self - gravitating .
we have applied a similar analysis to the galaxies in our sample after rewriting the expression for @xmath39 in the form : @xmath40 where @xmath41 is the sound speed , @xmath42 is the azimuthally - averaged disk surface density in @xmath43 pc@xmath35 , and @xmath44 is the disk scale height in pc .
this expression for @xmath39 assumes that the nuclear disks are in solid body rotation , as is the case in ngc2207 ( rubin & ford 1983 ) and in the circumnuclear gas disks in galaxies with rapidly rotating cores in the virgo cluster ( rubin et al .
we have estimated @xmath41 to be 10 km s@xmath9 , which is a reasonable value for the inner disks of spirals ( e.g. spitzer 1978 ; elmegreen et al .
1998 ) , although some authors report values of @xmath45 km s@xmath9 for the main disks of spirals ( e.g. kennicutt 1989 ; van der kruit & shostak 1984 ) .
we are able to place an upper limit on @xmath44 by assuming that as the disks are thinner than the width of the spiral arms , as is true for spiral arms in galaxy disks proper .
most of our estimates of the arm widths are also upper limits , as at their narrowest points nearly all of the nuclear spiral arms are unresolved .
our estimates of @xmath46 , and @xmath39 are summarized in table 3 .
values of @xmath47 would imply that the spiral arms are formed by gravitational instabilities in self - gravitating disks ( e.g. binney & tremaine 1987 ) , whereas values of @xmath48 would correspond to spiral arms formed in non - selfgravitating disks by hydrodynamic instabilities .
the @xmath39 values we find for the spiral arms in our seyfert galaxies range from @xmath49 ( table 3 , column 4 ) .
we thus conclude , as did elmegreen et al .
( 1998 ) for ngc2207 , that these nuclear spiral arms do not occur in self - gravitating gas disks .
since @xmath39 scales as the square of the sound speed , a lower value of @xmath41 could greatly reduce @xmath39 .
as most of our estimates for @xmath39 are significantly above 1 , @xmath41 would have to be quite small to change our results in the direction of self - gravitating @xmath39 values .
faster sound speeds , in contrast , would increase @xmath39 as @xmath50 .
our estimates of @xmath42 are rough values that could vary due to several competing effects .
as mentioned above , we may be underestimating the column density in these dust lanes due to scattering and the effects of unresolved dust lanes .
in contrast , by measuring @xmath26 from the maximum arm - interarm @xmath15 contrast , we may be overestimating the disk surface density .
in addition , if the spiral arms are shock fronts they may correspond to density enhancements of up to a factor of 4 ( for nonradiating , adiabatic shocks ) or higher ( for radiating shocks ; e.g. spitzer 1978 ) over the azimuthally - averaged disk surface density .
in addition to the uncertainty in @xmath41 and @xmath42 , our estimate of the scale height is an upper limit as the spiral arms are generally unresolved .
however , decreasing @xmath44 would increase @xmath39 and the upper limit on @xmath44 thus only strengthens the argument that these disks are not self - gravitating .
the fact that many of these nuclear spirals are multi - arm spirals further supports the notion that they are unlikely to be due to gravitational instabilities in self - gravitating disks ( lin & shu 1964 ) .
spiral arms can form in gaseous disks that are not self - gravitating through hydrodynamic shocks .
one possibility is that these shocks could be caused by turbulent gas motions driven by the inflow of material from larger radii ( struck 1997 ) .
englmaier & shlosmann ( 1999 ) have also shown that shocks in non - selfgravitating , nuclear disks can form grand - design , two - arm spiral structure in galaxies with a large - scale stellar bar . since the orbital speeds in the inner regions of galaxies are of order the sound speed , it is also possible that the arms are formed by acoustic shocks ( elmegreen et al . 1998 ; montenegro , yuan , & elmegreen 1999 ) .
shocks in these gaseous disks could be a viable means of removing angular momentum and energy from the gas and thus these nuclear spirals could be signposts of the fueling mechanisms for seyfert galaxies .
while investigations have traditionally looked for large - scale signatures of interaction or bars in the host galaxies , the actual funneling of gas the last few hundred parsecs into the central black hole may be mediated by the hydrodynamics of non - selfgravitating gas disks on small scales .
this mechanism could be common , regardless of whether large - scale bars or interactions transported the gas from large radii to 100 s of parsec scales .
these @xmath13-band images and @xmath15 color maps of the centers of a representative sample of 24 seyfert 2 galaxies rule out the presence of nuclear bars in all but at most 5 of the 24 galaxies in our sample . while this is not to say that nuclear bars can not play some role in fueling an agn if present , they are found in only a minority of seyferts .
this strengthens and extends similar results found with a smaller sample of seyferts by regan & mulchaey ( 1999 ) . while this sample is composed of only seyfert 2s , a study of all of the seyfert 1s and 2s in the cfa seyfert sample with _
wfpc2 imaging did not reveal an excess of bars among seyfert 1s relative to seyfert 2s ( pogge & martini 2000 ) .
we note , however , that the circumnuclear environments of seyfert 1s are more difficult to study given their strong nuclear psfs and they have not been imaged in the near - infrared , which was needed to find 2 of the 5 possible bars in our sample .
our @xmath15 color maps of the centers of these seyfert galaxies have shown that 20 of 24 have nuclear spiral structure .
of the 4 that do not show nuclear spiral structure , 2 of these ( ngc4388 and ngc5033 ) are nearly edge - on systems and thus the geometry is unfavorable for viewing any kind of nuclear spiral structure .
the remaining two , mrk266sw and ngc5929 , are both in strongly interacting systems and have very chaotic dust morphologies , perhaps reflecting disordered delivery of host galaxy gas due to the tidal interaction .
we have used the @xmath15 color maps to estimate the amount of material in these spiral arms and have found that the nuclear disks in which they reside contain enough gas to serve as a fuel reservoir for the agn .
the surface densities of these disks , along with our estimates of the disk scale height and surface density , also suggest that these disks are not self - gravitating .
the nuclear spiral structure is therefore probably formed by shocks propagating in the disk , via either an acoustic mechanism or other hydrodynamic process , and these shocks could dissipate energy and angular momentum from the gas .
our observations strongly rule out small - scale nuclear bars as the primary means of removing angular momentum from interstellar gas to fuel the agn in our representative sample of seyfert 2s .
instead , we find that the most common morphological features in the centers of agn are nuclear spirals of dust .
these spirals may be caused by shock waves in a disk or may be streamers of mass falling inwards from the innermost orbital resonance .
the large fraction of seyfert 2s with dusty nuclear spirals demonstrates the importance of shocks and gas dynamical effects in removing angular momentum from the gas and feeding this material into the nucleus and hence into the supermassive black hole .
an outstanding question posed by these nearly ubiquitous nuclear spirals is how their host nuclear disks formed . as proposed by rubin et al .
( 1997 ) for nuclear disks in some virgo cluster galaxies , we conclude that two likely scenarios are the standard processes previously invoked for fueling agn : bars and interactions .
both large - scale bars and interactions can remove angular momentum from gas and drive it into the inner 100 s of parsecs , where it settles into a nuclear disk .
shocks or other hydrodynamic effects in these disks then complement the action of bars and interactions by removing angular momentum from this accumulated material and feeding the fuel into the active nucleus . two future areas of work are suggested by our results .
first , similar observations of a large sample of quiescent spiral galaxies should be obtained to determine if the nuclear spirals found in seyferts are present in all spirals .
this work should also be augmented by spectroscopic data to provide kinematic information on the gas and dust in the inner regions .
second , theoretical predictions of accretion rates from spiral structure in both self - gravitating and non - selfgravitating disks are needed . answers to these questions are needed before we can understand the role played by dusty nuclear spirals in fueling agn .
we would like to thank alice quillen for many useful discussions and comments and brad peterson and barbara ryden for their helpful comments on the manuscript .
we would also like to thank the referee , mike regan , for his helpful comments that have improved and clarified this paper .
support for this work was provided by nasa through grant numbers go-07867.01 - 96a and ar-06380.01 - 95a from the space telescope science institute , which is operated by the association of universities for research in astronomy , inc .
, under nasa contract nas5 - 26555 .
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 .
adams , t.f .
1977 , , 33 , 19 alonso - herrero , a. , ward , m.j . , & kotilainen , j.k .
1996 , , 278 , 902 alonso - herrero , a. , simpson , c. , ward , m.j . , &
wilson , a.s .
1998 , , 495 , 196 athanassoula , e. 1992 , , 259 , 345 bahcall , j.n .
, kirhakos , s. , & schneider , d.p .
1995a , , 447 , l1 bahcall , j.n . , kirhakos , s. , & schneider , d.p .
1995b , , 450 , 486 bahcall , j.n . ,
kirhakos , s. , saxe , d.h . , & schneider , d.p .
1997 , , 479 , 642 .
barnes , j.e . & hernquist , l.e .
1991 , , 370 , l65 bekki , k. & noguchi , m. 1994 , , 290 , 7 bekki , k. 1995 , , 276 , 9 berlind , a. , quillen , a.c . ,
pogge , r.w . , & sellgren , k. 1997 , , 114 , 107 bertin , g. , lin , c.c . ,
lowe , s.a . , &
thurstans , r.p .
1989 , , 338 , 78 biretta , j.a .
, et al . 1996 , wfpc2 instrument handbook , version 4.0 ( baltimore : stsci ) bohlin , r.c . , savage , b.d . , &
drake , j.f .
1978 , , 224 , 132 bower , g.a . ,
wilson , a.s . ,
mulchaey , j.s . ,
miley , g.k . ,
heckman , t.m . , &
krolik , j.h .
1994 , , 107 , 1686 buta , r. 1986a , , 61 , 609 buta , r. 1986a , , 61 , 631 capetti , a. , axon , d.j . , macchetto , f. , sparks , w.b . , & boksenberg , a. 1996 , , 469 , 554 carollo , c.m . , stiavelli , m. , & mack , j. 1998 , , 116 , 68 dahari , o. 1984 , , 89 , 966 derobertis , m.m . , hayhoe , k. , & yee , h.k.c .
1998 , , 496 , 93 de vaucouleurs , g. 1975 , , 29 , 193 elmegreen , b.g . ,
elmegreen , d.m . , &
montenegro , l. 1992 , , 79 , 37 elmegreen , b.g . ,
elmegreen , d.m . , brinks e. , yuan , c. , kaufman , m. , klari , m. , montenegro , l. , struck , c. , & thomasson , m. 1998 , , 503 , l119 englmaier , p. & shlosman , i. 1999 , , _ in press _
, astro - ph/9908256 erwin , p & sparke , l.s .
1998 , in galaxy dynamics , proceedings of a conference held at rutgers university , 8 - 12 august 1998 , ed .
merritt , m. valluri , and j.a .
sellwood , asp conf .
falcke , h. , wilson , a.s . , & simpson , c. 1998 , , 502 , 199 friedli , d. & benz , w. 1993 , , 268 , 65 fuentes - williams , t. & stocke , j. 1988 , , 96 , 1235 glass , i.s .
& moorwood , a.f.m .
1985 , , 214 , 429 glass , i.s .
1998 , , 297 , 18 gonzlez - delgado , r. , heckman , t. , leitherer , c. , meurer , g. , krolick , j. , wilson , a.s . , kinney , a. , & koratkar , a. 1998 , , 505 , 174 heckman , t. , krolick , j. , meurer , g. , calzetti , d. , kinney , a. , koratkar , a. , leitherer , c. , robert , c. , & wilson , a. 1995 , , 452 , 549 heckman , t.m .
, gonzlez - delgado , r. , leitherer , c. , meurer , g.r . ,
krolick , j. , wilson , a.s . ,
koratkar , a. , & kinney , a. 1997 , , 482 , 114 hernquist , l.e .
1989 , , 240 , 687 hernquist , l.e . &
mihos , j.c .
1995 , , 448 , 41 huchra , j. & burg , r. 1992 , , 393 , 90 huchra , j. , davis , m. , latham , d. , & tonry , j. 1983 , , 52 , 89 hutchings , j.b .
& neff , s.g .
1988 , , 96 , 1575 hutchings , j.b .
& neff , s.g . 1992 , , 104 , 1 hutchings , j.b .
& neff , s.g .
1997 , , 113 , 1514 keel , w.c .
, kennicutt jr . , r.c . , van der hulst , j.k . , & hummel , e. 1985 , , 90 , 708 keel , w.c .
1996 , , 111 , 696 kennicutt , r. 1989 , , 334 , 685 kormendy , j. 1979 , , 227 , 714 kotilainen , j.k . ,
ward , m.j . ,
boisson , c. , depoy , d.l . ,
smith , m.g . , & bryant , l.r .
1992 , , 256 , 125 kuchinski , l.e . & terndrup , d.m .
1996 , , 111 , 1073 kuchinski , l.e .
, terndrup , d.m .
, gordon , k.d . , & witt , a.n .
1998 , , 115 , 1438 laine , s. , knapen , j.h .
, prez - ramrez , d. , doyon , r. , & nadeau , d. 1999 , , 302 , l33 lin , c.c . & shu , f.h .
1964 , , 140 , 646 mcleod , k.k . & rieke , g.h .
1995 , , 441 , 96 mackenty , j.w .
, skinner , c. , calzetti , d. , & axon , d.j .
1997 , nicmos instrument handbook , version 2.0 , ( baltimore , stsci ) maciejewski , w. & sparke , l.s .
1997 , , 484 , l117 maiolino , r. & rieke , g.h .
1995 , , 454 , 95 malkan , m.a . ,
gorjian , v. & tam , r. 1998 , , 117 , 25 mihos , j.c . ,
richstone , d.o . , & bothun , g.d .
1992 , , 400 , 153 mihos , j.c . & hernquist , l.e .
1995 , , 425 , l13 mulchaey , j.s . & regan , m.w .
1997 , , 482 , l135 osterbrock , d.e .
1981 , , 249 , 462 osterbrock , d.e . &
martel , a. 1993 , , 414 , 552 peterson , b.m .
1997 , an introduction to active galactic nuclei , ( cambridge university press : cambridge ) piner , b.g . ,
stone , j.m . , & teuben , p.j .
1995 , , 449 , 508 phillips , a.c . ,
illingworth , g.d .
, mackenty , j.w . , & franx , m. 1996 , , 111 , 1566 pogge , r.w . & derobertis , m.m .
1993 , , 404 , 563 pogge , r.w .
1996 , in emission lines in active galaxies : new methods and techniques , ed .
peterson , f .- z .
cheng , & a.s .
wilson , asp conf .
, 113 , 378 pogge , r.w .
& martini , p. 2000 , , _ in preparation _ pogge , r.w . ,
martini , p. , & an , j.h .
, , 193 , 0604 quillen , a.c . , frogel , j.a . , kenney , j.d.p .
, pogge , r.w . , & depoy , d.l .
1995 , , 441 , 549 quillen , a.c . , alonso - herrero , a. , rieke , m.j . ,
mcdonald , c. , falcke , h. , & rieke , g.h .
1999 , , _ in press _ , astro - ph/9902233 regan , m.w . ,
vogel , s.n .
, & teuben , p.j .
1997 , , 482 , l143 regan , m.w . & mulchaey , j.s .
1999 , , 117 , 2676 rubin , v.c . & ford jr . ,
1983 , , 271 , 556 rubin , v.c . ,
kenney , j.d.p .
, & young , j.s .
1997 , , 113 , 1250 sandage , a. & tammann , g.a .
1987 , a revised shapley - ames catalogue of bright galaxies ( washington : carnegie institution of washington ) shaw , m. , axon , d. , probst , r. , & gatley , i. 1995 , , 274 , 369 shlosman , i. , frank , j. , & begelman , m.c .
1989 , , 338 , 45 shlosman , i. , begelman , m.c .
, & frank , j. 1990 , , 345 , 679 schwartz , m. 1981 , , 247 , 77 simpson , c. , wilson , a.s . , bower , g. , heckman , t.m . , krolik , j.h . , & miley , g.k . 1997 , , 474 , 121 spitzer jr . , l. 1978 , physical processes in the interstellar medium ( john wiley & sons : new york ) stephens , a.w . , frogel , j.a . , et al .
2000 , , _ submitted _ , astro - ph/9909001 struck , c. 1997 , , 113 , 269 toomre , a. & toomre , j. 1972 , , 178 , 623 toomre , a. 1981 , in the evolution of galaxies and stellar populations , ed .
fall & d. lynden - bell , ( cambridge university press : cambridge ) p. 111
van der kruit , p.c . & shostak , g.s .
1984 , , 134 , 258 white , r.e .
& keel , w.c .
1992 , , 359 , 129 white , r.e . ,
keel , w.c . , & conselice , c.j .
1996 , , astro - ph/9608113 witt , a.n . , thronson jr . ,
& capuano jr . , j.m .
1992 , , 393 , 611 lccccccc mrk 334 & & 1.8 & pec & disturbed & f606w & 91.5 & 444mrk 471 & & 1.8 & sba & & f606w & 136.9 & 664mrk 744 & ngc 3786 & 1.8 & sab(rs)a pec & interacting & f606w & 36.1 & 175ugc 12138&2237 + 07 & 1.8 & sba & & f606w & 102.8 & 498ngc 5033 & & 1.9 & sa(s)c & edge on & f547 m & 12.1 & 59ngc 5252 & & 1.9 & so & & f606w & 90.7 & 440ngc 5273 & & 1.9 & sa(s)0@xmath51 & & f547 m & 14.9 & 72ngc 5674 & & 1.9 & sabc & & f606w & 98.1 & 476um 146 & 0152 + 06 & 1.9 & sa(rs)b & & f606w & 71.6 & 347mrk 461 & & 2 & & & none & 65.6 & 318mrk 266 & ngc5256 & 2 & comp pec & & f606w & 111.0 & 538mrk 270 & ngc5283 & 2 & so ? & & f606w & 38.2 & 185mrk 573 & & 2 & ( r)sab(rs)o+ : & & f606w & 71.0 & 344ngc 1068 & & 2 & & & f606w & 16.2 & 79ngc 1144 & & 2 & ringb & & f606w & 116.5 & 565ngc 3362 & & 2 & sabc & & f606w & 108.8 & 527ngc 3982 & & 2 & sab(r)b : & & f606w & 17.1 & 83ngc 4388 & & 2 & sa(s)b : sp & edge on & none & 32.2 & 156ngc 5347 & & 2 & ( r)sb(rs)ab & & f606w & 31.4 & 152ngc 5695 & mrk 686 & 2 & sbb & & f606w & 56.9 & 276ngc 5929 & & 2 & sab : pec & interacting & f606w & 34.9 & 169ngc 7674 & mrk 533 & 2 & sa(r)bc pec & & f606w & 118.5 & 575ngc 7682 & & 2 & sb(r)ab & & f606w & 70.8 & 343ugc 6100 & a1058 + 45 & 2 & sa ? & & f606w & 117.6 & 570 lll mrk 334 & none & nuclear spiral mrk 471 & nuclear & nuclear spiral mrk 744 & host & nuclear spiral ugc 12138 & host & nuclear spiral ngc 5033 & none & edge on ngc 5252 & none & nuclear spiral ngc 5273 & none & nuclear spiral ngc 5674 & both & nuclear spiral um 146 & none & nuclear spiral mrk 461 & none & no visible mrk 266 & none & irregular mrk 270 & nuclear & nuclear spiral mrk 573 & both & nuclear spiral ngc 1068 & host & nuclear spiral ngc 1144 & ring & nuclear spiral ngc 3362 & host & nuclear spiral ngc 3982 & host & nuclear spiral ngc 4388 & none & edge on ngc 5347 & host & nuclear spiral ngc 5695 & host & nuclear spiral ngc 5929 & nuclear & nuclear spiral ngc 7674 & none & nuclear spiral ngc 7682 & host & nuclear spiral ugc 6100 & none & nuclear spiral lccc mrk 334 & 20 & 141 & 5 mrk 471 & 16 & 60 & 15 mrk 744 & 19 & 16 & 48 ugc 12138 & 14 & 113 & 10 ngc 5252 & 9 & 40 & 41 ngc 5273 & 17 & 7 & 130 ngc 5674 & 15 & 87 & 11 um 146 & 18 & 32 & 26 mrk 270 & 6 & 17 & 137 mrk 573 & 9 & 63 & 26 ngc 1144 & 15 & 51 & 19 ngc 3362 & 9 & 48 & 34 ngc 3982 & 9 & 8 & 215 ngc 5347 & 18 & 42 & 20 ngc 5695 & 14 & 25 & 43 ngc 7674 & 12 & 105 & 12 ngc 7682 & 9 & 31 & 52 ugc 6100 & 12 & 52 & 24
|
we present an investigation of possible fueling mechanisms operating in the inner kiloparsec of seyfert galaxies .
we analyze visible and near - infrared _ hubble space telescope _ images of 24 seyfert 2s from the cfa redshift survey sample .
in particular , we are searching for the morphological signatures of dynamical processes reponsible for transporting gas from kiloparsec scales into the nucleus .
the circumnuclear regions are very rich in gas and dust , often taking the form of nuclear spiral dust lanes on scales of a few hundred parsecs .
while these nuclear spirals are found in 20 of our 24 seyferts , we find only 5 nuclear bars among the entire sample , strongly reinforcing the conclusions of other investigators that nuclear bars are not the primary means of transporting this material into the nucleus .
an estimate of the gas density in the nuclear spirals based on extinction measurements suggests that the nuclear spiral dust lanes are probably shocks in nuclear gas disks that are not strongly self - gravitating .
since shocks can dissipate energy and angular momentum , these spiral dust lanes may be the channels by which gas from the host galaxy disks is being fed into the central engines .
|
one of the most extensively studied phases of liquid crystals , both for basic research and for applications , is the smectic - c * ( smc * ) phase of chiral molecules . in this phase ,
the molecules lie in layers and are tilted with respect to the layer normal direction .
the combination of molecular chirality , smectic layering , and tilt order leads to two effects : a ferroelectric polarization within the smectic layer plane and a helical modulation in the orientation of the molecular tilt from layer to layer @xcite .
the ferroelectric polarization is useful for display devices , which use an applied electric field to switch the molecular orientation @xcite .
it is also useful for thermal sensors , which measure the temperature variation of the polarization , known as the pyroelectric effect @xcite .
both of these applications require a uniform orientation of the molecules .
hence , the helix must be suppressed , or unwound , by an applied electric field or by boundary effects .
helix unwinding has been modeled through continuum elastic theory . in a bulk
smc * phase , unwinding induced by an applied electric field can be described by the sine - gordon equation , presented below @xcite . under an electric field ,
the helix distorts and the helical pitch increases .
if the field is increased above a critical threshold , the pitch diverges and the helix is suppressed . by contrast ,
helix unwinding induced by boundary effects in a narrow cell is more complex @xcite . in this case
, a helix must have a series of disclination lines near the cell surfaces , which separate a twisted interior region from a uniform surface region .
continuum elastic theory shows that the unwinding transition is controlled by an energetic competition between the helical state with disclinations and the uniform untwisted state . if the cell thickness is below a critical threshold comparable to the helical pitch , the uniform state is favored and the helix is suppressed . (
a third possible mechanism for helix unwinding is shear flow @xcite , but this is not often used in smc * liquid crystals , and we will not discuss it here . ) in this paper , we investigate the unwinding of a smc * helix by boundary effects , or by combined boundary and electric - field effects , through a series of monte carlo simulations of a continuum free energy discretized on a 2d lattice .
these simulations serve two purposes .
first , they allow us to visualize the complex director configuration within a narrow cell as a function of cell thickness and electric field .
we obtain snapshots of the molecular tilt profile through the helix unwinding process .
second , they allow us to relate the microscopic director configuration to two macroscopic variables , an average chiral order parameter and the net electrostatic polarization of the cell . the latter variable can be compared with experimental measurements of smc * cells . in these simulations , we first consider boundary effects alone .
we simulate the smc * phase in narrow cells , and determine the structure of the helix as a function of cell thickness .
we confirm that the helix unwinds at a threshold thickness approximately equal to the helical pitch , in agreement with continuum elastic theory .
we then use the same monte carlo approach to simulate helix unwinding due to the combined effects of cell boundaries and electric field . in these simulations , we calculate both the molecular tilt profile across the cell and the net electrostatic polarization . we observe three distinct regimes of response , in which the helical pitch is first distorted and then expelled as the applied field is increased . by varying the cell thickness ,
we determine how boundary effects shift the critical field for helix unwinding and lower the saturation polarization .
these simulation results are consistent with trends observed in experiments .
the plan of this paper is as follows . in sec .
ii , we review the continuum elastic theory of the smc * phase .
we show how an applied electric field unwinds the smc * helix in a bulk system , and sketch an approximate energetic argument for a finite cell . in sec .
iii , we use monte carlo simulations to investigate the effects of finite cell thickness under zero applied electric field . in sec .
iv , we combine the effects of cell boundaries and applied electric field to determine the tilt profile and the polarization , and compare the polarization with experimental measurements .
-axis , perpendicular to the smectic layers .
an electric field is applied along the @xmath0-axis.,height=288 ] a smc * liquid crystal in a narrow cell has the idealized bookshelf geometry shown in fig . [ geometry ] .
the molecules lie in layers , and they are tilted with respect to the smectic layer normal .
the 3d orientation is represented by the director @xmath1 .
this unit vector can be written in spherical coordinates as @xmath2 , where @xmath3 is the polar angle of the tilt and @xmath4 is the azimuthal angle .
the molecular tilt is conventionally described in terms of the projection of @xmath1 into the layer plane , @xmath5 . in this geometry , we expect the director to depend on the @xmath0 and @xmath6 coordinates .
the @xmath6 coordinate is along the smectic layer normal . because of the molecular chirality
, the molecular orientation rotates in a helix from layer to layer , which makes the director a periodic function of @xmath6 .
the @xmath0 coordinate is the narrow dimension of the cell , with a thickness @xmath7 of order microns , across which an electric field @xmath8 is applied .
the molecules may interact strongly with the front and back surfaces of the cell , at @xmath9 and @xmath7 . as a result ,
the director may rotate as a function of @xmath0 .
by contrast , we do not expect the director to depend on the third coordinate @xmath10 . in this bookshelf geometry ,
the system is uniform as a function of @xmath10 .
( in certain liquid crystals , the smectic layers buckle as a function of @xmath10 @xcite , but we do not consider that effect here . ) for that reason , we write the angles @xmath3 and @xmath4 as functions of @xmath0 and @xmath6 .
we can now construct the simplest model free energy to describe variations in the director . this model must include four interactions .
first , the smectic layer order interacts with the molecular orientation , and favors a particular tilt of the molecules with respect to the layers .
this interaction can be expanded as a power series in the tilt magnitude @xmath11 , which gives @xmath12 , for some series coefficients @xmath13 and @xmath14 . in terms of these coefficients ,
the favored tilt is @xmath15 .
second , there is an electroclinic interaction of the molecules with the applied electric field @xmath8 .
because of the molecular chirality , this interaction couples the field in the @xmath0 direction with the tilt in the @xmath10 direction , so it can be written as @xmath16 . in other words ,
the electrostatic polarization is @xmath17 , or @xmath18 .
third , there is a chiral interaction that favors a variation of the director from layer to layer , which can be written as @xmath19 .
fourth , there is the frank free energy for elastic distortions of the director , which limits the variations from layer to layer .
this contribution can be written as @xmath20 , summed over @xmath21 and @xmath22 .
putting these pieces together , the total free energy density becomes @xmath23 this free energy is invariant under rotations in the @xmath24 plane , but it is not invariant under reflections , because reflection symmetry is broken by the molecular chirality . .
solid line : perfect helix for @xmath25 . dashed line : distorted helix under moderate @xmath26 .
dot - dashed line : single soliton in the director , near the threshold for helix unwinding .
dotted line : unwound uniform director.,width=324 ] we can see immediately that this minimal model leads to a helix in the @xmath6 direction .
consider a director of the form @xmath27 , where the magnitude @xmath28 is constant and the azimuthal angle @xmath4 depends on position .
the free energy density simplifies to @xmath29 in the limit of zero electric field , @xmath25 , we minimize this free energy over @xmath4 to obtain @xmath30 hence , the azimuthal angle increases linearly as @xmath31 , with @xmath32 , as shown by the solid line in fig .
[ helix ] .
this linear increase in @xmath4 implies a perfect sinusoidal helix in @xmath33 .
this minimal model also predicts unwinding of the helix under an applied electric field .
this behavior is analogous to the theory of meyer for a cholesteric phase in an electric or magnetic field @xcite . for @xmath34 , minimizing the free energy over @xmath4 gives @xmath35 this is the standard sine - gordon equation .
the form of the solutions depends on the value of @xmath26 , as shown in fig .
[ helix ] . for low @xmath26 ,
the helix is distorted , so that the director is approximately aligned with the field in most of the system . as @xmath26 increases
, the helix becomes even more distorted , with sharper steps between domains where @xmath4 is approximately a multiple of @xmath36 .
eventually the system crosses over into regime of uniform domains separated by sharp domain walls , or solitons . in that regime , a single soliton has the profile @xmath37 $ ] , where @xmath38 .
as @xmath26 continues to increase , the spacing between the domain walls increases , or equivalently their density decreases . at the critical threshold @xmath39 ,
the last domain wall vanishes and the system becomes uniform .
the question is now : what other types of influences can also unwind a smc * helix ? clearly one possibility is surface effects .
interactions along the front and back surfaces of a finite cell , at @xmath9 and @xmath7 , can anchor the director at those surfaces .
if the elastic interactions described by the parameter @xmath40 are sufficiently strong , and the cell thickness @xmath7 is not too big , then this anchoring may extend throughout the interior of the cell , giving a uniform director .
this is the basis of surface - stabilized ferroelectric liquid crystal cells @xcite .
the threshold thickness for unwinding a helix is not obvious . as shown in fig .
[ geometry ] , the helical pitch is along the @xmath6 direction , but the narrow dimension of the cell is along the @xmath0 direction . because these directions are perpendicular , there is no simple geometric reason why a helix must unwind when the cell thickness is less than the pitch .
rather , there must be some energetic argument that relates these two length scales .
an energetic argument has been developed for narrow cells of the cholesteric phase @xcite , and has been extended to narrow cells of the smc * phase @xcite . in its simplest form ,
the argument can be stated as follows @xcite .
if a cell has a helix in the interior , but a uniform director along the front and back surfaces , then it must have a series of disclination lines running along the @xmath10 direction near the surfaces .
there must be one disclination line for each helical pitch .
we can compare the energy of the helix ( with disclinations ) with the energy of the uniform state to find the threshold thickness for helix unwinding .
the energy of the helix ( with disclinations ) is the negative energy gained from the helix plus the positive energy lost to the disclinations , @xmath41 where @xmath42 is the disclination line energy per unit length .
the helix unwinds if @xmath43 , which implies @xmath44 since the line energy @xmath42 should be of order @xmath40 , the threshold thickness should be comparable to the pitch . in the following sections , we test this argument through a series of monte carlo simulations .
in these simulations , we obtain snapshots of the director configuration for different cell thicknesses , both at zero field and under an applied electric field .
these snapshots provide specific illustrations of the disclinations discussed above . for zero field ,
the simulations confirm that the helix unwinds at a critical thickness approximately equal to the helical pitch . for finite field ,
the simulations show helix unwinding induced by the combined effects of boundaries and electric field in a cell above the critical thickness . in both cases , we relate the microscopic snapshots of the director configuration to macroscopic variables . one of these variables , the net electrostatic polarization , can be compared with experimental measurements .
to model helix unwinding in a finite cell of the smc * phase , we map the system onto a 2d square lattice .
the lattice dimensions represent the @xmath45 plane shown in fig .
[ geometry ] : @xmath0 is the narrow dimension of the cell and @xmath6 is the smectic layer normal .
we assume the system is uniform in the @xmath10 direction . on each lattice site
@xmath46 there is a 3d unit vector @xmath47 representing the local molecular director .
this vector can be parametrized in terms of the polar angle @xmath48 and azimuthal angle @xmath49 , or equivalently in terms of the projection @xmath50 into the smectic layer plane . for the lattice simulations , we discretize the free energy of eq .
( [ freeenergy ] ) to obtain @xmath51\\ & & + \frac{1}{2}k(|\mathbf{c}(y+1,z)-\mathbf{c}(y , z)|^2 + |\mathbf{c}(y , z+1)-\mathbf{c}(y , z)|^2 ) \biggr ] , \nonumber \label{latticefreeenergy}\end{aligned}\ ] ] with @xmath52 in this section .
this free energy is similar to the free energy studied in ref . @xcite but with one important difference : that paper simulated the @xmath24 plane , but we now simulate the @xmath45 plane . a further consideration for the simulations is boundary conditions .
experimental cells may be symmetric , with the local director aligned along the same direction on both front and back confining walls , or they may be asymmetric , with the director aligned along one wall and an open boundary on the other side . in our simulations we use an aligning boundary condition with the director fixed on the wall at @xmath9 with a specified tilt angle . on the other wall at @xmath53 we use the boundary condition
@xmath54 . this arrangement can represent an asymmetric cell , or one half of a symmetric cell , with the other half a mirror image of the first .
experimental cells are very large in the @xmath6 direction so that the top and bottom boundaries should not affect the physics of the interior . in the simulations , we use the boundary condition @xmath55 for the top and bottom boundaries .
we simulate the system with the parameters @xmath56 , @xmath57 , @xmath58 , @xmath25 , @xmath59 , and @xmath60 and @xmath61 .
the small value of @xmath13 corresponds to a tilt angle of approximately @xmath62 .
we use the large @xmath6 dimension of 160 , and several values of the thickness @xmath7 in the @xmath0 direction . for each set of parameters and system size ,
we begin the simulations in a disordered state at a high monte carlo temperature , and then slowly reduce the temperature until the system settles into a single ground state and the fluctuations in @xmath33 become negligible .
this procedure can be regarded as a simulated - annealing minimization of the lattice free energy of eq .
( [ latticefreeenergy ] ) .
( a ) plane for simulations with @xmath60 , leading to a pitch of approximately 38 .
the system has an aligning boundary along the left side ( @xmath9 ) .
note that the @xmath0 axis is exaggerated compared with the @xmath6 axis .
( a ) for a thickness of 10 ( @xmath9 through @xmath63 ) , the system is uniform , except for some edge effects at the top and bottom .
( b ) for a thickness of 12 , the system has a clear helix.,title="fig:",height=217 ] ( b ) plane for simulations with @xmath60 , leading to a pitch of approximately 38 .
the system has an aligning boundary along the left side ( @xmath9 ) .
note that the @xmath0 axis is exaggerated compared with the @xmath6 axis .
( a ) for a thickness of 10 ( @xmath9 through @xmath63 ) , the system is uniform , except for some edge effects at the top and bottom .
( b ) for a thickness of 12 , the system has a clear helix.,title="fig:",height=217 ] ( a ) plane for simulations with @xmath64 , leading to a pitch of approximately 75 .
( a ) for a thickness of 23 , the system is uniform , except for some edge effects .
( b ) for a thickness of 25 , there is a clear helix.,title="fig:",height=217 ] ( b ) plane for simulations with @xmath64 , leading to a pitch of approximately 75 .
( a ) for a thickness of 23 , the system is uniform , except for some edge effects .
( b ) for a thickness of 25 , there is a clear helix.,title="fig:",height=217 ] to visualize the simulation results , we draw the @xmath45 plane in figs . [ lambda25 ] and [ lambda125 ] .
the director configuration is represented by short lines that show the projection of the 3d director into the @xmath45 plane .
hence , vertical lines indicate @xmath65 , and tilted lines indicate @xmath66 . because the lines representing the director are drawn close together , helical regions resemble twisted ribbons .
figure [ lambda25 ] shows the simulation results for @xmath60 and @xmath59 . for these parameters ,
the favored wavevector is @xmath67 , and hence the unperturbed pitch is @xmath68 . for a thickness of 10 , the system is uniform , with @xmath65 everywhere except near the top and bottom surfaces .
those distortions are edge effects within a fractional pitch of the free surfaces , which do not affect the bulk behavior inside the cell .
hence , we see that the system of thickness 10 is unwound . by contrast , for a thickness of 12 , the system shows a well - defined helix , with a periodic modulation of @xmath69 throughout the cell , except very close to the aligning surface at @xmath9 .
thus , there is a clear helix winding / unwinding transition as a function of thickness .
the transition occurs at a critical thickness between 10 and 12 for a cell with asymmetric boundary conditions ( or a half - thickness between 10 and 12 if the simulation is regarded as half of a symmetric cell ) .
this critical thickness is not equal to the pitch , but it is certainly the same order of magnitude , in agreement with the theoretical expectation of eq .
( [ thicknessthreshold ] ) . figure [ lambda125 ] shows the corresponding results for @xmath64 . in this case , the favored wavevector is @xmath70 , so the unperturbed pitch is @xmath71 . for a thickness of 23 ,
the system is uniform , again except for some edge effects near the top and bottom surfaces . by comparison , for a thickness of 25
, there is a distinct helix throughout the interior of the cell .
thus , there is a helix winding / unwinding transition with a critical thickness between 23 and 25 .
this critical thickness is approximately twice the critical thickness of the previous case .
hence , we see that the critical thickness is approximately proportional to the pitch , again in agreement with the theoretical expectation .
defined in eq .
( ) , as a function of the system thickness @xmath7 . for @xmath60 ( pitch @xmath72 )
, the unwinding transition occurs at a thickness between 10 and 12 . for @xmath64 ( pitch @xmath73 ) ,
the transition occurs at a thickness between 23 and 25.,width=324 ] for a quantitative measurement of helix winding and unwinding , we must define a chiral order parameter .
one simple choice of a chiral order parameter is just the chiral term of the free energy ( [ latticefreeenergy ] ) , without the factor of @xmath74 itself , @xmath75 .
\label{orderparameter}\end{aligned}\ ] ] figure [ orderparameterfig ] shows this order parameter as a function of the system thickness @xmath7 for both series of simulations , with @xmath60 and @xmath61 . for @xmath60
, the plot shows a sharp transition between thickness 10 and 12 , as @xmath76 jumps from 0.022 to 0.24 . for @xmath64 , there is a distinct transition between thickness 23 and 25 , as @xmath76 jumps from 0.026 to 0.13 .
this analysis confirms that the winding / unwinding transition occurs at a thickness that is proportional to , and the same order of magnitude as , the helical pitch .
in the previous section , we showed that a smc * helix can be unwound by surface effects in a finite cell .
if the thickness is greater than the critical threshold , the helix is present at zero electric field .
however , when an electric field is applied , the helix can be unwound by the combined effects of the surfaces and the electric field . in this section
, we simulate that combination of surface and field effects . for these simulations
, we use the same monte carlo approach as in the previous section .
we use the discretized free energy of eq .
( [ latticefreeenergy ] ) with the parameters @xmath77 , @xmath57 , @xmath58 , @xmath59 , and @xmath60 .
this value of @xmath13 corresponds to a tilt angle of approximately @xmath78 .
we perform the simulations for four values of the cell thickness , @xmath79 , 20 , 40 , and 60 , and scan through the electric field @xmath26 at each thickness . for these parameters , the unperturbed pitch is @xmath80 and hence , based on the results of the previous section , the zero - field winding / unwinding transition occurs at a thickness between 10 and 12 .
hence , the simulations at thickness @xmath81 should be unwound at all values of the electric field , while the simulations at larger thickness should go through the winding / unwinding transition as a function of electric field .
figure [ varyfield ] shows the simulation results for the system of thickness 20 . for @xmath25 ,
the system has a helix everywhere in the cell , except a narrow region near the aligning boundary at @xmath9 . by comparison , for @xmath82 ,
the helix is suppressed in much of the cell .
it persists only in regions of the cell near the free surfaces at the top and bottom .
this behavior near the free surfaces is consistent with the previous section , which showed that free surfaces tend to favor the helical modulation .
when the field increases to @xmath83 , the helix is suppressed in more of the cell , and it persists only in smaller regions near the top and bottom surfaces .
once the field reaches @xmath84 , the helix is suppressed throughout the interior of the cell .
the director in the cell is now uniform , except for very narrow edge effects at the top and bottom .
hence , the electric field has driven the finite cell through the helix winding / unwinding transition .
as in the previous section , we need an order parameter to describe the winding / unwinding transition quantitatively . in this case , the electrostatic polarization provides an experimentally relevant order parameter , which shows how the net polar order of the cell couples to the electric field . as argued in sec .
ii , the polarization is the quantity conjugate to the electric field in the free energy , and hence @xmath17 , or @xmath85 .
we average this quantity over the cell to obtain @xmath86 figure [ pvse ] shows the simulation results for the polarization as a function of electric field for each cell thickness .
figure [ pvse](a ) presents the results on a linear scale , and fig .
[ pvse](b ) presents the same results on a logarithmic scale .
note that the polarization is not zero at zero field because of the symmetry - breaking surface alignment at @xmath9 . from the plots in fig .
[ pvse ] , we can see that the system has three distinct regimes in its response to an electric field . for low field @xmath87
, there is a regime of helix unwinding . in this regime
, an increasing electric field gradually aligns the directors , suppresses the helix , and prevents the local polarization from averaging to zero . as a result
, the net polarization increases rapidly as a function of field .
( this low - field regime does not occur for the lowest thickness @xmath79 , because the helix is suppressed by surface effects even without a field . ) for intermediate field @xmath88 , there is a regime of suppressed twist . in this regime ,
the helix is already unwound , so the only effect of electric field is to increase the local tilt and polarization .
hence , the net polarization increases more slowly , roughly as @xmath89 .
finally , for high field @xmath90 , there is a saturated regime . here
, the helix is already unwound , the local tilt is at its maximum value @xmath91 , and the polarization is at its maximum value of @xmath92 .
although the simulations only go to @xmath93 , we see that the polarization can not increase at higher field because it is already saturated . in fig .
[ pvse ] , we can also compare the relative polarization of thinner and thicker cells . for low field ,
thinner cells have a higher polarization than thicker cells , because the helix reduces the polarization in thicker cells but the aligning boundary at @xmath9 unwinds the helix in thinner cells .
by contrast , for high field , thicker cells have a higher polarization than thinner cells , because there is no helix at any thickness , and surface effects at @xmath53 suppress the polarization in thinner cells .
a similar high - field limit has been discussed by shenoy _
et al . _
@xcite , who see the same effect experimentally in cells with different boundary conditions . for intermediate field ,
the polarization of thinner and thicker cells must cross . to compare our results with experimental measurements of the polarization as a function of electric field
, we must take into account one subtlety of the experiments . as shown by ruth _
et al . _
@xcite , the polarization observed experimentally ( by the triangle - wave technique or other techniques ) is not the total polarization conjugate to the electric field .
rather , it is a specific nonlinear component of the polarization , which can be written as @xmath94 the difference between the total polarization @xmath95 and the observable polarization @xmath96 is small when @xmath95 is saturated , but it is significant whenever @xmath95 varies with @xmath26 .
hence , we must extract @xmath96 from the simulations and compare that quantity with experiments . to extract @xmath96 from the simulations ,
we need to calculate the derivative @xmath97 .
for that reason , we fit the simulation results for the polarization to the function @xmath98 .
this is just an empirical fitting function , with no theoretical basis , but it gives a fairly good fit to the data in fig .
[ pvse](a ) .
we can then differentiate this function and calculate the observable polarization @xmath96 that corresponds to the simulation results .
( a ) in cells of four thicknesses .
( b ) experimental data for @xmath99 in three cells .,title="fig:",width=307 ] ( b ) in cells of four thicknesses .
( b ) experimental data for @xmath99 in three cells .,title="fig:",width=307 ] figure [ pobsvse](a ) shows the results for @xmath96 extracted from the simulations at each system size .
this function is small for low @xmath26 , then increases towards its saturation level @xmath100 at high @xmath26 .
as discussed above , thinner cells have a higher value of @xmath99 at low field , and thicker cells have a higher value of @xmath99 at high field . by comparison , fig .
[ pobsvse](b ) shows sample experimental measurements of @xmath99 as a function of applied electric field in three cells with asymmetric boundary conditions @xcite .
the material is 10ppbn4 ( described in ref .
@xcite ) , and the temperature is 7 @xmath101c below the sma - smc transition .
note that the experimental data show the same general features as the theoretical plots .
in particular , we see the same crossover between higher @xmath99 in thinner cells at low field and higher @xmath99 in thicker cells at high field .
thus , the simulation results are consistent with the trends seen in experiments . in conclusion
, we have developed an approach for simulating the helix winding / unwinding transition in smc * liquid crystals .
this approach is based on a minimal model for the free energy , which includes a chiral term that favors a helical modulation of the director from layer to layer .
this bulk free energy competes with surface effects and electric - field effects , which both favor a uniform alignment of the director . in zero field , the competition between the bulk helix and the surface alignment leads to helix unwinding at a critical thickness approximately equal to the helical pitch .
when an electric field is applied , the field - induced alignment adds to the surface effects , and induces helix unwinding even for thicker cells .
the electrostatic polarization is an appropriate order parameter to quantify this field - induced unwinding , and our simulation results for the polarization are consistent with experimental measurements .
we would like to thank r. b. meyer for helpful discussions and k. a. crandall for sharing his unpublished data with us .
this work was supported by the u. s.navy through contract no . n00173 - 99 - 1-g015 , and by the national science foundation through grant no .
dmr-9702234 and dmr-0116090 .
j. w. goodby , r. blinc , n. a. clark , s. t. lagerwall , m. a. osipov , s. a. pikin , t. sakurai , k. yoshino , and b. ek , _ ferroelectric liquid crystals : principles , properties and applications _ ( gordon and breach , amsterdam , 1991 ) .
|
in bulk ferroelectric liquid crystals , the molecular director twists in a helix . in narrow cells , this helix can be unwound by an applied electric field or by boundary effects . to describe helix unwinding as a function of both electric field and boundary effects
, we develop a mesoscale simulation model based on a continuum free energy discretized on a two - dimensional lattice . in these simulations ,
we determine both the director profile across the cell and the net electrostatic polarization . by varying the cell size ,
we show how boundary effects shift the critical field for helix unwinding and lower the saturation polarization .
our results are consistent with experimental data .
|
the theory of the cuprate high temperature superconductors remains one of the most important and daunting problems in condensed matter physics .
the high @xmath2 cuprate superconductors are layered materials with relatively complex structures and chemical composition .
they are highly correlated , with an effective bandwidth roughly equal to the effective local coulomb interaction .
the short - range correlations are known to play a paramount role in these materials .
therefore , the dynamical cluster approximation ( dca ) @xcite , which treats short - range correlations explicitly and the long - range physics at the mean - field level , is an ideal tool for the investigation of these systems . a common characteristic all the cuprate materials share is the presence of quasi - two - dimensional @xmath3 planes .
these planes are commonly believed to contain the low - energy physics . however , the full complexity of the orbital chemistry of just the @xmath4 planes and the strong coulomb repulsion on the cu ions would lead to models which are very difficult to study with conventional techniques .
the cuprates are characterized by a very rich , but also in many respects very intriguing physics .
the undoped materials are antiferromagnetic ( afm ) insulators with a gap of approximatively @xmath5 . upon doping the afm
is destroyed and the system becomes superconducting ( sc ) . at small doping , in the proximity of the afm phase , the normal state physics can not be described in terms of fermi liquid theory and is characterized by the presence of a pseudogap .
an essential demand of every successful theory is to capture all these fundamental features at the same time .
experimental data show that the phase diagram and other physical characteristics like the density of states ( dos ) near the fermi level of the hole - doped and electron - doped materials are very different @xcite
. there could be many reasons for this asymmetry .
the electron and the hole doped materials are physically different and apart from the @xmath4 planes , they contain different elements and chemical structures . these structural and compositional differences can influence the low - energy physics . therefore in this paper
, we use dca to address whether the physics of a pure @xmath4 plane contains this asymmetry or if the origin of the asymmetry in real materials comes from other influences .
different models for describing the physics of a @xmath3 plane were proposed by various authors .
photoemission experiments in the insulating parent material show that the first electron - removal states have primarily oxygen character ; whereas , the first electron - addition states have @xmath6 character , already suggesting a strong asymmetry .
this places these materials in the charge - transfer gap region of the zaanen - sawatzky - allen scheme @xcite .
early on , considering the ligand field symmetry and band structure calculations @xcite , it was realized that the most important degrees of freedom are the cu @xmath7 which couple with the in - plane o @xmath8 orbitals .
therefore , one of the first models proposed to describe the physics of high @xmath2 materials was the so called _ three - band hubbard model _ introduced by varma _
et al . _
@xcite and emery _ et al . _
@xcite , which considers explicitly both the oxygen @xmath9 and the cooper @xmath7 orbitals .
in fact , because the direct oxygen - oxygen hopping is neglected , only the combination of oxygen orbitals with @xmath10 symmetry couples with the @xmath6 orbitals , and the above proposed three - band model reduces to a two - band one .
however zhang and rice ( zr ) @xcite argue that the low - energy physics of the _ hole doped _ superconductors can be described by a single - band model . starting from the two - band model , zhang and rice claim that an extra hole added into the oxygen band binds strongly with a hole on the @xmath11 , forming an on - site singlet .
this singlet state , which has zero spin can be thought as moving through the lattice like a hole in an antiferromagnetic background .
consequently , the physics can be described by a one - band t - j model .
pertinent criticism to these simplified models were raised by various authors . with respect to cu degrees of freedom , h. eskes _ et al . _
@xcite stressed the possible importance of the other @xmath6 orbitals , showing that they should be explicitly considered when physics which implies excitations with energy larger than @xmath12 is involved .
however , these criticisms do not concern us for the present study since we are interested only in physics at energies lower than @xmath13 . investigating the relative importance of various parameters describing the @xmath3 planes it was realized early on that in addition to the cu on - site coulomb repulsion ( @xmath14 ) and cu - o hopping integral ( @xmath15 ) the o - o hopping integrals result in a large o @xmath16 band width ( @xmath17 ) , indicating that these should be included explicitly in any theory @xcite
therefore , using the dca technique as a means of including all these most important parameters and bands , we address two major problems in this paper : the physics of the @xmath3 plane ( including a detailed study of the electron - hole asymmetry ) and the reduction of the multi - band model to a single - band model .
regarding the reduction to a one - band model , one of the most serious criticisms to zr theory is the neglect of the o @xmath16 band structure @xcite .
the natural tendency of the finite oxygen bandwidth is to delocalize and destabilize the zr singlets .
the question arises whether the low - energy states ( i.e. the zr singlets ) are still well separated from the higher energy states ( i.e. the non - bonding oxygen states ) .
otherwise , the reduction to a single - band model which neglects these high energy states is not possible .
this problem was previously considered by eskes and sawatzky @xcite within an impurity calculation approach , but there , unlike in the dca approach , both the spatial correlation effects and the dispersion of the low - energy states were neglected .
another important objection to zr theory was first raised by emery and reiter @xcite , and regards the nature of the low - energy states .
are these states real singlets which can be mapped onto holes , or does the hole on the o bind into a more complicated state which involves more than one cu hole ? choosing a particular solvable example , which considers the cu spins arranged ferromagnetically , they showed that the low - energy states are in fact an admixture of the zhang - rice singlets and the corresponding triplets .
this implies a nonzero value for the oxygen spin , and destroys the equivalence of these states to holes .
however , it is not clear if the situation is similar in the cuprates , i.e. if the zr singlet - zr triplet admixture is significant .
but the merit of emery and reiter is to emphasize that the fact that , as a consequence of the strong cu - o hybridization low - energy states well separated from the non - bonding oxygen band states appear , does not necessarily mean that the physics can be reduced to a single - band model .
the third problem we address regarding the reduction to a single - band model is the estimation of the single - band parameters .
we notice that different approximations result in different values of the parameters .
especially the magnitude of the next - nearest - neighbor hopping is very dependent of the initial assumptions .
for example , if we assume that the hole addition low - energy states are genuine zr singlets , i.e bound states between a cu hole and a orthogonal wannier oxygen orbital , we obtain a negligible next - nearest - neighbor hopping @xcite . on the other hand , if we consider the low - energy states to be plaquette singlets , i.e. bound states between a cu hole and a hole on the state formed by the four oxygen around the cu , the value of the next - nearest - neighbor hopping is significant @xcite . of course , because of the nonorthogonality of the plaquette states , the plaquette singlets are not genuine singlets and therefore they can not be rigorously mapped into holes .
however , because their overlap with the local singlets is large ( @xmath18 ) @xcite , it is still possible that this approximation is good .
our calculations show that a multi - band model and a single - band t - t-u hubbard model with a significant value of the next - nearest - neighbor hopping exhibit a similar low - energy physics .
the essential parameter needed for the agreement is the next - nearest - neighbor hopping , @xmath0 .
this parameter is also the main culprit for the observed electron - hole asymmetry .
however , as mentioned above , the large value of @xmath0 can not be obtained in a strict zr picture .
thus our results also implicitly indicate that the multi - band model can not be rigorously reduced to a single - band model .
therefore , besides showing the similarities between the two models , we also point out their significant differences in this paper .
the final conclusion is that a single - band t - t-u hubbard model , with a significant value of @xmath0 , captures the basic physics of the cuprates , and thus is suitable to describe the afm , pseudogap and sc physics together with the relevant asymmetries observed in the phase diagram , in the one - particle spectra and in the two - particle response functions .
however , we believe that it is not suitable for a quantitative material specific analysis , for describing the higher energy spectroscopic features as in optical spectroscopy or resonant inelastic x - ray scattering , or for studying more subtle features related to the finite value of the spin on the oxygen . this paper is organized as follows . in section [ sec : form ] the two - band hubbard model and the dca technique
is introduced .
our two - band model takes fully into account the oxygen dispersion and considers only the oxygen degrees of freedom which couple directly to the cu @xmath7 orbitals .
the results of the dca calculation applied to the two - band hamiltonian are presented in section [ sec:2bm ] .
the possible reduction of the two - band model to a single - band one , together with a detailed analysis of the single - band t - t-u hubbard model , is addressed in section [ sec:1bm ] . a discussion regarding the similarities and the differences between the two - band model and the single - band one is given in section [ sec : disc ] .
the conclusions of our study are reviewed in section [ sec : conc ] .
band structure calculations @xcite , cluster calculation @xcite , photoemission @xcite and other experiments show that the relevant cu degrees of freedom are the @xmath7 orbitals which couple with the in - plane @xmath19 and @xmath20 o orbitals .
all these degrees of freedom result in a five - band ( four oxygen and one copper band ) hamiltonian in general .
we have studied the five - band model in detail @xcite and have found that due to the strong cu - o hybridization , only the oxygen degrees of freedom which couple directly with cu are relevant for the low - energy physics .
consequently , to a very good approximation , the five - band model can be reduced to a two - band one .
the two - band model contains one cu @xmath7 correlated band and one oxygen band . at every site
the oxygen states are obtained by taking a linear combination with @xmath21 symmetry of the four o @xmath22 orbitals which form a plaquette around the cu ion .
these are the only oxygen states which hybridize directly with cu .
however , it should be mentioned that these plaquette states are not orthogonal , two neighboring states sharing a common oxygen atom .
an orthogonal basis can be obtained by the procedure described in the original zr paper @xcite .
first , applying a fourier transform , translational invariant ( bloch ) states are constructed .
the bloch states are orthogonal but not normalized , so they should be multiplied by a normalization factor @xmath23^{-1/2}~.\ ] ] after normalization a complete and orthonormal set of oxygen states is obtained . in this basis
the two - band hubbard hamiltonian can be written as : @xmath24 we work in the hole representation and @xmath25 ( @xmath26 ) represents the creation operator of a cu ( o ) hole with spin @xmath27 and momentum @xmath28 .
the o band dispersion and the cu - o hybridization are given by @xmath29 @xmath30 with @xmath31 being the o - o hopping integral .
the last term in eq.[eq : ham ] represents the coulomb repulsion between two holes on the same @xmath6 orbital .
we choose the commonly accepted values of the parameters , based on the band structure calculations of mcmahan _ et al .
_ @xcite and hybertsen _ et al .
_ @xcite because of the low density of oxygen holes ( @xmath32 ) , we treat the coulomb repulsion on @xmath33 ( given by @xmath34 ) and the repulsion between nearest - neighbor cu and o holes ( given by @xmath35 ) at the mean field level as a reasonable approximation .
the effect will be an increases of our estimation for @xmath36 by @xmath37 , where @xmath38 and @xmath39 are the average occupation of cu and respectively o bands . a choice of @xmath40 , @xmath41 and @xmath42 results in a increase of @xmath43 by @xmath44 . to conclude , we take in eq . [
eq : ham ] , @xmath45 and @xmath46 . ) around @xmath47=@xmath48 and @xmath49.,width=240 ] the dca is an extension of the dynamical mean field theory ( dmft)@xcite . the dmft maps the lattice problem to an impurity embedded self - consistently in a host and therefore neglects spatial correlations .
the dca maps the lattice to a finite - sized cluster embedded in a host .
non - local correlations up to the cluster size are treated explicitly , while the physics on longer length scales is treated at the mean field level .
this technique is ideal for the problems where short - ranged correlations are predominant as in the high-@xmath2 materials . here
we calculate the properties of the embedded cluster with a quantum monte carlo ( qmc ) algorithm .
the cluster self - energy is used to calculate the properties of the host , and this procedure is repeated until a self - consistent convergent solution is reached .
the self - energy and vertex functions of the cluster are then used to calculate lattice quantities .
a detailed description of the qmc - dca algorithm is given in ref .
@xcite in the two - band model the oxygen degrees of freedom are not correlated , and therefore they are not included explicitly in the cluster .
their effect is fully contained in the cluster - host hybridization function and in the host green s function .
here we consider a @xmath50 cluster of cu ions , which we believe to be large enough to capture the essential physics of hubbard - type models .
the @xmath50 cluster will result in a coarse - graining of the bz in four cells , as shown in fig .
[ fig : coarse ] .
the solid line is the @xmath6 part of the dos calculated at @xmath51 .
the value of the parameters is @xmath52 .
the dashed line shows the dos when @xmath53 .
the chemical potential @xmath54.,width=316,height=211 ] the undoped materials have one hole per @xmath55 unit . for @xmath53 the dos is given by the dashed line in fig .
[ fig : zr ] and the hole addition states would be of pure o character . when the cu - o hybridization @xmath56 is switched on , the extra holes added to the oxygen band will scatter with the @xmath11 spins and bound states will appear at the bottom of the oxygen band .
this is illustrated by the solid line , which plots the partial @xmath6 dos which was obtained using the maximum entropy method ( mem)@xcite for the analytic continuation of the qmc data to real frequencies .
it can be noticed that the first hole addition states have a strongly mixed @xmath6 and @xmath8 character ( the @xmath6 character in the spectrum is large now ) and an energy pushed well below the edge of the initial non - bonding oxygen band .
therefore only these states are relevant for the low - energy physics up to @xmath57 hole doping . ] . in the zr theory these low - energy states which appear as a consequence of the strong cu - o hybridization are considered to be local singlets which move through the lattice like holes in an afm background .
consequently the claim is that the physics can be described by a one - band t - j model . in order to determine the phase diagram we calculate a large number of susceptibilities which are relevant for spin , charge and superconducting ordering , both at the center and the corner of the bz .
for example , the nel and sc critical temperatures , @xmath58 and respectively @xmath2 in the phase diagram presented in fig .
[ fig : phase ] are determined from the divergence of the corresponding susceptibilities .
the pseudogap crossover temperature @xmath59 is obtained from the maximum in the uniform magnetic susceptibility when accompanied by a suppression of spectral weight in the dos .
similar to what was found in the single - band hubbard model @xcite , we find afm and d - wave sc for both electron and hole doped regimes . however the electron - hole symmetry is broken . in the electron doped case afm
persists to a much larger doping . on the contrary
, sc disappears at a smaller critical doping clusters are considered . ] .
these features of the phase diagram are in qualitative agreement with the experimental findings @xcite .
the one - electron spectral functions as measured with photoemission are also different .
our @xmath60 cluster divides the bz into four cells around @xmath47=@xmath48 and @xmath49 ( see fig .
[ fig : coarse ] ) and approximates the lattice self - energy by a constant @xmath61 within a cell . because of this coarse - graining , a comparison with arpes is not possible apart from gross features .
however , as the phase diagram shows , we believe that even our small cluster captures much of the physics of the cuprates . here
we want to stress the difference between the electron and the hole doped case within our @xmath60 cluster approximation . in fig .
[ fig : dos ] a ) and b ) we show the total @xmath6 states dos and the @xmath6 coarse - grained @xmath47 dependent dos ( which would correspond to the average over all @xmath28 belonging to a coarse - graining cell of the single particle spectra @xmath62 ) for the hole and respectively for the electron doped case , at @xmath63 doping .
the total dos looks qualitatively similar and at the chemical potential we see in both cases a depletion of states which indicates the presence of the pseudogap .
the @xmath47 dependent dos is very different .
the important feature which we want to stress is the location of the pseudogap in the bz . in the hole doped case
the pseudogap appears around @xmath64 . for the electron doped case we do not detect any suppression of states around @xmath64 even though the pseudogap is clearly present in the total dos .
these features are in agreement with the photoemission experiments .
the hole doped materials show fermi pockets around @xmath65 and gapped states around @xmath64 @xcite . for the electron doped materials the photoemission spectra @xcite exhibit a gap near @xmath65 and fermi surface pockets around @xmath64 . with the present cluster size
the dca can not determine where in @xmath28-space the pseudogap is , but it is interesting that it is not at @xmath64 .
the presence of the pseudogap at @xmath65 for the electron doped system can only be checked by increasing the cluster size , and this work is in progress .
dos and coarse - grained @xmath47 dependent @xmath6 dos at @xmath63 doping .
a ) hole doping case .
b ) electron doping case.,width=316,height=240 ] ( upper part ) and charge @xmath66 ( lower part ) susceptibilities versus temperature for different hole densities . @xmath67 in the legend represents the number of holes per unit cell.,width=316,height=240 ] -wave pairing susceptibility @xmath68 versus temperature for different hole densities.,width=316,height=240 ] , the unscreened cu moment , @xmath69 ( eq . [ eq : uscr ] ) versus hole filling .
b ) the screened cu moment , @xmath70 ( eq . [ eq : scr ] ) versus hole filling.,width=316,height=240 ] the electron and the hole doped susceptibility functions are also different both for the divergence temperatures and the temperature and doping dependence . in fig .
[ fig : sus ] we show the uniform spin and charge susceptibilities versus temperature at @xmath63 and @xmath71 doping .
a common feature for all cases is the existence of a characteristic temperature @xmath72 below which the spin response is suppressed and the charge one is enhanced .
@xmath72 corresponds to the pseudogap ( seen in the dos ) onset temperature .
the suppression of the spin excitations below @xmath72 was also seen in nmr @xcite experiments and it was associated with the pseudogap . besides these common features the electron and the hole doped susceptibilities behave differently .
generally , the maximum value of the spin susceptibility increases with hole filling .
this means that in the hole doped case , the spin susceptibility at the pseudogap temperature is strongly increasing with doping unlike in the electron doped case where it decreases upon doping . at the same doping the hole doped spin susceptibility is much larger than the electron doped one .
another interesting feature is the very strong increase of the charge susceptibility for the electron doped case in the underdoped region ( @xmath63 doping ) , suggesting a tendency towards phase separation @xcite .
asymmetric behavior can also be noticed in fig .
[ fig : scsus ] , where we plot the _ inverse _ of the @xmath6-wave paring susceptibility .
above @xmath73 the pairing susceptibility increases with doping in the electron doped case and remains more or less constant in the hole doped case . because of the large cu - o hybridization the system is strongly covalent .
for example in the undoped regime the cu occupation number is only @xmath74 .
the fact that the cuprates are strongly covalent was also observed in nmr measurements @xcite .
we notice that the system exhibits a slightly doping dependent covalency .
this is shown in fig .
[ fig : ndmiu]-a where the cu occupation number versus hole density is plotted . a constant covalency , equal to the one in the undoped regime ( i.e. @xmath75 cu holes and @xmath76 o holes per site )
, would correspond to the dashed line .
it can be noticed that , for the electron doped regime , the cu hole occupation number is decreasing faster than the hole concentration , which indicates an increasing covalency with increasing electron doping .
this happens because at large electron doping , i.e. when the hole - filling of the @xmath3 plane is small , the effective hybridization is a result of a large @xmath77 in the bz ) is strongly @xmath28 dependent , its value taking values from @xmath78 at @xmath49 point to zero at @xmath79 point in the bz . ] . increasing the number of holes , the bz starts to fill up and a smaller @xmath77 will be responsible for the hybridization , and consequently the covalency decreases . for the hole doped regime ,
the extra holes go primarily on the oxygen band , and therefore we do not have a direct measure of the covalency . in fig .
[ fig : ndmiu]-a the unscreened moment on the cu orbitals is shown .
it is defined as : @xmath80 the difference between @xmath81 and @xmath82 is a measure of the double occupancy with holes on cu sites . in the electron doped regime
the double occupancy is very small , but it increases substantially in the hole doped regime , which indicates that the low - energy hole addition states contain double occupied cu configurations in a significant measure . in fig .
[ fig : ndmiu]-b the screened moment on cu , defined as @xmath83 where @xmath84 is cu spin operator at site @xmath85 , is shown . the main effect of the extra holes is to screen the spins on the cu sites .
the screening starts to be effective bellow temperatures of about @xmath13 ( not shown ) . in the zhang - rice scenario
an extra hole perfectly screens one spin on cu forming a strongly bound on - site singlet which would contain a significant amount of the double occupied cu configuration .
so , our results do not contradict the zr theory , but also do not exclude other scenarios where the extra holes form more complicated bound states which involve more than one cu spin .
quantitative analysis based on the amount of screening as function of hole doping can not give an answer to the validity of the zr assumption , because , aside from the screening due to the oxygen holes , there are also non - local processes which contribute to the screening of cu moments ( for example , a possibility is the formation of inter - site spin singlets associated with the resonance valence bond scenario ) .
concluding that the electron - hole asymmetry is an intrinsic property of the @xmath3 plane , we next address the cause of this asymmetry and the possible reduction to a one - band model . in sec .
[ sec:2bm ] we showed that , due to the cu - o hybridization , the addition of holes results in the formation of low - energy states , with an energy well bellow ( @xmath12 ) the initial oxygen band ( see fig . [
fig : zr ] ) .
the reduction to a one - band model is based on the zr claim that these states are singlets , i.e. spinless entities which can be regarded as holes moving in an antiferromagnetic background . due to the monte carlo nature of our calculation , which does not provide a wave function for the ground state
, we can not determine directly the exact nature of these states .
the most we can do is to compare the results of a two - band hubbard model calculation with those of a one - band hubbard model , and based on the similarities and differences that we might find to decide about the validity of the single - band approach . in order to compare the two - band and the one - band models
, we should first get an idea about the possible single - band model effective parameters .
we discuss here two different approaches for calculating these parameters , both based on the assumption that the low - energy states are localized , and close to the zr proposed singlets . [ cols="<,<,<,<,<,<,<,<,<,<",options="header " , ] the cell - perturbation theory @xcite assumes that the zr mapping is _ strictly _ true and therefore the low - energy states are genuine local singlets . here and everywhere in the paper by local we refer to the oxygen _ orthogonal _ wannier states which are different from the non - orthogonal plaquette states around the cu ions . to deduce the one - band model parameters we work in the site representation .
we can fourier transform eq .
[ eq : ham ] and write it as @xmath86 where @xmath87 here @xmath85 represents the site index .
the oxygen operators @xmath88 describe the orthogonal wannier states .
the zr assumption implies that @xmath89 is responsible for the formation of the low - energy states ( local singlets ) , and @xmath90 will determine the hopping parameters .
therefore the cell - perturbation theory provides a means to determine the one - band parameters provided that the zr assumption is correct .
elaborate calculations along this line were done in @xcite for a variety of multi - band parameters . in a first order approximation in @xmath91 ,
the effective u is given by @xmath92 where @xmath93 , @xmath94 and @xmath95 represent the energies of the two ( i.e. the zr singlet ) , one ( i.e. the bonding state ) and respectively zero hole states of eq .
[ eq : ham1 ] .
an important point is that @xmath90 introduce three types of hoppings . if we denote with @xmath96 , @xmath97 and @xmath98 the lowest energy states of @xmath99 corresponding to two , one and respectively zero holes , we have the following hopping integrals @xmath100 @xmath101 ( eq .
[ eq : hopph ] ) describes the hopping of the zr singlet , @xmath102 ( eq . [ eq : hoppe ] ) the hopping of the electron , and @xmath103 produces the exchange interaction @xmath104 the cell - perturbation theory applied to our model gives the parameters shown in the first row of table .
[ tab : zrparam ] .
we want to point out two things .
first , the reduced hamiltonian in the cell - perturbation theory is a t - t-j model , @xmath105 with different hopping parameters for the electron and the hole doped regions and with a value of the exchange interaction not determined by the quasiparticle s hopping ( @xmath101 or @xmath102 ) , but , as it is shown in eq .
[ eq : jj ] by @xmath103 .
therefore , a comparison with a one - band hubbard model , should be done cautiously .
second , we want to stress that the value of the next - nearest - neighbor hopping terms ( @xmath106 and @xmath107 ) is very small compared to the nearest - neighbor terms .
the reason is that the initial oxygen - oxygen hybridization , @xmath31 , results in an effective hopping term comparable in magnitude with the one resulting from the copper - oxygen hybridization , but with a different sign .
this was also remarked in @xcite , and turns out to be an important observation for our final conclusions .
the other approach used for determining the parameters of the one - band model is based on a cluster calculation . in order to estimate the nearest - neighbor hopping , the next - nearest - neighbor hopping and the exchange terms , eskes _
@xciteconsidered two clusters , @xmath108 ( which contains two nearest - neighbor cu ions ) and respectively @xmath109 ( which contains two next - nearest - neighbor cu ions ) .
the exchange term is determined as the energy difference between the singlet and the triplet state of two holes on a cluster . for three holes on a cluster , the two energetically lowest states
can be very well ( @xmath110 ) approximated with the bonding and anti - bonding states of a _ plaquette zr singlet _ hopping between the two cells .
therefore , the differences between these two levels is two times the zr singlet hopping , @xmath101 . in an analogous way ,
considering only one hole on a cluster , the electron hopping , @xmath102 , is determined . using the cluster approach , our two - band model results in the effective parameters shown in the second row of the table [ tab : zrparam ] .
it can be immediately noticed that the two approaches produce different parameters , especially regarding the value of the next - nearest - neighbor hoppings . in the cluster calculation
we obtain significant next - nearest - neighbor hoping terms , @xmath111 and @xmath112 with different sign for the hole and respectively electron doped case ( @xmath113 , @xmath114 ) .
the reason for the discrepancy between the two approaches is that , unlike the cell - perturbation method which considers local singlets , the cluster approach considers singlets between a cu hole and a oxygen state formed on the plaquette around the cu ion .
since the oxygen plaquette states are nonorthogonal it is possible to write them as a linear combination of many orthogonal oxygen states at different sites , i.e. the plaquette singlets are nonlocal states ( in the orthogonal base ) .
it should be pointed out that at first glance this non - locality seems irrelevant ( the overlap of the local oxygen states with the plaquette states @xcite is about @xmath18 ) , but apparently it turns out to influence the value of the next - nearest - neighbor hopping of the reduced hamiltonian considerably .
as pointed out by emery and reiter @xcite , the plaquette singlets are in fact an admixture of zr singlets and zr triplets and this can result in a significant value of the spin - spin correlation on the oxygen sites .
of course , if the zr singlet - zr triplet admixture is significant , this will make a rigorous reduction to a one - band model impossible , because now the low - energy states have also a spin component and therefore can not be mapped into holes ( spinless entities ) .
it is worth pointing out that , in the cluster approach , the large value of the next - nearest - neighbor hopping terms results solely from the finite oxygen dispersion and the lack of hopping between the copper and the next - nearest neighbor oxygen plaquette state . on the other hand , in the cell - perturbation theory a copper - next - nearest neighbor oxygen hopping term is present .
it results in an effective next - nearest - neighbor hopping with a sign different than the one produced via oxygen - oxygen hopping .
we believe that a comparison between the two - band hubbard model and a single - band hubbard model should be done with extreme caution .
we want to stress the possible problems here .
first , the reduction based on the zr approximation , which results in a single - band t - j ( or t - t-j ) model assumes the strong - coupling limit , i.e. a ratio @xmath115 ( the two - dimensional bandwidth is @xmath116 ) . the low - energy density of states of the two - band model shown in fig .
[ fig : zr ] and fig .
[ fig : dos ] indicates a bandwidth of the order of the gap , showing that we are rather at the intermediate - coupling than at strong - coupling .
in the cell - perturbation theory we get @xmath117 , which also suggests intermediate - coupling physics . therefore , the question to be asked is whether the intermediate coupling regime , characterized by an effective repulsion of the same order of magnitude as the bandwidth , can still be well approximated by a second - order perturbation reduced t - j model .
second , considering the previous objection , one may think that a reduction to the single - band hubbard model in intermediate coupling regime , rather than to a t - j model , is more appropriate . however , serious problems arise from the fact that , in the zr theory the nature of the antiferromagnetic correlations is different from that in the single - band hubbard model , i.e. it is not directly related with the quasiparticle ( zr singlet or electron ) hopping . therefore ,
unless both the two - band hubbard model and the one - band hubbard model can be reduced to a t - j model , a comparison between them does not make much sense . nevertheless , we believe that even when the effective repulsion is comparable with the bandwidth the second order perturbation theory which produce the t - j model can be used successfully . we are going to discuss this at the beginning of the next section ( sec . [ ssec : ttu ] ) .
third , the non - locality of the low - energy states ( in the sense discussed in sec .
[ sssec : comp ] ) can have very serious consequences beyond determining the value of the hopping parameters , making the single - band approach to fail completely . , for different values of the ratio @xmath118 of the single - band hubbard model .
, width=316,height=211 ] the t - j model results as a low - energy effective hamiltonian from the hubbard model by projecting out the doubly occupied states .
therefore , the double occupancy of the site orbitals constitutes a measure of the validity of this approximation . in fig .
[ fig : docc ] we plot the double occupancy of the site orbitals for different values of the ratio @xmath118 .
it can be noticed that for @xmath119 the double occupancy is always smaller than @xmath120 beyond this value , the double occupancy increases fast , being about @xmath121 for @xmath122 . ] .
this indicates that , even in the intermediate coupling regime the low - energy physics of the one - band hubbard model can be well described by a t - j model . even if , it is more natural to compare the two - band model with a t - t-j ( or a t - t-j - j ) model , this turns out from our perspective to be rather inconvenient due to the technical difficulties encountered by the qmc when applied to such models .
therefore , we proceed by comparing the two - band model with a t - t-u hubbard model , focusing on the qualitative features rather than on a quantitative comparison . in the strong - coupling limit , the t - t-u model reduces to a t - t-j - j model , with the constraint @xmath123 .
therefore , it is reasonable to assume that if the value of @xmath124 is not too large and the reduction of the two - band model to a single - band model is valid , the two models should exhibit similar physics . assuming that the reduction to a one - band model in the spirit of the zr approximation is possible
, we should expect from table .
[ tab : zrparam ] the hopping parameters to be different in the hole doped and in the electron doped regions .
on the other hand the exchange interaction , @xmath125 should be the same .
therefore , we study the single - band t - t-u hubbard model and address the following questions : * how do the system properties depend on the ratio @xmath126 ? * what is the role of the next - nearest - neighbor hoping t ?
doping , for three different values of @xmath127 , when @xmath128 is constant.,width=316,height=211 ] ( black ) and @xmath129 ( grey ) doping for @xmath130 ( circle ) and @xmath131 ( square).,width=316,height=211 ] at @xmath63 ( circle ) and @xmath71 ( squares ) doping .
inset -b ) inverse of d - wave pairing susceptibility versus temperature at @xmath132 doping , for @xmath130 ( circles ) and @xmath131 ( diamonds ) . ,
width=316,height=211 ] dependent dos at @xmath63 doping .
@xmath133 , @xmath134 , width=316,height=192 ] the values of the parameters in table .
[ tab : zrparam ] show that in general the ratio @xmath135 is larger in the hole - doped regime than in the electron - doped case . in order to address the electron - hole asymmetry observed in the two - band model , in this section we study the properties of the single - band hubbard model as a function of @xmath126 , by keeping @xmath128 ( given by eq .
[ eq : jju ] ) constant and varying the hopping @xmath127 .
the next - nearest - neighbor hopping @xmath0 is set to zero .
with respect to antiferromagnetism , with increasing @xmath127 the nel temperature at small doping and the critical doping where the antiferromagnetism disappear decrease .
for example , at @xmath63 doping , the antiferromagnetic susceptibility is diverging only for the small value of @xmath127 shown in fig .
[ fig : jtafm ] . assuming that the hole doped regime is characterized by a larger value of @xmath126 , this feature is in agreement with the two - band model asymmetric behavior ( see fig . [
fig : phase ] ) .
the uniform spin susceptibility is shown in fig .
[ fig : jtssus ] .
one can notice that an increase of @xmath127 results in an increase of @xmath59 and a decrease of the spin susceptibility at @xmath59 .
this together with the behavior of the susceptibility as a function of doping is in contrast to what was observed in the two - band model ( see fig .
[ fig : sus ] ) where the spin susceptibility is larger in the hole doped case and an increase ( decrease ) with doping of the susceptibility at @xmath59 for the hole ( electron ) doped regimes is found . the behavior of the @xmath6-wave pairing susceptibility as a function of @xmath127 is shown in fig . [ fig : jtdsus ] . the critical temperature increases with increasing @xmath127 ( the increase of @xmath2 is about @xmath71 of the increase of @xmath127 ) , as can be seen in the inset ( a ) .
this increase is much too large to be in agreement with the two - band model results even if , actually , for the two - band model we obtained a hole - doped @xmath2 larger with about @xmath136 than the electron - doped one ) . ] . by extrapolating the inverse of the @xmath6-wave pairing susceptibility at @xmath132 doping ( see inset ( b ) ) it can be concluded that an increase of @xmath127 results in an increase of the critical doping where sc disappears .
we also notice that , at small doping and above @xmath2 , a large @xmath127 suppresses the pairing correlations .
these features are in agreement with the asymmetry of the two - band model phase diagram .
nevertheless , we notice that , above @xmath2 and for both values of @xmath127 , with increasing the doping the pairing correlations increase too .
this behavior is characteristic in the electron doped regime of the two - band model , but can not explain the hole doped regime where the pairing does not depend on the doping ( see fig .
[ fig : scsus ] ) .
the other difference between the two - band and the single - band hubbard model is the value of the sc susceptibility critical exponent @xmath137 , which is much smaller in the two - band model case .
the density of states and the @xmath47 dependent dos for the one - band t - u hubbard model at @xmath63 doping is shown in fig .
[ fig : jtdos ] .
the one - particle spectra exhibit a pseudogap in the total dos and in the @xmath47 dependent dos at @xmath64 point in bz , similar to the hole doped spectra of the two - band hubbard model .
the single - band t - u hubbard hamiltonian is particle - hole symmetric and therefore can not explain the one - particle spectra in the electron - doped regime of the two - band hubbard model . at the end of this section
we conclude the following : a single - band t - u hubbard model ( i.e. @xmath138 ) with a larger value of the hopping parameter for the hole doped regime _ can not _ explain the electron - hole asymmetries observed in the two - band hubbard model , especially the ones which characterize the one - particle spectral functions and the susceptibility functions . .
for the t - t-u hubbard model @xmath139 .
, width=316,height=211 ] ( upper part ) and charge @xmath66 ( lower part ) susceptibilities versus temperature for different hole densities.,width=316,height=211 ] -wave pairing susceptibility @xmath68 versus temperature for different hole densities.,width=316,height=211 ] total dos and coarse - grained @xmath47 dependent dos at @xmath63 doping for @xmath140 . a ) hole doping case .
b ) electron doping case . , width=316,height=211 ] in this section we study the role of the next - nearest - neighbor hopping , @xmath0 , in the single - band hubbard model @xmath141 we choose the following parameters , @xmath142 , @xmath143 and @xmath144 .
these parameters are close to the ones in table .
[ tab : zrparam ] , resulting in @xmath145 and @xmath146 . as for the two - band hubbard model , we work in the hole representation , defined as the one where the filling @xmath147 corresponds to a hole doping @xmath148 .
values of the filling smaller than one correspond to the electron doped regime .
we keep the sign of @xmath0 always positive . in order to avoid confusion
we want to point out that in a t - j model the filling is always smaller than one .
therefore , in order to describe the electron and the hole doped regimes one has to employ the hole and , respectively , the electron representation .
accordingly , the sign of @xmath0 has to be chosen negative in the hole doped regime and positive in the electron doped case and @xmath0 change sign .
a change of @xmath127 sign has no physical consequences aside from a translation with @xmath49 in the bz corresponding to a canonical transformation which changes the sign of site orbitals on one sublattice .
an equivalent statement is that the particle - hole transformation , as defined in ref .
@xcite , change the sign of @xmath0 but not of @xmath127 . ] . in fig .
[ fig : ttuphase ] the phase diagram of the t - t-u model is shown with a solid line .
for comparison , the phase diagram of t - u hubbard model ( i.e. @xmath138 case ) , which is symmetric with respect to hole and electron doping , is shown with dashed line . at half
filling @xmath0 introduces an effective antiferromagnetic exchange @xmath149 between the same sublattice spins and subsequently frustrates the lattice .
however at finite electron doping , @xmath0 favors the antiferromagnetism , making it persist up to a larger doping . on the other hand , in the hole doped case , the antiferromagnetism is always suppressed by @xmath0 . with respect to superconductivity , the presence of @xmath0 results in a smaller ( larger ) critical electron ( hole ) doping at which the superconductivity disappears .
the asymmetry introduced by @xmath0 is in agreement with the one observed in the two - band model phase diagram .
we find that @xmath0 has no major influence on the maximum superconductivity critical temperature @xmath150 .
the uniform spin and charge susceptibilities are shown in fig .
[ fig : ttuspinsus ] .
the spin susceptibility at the pseudogap temperature @xmath72 is strongly increasing with doping for the hole doped case and an opposite effect is seen for the electron doped case .
the downturn at @xmath72 in the spin susceptibility is much sharper for the hole doped regime indicating a fast transition to the pseudogap physics .
all these features are in very good qualitative agreement with the ones corresponding to the two - band hubbard model .
because of the similarity with the two - band model , it is also worth mentioning that in the electron - doped regime the charge susceptibility is strongly increased below @xmath72 in the underdoped region .
the @xmath6-wave paring susceptibilities shown in fig .
[ fig : ttuscsus ] exhibit asymmetric features , also in a qualitative agreement with those in the two - band model . in the electron doped regime , with increasing the doping , the pairing correlations above @xmath2 increase . in the hole - doped regime close to @xmath2
, the pairing correlations do not significantly depend on the doping .
however , contrary to the two - band model behavior , at larger temperature an increase of pairing correlations with doping is observed .
the magnitude of this increase is smaller than in the electron - doped case and a larger value of @xmath0 ( e.g. @xmath151 , not shown ) will reduce it further , improving the resemblance with the two - band model . in fig .
[ fig : ttudos ] we present the dos of the t - t-u hubbard model at @xmath63 doping .
the one - particle spectral functions resemble the corresponding two - band hubbard model ones .
the presence of the @xmath0 parameter is responsible for the location of the pseudogap in the bz .
the necessity of the @xmath0 in explaining the measured arpes line shape and the electron - hole asymmetry was realized early on @xcite .
representing hoppings in the same sublattice , this parameter is not severely renormalized by the afm background , and consequently its influence turns out to be important .
exact diagonalization results @xcite of a t - t-j model are in agreement with ours .
the @xmath0 hopping process lowers the kinetic energy and moves the quasiparticle position from @xmath65 to @xmath64 in the electron doped case .
the nel - like configurations , which do not hinder this process , are stabilized . in the hole doped case
the @xmath0 hopping does not lower the kinetic energy of quasiparticles and it is not energetically favorable , therefore leading to a suppression of afm at all dopings .
the main conclusion of this section is that a one - band t - t-u hubbard model describes qualitatively well the physics ( i.e. the phase diagram , the one - particle spectra and the two - particle response functions ) of the two - band hubbard model , provided a significant value of the next - nearest neighbor hopping ( @xmath152 ) is considered .
however , besides all these similarities there are also some important differences which we are going to emphasize in the next section .
in general , the deduction of an effective low - energy hamiltonian implies two steps .
first , defining the low - energy states and second , projecting the resolvent operator , @xmath153 , on the subspace spanned by these low - energy states @xcite .
the inverse of the projected operator can be written as @xmath154 , where @xmath155 is the low - energy hamiltonian can be considered independent of @xmath156 if the high - energy and the low - energy states are well separated . ] .
this procedure is equivalent to finding an hamiltonian which produces the same one - particle , two - particle , three - particle , etc . , spectral functions on the energy range considered to be `` low - energy '' . rigorously , in order to prove that the one - band model is the effective hamiltonian which describes the two - band hubbard model low - energy physics we should compare not only the one - particle and the two - particle spectra , but also all higher order correlation functions .
however , we believe that the comparison of only the one - particle and the two - particle spectral functions is compelling enough , especially since the experimental information is also obtained by measuring the response functions behavior ( and in almost all cases the two - particle operators or the one - particle ones , as in photoemission , are involved ) .
it is also true that a comparison of the dynamic susceptibilities would be required , but with our quantum monte carlo based algorithm the calculation of these quantities for the two - band model is extremely computational resources consuming and has not been done yet .
however partial information about the relevant excited states is contained in the temperature behavior of the static susceptibilities . the main conclusion of sec .
[ sec:1bm ] is that a t - t-u hubbard model describes qualitatively well the physics of the two - band hubbard model , but only if a substantial next - nearest - neighbor hopping is considered .
however , the calculation in sec .
[ sssec : cell ] ( first row of table .
[ tab : zrparam ] ) and the more rigorous results by feiner _
et al . _
@xcite , show that in a strict zr picture the next - nearest - neighbor hopping is negligible . therefore it is difficult to explain the two - band hubbard model physics assuming the formation of local zr singlets . for hole doped systems ,
a significant value of @xmath0 can be obtained only if the extra holes form nonlocal bound states with the existing cu holes , presumably something close to the plaquette singlets .
of course we have no reasons to discard other states spread over even more oxygen sites , which can result in a magnitude of the hopping parameters different ( probably not too much ) form the one obtained by cluster calculation ( second row of table . [
tab : zrparam ] ) . in the electron doped systems , the doping dependent
covalency shown in fig .
[ fig : ndmiu]-a , clearly indicates that the hybridization of the cu with the o states at different sites is important .
a doping dependent covalency should also imply doping dependent parameters .
the cluster calculation which allows the formation of non - local ( plaquette ) low - energy states , unlike the cell - perturbation ( or strict zr ) approach , provides a value of the hopping parameters which captures qualitatively the physics of the two - band model .
however , we do not believe that finding the exact value of the one - band hubbard model parameters is a relevant or even a well addressed problem , because the non - locality of the low - energy states implies that the two models are not equivalent , as was pointed out by emery _ et al . _ @xcite . aside from the similarities between the two - band and the t - t-u hubbard models discussed in sec.[sssec : tprim ]
we also find some differences .
for example , one important difference can be observed in the @xmath6-wave pairing susceptibility ( fig .
[ fig : ttuscsus ] and fig .
[ fig : scsus ] ) . in the two - band hubbard model
the critical exponent @xmath137 which defines the divergence of the susceptibility at @xmath2 , is much smaller ( around @xmath157 at finite hole doping ) than the one characteristic to the one - band model ( around @xmath158 ) , indicating larger fluctuations from the mean - field value @xmath159 is a measure of fluctuation . ] .
both the cell - perturbation and cluster calculation provide a larger nearest - neighbor hopping @xmath127 for the hole doped region . according to the analysis presented in sec .
[ sssec : tj ] this should result in both larger @xmath59 and @xmath2 .
however the two - band model results does not indicate that this is the case , the respective critical temperatures being not very different in the electron and hole doped regimes . based on our comparison
we can draw the following conclusions .
the one - band hubbard model retains much of the two - band hubbard model physics , but a significant next - nearest - neighbor hopping ( @xmath1 ) should be provided .
if the purpose of the investigation is the study of the basic physics like the sc mechanism , the proximity of afm , sc and pseudogap we believe that a one - band t - t-u hubbard model should be good enough . on the other hand , if the purpose is to describe more subtle features like the ones which may result from the finite value of the spin correlation on oxygen , or if a quantitative material specific calculation is desired , the single - band model approach fails .
obviously also the single - band model should not be used to describe spectral features at energies above @xmath160 such as the optical , electron energy loss and inelastic x - ray scattering results .
in this paper we use the dca to calculate the properties of the two - band hubbard model .
the 2@xmath1612-site cluster phase diagram resembles the generic phase diagram of the cuprates and exhibits electron - hole asymmetry .
we also find asymmetric features for the one - particle spectral functions and for the relevant susceptibility functions .
these characteristics are in qualitative agreement with experimental findings .
we address the validity of the single - band hamiltonian as the effective low - energy model for the cuprates .
we discuss the possible problems which may cause the failure of the reduction from two - band to one - band and also show that , depending on the approximations involved , the value of the one - band hubbard parameters ( especially the next - nearest - neighbor hopping ) can be significantly different .
we use dca to study the role of the different parameters in the single - band t - t-u hubbard model and compare the phase diagram , the one - particle and the two - particle response functions with the corresponding two - band hubbard model ones .
we conclude that the two models exhibit similar low - energy physics provided that a significant next - nearest - neighbor hopping @xmath0 is considered .
the parameter @xmath0 is also the main culprit for the electron - hole asymmetry of the cuprates .
the large value of @xmath0 needed for a qualitative agreement between the two models can not be obtained in a strict zr picture where the extra holes form local singlets with the existing cu holes .
plaquette singlets , which in the oxygen wannier representation are not local , and presumably other spatially extended states can provide a larger value of @xmath0 .
the doping dependent covalency in the electron doped case also indicates that the non - local cu - o hybridization is important .
however , the formation of non - local low - energy states also implies that they are not real singlets and consequently can not be rigorously mapped into holes and therefore the two models are not equivalent .
we also point out some differences between the two models . in the two - band hubbard model the fluctuations in the @xmath6-wave pairing channel above @xmath2 is much stronger . the deduction of the parameters both in cell - perturbation and cluster approach results in a larger nearest - neighbor hopping @xmath127 for the hole doped regime .
however the critical temperatures @xmath59 and @xmath2 in the two - band hubbard model are approximatively the same in both regimes , quite different from what should be expected .
the conclusion is that a single - band hubbard model with a significant value of the next - nearest - neighbor hopping ( @xmath1 ) , captures the basic physics of the two - band hubbard model , including the proximity of antiferromagnetism , superconductivity and pseudogap and explaining the electron - hole asymmetry seen in the phase diagram , one - particle and two - particle spectral functions .
however , the single - band hubbard model is not entirely equivalent to the two - band hubbard one and we believe that it is not suitable for quantitative material specific studies or for describing more subtle features which may results from the non - locality of the low - energy states .
it is also not suitable to describe physics which implies excitations with energy scales larger than @xmath13 .
we thank f.c . zhang and paul kent for useful discussions . the work was supported by nsf grant dmr-0073308 , by cmsn grant doe de - fg02 - 04er46129 and by the netherlands foundation for fundamental research on matter ( fom ) with financial support from the netherlands organization for scientific research ( nwo ) and the spinoza prize program of nwo .
the computation was performed at the pittsburgh supercomputer center , the center for computational sciences at the oak ridge national laboratory and the ohio supercomputer center .
part of this research was performed by tm as a eugene p. wigner fellow and staff member at the oak ridge national laboratory , managed by ut - battelle , llc , for the u.s .
department of energy under contract de - ac05 - 00or22725 .
99 m. h. hettler , a. n. tahvildar - zadeh , m. jarrell , t. pruschke and h.r .
krishnamurthy , phys .
b * 58 * , r7475 ( 1998 ) ; m. h. hettler , m . mukherjee , m. jarrell and h.r .
krishnamurthy , phys .
b * 61*,12739 ( 2000 ) .
c. almasan and m.b .
maple , _ chemistry of high - temperature superconductors _ , 1991 edited by c.m.r .
rao ; e. dagotto , rev .
phys * 66 * 763 ( 1994 ) b.o .
lett . * 74 * , 964 ( 1995 ) n.p .
et al . _ ,
lett . * 88 * , 257001 ( 2002 ) j. zaanen , g.a.sawatzky and j.w.allen , phys.rev.lett . *
55 * , 418 ( 1985 ) j. d. jorgensen , h. b. schuttler , d.g .
hinks , d. w. ii capone , k. zhang , m. b. brodsky , and d. j. scalapino , phys.rev.lett . * 58 * , 1024 ( 1987 ) l. f. mattheiss , phys.rev.lett . * 58 * , 1028 ( 1987 ) j. yu , a. j. freeman and j. h.
xu , phys.rev.lett .
* 58 * , 1035 ( 1987 ) c. m. varma and s. schmitt - rink , solid state com . *
62 * , 681 ( 1987 ) v.j .
emery , phys.rev.lett . * 58 * , 2794 ( 1987 ) ; v.j .
emery and g. reiter , phys . rev . *
b 38 * , 4547 ( 1988 ) f.c .
zhang and t.m .
rice , phys . rev . * b 37 * , r3759 ( 1988 ) ; phys .
rev.*b 41 * , 7243 ( 1990 ) h. eskes , l.h.tjeng and g.a.sawatzky , phys.rev .
* b 41 * , 288 ( 1990 ) e.b .
stechel and d.r .
jennison , phys.rev . *
b 38 * , 4632 ( 1988 ) m.s .
hybertsen , m. schluter and n.e .
christensen , phys.rev .
* b 39 * , 9028 ( 1989 ) j. zaanen and a.m. oles , phys . rev . *
b 37 * , 9423 ( 1988 ) h. eskes and g.a .
sawatzky , phys .
61 * , 1415 ( 1988 ) v.j .
emery and g. reiter , phys .
* b 38 * , r11938 ( 1988 ) ; phys .
rev.*b 41 * , 7247 ( 1990 ) j. h. jefferson , h. eskes and l. f. feiner , phys .
rev.*b 45 * , 7959 ( 1992 ) ; l. f. feiner , j. h. jefferson , and r. raimondi , phys .
rev.*b 53 * , 8751 ( 1996 ) h. eskes , g.a.sawatzky and l. feiner , physica * c 160 * , 424 ( 1989 ) a. k. mcmahan , j.f .
annett , r.m .
martin , phys . rev . *
b 42 * , 6268 ( 1990 ) a. macridin , _ ph.d . thesis _
, university of groningen , http://www.ub.rug.nl/eldoc/dis/science/a.macridin .
g.kotliar , w.krauth and m.rozenberg , rev.mod.phys .
* 68 * , 13 ( 1996 ) m. jarrell , th.maier , c.huscroft , and s.moukouri , phys.rev . * b 64*,195130 ( 2001 ) m. jarrell and j. e.gubernatis , phys.rep . *
269 * , 135 ( 1996 ) m. jarrell , th.maier , m.h.hettler , a.n.tavildarzadeh , europhys.lett . *
56 * , 563 ( 2001 ) w.w .
et al . _ ,
lett . * 62 * , 1193 ( 1989 ) ; m. takigawa _
et al . _ , phys.rev.*b 43 * , 247 ( 1991 ) ; h. alloul , a. mahajan , h.casalta , o. klein , phys .
lett . * 70 * , 1171 ( 1993 ) a. macridin _
et al . _ , in preparation .
m. takigawa , _ et al .
_ physica c * 162 - 164 * , 853 ( 1989 ) ; m. takigawa _ _ , edited by k.bedell , d. coffey , d.e .
meltzer , d. pines , and j.r .
schriffer ( , ) , p. 236
s.moukouri and m.jarrell , phys.rev.lett .
* 87 * , 167010 ( 2001 ) eduardo fradkin , _ _ , ( , 1991 ) , p. 9
- 10 a. nazarenko , k.j.e .
vos , s. haas , e. dagotto , r.j .
gooding , phys . rev . *
b 51 * , r8676 ( 1995 ) ; r. eder , y. ohta , g. a. sawatzky , phys . rev . *
b 55 * , 3414 ( 1997 ) ; p.w .
leung , b.o .
wells , r.j .
gooding , phys . rev . *
b 56 * , 6320 ( 1997 ) ; o. p. sushkov , g. a. sawatzky , r. eder , h. eskes , phys .
* b 56 * , 11769 ( 1997 ) t. tohyama and s. maekawa , phys .
rev.*b 64 * , 212505 ( 2001 ) ; supercond .
technol.*13 * , r17 a.auerbach , _ _ , ( , ) , p. 25
|
we calculate the properties of the two - band hubbard model using the dynamical cluster approximation .
the phase diagram resembles the generic phase diagram of the cuprates , showing a strong asymmetry with respect to electron and hole doped regimes , in agreement with experiment .
asymmetric features are also seen in one - particle spectral functions and in the charge , spin and d - wave pairing susceptibility functions .
we address the possible reduction of the two - band model to a low - energy single - band one , as it was suggested by zhang and rice . comparing the two - band hubbard model properties with the single - band hubbard model ones , we have found similar low - energy physics provided that the next - nearest - neighbor hopping term @xmath0 has a significant value(@xmath1 ) .
the parameter @xmath0 is the main culprit for the electron - hole asymmetry .
however , a significant value of @xmath0 can not be provided in a strict zhang and rice picture where the extra holes added into the system bind to the existing cu holes forming local singlets .
we notice that by considering approximate singlet states , such as plaquette ones , reasonable values of @xmath0 , which capture qualitatively the physics of the two - band model can be obtained .
we conclude that a single - band t - t-u hubbard model captures the basic physics of the cuprates concerning superconductivity , antiferromagnetism , pseudogap and electron - hole asymmetry , but is not suitable for a quantitative analysis or to describe physical properties involving energy scales larger than about 0.5 ev .
|
the design of the irs was driven by the objective of maximizing sensitivity given the 85 cm aperture of the spitzer space telescope @xcite and the then - available detectors .
dividing the optical trains of the irs into four separate spectrographs substantially reduced the complexity and overall cost of the system .
the result is four separate modules , known by their wavelength coverage and resolution as short - low ( sl ) , short - high ( sh ) , long - low ( ll ) , and long - high ( lh ) .
the slit widths are set to @xmath2/85 cm , where @xmath3 is the _
wavelength for the module .
further in the geometric limit the monochromatic slit image covers two pixels .
two si : as detectors , 128@xmath4128 pixels in size , collect the light in the sl and sh modules , while two si : sb detectors of the same number of pixels are used in the ll and lh modules .
in addition to its spectrographs , the irs contains two peak - up imaging fields , which are built into the short - low module , and have bandpasses centered at 16 ( `` blue '' ) and 22 ( `` red '' ) .
a picture of irs is presented in figure 1 , and table 1 gives the parameters of each module .
the two long - slit low - resolution modules were designed for optimum sensitivity to dust features in the local and distant universe and are effectively limited in sensitivity by the zodiacal and/or galactic backgrounds .
the two cross - dispersed high - resolution echelle modules were designed to achieve the highest possible resolution for the given array dimensions , and they were optimized for sensitivity to emission lines .
lcccccc short - low & si : as & 1.8 & sl2 & 3.6@xmath457 & 5.27.7 & 80128 + & & & sl1 & 3.7@xmath457 & 7.414.5 & 64128 + & & & `` blue '' peak - up & 56@xmath480 & 13.318.7 & @xmath53 + & & & `` red '' peak - up & 54@xmath482 & 18.526.0 & @xmath53 + long - low & si : sb & 5.1 & ll2 & 10.5@xmath4168 & 14.021.3 & 80128 + & & & ll1 & 10.7@xmath4168 & 19.538.0 & 64128 + short - high & si : as & 2.3 & 1120 & [email protected] & 9.919.6 & @xmath5600 + long - high & si : sb & 4.5 & 1120 & [email protected] & 18.737.2 & @xmath5600 + in the following sections we describe the instrument , its operation and calibration , and the reduction of its data using the pipeline developed at the spitzer science center ( ssc ) .
the spitzer observer s manual ( som ) provides a more thorough discussion , and it is updated frequently .
the irs contains no moving parts .
the diamond - machined optics are bolted directly to the precision - machined module housings , and all are constructed of aluminum so that the assembled modules retain their focus and alignment from room temperature to their operating temperature near 1.8 k without any further adjustment .
the focal plane assemblies are equipped with custom plates to interface with the module housing and account for the individual dimensions and locations of the arrays .
each module contains two flood - illuminating stimulators to monitor the performance of the arrays .
the low - resolution modules each contain two sub - slits , one for the first - order spectrum and one for the second - order spectrum .
when a source is in the sub - slit for the second order , a short piece of the first order appears on the array ; this `` bonus '' order improves the overlap between the first and second orders .
the high - resolution modules are cross dispersed so that ten orders ( 1120 ) fall on the array .
the modules were extensively tested at operating temperature at ball aerospace @xcite .
light from external sources passed through the dewar window and a series of neutral density filters at liquid n@xmath6 and he temperatures .
an external optical system projected both point and extended images from either a blackbody or a monochromator onto each slit or sub - slit .
these tests verified the resolution , internal focus and alignment as well as the correct centering of the spectrum on the arrays . from the beginning of testing through integration and launch , the
focus and alignment of the modules has remained constant on a scale of a tenth of a pixel .
the monochromator tests allowed us to map the wavelength positions roughly on the arrays in preparation for flight .
uncertainties in the transmission of the neutral - density filter stacks prevented a useful determination of the overall sensitivities prior to launch .
instead , the sensitivities of the modules were predicted by an analytic model of the system .
extensive observations using a prototype of the short - high module at the hale 5 m telescope verified the procedures used in the sensitivity predictions to the 20% level @xcite .
when the irs is operating , all four of its detector arrays are clocked simultaneously , but it is only possible to capture data from one array at a time .
two techniques are used in data collection , the double correlated sampling ( dcs mode ) and raw data collection ( raw mode , or `` sample up the ramp '' ) .
science data are collected in raw mode while peak - up employs dcs . in dcs mode following an initial series of bias boost and reset frames ,
each pixel is sampled , and then after a number of non - destructive spins through the array each pixel is sampled again via a destructive read and the difference between the two samples is stored as an 128@xmath4128 pixel image . in the raw mode after the same initial bias boost , reset frames and first sampling of pixels there are a number of spin frames followed by a non - destructive read .
this pattern is repeated @xmath7 times .
when all pixels are sampled again the result is a 128@xmath4128@xmath8 cube , where @xmath7 is the number of non - destructive reads , and the values are @xmath7=4 , 8 , or 16 .
a final 128@xmath4128 image is created by calculating the signal slope of each pixel of the raw cube .
mapping the relative positions of each field of view and the spacecraft s pointing calibration reference sensor ( pcrs ) was the most critical step of the in - orbit checkout ( ioc ) phase .
the positions must be measured to an accuracy of better than 0@xmath914 radial ( 0@xmath928 for the long - wavelength modules ) to meet the 5% radiometric requirements .
the positions were measured iteratively , starting with ground - based estimates and proceeding successively through ultra - coarse , coarse , and fine focal - plane surveys . combined with the determination of the focus between the telescope and the slits , this process took place over nearly six of the eight weeks of the ioc period .
the estimated uncertainties in the final measured positions of the slits and the peak - up arrays are better than the requirements in all cases , ranging from 0@xmath909 to 0@xmath912 .
our knowledge of the internal focus and optical calibration of the spectral orders were updated in - flight using a combination of photometric standard stars and emission line objects .
the widths of the orders were derived from the zodiacal light at maximum intensity , while order curvatures , tilts , and wavelength solutions were re - derived from spectral maps of emission line stars such as p cygni and planetary nebulae such as ngc 6543 , ngc 7027 , and smp 083 .
figure 2 presents the hydrogen recombination spectrum of the be star @xmath10 cas , which provides a good check of the wavelength calibration for short - high .
flatfielding and spectrophotometric calibration are determined from mapping and staring observations of well known standards , including @xmath11 dra ( hr 6688 , k2 iii ) , hr 7310 ( g9 iii ) and hr 6606 ( g9 iii ) .
calibration of the low - resolution modules also used fainter standards ( e.g. hd 42525 , a0 v ) . @xcite
provide more details about the calibration scheme . @xcite
discuss the stars and synthetic spectra used for in - flight spectrophotometric calibration .
we are also verifying the calibration using spectral templates generated as described by @xcite and observations of additional standard stars . the overlaps in wavelength coverage between the various modules and in orders within each module aid the internal cross - calibration of the irs .
these overlaps were also used to search for leaks in the order - sorting filters by observing `` cold '' and `` hot '' sources in two overlapping orders . in the case of the first - order filter for long - low ( ll1 ) , which had partially delaminated prior to launch
, we used the combination of uranus and neptune as the cold sources and spectrophotometric standard stars as the hot sources . at this time the analysis limits any possible filter leak in ll1 to a maximum of @xmath55% for a rayleigh - jeans spectrum .
further on - going efforts will refine this limit .
the third largest solar proton flare in the past 25 years began on 2003 october 28 , resulting in an integrated proton flux through the arrays of 1.9@xmath410@xmath12 protons @xmath13 , equivalent to the 50 percentile dose expected during the first 2.6 years of the mission .
as a result , the sensitivity of @xmath54% of the pixels in long - high and @xmath51% of the pixels in the other modules have been degraded .
the degree of damage is consistent with the pre - launch damage experienced by non - flight arrays that were exposed to a beam of 40 mev protons at the harvard cyclotron facility .
the irs is more sensitive to damage than the other science instruments on board spitzer because it is operating at lower background conditions where even a small increase in dark current has a significant effect .
the measured responsitivity of the system in flight was on average two times better than the pre - launch model prediction .
the detector noise measured in the unilluminated parts of the arrays are the same as measured pre - launch .
therefore , potentially in the limit of small signals the sensitivity is on average two times better than the model predictions .
most of this increase is due to design margin , an assumed factor to account of errors and non - modeled effects .
the assumed margins were a factor of @xmath14 for the short wavelength modules and a factor of 2 for the long wavelength modules .
updated estimates of sensitivity are available in the current version of the som .
the ssc provides a sensitivity calculator for the irs known as spec - pet that includes detector noise , source shot noise , and background shot noise . at the current time
the sensitivity is limited by systematic effects , such as noise from the flatfield , fringing , and/or extraction edge effects , which are not included in the calculations of spec - pet .
the observed sensitivities are approximately three times worse than would be achieved by the above responsivities and detector noise .
it is anticipated that the realized sensitivities will improve as the system calibration improves .
the irs currently has two modes of operation , the spectral `` staring '' and `` mapping '' modes . in both cases the target
is acquired and then observed in a fixed sequence starting with sl2 ( short - low second - order ) , sl1 , sh , ll2 , ll1 , lh , skipping any unrequested slits . the target acquisition method is selected by the observer .
one may simply rely on the blind telescope pointing to point the slits to the specified position on the sky .
a second option is to use the pcrs @xcite to calculate the centroid of a reference star , use the offset between the star and the science target and move the science target into the slit .
a third and more accurate option is to use the irs peak - up cameras to centroid on a 16 or 22@xmath15 m image of the science target or a reference star before commencing the requested spectroscopic observations ( see 4.2 ) . using the the pcrs or irs peak - up on a reference star requires accurate coordinates for both the offset star and the science target .
the irs staring is the more basic mode of operation , and it results in observing the target at the 1/3 and 2/3 positions along the slit ( its two nod positions ) , each with the integration time specified for that slit .
the mapping mode steps the slit parallel and/or perpendicular to the slit according to the number and the size of the steps specified by the observer for each slit .
the mapping mode does not perform the 1/3 and 2/3 nod observations that are automatically done in the staring mode .
both the staring and mapping modes accept multiple target / position inputs , as long as all of the positions are within a region two - degree radius in the sky .
multiple targets are specified in a `` cluster '' list by either their 1 ) absolute positions , 2 ) right ascension and declination offsets or 3 ) parallel / perpendicular slit offsets with respect to an absolute fiducial position . if using irs in spectral mapping mode with cluster inputs , parallel / perpendicular slit offsets can not be used .
if the cluster specification is selected , all sources / offsets in the cluster are observed with the same slit before proceeding to the next slit . use of the cluster specification , as opposed to repeating the same observation one source at a time , reduces the observatory overhead and can lead to substantial savings in total `` wall clock '' time .
the som describes the irs observing modes and the various methods of target acquisition in more detail . in the sections below , we concentrate on using the irs peak - up cameras to place a target in the desired slit , and describe how an observer can use the peak - up fields for imaging at 16 and 22 .
the irs sl module contains two peak - up imaging fields .
their field of view is @xmath16 , with a scale of 18 per pixel , and their bandpasses are centered at 16 and 22 for the `` blue '' and `` red '' camera respectively ( see table 1 ) .
the irs peak - up mode enables the placement of a source on a spectrographic slit or series of slits more accurately than just using blind pointing of the spacecraft alone .
the telescope blind pointing has a positional accuracy of @xmath17 ( 1@xmath18 rms radial ) .
an on - board algorithm determines the centroid of the brightest source in the specified peak - up field and communicates the offsets required to accurately position the target in the requested slit to the spacecraft . as long as the coordinates of a target are accurate enough to place it on the peak - up imaging field and it is the brightest object in the field
, the irs will accurately offset to the selected slits .
the peak - up images are supplied to the observers and can also be used scientifically if desired .
the allowed ranges of flux densities for blue and red peak - up point sources are f@xmath19=0.8150mjy and f@xmath20=1.4340mjy , respectively .
however , to avoid excessive integration times and a higher probability of failure , we recommend a flux of at least 2mjy for the blue and 5mjy for the red .
the irs peak - up algorithm has been optimized for point sources ( and indeed this mode has been the most extensively tested and verified in - orbit ) , but an `` extended source '' mode is available for sources with diameters between 5@xmath21 and 20@xmath21 ( e.g. comets ) .
the allowed flux density range for extended sources , for either the blue or red mode , is f@xmath22=15340mjysr@xmath23 .
the observer can specify two peak - up accuracies : `` high '' and `` medium '' . in the simplest case , when peaking up on the science target itself , the first move to a slit after peak - up insures placement to within 0@xmath94 ( 1@xmath18 rms radial ) of the slit center for high accuracy and 1@xmath90 for medium accuracy .
the high accuracy value is driven by the 5% radiometric accuracy requirement for the short - high slit ( 4@xmath97 width ) .
the irs section of the som provides details on how these accuracy options apply to multiple slit positioning after peak - up ( e.g. , for targets in irs cluster mode ) . in the event that the science target can not also serve as the peak - up target ( e.g. if its flux is too low or too high ) , the `` offset peak - up '' mode gives the option of peaking - up using a nearby source .
the observer is free to use any suitable source within 30@xmath24 of the science target , but the spitzer planning tool ( spot ) also provides a list of recommended candidates . note that this mode requires accurate coordinates for both the offset and science target .
an irs peak - up returns the image of the peak - up target .
the pcrs extends the irs peak - up capability to optical point sources with visual magnitudes ranging between @xmath25 and @xmath26 .
the only mode for pcrs currently available provides pointing equivalent to the high accuracy mode of the irs .
the pcrs peak - up does not produce or return an image .
the peak - up fields in the short - low module provide a means of obtaining images at 16 and 22 ( 5.4 and 7.5 fwhm respectively ) . the 16 window in particular
is interesting because the other cameras on spitzer do not cover this wavelength region , and because this wavelength region was used by the _ infrared space observatory _
( _ iso _ ) for most of its deep extragalactic surveys .
since these surveys probed the properties of galactic evolution only up to @xmath27 , irs can provide the link between these past results and the new discoveries which will be made by spitzer at @xmath28 .
imaging with the irs peak - up cameras has not been fully supported for the first year of operations , but the ssc will provide the ability to obtain large mosaics in a manner similar to irac @xcite before the end of 2005 .
however , an interim method called cheap ( cornell high - efficiency advanced peak - up ) has been used extensively to obtain mid - infrared images .
cheap is based on an irs short - low staring observation and uses our knowledge of the offsets between the center of the peak - up windows and the short - low slits in spacecraft coordinates , along with the standard two - position nodding along the slit , to place a source target in various positions on the two peak - up windows .
the user may select any combination of exposure times and cycles of the short - low to obtain an image .
the data are processed by the standard ssc pipeline as though they are standard short - low spectroscopic observations .
the software that combines and calibrates the produced images is already available . as an example , taking a cheap image using a standard 60sec short - low staring exposure will produce two images at 16 and 22 with an rms noise of @xmath29 in only @xmath5720 sec of total time .
it is interesting to note that the rms noise of the _ iso _ observations of the hubble deep field ( hdf ) was just @xmath30 at 15 and that the faintest source detected in the hdf at 15 was 50% brighter than the sensitivity in the above example @xcite .
the use of cheap to obtain 16 and 22@xmath15 m images of a sample of high redshift ( @xmath311.5 ) submillimeter galaxies is presented by @xcite and the first 16@xmath15 m imaging of a lyman break galaxy at @xmath32=2.79 is discussed by @xcite .
a number of issues should be considered in the design of spectroscopic observations using irs .
these are addressed in detail in the som , but we briefly mention a few of the more important ones in this section .
although the infrared background is fainter than the ground based background by a factor of 10@xmath33 , it may still be necessary to take an `` off source '' measurement of the background .
the low resolution modules do this automatically in the sense that both sub - slits are exposed at the same time so the spectrum from the `` off '' slit can be subtracted from the `` on '' slit to remove the background .
alternately , the low - resolution slits are long enough to do the background subtraction on the slit itself . however , the high - resolution slits are too short to subtract the background by differencing the nod positions .
if the continuum level is important , off source integrations using the same module are required for the purpose of background subtraction .
spot provides an estimate of the background that is helpful in planning for background subtraction . as a general rule
, the observers should use the low - resolution modules to measure continuum and broadband features , and the high - resolution modules to measure unresolved lines .
the peak - up system works extremely well over its range of parameters .
however , the observer needs to carefully check that the flux of the peak - up candidate falls within the specified limits of the cameras and correctly enter it into spot .
furthermore , as already mentioned in section 4.2 the candidate must be the brightest object within @xmath5120@xmath34 , to ensure that it will be the one selected by the peak - up algorithm .
there have been several peak - up failures due to violations of one or the other of above the requirements .
observing with spitzer is unlike most ground based infrared observing . on the ground
the background signal overwhelms the source by many orders of magnitude .
therefore , the system noise is set by the shot noise from the background . on spitzer , the noise ( n )
is dominated by detector noise for small signals ( s ) , so n@xmath5constant . however , as the signal level increases the shot noise in the signal eventually dominates : n@xmath35s@xmath36 . at the highest signal levels fixed pattern noise ( flat fielding errors , etc . )
are dominant and consequently n@xmath35s . at the present time
it appears that the flat fielding term is 1 to 2% , effectively limiting the maximum s / n ratio to @xmath550 to 100 .
similarly at very low signals the flatfield uncertainty limits our 1@xmath18 detection to @xmath50.3mjy at 16@xmath15 m for low backgrounds ( see * ? ? ? * ) .
the online s /
n estimator includes all but this last term .
as described in detail in the som , irs observations , either in dcs or raw mode , are stored as fits files of what is called a data collection event ( dce ) .
a dce contains all the data obtained by an irs module since the most recent destructive read .
the irs science pipeline at the ssc treats each dce independently .
the processing of dcs data , such as those collected during an irs peak - up , by the pipeline is minimal and all processing is performed on - board the spacecraft .
however , for data obtained in raw mode the pipeline removes basic instrumental signatures and corrects for variations in spectral response within and between spectral orders . a number of these steps , such as corrections / checks for saturation , cosmic rays , dark current subtraction as well as linearization are performed in the cube level prior to fitting a slope to the sampled up the detector charge integration ramp .
others , such as correction for drifts in the dark current , and stray light or crosstalk between the orders are applied once the slope image has been created .
the irs array data supplied in fits files to the observer are organized into four categories : engineering pipeline data , basic calibrated data ( bcd ) , browse - quality data ( bqd ) , and calibration data .
additional files are also available of the masks used in the pipeline processing as well as processing log and quality assurance files .
the engineering pipeline data consist of the raw detector sample images , with some descriptive information in the fits file headers .
the bcd files are the fundamental basis for science analysis , with the primary product being a two - dimensional slope image of each dce in units of electrons sec@xmath23 pixel@xmath23 , accompanied by a header containing the essential programmatic information and the processing , calibration , and pointing history .
additional bcd data files include uncertainty images that accompany each slope image and various intermediate images that do not have all of the pipeline processing corrections applied .
the bcd also include coadded image data for those observations with more than one ramp at a sky location .
it is expected that the bcd files will be used by the irs observers as the standard data input for subsequent publishable science data analysis . as an example , in a standard irs staring observation in which the observer has requested 3 cycles of 30sec integrations on a science target with the sh module will result in 6 dces in raw mode .
the engineering data pipeline will process the data , which will have the form of six data cubes ( 3 integration cycles at each of the two nod positions along the slit ) , each 128@xmath4128 pixels @xmath4 16 detector reads in size . the bcd pipeline will produce six 128x128 bcd slope images as well as two 128@xmath4128 co - added slope images .
the calibration data include measurements of standard calibration objects and can be used with the bcd to further refine the quality of the science data .
the bqd is designed to provide the observer with an early look at spectra extracted from 2-dimensional bcd images of point sources . in general , the bqd spectra are designed to convey the richness of the observed data and in some cases may be of publishable quality . to extract spectra from each individual bcd image the
post - bcd pipeline first applies a search algorithm to trace the spectrum in the two dimensional 128@xmath4128 pixel images .
extraction is performed using an aperture that follows the linear expansion of the point spread function ( psf ) with wavelength .
all spectra are extracted as point sources and no sky background is subtracted before the extraction is made ( see discussion in 4.1 on how this may affect the extracted spectrum ) .
the spectra are then photometrically calibrated ( in janskys ) to remove the effects of truncation of the psf by the slit aperture and high - order systematic residuals from flatfielding .
most pipeline operations are now routine , but the need to characterize the performance of the instrument in - orbit has led to numerous modifications to the algorithms and adjustments to parameters .
most important are ongoing efforts to refine the detection and mitigation of latencies and cosmic ray hits , correct for stray light between the echelle orders , and remove stray light in short - low module originating from the peak - up apertures .
extraction methods and the final spectrophotometric calibration of the spectra are also very sensitive to the pixelation of psf at each wavelength , and as a result both the extraction and calibration will require some fine - tuning as a result of what has been learned from in - flight measurements
. a detailed user s guide to the ssc pipeline and its products is in preparation .
the irs spectroscopy modeling , analysis and reduction tool ( smart ) is an idl software package used to reduce and analyze irs data from the four modules and the two peak - up arrays .
smart is designed to operate on the basic calibration data ( bcd ) delivered by the ssc pipeline and can be run both interactively and in batch mode .
it provides a suite of quick - look routines that enable observers to assess the quality of their data .
smart extracts spectra from either individual or coadded bcds , and these can be weighted by a pixel mask which characterizes the noise in the data .
the extraction routines include column extraction of a point source scaled to the instrumental psf , fixed - aperture extraction for extended sources and a weighted gaussian extraction for point sources .
several options are available to optimize the sky subtraction .
the spectra from the individual modules can be combined to produce a single spectrum from 538 .
spectral analysis , including line , black - body , template and zodiacal light fitting , photometry , defringing and dereddening are also available .
the spectral analysis package is based on the _ iso _ spectral analysis package ( isap , * ? ? ?
smart includes packages developed by the spitzer legacy teams .
the c2d team provided a number of defringing tools while the feps team has contributed idp3 , an image analysis software package originally developed for the nicmos instrument of the hubble space telescope .
@xcite provide more details about the smart package .
it will become publically available in 2004 september .
the articles in this special spitzer issue provide many examples of the quality of the spectra from the irs .
in addition to the spectrum of @xmath10 cas in figure 2 , we show examples of one low - resolution spectrum and one high - resolution spectrum , both of galaxies .
all data presented have been processed and extracted from individual dces using the irs pipeline at the ssc .
figure 3 presents a spectrum of ngc 7714 using short - high and long - high ( 240 seconds integration each ) .
@xcite describe this spectrum in more detail , but one can immediately note the many forbidden lines and dust features present .
figure 4 presents the spectrum of ugc 5101 obtained using the low - resolution modules .
the integration times were 12 seconds in each short - low sub - slit and 28 seconds in each long - low sub - slit , and this was sufficient to show a rich spectrum with gas , dust , and ice features .
see @xcite for more information on this spectrum .
the irs is a major step forward in speed and sensitivity .
it clearly enables the extension of infrared spectroscopy to a very large number of extragalactic sources . the designed spectral resolution is well matched to the expected line widths from these objects .
the irs also allows mid - infrared spectroscopy of objects such as brown dwarfs , individual stars in neighboring galaxies , and a wide variety of other sources previously difficult or impossible to study spectroscopically in the mid - infrared .
this description of the irs gives the potential observer a brief review of its design , its capabilities , and how to use it .
now it is up to the astronomical community to fully exploit what the irs can do .
the design , fabrication , and testing of the irs at ball aerospace corporation was supported by a nasa contract awarded by the jet propulsion laboratory to cornell university ( jpl contract number 960803 ) .
the irs pipeline was developed at the ssc at the california institute of technology under contract to jpl .
|
the infrared spectrograph ( irs ) is one of three science instruments on the spitzer space telescope .
the irs comprises four separate spectrograph modules covering the wavelength range from 5.3 to 38 with spectral resolutions , @xmath0 and @xmath1 , and it was optimized to take full advantage of the very low background in the space environment .
the irs is performing at or better than the pre - launch predictions .
an autonomous target acquisition capability enables the irs to locate the mid - infrared centroid of a source , providing the information so that the spacecraft can accurately offset that centroid to a selected slit .
this feature is particularly useful when taking spectra of sources with poorly known coordinates .
an automated data reduction pipeline has been developed at the spitzer science center .
|
the phenomenon of entanglement has drawn considerable attention since its inception in einstein , podolsky and rosen s ( epr ) thought experiment @xcite .
entanglement describes a system of particles that have one or more highly correlated quantum properties , such as position , momentum , spin , etc .
specifically , two sub - systems that are entangled can not be described by independent wavefunctions .
instead the quantum state of the combined system is described by a single wavefunction .
for example , in the initial experiments on entanglement conducted with optical parametric amplifier ( opa ) @xcite , a linear intensity dependence was observed in the coincident absorption probability .
this has been explained @xcite by considering the absorption of the signal photon together with the absorption of the idler counterpart , as they travel together .
this explanation is based on the correlated nature of two photon state , though the usual quadratic dependence signature for a two - photon absorption process is absent in the observation .
however , this is simple enough and deficient as it neither distinguishes between the classical correlation and entanglement , nor does it say anything about the higher order entanglement .
later , first part of the deficiency was qualitatively addressed in ref .
@xcite . here
, the time asymmetry , which is intrinsic to the two - photon state vector produced by successive decay of a three - level cascade system , is held responsible for the distinction between classical correlation and entanglement .
still the order of entanglement obtainable from a cascade system remains unanswered . in this paper , we have attempted to address this problem by taking recourse to four wave mixing ( fwm ) process in a cascade system .
the fwm process intrinsically acts as a generator of nonclassical states , especially as a generator of entangled photons @xcite and offers a unique scope for analyzing the existence of different orders of intermodal entanglement . in the quantum description of the fwm process
, simultaneous annihilation of two pump photons ( which may have different frequencies ) creates a signal - idler photon pair .
nonclassical properties associated with the fwm process have been studied almost since the inception of quantum optics .
in fact , squeezed light was first experimentally generated using fwm @xcite . in the last few decades , nonclassical properties associated with fwm process
have been studied in various ways ( @xcite and references therein ) .
applications of fwm have also been reported in various contexts ( @xcite and references therein ) .
specifically , applications of fwm have been reported for optical parametric oscillators ( opos ) @xcite , frequency - comb sources @xcite , single photon sources for quantum cryptography @xcite , stimulated generation of superluminal light pulses @xcite , optical filtering @xcite , low noise chip - based frequency converter @xcite , etc .
further , in silicon nanophotonic waveguides , several useful optical phenomena related to telecom - band ( @xmath2 all - optical functions ( such as , wavelength conversion , signal regeneration and tunable optical delay ) have been demonstrated using fwm ( see @xcite and references therein ) .
in addition to these , fwm has recently been used to develop fwm microscopy @xcite , which is found to be very useful for the study of the nonlinear optical response of nanostructures @xcite ; enhancement of fwm ( i.e. , larger value of third order susceptibility @xmath3 in comparison to existing optical materials ) has been observed in plasmonic nanocluster @xcite .
thus , we may comment that fwm is an extremely important process , which acts as a test bed for studying non - classicality of photons @xcite .
this fact and the above mentioned applications have motivated us to investigate a particular aspect of fwm for a cascade system : intermodal entanglement . specifically , in this paper , we investigate the possibilities of generation of lower order and higher order entanglement in fwm process associated with a cascade system because entanglement has been established as one of the most important resource for quantum information processing and quantum communication @xcite . to be precise , with the advent of quantum information theory , several interesting phenomena ( e.g. , quantum teleportation @xcite , dense coding @xcite , etc . )
are reported which do not have any classical analogue and which require entanglement as an essential resource .
consequently , several systems have already been investigated as sources of entanglement ( see @xcite and references therein ) .
however , it is still interesting to find experimentally realizable simple systems that can produce entanglement . in what follows
, we will show that fwm process associated with a cascade system can provide us one such experimentally realizable and relatively simple system .
it would be apt to note that some efforts have already been made to investigate the existence of intermodal entanglement in fwm process , both theoretically and experimentally ( @xcite and references therein ) .
however , to the best of our knowledge , higher order entanglement is not studied in any of the existing works .
although , studies on higher order nonclassicalities @xcite have become relevant in the recent past .
these works showed that there is indeed a dire necessity to introduce a higher order nonclassical criterion to detect weak nonclassicalities in a relatively easy manner . keeping these facts in mind , in the present paper
, we investigate the possibilities of observing lower order and higher order intermodal entanglement in fwm process associated with a cascade system under the framework of sen - mandal perturbative approach @xcite that is known to provide analytic expressions for time evolution of field operators with greater accuracy compared to the traditionally used short - time solution @xcite .
this is well established in earlier works ( @xcite and references therein ) . in what follows ,
we report a perturbative solution ( using the sen - mandal approach ) for the heisenberg s equations of motion for various modes present in the hamiltonian of the fwm process .
the perturbative solution obtained here is subsequently used to investigate the existence of lower order and higher order entanglement using a set of inequalities that can be expressed as moments of annihilation and creation operators . to be precise ,
we have used here duan et al.s criterion @xcite and hillery zubairy s criteria @xcite to investigate the existence of intermodal entanglement .
the investigation has revealed the signatures of the existence of lower order and higher order intermodal ( two - mode ) entanglement for all possible combinations of modes ( i.e. , entanglement is observed between ( i ) pump and idler modes , ( ii ) pump and signal modes , ( iii ) idler and signal modes ) .
not only that the possibility trimodal entanglement is also investigated here , and it is found that the appropriate choice of parameters yields trimodal entanglement involving pump , signal and idler modes .
remaining part of the paper is organized as follows . in section [ sec : the - cascade - level - system ] , we describe the relevance of using the @xmath0 @xmath1 hyperfine manifold as a test bed for entanglement study by exercising fwm .
the model hamiltonian for the fwm process described in section [ sec : the - cascade - level - system ] , is described in section [ sec : the - model - hamiltonian ] .
this takes into consideration all the four modes as weak and thus quantum mechanical and subsequently we report an operator solution of the heisenberg s equations of motion corresponding to the each mode of fwm process .
the solution is obtained using the sen - mandal perturbative approach . in section [ sec
: intermodal - entanglement ] , possibilities of generation of lower order and higher order ( including tri - modal entanglement ) in fwm process are studied using the operator solutions obtained in section [ sec : the - model - hamiltonian ] . finally , the paper is concluded in section [ sec : conclusion ] .
a practical level scheme , where cascade decay can be observed , is @xmath0 @xmath4 route ( cf . fig .
[ fig : scheme - of - fwm ] a ) .
the excitation of atoms to @xmath5 can be done in a two photon process either through an intermediate state , i.e. , ( i ) @xmath6 @xcite or through ( ii ) a virtual level by using photons of @xmath7 wavelength @xcite .
however ( i ) is a more practised option to populate @xmath5 as for ( ii ) the absorption cross - section is much smaller .
option ( i ) is used to describe several important optical processes , e.g. , fwm @xcite , electromagnetically induced transparency ( eit ) @xcite , double resonance optical pumping ( drop ) @xcite , optical switching @xcite , etc .
[ fig : scheme - of - fwm ] a illustrates the level scheme under consideration where both ( i ) and ( ii ) pathways are clearly shown .
the life times ( @xmath8 ) are in order : @xmath9 .
hence @xmath10 acts as a leaky reservoir @xmath11 w.r.t .
, @xmath12 with a leakage rate slow enough to satisfy @xmath13 so faster optical pumping and decay cycles centered on @xmath14 can be averaged over a single @xmath15 this does not violate the steady state condition ; further @xmath11 is decoupled from the main system .
so it bears negligible influence on cascade excitations . on the other hand , @xmath16 ; hence spontaneous decay @xmath4 would replicate hyperfine structure of @xmath5 level ( cf .
[ fig : scheme - of - fwm ] a ) . similarly , fig .
[ fig : scheme - of - fwm ] b illustrates a typical experimental spectrum obtained under pump - probe rabi frequency combination of @xmath17 both lasers are plane polarized and satisfy the @xmath18 connection .
the x - axis of fig .
[ fig : scheme - of - fwm ] b ( ii ) is calibrated with the saturation absorption spectrum of the probe laser ( inset ; fig .
[ fig : scheme - of - fwm ] b ( i ) ) . here
the probe laser is scanned and the pump is stationary .
the spectrum shows prominent signatures of drop @xcite on the doppler broadened background of probe absorption .
this situation is further clarified when the probe is locked to @xmath19 and pump is scanned and blue fluorescence of @xmath20 the decay channel is simultaneously monitored ( see fig .
[ fig : scheme - of - fwm ] b ( iii ) , ( iv ) ) . in cases of fig .
[ fig : scheme - of - fwm ] b ( ii ) , ( iv ) the i , ii , iii represent drop signals of @xmath21 two photon transitions . unlike fig .
[ fig : scheme - of - fwm ] b ( ii ) , the spectrum in ( iv ) has no doppler background . here
the locked probe laser in principle addresses small velocity groups of atoms resonant with the same .
these atoms further reach @xmath22 by the pump laser itself . due to participation of highly selective velocity groups of atoms the resultant doppler background is largely reduced .
though the counter propagating pump - probe combination is favorable for observing strong eit under two photon resonance ( @xmath23 ) condition ( i.e. , @xmath24 and @xmath25 ) ; only trace of eit is observed to be present at the tip of drop profiles
. this may be explained by considering the decay route @xmath26,which forms a pseudo - closed absorption - emission cycle . however , due to velocity selective nature ( considering @xmath27 is the generalized rabi frequency where , @xmath28 is laser detuning ) of optical pumping , it is apparent that there are also other decay routes , which remain active , resulting in a sufficient population of @xmath29 state facilitating drop condition @xcite and the eit is almost obscured . the striking feature of fig .
1b ( iii ) , i.e. the blue fluorescence , is its one - to - one correspondence with the two photon absorption .
this is because blue photons mainly originate from @xmath30 @xcite decay channel . in a simplistic manner
we may think that under relatively higher pump power @xmath31and relatively lower probe power @xmath32,the blue light intensity ( @xmath33 ) bears a correlation of @xmath34 it shows that there exists finite possibility to produce spontaneous emission in a cascade decay , which at least remain intensity correlated to the probe .
this intrinsic capability of cascade emission deserves further attention to explore if any kind of phase correlation is obtainable . to generate phase correlation , the co - propagating pump - probe configuration holds edge over the counter - propagating one .
this is because fwm is realizable under the first kind of alignment where the phase matching condition demands : @xmath35 this indeed offers an unique scope where the phase and intensity correlation can be simultaneously obtained . by definition ,
two beams of light can be made quantum mechanically entangled through correlations of their phase and intensity fluctuations .
hence the cascade system under co - propagating laser action may act as a source of non - classical light .
for a two photon cascade decay@xmath36 the situation may be understood as follows : ( i ) initially the atom occupies highest excited state @xmath22 while the field remains in vacuum @xmath37 ( ii ) the excited atom decays to intermediate state @xmath38 by emitting a photon and makes a final dash to @xmath39 with emission of another photon . to consider the total process
, it is not possible to distinguish between the photon emission sequences like @xmath40 followed by @xmath41 or vice versa .
in such situation , indistinguishablity of photon comes into play . for a cascade system ,
the product state for ( i ) is @xmath42 while for ( ii ) they are @xmath43 and @xmath44 the total state vector becomes a linear superposition of these individual state vectors . at the very beginning
the atom and the field remains entangled .
but at later times the decoherence of the excited level population destroys the entanglement . as a result the final product state appears as : @xmath45 . in this case
@xmath46 which is a two photon state , may be considered to be an entangled state if it is not separable ( i.e. , @xmath47 ) ; where are wave vector ( polarization ) of radiation . for a cascade medium
the degree of entanglement is determined by the ratio of linewidths of upper and intermediate excited states @xcite . since in the cascade emission of fig .
[ fig : scheme - of - fwm ] a @xmath48 ; it is in principle possible to generate entangled light through non - degenerate fwm process .
the quantum mechanical hamiltonian for fwm process shown in fig .
[ fig : scheme - of - fwm ] is @xmath49 where @xmath50 is the interaction constant and the @xmath51 @xmath52 and @xmath53 are annihilation ( creation ) operators for two degenerate pump modes , signal mode and idler mode , respectively . now from the fig .
[ fig : scheme - of - fwm ] @xmath54 , @xmath39 - @xmath22 is populated using two - photon process with the pump mode @xmath54 and the transition from @xmath22 to @xmath55 represents the mode @xmath56 of the hamiltonian ( [ hamil . ] ) .
finally , mode @xmath57 corresponds to the transition from @xmath55 to @xmath58 here we consider all the modes as weak and that requires a completely quantum mechanical treatment . to obtain the time evolution of the annihilation operators of different modes , we first obtain the heisenberg s equations of motion for various field operators as @xmath59 these equations
are coupled , nonlinear differential equations of field operators and are not exactly solvable in closed analytical forms .
consequently , it is required that we follow a perturbative approach . here
, we have used the sen - mandal perturbative technique @xcite , which is already known to be more general than the well - known short - time approximation approach @xcite .
now , following sen - mandal s perturbative technique , we can write assumed solutions ( assumed analytic forms of the the time evolution of annihilation operators of various modes ) as @xmath60 where @xmath61@xmath62 and @xmath63 are time dependent parameters . _
( color online ) level scheme with result for counter propagating wave . * ( a ) * level scheme in @xmath64 transition ( i ) of rubidium atom ( @xmath65 ) , relevant for drop ( eit ) experiments . the pump ( probe ) laser beams are linearly polarized .
dephasing between ground state ( @xmath39 ) hyperfine components ( @xmath66 ) is @xmath67 , governed by the transit time broadening .
spontaneous decay rate from @xmath68 is @xmath69 ( dotted arrows ) . here , @xmath70
@xmath71 @xmath72 are the natural linewidths .
the coherent dephasing rate @xmath73 is @xmath74 the decay route @xmath75 is used to monitor population history at @xmath76 lifetimes of other states are also mentioned below respective level captions .
the direct excitation @xmath77 ( ii ) is also possible through virtual level . *
( b ) * here , the saturation absorption spectrum ( i ) of rb is shown where 1,4,5 correspond to @xmath78 hyperfine components , whereas 2,3 correspond to crossover transitions @xmath79 and @xmath80 .
the drop spectra @xmath81 are presented under ( ii ) probe laser frequency scan ( pump stationary ) with doppler background and ( iv ) pump laser detuning ( probe static ) without doppler background .
the plot in ( iii ) presents the blue fluorescence spectra mimicking the hyperfine separation of @xmath5 state.,title="fig : " ] _ the above mentioned assumed solution is obtained by using the fact that the time evolution of the annihilation operator @xmath82 under hamiltonian @xmath83 can be expressed as @xmath84 and the same ( i.e. , eq . ( [ timeevolution ] ) ) , can be expanded as @xmath85+\frac{\left(it\right)^{2}}{2!}\left[h,\left[h , a\left(0\right)\right]\right]\\ & + & \frac{\left(it\right)^{3}}{3!}\left[h,\left[h,\left[h , a\left(0\right)\right]\right]\right]+\cdots.\end{array}\label{eq : taylorseries}\ ] ] it is to be noted that we have neglected the terms beyond @xmath86 , but have not imposed any restriction on time @xmath87 provided @xmath88 .
specifically , the assumed solution is obtained by keeping all the terms that arise from the infinite series ( [ eq : taylorseries ] ) , provided that the terms are not of higher power ( higher than quadratic ) in @xmath50 .
subsequently , the assumed solution for a specific mode is substituted in the heisenberg s equation of motion for that particular mode which is obtained using the given hamiltonian .
now , the coefficients of the similar terms are compared to obtain a set of coupled ordinary differential equations involving @xmath89 and @xmath90 .
finally , this set of coupled differential equations is solved to obtain the final analytic solution .
this process leads to some additional terms that are not obtained in the conventional short - time solution.these extra terms provide an edge to the sen - mandal method in comparison to the conventional short - time method .
now , following the prescription described above in general and using eqns .
( [ eom ] ) and ( [ assumedsolution ] ) in particular , we can obtain the functional forms of these parameters as @xmath91 @xmath92 @xmath93 where@xmath94 .
thus , we obtain a perturbative solution for the equations of motion corresponding to the hamiltonian of the fwm process .
the correctness of the above solutions can can be checked by the equal time commutation relation ( etcr ) i.e. , by verifying that @xmath95=[b(t),\,b^{\dagger}(t)]=[b(t),\,b^{\dagger}(t)]=1.$ ] the obtained solutions may now be used to investigate the existence of lower order and higher order entanglement involving various modes by using a set of moment - based criteria of entanglement .
the same is done in the following section .
we have already mentioned that entanglement plays very crucial role in quantum information processing . there
exist several inseparability criteria @xcite which may be used to investigate the possibility of the existence of entanglement in the physical systems . many of these inseparability criteria ( e.g. , hillery and zubairy s criteria @xcite , duan et al.s criterion @xcite , etc . )
are moment based ( i.e. , they are expressed in terms moments of the annihilation and creation operators of two or more modes ) , and thus suitable for the present study as we already have closed form analytic expressions for the time evolution of various modes .
interestingly , most of these inseparability criteria are only sufficient and not necessary . thus ,
if inseparability condition is found to satisfy , we know with certainty that the investigated state is entangled , but if the condition is not satisfied we can not conclude anything about the separability . keeping this in mind , we usually investigate the existence of entanglement using two or more inseparability criteria , so that if one criterion fails to detect entanglement for a specific state and specific parameters , the other criterion ( criteria ) may succeed to detect it . in what follows
, we will use two criteria of hillery and zubairy and one criterion of duan et al .
further , we would like to note that these inseparability criteria can be classified as : ( i ) lower order criteria : if an inseparability criterion ( inequality ) contains terms only up to fourth order in the annihilation and/or creation operators of different modes , the criterion is referred to as a lower order criterion .
this is so , as to study the correlations between two modes we need at least fourth order terms .
consequently , all criteria that involve terms up to 4th order in annihilation / creation operators are known as lower order criteria .
( ii ) higher order criteria : if an inseparability criterion ( inequality ) contain terms of order higher than the fourth order in the annihilation and/or creation operators of different modes , the criterion is referred to as a higher order criterion .
we can easily see that by this definition all inseparability criteria for three or more modes ( or involving three or more particles ) must be higher order criteria .
interestingly , one can also construct higher order criteria for the investigation of entanglement in two - mode case @xcite . in what follows
, we will study higher order entanglement from both the perspectives ( i.e. , three - mode cases and two mode higher order cases ) , but to begin with , in the next subsection we investigate the possibility of observing lower order entanglement between two modes of fwm process . to obtain the signature of entanglement in fwm process
, we have used the following inseparability criteria introduced by hillery and zubairy @xcite @xmath100 and @xmath101 where @xmath54 and @xmath56 represent two arbitrary modes . throughout our present paper ,
we refer to these inequalities ( [ hz1critreria ] ) and ( [ hz2criteria ] ) as hz1 and hz2 criteria , respectively . as these two criteria
are only sufficient not necessary , we also use another moment based inseparability criterion which is referred to as duan et al.s criterion @xcite for any two arbitrary mode @xmath54 and @xmath56 , duan et al.s criterion describes the condition of inseparability as @xmath102 where @xmath103 now using eqns .
( [ assumedsolution ] ) , ( [ compositecoherentstate ] ) and hz1 criterion ( [ hz1critreria ] ) , we obtain @xmath104 @xmath105+\left(h_{1}h_{2}^{*}\alpha^{*2}\beta\gamma+{\rm c.c.}\right ) \end{array},\label{bc - hz1}\ ] ] @xmath106 where we have assumed that the initial state is described by ( [ compositecoherentstate ] ) .
the same initial state is used in the entire paper .
now , we plot right hand sides of ( [ ab_hz1])-([ac_hz1 ] ) in fig .
[ fig : lower - order - hz1 and hz2](a)-(c ) with three values of the phase @xmath107 of input pump mode @xmath54 .
precisely , we have considered @xmath108 and plotted right hand sides of ( [ ab_hz1])-([ac_hz1 ] ) using @xmath109 and @xmath110 .
negative regions of the plots clearly illustrate bi - modal entanglement in @xmath111 modes for all values @xmath107 and in @xmath112 modes for @xmath113 .
we could not find any signature of entanglement for @xmath114 mode .
however , we can not conclude anything about the separability / inseparability in those cases where negative regions are not found .
this is so because the hz1 criterion and other similar criteria of inseparability used in this paper are only sufficient and not necessary .
similarly , we may use ( [ assumedsolution ] ) , ( [ compositecoherentstate ] ) and ( [ hz2criteria ] ) to obtain @xmath115 @xmath116- \left(h_{1}h_{2}^{*}\alpha^{*2}\beta\gamma+c.c.\right ) \end{array},\label{bc_hz2}\ ] ] @xmath117 as before , we plot right hand sides of eqns .
( [ ab_hz2])-([ac_hz2 ] ) in figs.[fig : lower - order - hz1 and hz2](d)-(f ) and it is clear from the figures that the intermodal entanglement is observed only in @xmath112 mode for the phase angle @xmath118 no signature of intermodal entanglement is observed in the remaining two cases . as discussed above ,
in these cases we are nonconclusive about the inseparability .
now , we may extend our investigation on inseparability using duan et al.s criterion described in eq .
( [ duanetalcriteria ] ) and obtain the following expressions with the help of the solutions ( [ assumedsolution ] ) , ( [ coeff . of a])-([coeff . of c ] ) described in the previous section : @xmath119 @xmath120 right hand sides of eqns .
( [ ab_duan ] ) - ( [ bc_duan ] ) are clearly positive and thus we may conclude that for the physical system studied here , duan et .
al.s criterion can not identify any signature of entanglement . in order to investigate the existence of bi - modal higher order entanglement in fwm process , we may use the following two criteria introduced by hillery and zubairy @xcite : @xmath121 where @xmath54 and @xmath56 are two arbitrary modes and @xmath122 and @xmath123 are the positive integers . here
, @xmath124 gives the criteria for higher order entanglement .
the negativity of the right hand side would show the signature of the entanglement .
now , using the first higher order criteria of hillery and zubairy for various field modes we obtain @xmath125 , \end{array}\label{hz1abmn}\ ] ] @xmath126 } , } \end{array}}\label{eq : hz1bcmn}\ ] ] and @xmath127 .
\end{array}\label{eq : hz1acmn}\ ] ] ( color online ) plot of higher order entanglement using hz1 criterion using @xmath128 hz , @xmath129 hz , @xmath130 hz , @xmath131 @xmath132 and @xmath133 the solid ( blue ) line , dashed ( red ) line and dash - dotted ( green ) line represent the @xmath134 , and @xmath135 @xmath136 and @xmath137 , respectively . here , in all the plots @xmath134 and @xmath138 and @xmath137 are multiplied by 300 and 20 , respectively .
higher order intermodal entanglement is observed in ( a ) ab mode , ( b ) bc mode for phase angle @xmath113 and for ( c ) ac mode . ]
( color online ) higher order entanglement using hz2 criterion using @xmath128 hz , @xmath129 hz , @xmath130 hz , @xmath131 @xmath132 and @xmath133 the solid ( blue ) , dashed ( red ) , and dash - dotted ( green ) lines represent @xmath134 and @xmath135 @xmath136 and @xmath137 , respectively . here , in figure , @xmath139 and @xmath140 and @xmath137 are multiplied by 300 and 20 , respectively .
higher order intermodal entanglement is observed in ( b ) @xmath112 mode for phase angle @xmath141 and @xmath110 ( not shown ) only ; but no signature of intermodal entanglement is observed in ( a ) @xmath111 and ( c ) @xmath114 modes . ] to illustrate the fact that eqs .
( [ hz1abmn])-([eq : hz1acmn ] ) provide us signatures of higher order entanglement , we have plotted the right hand sides of these equations with the dimensionless interaction time @xmath142 in fig .
[ fig : hz1higher order ] a - c , where negative regions of the curves depict the existence of higher order entanglement .
these figure reveal that the signature of the higher order entanglement is observed in all three possible combinations in which two modes can be chosen . here , in fig.[fig :
hz1higher order ] , there are three lines in each plot , and they represent three cases for each choice of modes i.e. , for @xmath134 and @xmath135@xmath136 and @xmath137 respectively . plot for @xmath143 ( i.e. , blue smooth line ) corresponds to normal order entanglement , whereas the @xmath144 ( red dashed line ) and @xmath145 ( green dot dashed line ) correspond to the higher order entanglement .
it is clear from the figure that the higher order intermodal entanglement is observed for all three modes using hz1 criterion .
further , fig.[fig : hz1higher order ] c illustrates that for @xmath114 modes signature of lower order entanglement is not observed , but that of higher order entanglement is observed . in the similar manner , using the second criteria of the hillery - zubeiry we obtain @xmath146 , \end{array}\ ] ] @xmath147}{\normalcolor , } \end{array}}\ ] ] and @xmath148 \end{array}.\ ] ] right hand sides of the above set of equations are plotted in fig .
[ fig : hz2higherorder ] a - c , which clearly show the signature of the higher order intermodal entanglement for @xmath112 mode for the phase angle @xmath141 and @xmath110
. however , it does not show the same for other choices of modes .
thus , hz2 criterion could not detect the signature of higher order entanglement in @xmath111 and @xmath114 mode .
in other words , hz2 criterion fails to detect the higher order intermodal entanglement for @xmath111 and @xmath114 modes for any phase angle @xmath107 .
there is another way to investigate the higher order entanglement .
any multi - mode entangled state ( which involve more than two modes ) is considered to be higher order entangled . in what follows , we investigate the possibility of observing trimodal entanglement in fwm process using the following criterion @xcite @xmath149 where @xmath150 represents three different modes .
now , we may note that above criterion actually contain three criteria which represent three different bipartite cuts , and we may obtain analytic expressions for @xmath151 for various choices of @xmath152 mode as follows : @xmath153 @xmath154 and @xmath155 if any of the above quantity ( i.e. , @xmath156 ) is found to be negative , we will have a signature of trimodal entanglement . to obtain this signature eqs .
( [ abc_hz1])-([acb_hz1 ] ) are plotted in fig .
[ fig : tri - partaite - entanglement ] a - c , where we observe negative regions for appropriate choices of phase .
this indicates the existence of the trimodal entanglement . to further illustrate the existence of trimodal entanglement involving pump , signal and idler modes
, we further investigate its existence by using a symmetric criterion for inseparability of three modes .
the criterion describes the condition for trimodal entanglement as @xmath157 and using eqs .
( [ assumedsolution ] ) and ( [ compositecoherentstate ] ) , we obtian @xmath158 variation of rhs of this particular equation is plotted in fig .
[ fig : tri - partaite - entanglement ] d , which also illustrate the possible existence of trimodal entanglement ( for specific choice phase ) via its negative region .
traditionally , fwm process is viewed as a third order nonlinear optical phenomenon having applications in various fields as summarized in sec . [ sec : introduction ] . recently , the domain of the generation and applicability of fwm have been considerably amplified .
specifically , several new applications of fwm have been proposed . on the other hand ,
applications of entangled states have been reported in various areas of quantum information .
motivated by these facts , we have rigorously investigated the generation of lower order and higher order intermodal entanglement in fwm process using a set of moment - based criteria ( criteria based on moments of annihilation and creation operators ) and a physical system illustrated in fig.[fig : scheme - of - fwm ] , where fwm happens . to be precise ,
present discussion is focused on a cascade system shown in fig.[fig : scheme - of - fwm ] , where fwm process occurs for @xmath0 @xmath1 hyperfine manifold . considering experimentally achievable parameters , it s observed that the parameters whose negative values indicate the existence of higher / lower order entanglement show an oscillatory nature with variation of rescaled time .
further , it is observed that suitable choice of the phase of the initial coherent states plays a crucial role in the generation of entanglement . here , we observed bimodal entanglement between pump and signal modes and signal and idler modes , and have also observed trimodal entanglement using all these modes .
thus , the output of fwm process studied here appears to be highly entangled .
we conclude the paper with a hope that this highly entangled output of fwm process would find some applications in quantum communication and/or quantum information processing as fwm is a process that can be easily realized experimentally ( cf .
fig.[fig : scheme - of - fwm ] ) and as entanglement is one of the most important resource for quantum communication and computation . *
acknowledgment : * ap thanks department of science and technology ( dst ) , india for the support provided through the project number emr/2015/000393 . .
the authors also thank kishore thapliyal for his interest in the work and some fruitful discussions .
|
possibilities of generation of lower order and higher order intermodal entanglement in @xmath0 @xmath1 hyperfine manifold are rigorously investigated using the sen - mandal perturbative technique by showing the equivalence of the system with the four - wave mixing ( fwm ) process .
the investigation has revealed that for a set of experimentally realizable parameters we can observe lower order and higher order intermodal entanglement between pump and signal modes and signal and idler modes in a fwm process associated with the @xmath0 @xmath1 hyperfine manifold .
in addition , trimodal entanglement involving pump , signal and idler modes is also reported .
|
in both the weak value ( wv ) and null value ( nv ) measurement protocols there are two consecutive measurements . generally , using binary detectors ( `` click '' or `` no click '' )
, we would have four possible outcomes at any given attempt : ( `` click '' , `` click '' ) , ( `` click '' , `` no click '' ) , ( `` no click '' , `` click '' ) , ( `` no click '' , `` no click '' ) , where the notation here is ( first measurement outcome , second measurement outcome ) . taking
the outcome of a first weak measurement conditional to a strong second measurement ( i.e. , how the outcome of the first measurement is correlated with the given outcome of the second measurement ) would be proportional to the ( real or imaginary part of the ) standard weak value ( see e.g. , ref .
@xcite ) .
a partial - collapse measurement , employed in the nv protocol , is a measurement for which a `` click '' destroys the system .
hence , in our scheme ( `` click '' , `` click '' ) is not allowed , and there are only three possible outcomes : ( `` click '' , `` no click '' ) , ( `` no click '' , `` click '' ) , ( `` no click '' , `` no click '' ) .
note that having `` no click '' in the partial - collapse measurement still inflicts back action on the measured system .
for the reasoning above the standard correlation between a `` click '' in the first measurement conditional to a `` click '' in the second measurement is trivially zero .
by contrast , the correlation between a `` click '' in the first measurement conditional to `` no click '' in the second measurement is not trivial , which is the motivation for our _ null _ value scheme .
the experimental setup consists of an exact mapping of this idea .
the crucial point here , though , is to note what would be considered to be `` click '' and `` no click '' in either the first measurement or the second one . to be specific , impinging upon the window ( @xmath81 ) ( cf .
fig.2 ) , a photon can either be deflected to a photon counter @xmath82 ( `` click '' ) , or pass through the window ( `` no click '' ) . a photon that has passed through the window reaches the second polarizer @xmath83 , where the second measurement takes place .
it can either be absorbed by @xmath83 ( `` click '' in the second measurement ) or pass through it to the photon counter @xmath84 .
thus , counting a photon in @xmath84 will correspond to `` no click '' in the second measurement .
we provide here a description of the procedure leading to null values in terms of povms @xcite .
the procedure consists of two steps .
first , the state is measured using a partial - collapse measurement @xmath85 where @xmath86 are the probabilities for the states @xmath87 to `` click '' , respectively .
note that in the main text we have taken @xmath88 . following the partial - collapse step ,
the state is measured with a projective measurement @xmath89 .
consequently , we can represent the sequence of measurements with the following povm elements , @xmath90 where @xmath91 and @xmath92 .
indeed this set forms a povm set @xmath93 .
notably , unambiguous and intermediate state discriminations are performed by tuning such a three - component povm set @xcite .
there are several similar optical implementations of such three - component povm ( see e.g. ref .
@xcite and references therein ) .
the measurement probabilities in the manuscript can be expressed in terms of these povm operators , @xmath94 where we wrote here the joint probabilities @xmath95 and @xmath95 .
additionally , in table [ table1 ] we provide a translation of the povm operators to `` clicks '' at different components in the experimental setup .
.translation of the povm language to `` clicks '' in the experimental setup .
a `` click '' is denoted by @xmath96 and `` no click '' by @xmath97 .
note that while a `` click '' in the photon counters @xmath64 and @xmath65 is easily understood , a `` click '' in @xmath83 corresponds to a photon being absorbed by the polarizer .
hence , a `` no click '' at the postselection stage corresponds to a photon not being absorbed in @xmath83 . [ cols="^,^,^,^",options="header " , ] the nv is composed of a correlation between the different povm components but can not be written directly as a povm .
we study a branch of the quantum state discrimination problem where the observer prepares a device ( a protocol ) that should discriminate whether the provided state is equal to the known state @xmath5 or not , i.e. some other state @xmath6 .
we study two cases : discrimination using a single strong von neumann measurement generates erroneous results with the probability @xmath98\ , , \end{aligned}\ ] ] where @xmath99 is the probability of erroneously declaring the state to be @xmath6 when it was in fact @xmath5 and @xmath100 is the probability of the converse error .
let us choose a gauge such that @xmath101 , the measurement orientation is @xmath102 { \left|{0}\right > } + \sin[\theta_m]{\left|{1}\right>}$ ] , and the alternate state is @xmath103 { \left|{0}\right > } + \sin[\delta_1]e^{i\delta_2}{\left|{1}\right>}$ ] ( see fig . [ bloch1 ] ) .
if a `` click '' is obtained , we declare that the state @xmath6 was measured .
hence , @xmath104\ , , \\ p ( \psi_0 | \psi_\delta)=&\frac{1}{2}\big(1-\cos[2\theta_m]\cos[2\delta_1]\nonumber\\ & -\sin[2\theta_m]\sin[2\delta_1]\cos\left[\delta_2\right]\big)\ , .\end{aligned}\ ] ] a bloch sphere visualization of the chosen gauge for the analysis of the single von neumann measurement case . ]
the probability of error , @xmath105 , depends on the orientation of @xmath6 such that one can not find an optimal measurement orientation . nevertheless , assuming a probability distribution out of which @xmath6 is chosen , enables some optimization in the choice of the measurement direction .
we can assume , for example , that @xmath6 is uniformly distributed on the bloch sphere in some area around @xmath5 .
taking the mean over this distribution results in @xmath106\nonumber\\ & = \frac{1}{2}\left(1-\cos[2\theta_m]\cos^2\left[\delta\right]\right)\,,\end{aligned}\ ] ] where @xmath107 $ ] .
averaging over the entire bloch sphere amounts to taking @xmath108 .
thus , the optimal direction to measure ( minimizing @xmath105 ) would be in the direction orthogonal to @xmath5 , i.e. taking @xmath109 .
this optimal measurement direction is the same as that obtained in the main text using the snr analysis .
our type of intermediate discrimination using a three - component povm generates both inconclusive an erroneous results @xcite .
the probability for inconclusive results is @xmath110\ , .
\end{aligned}\ ] ] the probability of error in this case is @xmath111 where a `` click '' in @xmath112 , @xmath113 is erroneously interpreted as the state being @xmath114 , @xmath115 , respectively . a bloch sphere visualization of the chosen gauge for the analysis of the three - component povm case . ] let us choose a gauge such that @xmath101 , the measurement orientation is @xmath102 { \left|{0}\right > } + \sin[\theta_m]{\left|{1}\right>}$ ] , the alternate state is @xmath103 { \left|{0}\right > } + \sin[\delta_1]e^{i\delta_2}{\left|{1}\right>}$ ] , and the postselection orientation is @xmath116 { \left|{0}\right > } + \sin[\theta_f]e^{i\phi_f}{\left|{1}\right>}$ ] ( see fig .
[ bloch2 ] ) . in this gauge
we obtain the following probabilities : @xmath117-\cos [ 2\theta_f ] \left(1-p \cos [ 2\theta_m ] \cos^2\left[\theta_m\right]-(\sqrt{1-p}-1 ) \sin^2[2\theta_m]\right)\nonumber\\ & + \sin [ 2\theta_f ] \sin [ 2\theta_m ] \cos \left[\phi_f\right ]
\left(p \cos ^2\left[\theta_m\right]+\left(\sqrt{1-p}-1\right ) \cos [ 2\theta_m]\right)\bigg)\ , , \\ { \left<{\psi_\delta}\right|}\hat{\pi}_?{\left|{\psi_\delta}\right>}=&\frac{1}{4}\bigg(\left(1+\cos[2\theta_m]\cos[2\theta_f]+\sin[2\theta_m]\sin[2\theta_f]\cos\left[\phi_f\right]\right)\left(1-\cos[2\theta_m]\cos[2\delta_1]-\sin[2\theta_m]\sin[2\delta_1]\cos\left[\delta_2\right]\right)\nonumber\\ & ( 1-p)\left(1-\cos[2\theta_m]\cos[2\theta_f]-\sin[2\theta_m]\sin[2\theta_f]\cos\left[\phi_f\right]\right)\left(1+\cos[2\theta_m]\cos[2\delta_1]+\sin[2\theta_m]\sin[2\delta_1]\cos\left[\delta_2\right]\right)\nonumber\\ & \frac{1}{2}\sqrt{1-p}\big(-4 \sin[\delta_1 ] \cos^2[\delta_1 ] \sin^2[\theta_f ] \sin[4 \theta_m]-4 \sin [ 2 \delta_1 ] \sin [ 2 \theta_f ] \cos^4[\theta_m ] \cos[\delta_2+\phi_f]\nonumber\\ & -8 \sin^3[\theta_m ] \cos[\theta_m ] \left(\sin[2 \theta_f ] \left(\cos ^2[\delta_1 ] \cos[\phi_f]-\sin^2[\delta_1 ] \cos[2\delta_2-\phi_f]\right)+\sin[2\delta_1 ] \cos[\delta_2 ] \cos^2[\theta_f]\right)\nonumber\\ & + 8 \sin[\theta_m ] \cos^3[\theta_m ] \left(\sin[2 \theta_f ] \left(\cos^2[\delta_1 ] \cos[\phi_f]-\sin^2[\delta_1 ] \cos[2\delta_2+\phi_f]\right ) + \sin[2\delta_1 ] \cos[\delta_2 ] \cos^2[\theta_f]\right)\nonumber\\ & + \sin[2 \delta_1 ] \sin[2 \theta_f ] \left(4 \sin^2[\theta_m ] \cos[2 \theta_m ] \cos[\delta_1-\phi_f]+ \sin^2[2 \theta_m ] \cos[\delta_2+\phi_f]\right)\nonumber\\ & -2 \cos[2 \theta_f ] \sin^2[2 \theta_m ] \left(-2 \sin^2[\delta_1 ] \cos[2 \delta_2]+\cos[2 \delta_1]+1\right)\big)\bigg)\ , , \\ { \left<{\psi_0}\right|}\hat{\pi}_2{\left|{\psi_0}\right>}=&\frac{1}{2 } \bigg(1-p \cos^2\left[\theta_m\right]+\cos [ 2\theta_f ] \left(1-p \cos [ 2\theta_m ] \cos^2\left[\theta_m\right]-(\sqrt{1-p}-1 ) \sin^2[2\theta_m]\right)\nonumber\\ &
-\sin [ 2\theta_f ] \sin [ 2\theta_m ] \cos \left[\phi_f\right ] \left(p \cos ^2\left[\theta_m\right]+\left(\sqrt{1-p}-1\right ) \cos [ 2\theta_m]\right)\bigg)\ , , \\ { \left<{\psi_\delta}\right|}\hat{\pi}_1{\left|{\psi_\delta}\right>}=&\frac{p}{2}\left(1+\cos[2\theta_m]\cos[2\delta_1]+\sin[2\theta_m]\sin[2\delta_1]\cos\left[\delta_2\right]\right)\ , .\end{aligned}\ ] ] similar to the minimum error case , we can assume a probability distribution out of which @xmath6 is chosen in order to deduce some optimization in the choice of the measurement directions .
we can assume , for example , that @xmath6 is uniformly distributed on the sphere in some area around @xmath5 .
taking the mean over this distribution results in @xmath118)}\bigg\{4 - 2 p+\sin[2\theta_f ] \sin[2\theta_m ] \cos[\phi_f ] \left(2 p+(p-2 ) \sin^2[2 \delta ] \cos[2\theta_m]\right)\nonumber\\ & + \cos[2\theta_m ] \left(\sin^2[2 \delta ] ( ( p-2 ) \cos[2 \theta_f ] \cos[2 \theta_m]-p)+2 p ( 1-\cos [ 2 \delta ] ) \cos[2 \theta_f]\right)\nonumber\\ & + 2 \cos[2 \delta ] ( p-2-p \sin[2 \theta_f ] \sin[2 \theta_m]\cos[\phi_f])+\sqrt{1-p } \big(\frac{4}{\pi } \sin[2 \theta_f ] \cos[2 \theta_m ] \sin[\phi_f ] ( 2 \delta -\sin[2 \delta ] \cos[2 \delta])\nonumber\\ & + \sin ^2[\delta ] ( 3+\cos[2 \delta ] ) \left(\sin[2\theta_f ] \sin[4\theta_m ]
\cos[\phi_f]-2 \cos[2 \theta_f ] \sin^2[2 \theta_m]\right)\big)\bigg\}\ , , \\
\overline{{\left<{\psi_\delta}\right|}\hat{\pi}_1{\left|{\psi_\delta}\right>}}=&\frac{p}{2}\left(1+\cos[2\theta_m]\cos^2\left[\delta\right]\right)\ , , \end{aligned}\ ] ] where @xmath119 corresponds to the area of the probability distribution out of which @xmath6 is selected [ cf .
eq . ] .
keeping in mind the experiment discussed in the main text , we consider the specific case where both measurements are lying in the same plane ( @xmath120 ) .
we can see in fig .
[ plots ] that the optimal measurement orientations correspond to @xmath121 , i.e. an application of @xmath113 in a direction orthogonal to @xmath5 ( namely , @xmath122 ) .
the optimal orientation of @xmath112 depends on @xmath80 and @xmath119 .
importantly , for discrimination between nearby states ( requiring @xmath123 ) , the optimal orientation of @xmath112 is nearly orthogonal to @xmath5 , as in the case of a single von neumann measurement .
when the unknown state can lie anywhere on the bloch sphere ( @xmath124 ) the optimal orientation of @xmath112 is in the direction of @xmath5 itself .
in the main text , we introduce the snr analysis of both the standard strong measurement approach and the null values one . here , we provide more details of the calculations leading to eqs .
( 3 ) and ( 6 ) in the main text .
we also consider , here , general states on the bloch sphere , i.e. , states which are not confined to the great circle of polar angles .
the statistical measure for a multi - copy discrimination is defined by eqs .
( 1)-(3 ) in the main text .
here , we detail the derivation of the snr in eq .
( 3 ) , i.e. , the snr achieved through individual standard strong measurement , @xmath7 , performed on @xmath4 copies of a qubit . in this benchmark case
, the occupation of the state @xmath102 { \left|{0}\right > } + \sin[\theta_m]e^{i\phi_m}{\left|{1}\right>}$ ] is measured .
the probabilities to detect a qubit state in @xmath8 in any single attempt are @xmath125 \cos [ 2\theta_m ] + \sin [ 2\delta_1 ] \sin [ 2\theta_m ] \cos \left[\phi_m-\delta_2\right]\right)/2 $ ] , @xmath126 $ ] for the states @xmath6 , @xmath5 , respectively . the uncertainty in the signal [ cf .
( 2 ) in the main text ] is defined by @xmath127 where for the second equality we assumed poissonian noise ( which is dominant for coherent light experiments discussed below ) , i.e. @xmath18 and @xmath19 .
note that the obtained snr [ cf .
( 3 ) in the main text ] is in fact a student - t test @xcite @xmath128|\delta_1 \sqrt{n } > z_{1-\eta}\ , , \label{snr_simpsup}\ ] ] where the approximation is for @xmath22 , and @xmath129 is the critical value of the standard normal distribution function , @xmath130\equiv 1-\eta$ ] , i.e. eq .
describes the required snr for which the signal ( numerator ) measured in units of the standard deviation ( denominator ) , can be discerned with a given fidelity @xcite .
indeed , in this approach the maximal snr is obtained when the measurement orientation , @xmath8 , is orthogonal to @xmath5 .
this corresponds to the optimal measurement orientation obtained by the single - copy analysis ( see section [ single_copy ] ) . in the main text
, we wrote the statistical test for a multi - copy discrimination using null values [ cf .
( 4)-(7 ) in the main text ] . here
, we provide more details on how to obtain this result and how to calculate the nvs . in this case
fig . 1 in the main text )
, the qubit state is measured twice .
the first measurement , @xmath29 , is a partial - collapse measurement on the states @xmath8 with probability @xmath80 .
subsequently the qubit state is ( strongly ) measured a second time ( postselected ) , @xmath7 , to be in the state @xmath34 ( click ) or @xmath35 ( no click ) , where @xmath116 { \left|{0}\right > } + \sin[\theta_f]e^{i\phi_f}{\left|{1}\right>}$ ] .
we proposes to discriminate between the two possible initial qubit states via repeating the protocol for @xmath5 and @xmath6 and comparing the respective conditional outcomes of @xmath37 and @xmath38 , i.e. [ having a click the first time conditional to _ not _ having a click the second time ] .
in other words , we define our signal to be @xmath131 .
note that this procedure can also be written as a correlation between outcomes of povms ( see section [ povm ] ) .
our protocol takes advantage of the correlation between the two measurements . to shed some light on its outcome
we calculate explicitly the conditional probabilities following the measurement procedure sketched in fig . 1 in the main text .
for example , if the first measurement results in a `` click '' the system s state is destroyed and any subsequent measurement on the system results in a null - result , implying @xmath132 , and @xmath133 .
this represents a classical correlation between two measurements .
by contrast , @xmath40 embeds non - trivial quantum correlations .
the first partial - collapse measurement of a given preselected state @xmath6 results in the detector clicking with probability @xmath134 .
if no click occurs [ with probability @xmath135 , the qubit s state is modified by the measurement back action into @xmath136}{\sqrt{p(\bar{m}_{w,\delta})}}\,.$}\end{aligned}\ ] ] a second strong measurement , @xmath7 , yields a click [ no click ] with probability @xmath137 @xmath138 $ ] .
finally , using bayes theorem , we can write @xmath139}\nonumber\\ & = n_{w,\delta}/ ( n_{w,\delta}+n_{p,\delta})\,,\end{aligned}\ ] ] where the last equality is obtained by taking the measured estimator for the conditional probability [ theoretically @xmath140 is the number of clicks in the first measurement and @xmath141 is the number of no - clicks in the ( second ) postselection ] .
note that if the detector clicks in the first measurement , the protocol is truncated , and no second step is to be carried out .
this finally leads to the signal reported in eq .
( 5 ) in the main text , @xmath142 in complete analogy with the case of a single strong measurement , we define the uncertainty in the signal @xmath143 where @xmath144 ^ 2 n_w\,\\ \left(\frac{\partial s_{\text{nv}}}{\partial n_p}\right)^2\delta n_p^2=&\frac{n_w^2 n_p}{(n_w+n_p)^4}\,.\end{aligned}\ ] ] the obtained snr [ cf . eq . ( 6 ) in the main text ] is in fact also a student - t test @xmath145 for discriminating between the two states . in the main text
we show that it is preferable to use a tuned nv discrimination than a standard one .
the experimental setup is shown in fig . 2 .
a helium - neon laser at 633 nm is spatially filtered with a single mode fiber and attenuated to the picowatt level .
the resulting photons are passed through a fixed polarizer and then through a movable half - wave plate and polarizer ; this ensures a pure polarization state ( @xmath6 ) and relatively constant flux as the angle of the polarization state is varied .
the power at this stage is estimated by counting photons with a single photon detector when the removable mirror is in place .
we note that the coupling efficiency for each single photon detector is measured ; for the snr theory calculations in the text , we include this efficiency , losses from colored glass filters to block background , as well as the quantum efficiency of the detectors .
the removable mirror is then removed and the photons are sent to a glass window ( w ) that weakly reflects vertically polarized light . the reflection and transmission probabilities for 633 nm light were calculated ( and verified experimentally ) using the fresnel equations .
the reflected photons are counted ; photons that pass through the window are projected onto a linear polarization state ( @xmath35 ) with a second polarizer ( p2 ) and also counted . as stated previously
, each port has its own coupling efficiency and related losses .
photons which are reflected from the front face of the window are collected and counted as @xmath146 .
however , photons are also lost along the way . for example , they can be reflected from the back face of the window ( with @xmath147 ) , causing an additional backaction on the light which passes through the window .
we thus change the definition of our conditional @xmath148 to include the probability of being measured at the first measurement step [ being reflected from the window s front face ] , conditional on [ not being reflected off the back face ] _ and _ [ not being being absorbed by p2 ] , i.e. @xmath149 , where @xmath150 denotes the event of not being reflected off the back face .
notably , the measured estimator @xmath151 , in fact , takes into account all such sources of loss . from the recorded arrival times of the photons in each mode , we can separate photon detection events into time bins ; this allows for a determination of not only the average number of photons @xmath146 and @xmath152 but also their variances as we vary the input ( @xmath6 ) and post - selection ( @xmath35 ) states . to compare these values to theory , we plot the expected snr , incorporating the efficiencies of the experiment . for the data included here , we count photons for approximately 5s , with time bins of 150 @xmath153s or 25 ms for the null - wv or standard schemes , respectively . for technical reasons ,
the photon flux for the standard scheme was higher than for the null - wv scheme ; however , the time bins are chosen to ensure that each method uses an equal number of prepared photons per _
measurement_. we subtract dark counts , which constitute much less than 1% of the total counts on average .
the deviation of the data from the theory plot is due to a small degree of ellipticity induced by the glass window possibly arising from an unknown optical coating or stress - induced birefringence .
the linear polarization model giving rise to the theory plot of fig . 3 does not include complexity of this type .
inclusion of a phase shift is straightforward and leads to a nearly perfect fit between theory and experiment @xcite .
consequently , we expect that elimination of this ellipticity will result in an increased snr as suggested by the theory plot in fig . 3 .
we also note that the experimental data in fig .
3 has not been corrected for detector saturation effects .
implementing this correction is also straightforward and contributes to the nearly perfect fit mentioned above .
|
we present a measurement protocol for discriminating between two different quantum states of a qubit with high fidelity . the protocol , called _ null value _ , is comprised of a projective measurement performed on the system with a small probability ( also known as partial - collapse ) , followed by a tuned postselection .
we report on an optical experimental implementation of the scheme .
we show that our protocol leads to an amplified signal - to - noise ratio ( as compared with a straightforward strong measurement ) when discerning between the two quantum states .
the notion of `` measurement '' was part of the early framework of quantum mechanics . since early developments ,
the discord between information acquisition on the system and the corresponding disturbance of the system s state became clear .
the contest of obtaining information while keeping minimal disturbance is still an active and vibrant field of study that has branched off into many subtopics . of note and of great practical interest in quantum information processing is the study of quantum state discrimination @xcite .
the ability to optimally discriminate between nonorthogonal quantum states depends on the fidelity of the measurement apparatus and on the amount of prior knowledge one has on the states between which he wants to distinguish .
here we introduce a novel procedure to enhance the discrimination fidelity between two quantum states .
our procedure introduces the notion of quantum measurements with postselection in the field of quantum information processing .
our two - step measurement protocol is related , but differs from , the celebrated weak value ( wv ) measurement protocol , where postselected quantum measurements were first introduced @xcite .
we choose to demonstrate our new approach by focusing on a specific discrimination problem . in conjunction with our theoretical analysis
, we report on experimental results involving classical light , which demonstrate the practicality of our measurement protocol , denoted `` null value '' ( nv ) measurement protocol . in the original works on quantum state discrimination
, the observer is handed a single copy of the state to be discriminated , which may be either one of the _ a priori _ known pure states @xmath0 and @xmath1 .
well adapted to this task is the approach known as _ minimum error state discrimination _
@xcite , for which it was shown that the minimum error is obtained by optimizing the axis of a standard two - outcome measurement .
a second approach is the _ unambiguous state discrimination _
@xcite where the measurement produces either an error - free or an inconclusive result ; i.e. , the measurement apparatus is oriented such that it has three outcomes the state is either @xmath2 , @xmath3 , or unknown .
developments on the original works led to many variants of state discrimination , such as discrimination between two _ a priori _ unknown pure states @xcite , as well as discrimination between mixed states @xcite .
further works ( see e.g. , refs .
@xcite ) introduced also the notion of multicopy state discrimination ( employing a number of copies of the state to be discriminated ) . for such schemes , the notion of individual vs. collective measurements was introduced depending on whether the strategy consists of individual measurements each of which performed separately on a single copy , or a single measurement which is performed on all the copies as a whole .
notably , however , in most standard measurement procedures one performs individual measurements on @xmath4 single copies .
thus , it is necessary to define statistical tests to quantify the fidelity of the discrimination @xcite . in the present scheme
, we study a specific variant of the quantum state discrimination problem : the observer prepares a device ( a protocol ) that should discriminate whether the provided state is equal to the known state @xmath5 or not , i.e. is some other nearby state @xmath6 .
noting the context of earlier works on state discrimination , our variant applies to both single - copy and many - copy analyses @xcite . in the former ,
due to the _ a priori _ unknown orientation of @xmath6 , a minimum error state discrimination is underconstrained . additionally , an unambiguous state discrimination is impossible as the unknown state would generate both erroneous and inconclusive results .
such a case , dubbed `` intermediate discrimination scheme '' has been treated for discrimination between two different states ( see e.g. , refs .
@xcite ) .
we present our analysis henceforth for discrimination between two - level states ( qubits ) . assuming that the probability distribution of @xmath6 is uniform on the bloch sphere in some area around @xmath5 , we have analyzed the single - copy minimum error and the single - copy intermediate schemes vis - a - vis our discrimination problem @xcite . for the former ,
we obtain that , regardless of the area of the distribution , minimum error is obtained for a standard measurement in the direction orthogonal to @xmath5 . for the latter , we recall that a three - outcome measurement on a qubit can be realized by two - consecutive measurements @xcite .
their optimal orientations depend on the area of the distribution . for discrimination between nearby states ( @xmath5 or not @xmath5 ) , the optimal orientations of both measurements
are nearly orthogonal to @xmath5 .
by contrast , when the probability distribution of @xmath6 covers the entire bloch sphere the first of the two measurements is oriented in the direction of @xmath5 itself .
alas , a single - copy approach is unfit for most experimental situations due to measurement device imperfections and noise .
one then resorts to a multicopy approach .
here one considers a statistical test ( signal - to - noise ratio , snr ) that , given @xmath4 replicas of the state , would result in a discrimination outcome ( @xmath5 or not @xmath5 ) with some given fidelity . below
we define such snrs and employ them to compare a multicopy version of both the minimum error scheme and the intermediate scheme , focusing on a discrimination between nearby states .
for the intermediate scheme we consider a correlated signal , dubbed _ null value signal _ , for a reason that will be made clear below .
the snr obtained by a nv signal proves to be higher than that obtained by single von neumann measurements .
we further show that the analysis in terms of optimal snr fully agrees with a minimization of error probability in the single - copy cases @xcite .
let us begin with analyzing the snr of the discrimination , achieved through individual standard strong measurement , @xmath7 , on @xmath4 copies of a qubit . in this benchmark case ,
the occupation of the state @xmath8 , defined by polar angle @xmath9 , is measured @xcite .
the probabilities to detect the qubit states @xmath10 and @xmath11 with polar angles 0 and @xmath12 in @xmath8 in any single attempt are @xmath13 and @xmath14 , respectively .
we define a statistical measure to be the difference between the number of positive detections @xmath15 where the right - hand side is the measured estimator .
the signal is a function of two variables @xmath16 . the uncertainty in the signal is then given by @xmath17 we assume poissonian noise ( which is dominant for coherent light experiments discussed below ) , i.e. , @xmath18 and @xmath19 .
thus , @xmath20 , and the obtained snr is @xmath21|\delta \sqrt{n } \
, , \label{snr_simp}\ ] ] where the approximation is for @xmath22 .
indeed , in this approach the maximal snr is obtained when the measurement orientation , @xmath8 , is orthogonal to @xmath5 .
this corresponds to the optimal measurement orientation obtained by the single - copy analysis . a tree diagram of the qubit state evolution under subsequent partial - collapse measurements ;
the respective probabilities are indicated : @xmath23 [ @xmath24 is the probability that the detector `` clicks '' [ no `` click '' ] upon the first measurement . if it does `` click '' , the system is destroyed , hence there are no clicks upon further measurements [ this is marked by a ( red ) @xmath25 .
note that following @xmath26 ( null detection of the qubit ) , the back action rotates @xmath27 into @xmath28 . ]
turning to the multicopy intermediate discrimination , we define a snr by constructing a correlated outcome out of the three - outcome measurement .
recall that such a measurement is implemented by measuring the qubit state twice ( cf .
fig .
[ fig:1 ] ) @xcite .
the first measurement @xmath29 is a strong ( projective ) measurement which is performed on the system with small probability . here
the basis states @xmath30 are measured with probabilities @xmath31 , respectively . for simplicity ,
hereafter , we assume that only the state @xmath8 is measured with probability @xmath32 and @xmath33 .
if the detector `` clicks '' ( the measurement outcome is positive ) , the qubit state is destroyed . very importantly ,
having a `` null outcome '' ( no click ) still results in a back action on the system . we refer to this stage of the measurement process as `` partial - collapse '' @xcite .
subsequently the qubit state is ( strongly ) measured a second time ( postselected ) , @xmath7 , to be in the state @xmath34 ( click ) or @xmath35 ( no click ) , where @xmath34 has a polar angle of @xmath36 .
we propose to discriminate between the two possible initial qubit states by individual application of this measurement protocol on @xmath4 copies of @xmath5 and @xmath6 .
motivated by wvs , the compared observables are the countercausal conditional outcome of [ having a click the first time conditional to _ not _ having a click the second time ] , denoted by @xmath37 and @xmath38 , respectively .
events in which the qubit is measured strongly ( in the second measurement ) , @xmath7 , are discarded . in other words , we define our signal to be @xmath39 note that this procedure can also be written as a statistical correlation between outcomes of a positive - operator valued measure ( povm ) @xcite .
our protocol takes advantage of the statistical correlations between the partial - collapse and strong measurements . to shed some light on its outcome
we calculate explicitly the conditional probabilities following the measurement procedure sketched in fig .
[ fig:1 ] . for example , if the first measurement results in a `` click '' the system s state is destroyed and any subsequent measurement on the system results in a null result .
this represents a classical correlation between the two measurements .
by contrast , @xmath40 embeds nontrivial quantum correlations @xcite .
using bayes theorem , we can write @xmath41 , where we used the measured estimator for the conditional probability , namely , we denoted @xmath42 as the number of clicks in the ( first ) partial - collapse measurement and @xmath43 as the number of no - clicks in the ( second ) postselection @xcite . this finally leads to the measured signal @xmath44 in complete analogy with the case of a single strong measurement , the signal is now a function of four variables @xmath45 , and we can define the uncertainty , @xmath46 , in the statistical test [ cf .
eq . ] @xcite .
we focus on obtaining a large @xmath47 for discriminating between the two states .
it depends on the choice of the measurement orientations , @xmath8 and @xmath34 .
we propose to perform a first measurement that will have a back action on both states @xmath5 and @xmath6 but is nearby the optimal orientation of the single measurement case , i.e. taking @xmath48 .
we propose two possible measurement schemes for obtaining a large @xmath49 . in the first scheme we choose the postselection such that the reference state satisfies @xmath50 .
this means that the reference state @xmath5 would have always clicked in the second measurement had it not been first measured by the partial - collapse .
we call this _ scheme * a*_. alternatively , in _ scheme * b * _ , we choose the postselection such that @xmath51 : the null outcome rotates the reference state and it always clicks in the second measurement . for both schemes , we obtain @xmath52}{\sin[\delta_m+\delta]\sqrt{p}}\sqrt{n}\,,\ ] ] which becomes large for @xmath53 ( weak partial - collapse ) @xcite .
this is because the condition @xmath54 is satisfied vis - a - vis the nv signal of the reference state . varying @xmath12 such that @xmath55 decreases corresponds to a decrease of @xmath56 and an increase of @xmath57 .
a large @xmath49 is obtained when @xmath38 crosses to a regime where @xmath58 .
this happens first with scheme * _ a_*. hence , scheme * _ a _ * produces a larger @xmath49 for smaller @xmath12 ; scheme * _ b _ * leads to far larger @xmath49 for larger @xmath12 .
note , also , that taking the partial - collapse measurement to be more orthogonal to @xmath5 ( @xmath59 ) increases the snr for @xmath60 .
the postselection measurement orientations , which produce the high snr , coincide with those obtained in the single - copy analysis , i.e. @xmath61 , @xmath62 , for @xmath60 @xcite .
this suggests that though the spirit of the present multicopy analysis is quite different from the single - copy analysis , both analyses give similar guidance for optimally discriminating between nonorthogonal states .
we reiterate , however , that ( as compared with the single - copy approach ) the statistical snr approach ( based on nv ) is better suited to most experimental settings in which noise and experimental imperfections are present . a sketch of the experimental apparatus .
single spatial mode light from a helium - neon laser ( hene ) passes through a neutral density filter ( nd ) followed by a half - wave plate ( hwp ) and polarizer ( p1 ) to prepare the initial state . during data acquisition ,
the hwp is used to maintain a constant photon flux which is measured using a removable mirror ( rm ) . a glass window ( w )
weakly reflects vertically polarized light .
photons that pass through the window are then projected onto a linear polarization state with a second polarizer ( p2 ) .
the photons in each spatial mode are passed through colored glass filters to block background , collected via a multimode fiber and sent to single photon counting modules ( @xmath63 , @xmath64 and @xmath65 ) .
] we measure the nv signal and its amplified snr using an optical technique sketched in fig .
[ fig:2 ] . here
, the qubits are replaced by photons from a dramatically attenuated coherent beam , and the measurement device consists of polarization optics and single - photon detectors .
we encode the states in the polarization degree of freedom by passing the beam through a polarizer ( p1 ) , giving @xmath66 { \left|{0}\right > } + \sin[\delta-\delta_m]{\left|{1}\right>}$ ] , where @xmath67 correspond to the horizontal and vertical polarization states , respectively .
we perform a ( weak ) partial - collapse measurement by sending the photons through a glass window ( w ) set at the brewster angle .
the window therefore weakly reflects vertically polarized light , with probability @xmath68 , and passes horizontal light with near unit probability .
we set the second polarizer ( p2 ) in the transmitted arm to strongly project the photon into the state @xmath35 which is represented by scheme * _ a _ * or * _ b _ * , as desired @xcite . from the resulting photon detections we obtain the values of @xmath56 , @xmath57 , @xmath69 and @xmath70 and their variances @xcite .
we consider schemes * _ a _ * and * _ b _ * for @xmath71 rad and plot the results in fig . [ fig:3 ] .
we find that , for scheme * _ a _ * , we can discriminate between the two states with a higher snr than the standard scheme nearly over the whole range of angles considered . similarly , while the snr of the standard technique almost coincides with that of scheme * _ b _ * for small angles , we see that the sensitivity of the two schemes diverges quickly for larger angles ; in this regime ( @xmath72 ) , the nv scheme * _ b _ * is significantly better .
the discrepancy between theory and experiment is due to a small amount of ellipticity incurred from the glass window not included in the theory plot @xcite . a graph of the theoretical and experimental snr obtained for different measurement schemes .
scheme * _ a _ * ( red squares ) and * _ b _ * ( blue hollow circles ) correspond to the null value technique ( @xmath49 ) .
the parameter @xmath12 denotes the distance between the measured and the reference state ; it is varied by changing the angles for the input polarizer p1 . for a given p2 and w ( cf .
fig .
[ fig:2 ] ) the reference state is determined by finding p1 for which @xmath73 , @xmath74 is minimal for schemes * _ a _ * , * _ b _ * , respectively .
the standard scheme ( black circles ) is that defined by eq .
( [ snr_simp ] ) , and is represented by a single polarizer with no weak measurement .
dots correspond to calculations from data and lines correspond to the theoretical predictions .
each scheme used approximately the same number of photons , with @xmath75 per measurement . ]
the described nv procedure leading to large snr is based on the conditional outcome of a quantum measurement . as such , it resembles the well - established protocol of wv measurement @xcite .
the wv protocol consists of weakly measuring an operator @xmath76 of a _ system _ prepared in an initial state @xmath77 by weakly coupling it to a _
detector_. the detector output is kept only if the system is eventually measured to be in a chosen final state , @xmath34_postselection_. the obtained conditional average of @xmath76 , @xmath78 , is named _ weak value _ , and can be anomalously large @xcite .
this property has been exploited for amplifying small signals both in quantum optics @xcite and in solid state physics @xcite .
it is important to stress that the nv protocol is different from the wv protocol . the former makes use of a partial - collapse measurement of the operator @xmath76 , in which the system experiences back action only for a subset of all possible measurement outcomes , while a strong projection takes place for the remaining outcomes
this is not a weak measurement , which is used by the wv protocol .
the obtained conditional average of @xmath76 is now the nv , @xmath79 .
it is quantitatively different from the wv even when @xmath80 is explicitly `` weak '' @xcite .
moreover , while a large wv leads to an amplification of the snr for systems where the noise is dominated by an external ( technical ) component @xcite , the method presented here leads to high fidelity discrimination between quantum states on the background of quantum fluctuations . in conclusion
we have presented here a new protocol based on a partial - collapse measurement followed by a tuned postselection .
our protocol enables one to discern between quantum states with better accuracy than a standard measurement would allow .
by contrast to earlier protocols @xcite tuned to discriminate between two prescribed states , the present one facilitates the study of an amplified snr for a wide range of possible polarizations of one of the states , which is not _ a priori _ known .
we have demonstrated the feasibility and effectiveness of our protocol by employing an optical setup for discriminating between different polarization states of light .
notably , our present approach is based on a statistical analysis , which makes it particularly suitable for experiments .
we would like to thank n. katz , v. giovannetti , c. bruder , g. strbi , and a. ustinov for fruitful discussions , and n. gontmakher for the illustration of the setup .
this work was supported by gif , einstein minerva center , isf , minerva foundation , israel - korea most grant , and eu geomdiss .
j. c. h. was supported by army research office grants w911 nf-12 - 1 - 0263 and w911-nf-09 - 0 - 01417 . c. j.
b. acknowledges support from aro w911nf-09 - 1 - 0385 and nsf phy-1203931 . 30 natexlab#1#1bibnamefont # 1#1bibfnamefont # 1#1citenamefont # 1#1url # 1`#1`urlprefix[2]#2 [ 2][]#2 , _ _ ( , ) . , *
* , ( ) . , * * , ( ) . , _ _ ( , ) . , , , * * , ( ) . ,
* * , ( ) . , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , * * , ( ) .
, , , * * , ( ) . , , , , , , * * , ( ) . , , ,
, * * , ( ) . , , , ,
, * * , ( ) . , , , , , ,
, * * , ( ) . ,
* * , ( ) .
, * * , ( ) . , , , , , * * , ( ) . , * * , ( ) . , * * , ( ) . , , , , * * , ( ) . , , , , * * , ( ) . , * * , ( ) . , , , , ,
* * , ( ) . , , , * * , ( ) .
, , , * * , ( ) . ,
_ _ ( , ) .
* supplemental material * in this supplemental material , we provide additional technical details of our analysis and experimental methods .
section [ nvwv ] discusses the importance of the negative outcome postselection ( no click ) . in section [ povm ] ,
we provide a translation between the conditional probability formalism , used in the main text , and positive - operator valued measure ( povm ) language .
section [ single_copy ] describes the technical details of the single - copy minimum error scheme and the single - copy intermediate scheme vis - a - vis our discrimination problem .
section [ multi_copy ] presents additional technical details of the multi - copy minimum error scheme ( performed by individual standard strong measurements ) and the multi - copy intermediate scheme ( null values ) .
finally , in section [ experimental_methods ] , we provide additional details on our experimental methods with a focus on sources of loss .
|
graphene , an atomic layer of graphite that supports massless dirac fermions , displays remarkable and promising electronic properties .
recently there is increasing interest in bilayers and few layers of graphene , where the physics and applications of graphene become richer , with , e.g. , a tunable band gap @xcite for bilayer graphene .
there are some key signatures of dirac fermions that distinguish graphene from conventional electron systems .
( i ) in a magnetic field , graphene supports , as the lowest landau level ( lll ) , a special set of four zero - energy levels differing in spin and valley , as observed via the half - integer quantum hall effect .
( ii ) graphene is an intrinsically many - body system equipped with the valence band acting as the dirac sea .
quantum fluctuations of the filled valence band are fierce , even leading to ultraviolet divergences ; and one encounters such many - body phenomena as velocity renormalization , @xcite screening of charge , @xcite and nontrivial coulombic corrections to cyclotron resonance .
@xcite in multilayer graphene the zero - mode landau levels acquire a new aspect .
bilayer graphene supports eight such levels , with an extra twofold degeneracy @xcite in landau orbitals @xmath1=0 and 1 .
trilayer graphene has 12 such levels with threefold @xmath2orbital " degeneracy , and so on .
this orbital degeneracy is a new feature peculiar to the lll in multilayer graphene , and leads to intriguing quantum phenomena @xcite such as orbital mixing and orbital - pseudospin waves . in real samples
these zero - energy levels evolve , due to general interactions , into a variety of pseudo - zero - mode ( pzm ) levels , or broken - symmetry states within the lll , as discussed theoretically .
@xcite it has been unnoticed until recently that many - body effects work to lift orbital degeneracy . each zero - mode level , subjected to quantum fluctuations of the valence band , gets shifted differently within the lll , just like the lamb shift @xcite in the hydrogen atom .
this orbital lamb shift was first noted @xcite for bilayer graphene and is also realized in an analogous fashion @xcite in rhombohedral ( @xmath3-stacked ) trilayer graphene , which is a @xmath2chiral " trilayer generalization of bilayer graphene .
this orbital shift is considerably larger in scale than intrinsic spin or valley breaking , and one has to take it into account in clarifying the fine structure of the lll in multilayers .
graphene trilayers attracted theorists attention @xcite even before experiments , and it has been verified experimentally @xcite that the electronic properties of graphene trilayers strongly depend on the stacking order , with @xmath3-stacked trilayers exhibiting a tunable band gap and bernal @xmath4-stacked trilayers , the most common type of trilayers , remaining metallic
. currently trilayers are under active study both experimentally @xcite and theoretically .
@xcite the purpose of this paper is to examine the orbital lamb shift and its consequences in @xmath0-stacked trilayers , with focus on electron - hole and valley asymmetries due to general band parameters .
it turns out that @xmath0 trilayers critically differ in zero - mode characteristics from @xmath3 trilayers .
in particular , the way the coulomb interaction acts within the lll substantially differs between the two types of trilayers , leading to distinct basic filling - factor steps in which large level gaps appear in each of them .
in addition , the lll of @xmath0 trilayers , unlike that of @xmath3 trilayers , is greatly afflicted with interaction - enhanced electron - hole and valley asymmetries , which affect the sequence of broken - symmetry states within the lll , observable via the quantum hall effect . in sec . ii we examine the one - body spectrum of the pzm levels in @xmath0-trilayer graphene , and in sec .
iii show how the orbital lamb shift modifies their full spectrum . in sec .
iv we discuss how the level spectra evolve , with filling , via the coulomb interaction .
section vi is devoted to a summary and discussion on how @xmath0 trilayers differ in zero - mode characteristics from @xmath3 trilayers .
@xmath0-stacked trilayer graphene consists of three graphene layers with vertically - arranged dimer bonds @xmath5 and @xmath6 , where @xmath7 denote inequivalent lattice sites in the @xmath8-th layer .
the interlayer coupling @xmath9 ev is related to the fermi velocity @xmath10 m / s in monolayer graphene .
interlayer hopping via the nearest - neighbor dimer coupling @xmath11 0.4 ev leads to linear ( monolayer - like ) and quadratic ( bilayer - like ) spectra @xcite @xmath12 in the low - energy branches @xmath13 .
the effective hamiltonian for @xmath0-stacked trilayer graphene with general intra- and interlayer couplings is written as @xcite @xmath14 , \nonumber\\ { \cal h}_{k } & = & \left ( \begin{array}{lll } d & v & w \\ v^{\dag } & d & v^{\dag}\\ w & v & d \\ \end{array } \right ) + u,\ d = \left ( \begin{array}{cc } 0 & v\ , p^{\dag } \\ v\ , p & 0 \\ \end{array } \right ) , \nonumber\\ v & = & \left ( \begin{array}{cc } -v_{4}\ , p^{\dag } & v_{3}\ , p \\ \gamma_{1 } & -v_{4}\ , p^{\dag } \\ \end{array } \right),\ w= \left ( \begin{array}{cc } \gamma_{2}/2 & 0 \\ 0 & \gamma_{5}/2\\ \end{array } \right),\ \ \
\nonumber\\ u&= & { \rm diag}(u_{1 } , u_{1 } + \delta ' , u_{2}+ \delta ' , u_{2 } , u_{3 } , u_{3 } + \delta ' ) , \label{htri}\end{aligned}\ ] ] with @xmath15 and @xmath16 .
here @xmath17 stands for the electron field at the @xmath18 valley . @xmath19 and @xmath20 are related to the nonleading nearest - layer coupling @xmath21 and @xmath22 , respectively . @xmath23 and @xmath24 describe coupling between the top and bottom layers .
@xmath25 denote the on - site energies of the three layers ; we take @xmath26 without loss of generality and focus on the case of a symmetric bias @xcite @xmath27 .
such an interlayer bias leads to a tunable band gap for @xmath3-stacked trilayers , but not for @xmath0-stacked trilayers which involve monolayer - like subbands .
@xmath28 stands for the energy difference between the dimer and non - dimer sites .
@xmath29 is diagonal in ( suppressed ) electron spin .
the hamiltonian @xmath30 at another valley is given by @xmath29 with @xmath31 and @xmath32 , and acts on a spinor of the same sublattice content as @xmath33 .
@xmath30 is not linked to @xmath29 in a simple way and , in a magnetic field , their landau - level spectra significantly differ , @xcite especially for zero - mode levels , although they precisely but nontrivially @xcite agree when only the leading parameters @xmath34 are kept .
this is in sharp contrast to the case of bilayers and @xmath3-stacked trilayers , for which @xmath30 is linked to @xmath29 via unitary equivalence , @xcite such as @xmath35 . for the trilayer hopping parameters one may use , as typical values , those for graphite , @xcite @xmath36 in the present analysis we regard @xmath34 as the basic parameters and treat the nonleading ones @xmath37 and bias @xmath38 as perturbations .
we ignore @xmath39 from the start since its effect is negligible in high magnetic fields , as discussed later in this section .
let us place trilayer graphene in a strong uniform magnetic field @xmath40 normal to the sample plane ; we set , in @xmath29 , @xmath41 with @xmath42 , and scale @xmath43 so that @xmath44=1 $ ] .
it is easily seen that the eigenmodes of @xmath29 have the structure @xmath45 with @xmath46 , where only the orbital eigenmodes are shown using the standard harmonic - oscillator basis @xmath47 ( with the understanding that @xmath48 for @xmath49 ) .
the coefficients @xmath50 are given by the eigenvectors ( chosen to form an orthonormal basis ) of the reduced hamiltonian @xmath51 with @xmath52 where @xmath53 for short ; @xmath54 , @xmath55 .
@xmath56 , @xmath57 , @xmath58 and @xmath59 .
here @xmath60\ , \sqrt{b[{\rm t}]}\ { \rm mev}\ ] ] stands for the characteristic cyclotron energy for monolayer graphene , with @xmath61 in units of @xmath62 m/s and @xmath63 in teslas ; @xmath64 denotes the magnetic length .
note that eigenvectors @xmath65 can be taken real since @xmath66 is a real symmetric matrix .
solving the secular equation shows that there are 6 branches of landau levels for each integer @xmath67 , with two branches of monolayer - like spectra @xmath68 and four branches of bilayer - like spectra .
we denote the eigenvalues as @xmath69 , so that the index @xmath70 reflects the sign of @xmath71 .
the @xmath72 levels , e.g. , consist of the @xmath73 branches . as verified easily , with only @xmath74 and bias @xmath38 kept , the spectrum and eigenvectors of @xmath75 have the property @xmath76 for @xmath77 [ and each branch @xmath78 , where @xmath79 , e.g. , stands for @xmath80 with @xmath81 .
this structure @xcite is also seen from the fact that @xmath82 is unitarily equivalent to @xmath29 with the signs of @xmath83 reversed , @xmath84 where @xmath85 ; thus eq .
( [ vn ] ) is generalized to the full spectrum accordingly .
there are three zero - energy solutions ( per spin ) within the @xmath86 sector for @xmath87 . as seen from eq .
( [ psi_n ] ) , for @xmath88 , @xmath75 is reduced to a matrix of rank 1 , with an obvious eigenvalue @xmath89 and the eigenvector @xmath90 or @xmath91 for @xmath92 , @xmath75 has rank 4 , and we specify the four eigenmodes as @xmath93 and @xmath94 , with energy spectra @xmath95 and @xmath96 when only @xmath97 are kept , where @xmath98 @xmath99 and @xmath100 at @xmath101 t with @xmath102ev . for @xmath103 ,
in particular , the @xmath93 modes have zero energy with wave functions @xmath104 where @xmath105 and @xmath106 .
when bias @xmath38 and nonleading parameters @xmath107 are turned on , the zero - modes @xmath108 and @xmath109 in general deviate from zero energy and become the pseudo - zero - modes .
their spectra , to first order in such perturbations , can also be determined using this @xmath110 zero - mode basis @xmath111 .
writing @xmath112 in the @xmath113 matrix form @xmath114 yields the spectrum of the pseudo - zero - mode ( pzm ) sector , @xmath115 where @xmath116 this pzm spectrum @xmath117 , in the framework of degenerate perturbation theory , is correct to order linear in @xmath118 , which is sufficient for our present purpose .
diagonalizing @xmath119 by a rotation within the @xmath120 sector , @xmath121 yields the eigenspectrum @xmath122 with @xmath123 and @xmath124 ; note that @xmath125 and @xmath126 for the set ( [ triparameter ] ) of parameters and at @xmath12720 t. in particular , for @xmath103 @xmath128 ) the spectrum reads @xmath129 which , for @xmath130 , recovers an earlier result , @xcite with @xmath131 , @xmath132 and @xmath133 . here
we wish to discuss possible effects of the interlayer coupling @xmath134 .
it induces transitions that go outside the pzm sector , as one can verify using the solutions @xmath135 .
accordingly , its contributions to the spectra @xmath136 are only of second order in @xmath137 and are negligible in high magnetic fields .
the hamiltonian @xmath30 at another valley is given by @xmath29 with replacement @xmath138 .
as for its spectrum one readily finds the following : the associated eigenmodes @xmath139 take the form of @xmath140 in eq .
( [ psi_n ] ) , with replacement @xmath141 for @xmath142 and @xmath143 for @xmath144 .
the reduced hamiltonian @xmath145 is obtained from @xmath66 in eq .
( [ hn ] ) by replacing each @xmath146 by @xmath147 and each @xmath148 by @xmath149 .
one , of course , has to calculate the eigenvectors @xmath150 anew .
unlike @xmath66 , @xmath151 has rank 2 for @xmath88 and rank 5 for @xmath92 .
this already signals that the pzm spectra significantly differ between the two valleys . for @xmath88 one
considers the @xmath152 matrix hamiltonian @xmath153 , with eigenmodes ( denoted as @xmath154 , @xmath155 and the associated spectra @xmath156 where @xmath157 and @xmath158 . for @xmath92 , @xmath151 has rank 5 .
of its five eigenvalues , one belongs to the pzm sector , two are monolayer - like with @xmath159 and two are bilayer - like with @xmath160 . in the @xmath87 basis ,
the zero - energy mode is given by @xmath161 where @xmath162 and @xmath163 .
evaluating the expectation value @xmath164 yields the spectrum of the @xmath92 mode , @xmath165 correct to order linear in @xmath166 as well . the lll , i.e , the pzm sector , consists of @xmath167 at valley @xmath18 and of @xmath168 at valley @xmath169 ; there are thus twelve pzm levels differing in spin , valley and orbital .
it is interesting to look into their structure . for zero bias @xmath103 ( i.e. , @xmath170 , @xmath108 and @xmath171 at valley @xmath18
are predominantly composed of the orbital mode @xmath172 and @xmath173 , respectively , residing on the @xmath63 sites of the middle layer ; let us denote this feature as @xmath174 on @xmath175 and @xmath176 on @xmath175 .
one can further write @xmath177 on @xmath178 , @xmath179 on @xmath180 , @xmath181 on @xmath182 , and @xmath183 on @xmath184 .
this naturally explains why @xmath185 and @xmath186 are insensitive to the outer - layer coupling @xmath187 and are less sizable .
in this way , in @xmath0 trilayers the pzm levels show valley asymmetry in composition . as a function of @xmath63 for ( a ) @xmath188 and ( b ) @xmath189 mev .
real curves refer to valley @xmath18 and dashed ones to valley @xmath169 .
] the valley asymmetry is manifest in the one - body spectra @xmath190 , which , from now on , are denoted as @xmath191 to indicate that they come from @xmath112 .
numerically , for @xmath192 t and with the set of parameters in eq .
( [ triparameter ] ) taken , @xmath193 figure 1 shows the spectra @xmath194 for @xmath195 mev as a function of magnetic field @xmath63 , along with the @xmath196 bilayer - like spectra . the pzm sector is considerably spread in energy @xmath197mev , but , for large @xmath198 t , it is practically isolated from other levels .
the pzm spectra prominently differ between the two valleys ( real curves _
vs _ dashed ones ) .
in addition , they are highly electron - hole ( @xmath199 ) asymmetric ( i.e. , not symmetric about zero energy ) and this @xmath199 asymmetry comes from @xmath200 and @xmath201 , i.e. , primarily from @xmath202 at valley @xmath18 and @xmath203 at valley @xmath169 . note that @xmath204 and @xmath205 vary with interlayer bias @xmath38 .
practically only @xmath205 vary sensitively with @xmath38 while other levels are barely affected .
let us now make the landau - level structure explicit by passing to the @xmath206 basis @xmath207 with @xmath208 via the expansion @xmath209 , where @xmath1 refers to the level index , @xmath210 to the spin , and @xmath211 to the valley .
the charge density @xmath212 with @xmath213 is thereby written as @xcite @xmath214 where @xmath215 ; @xmath216 denotes the center coordinate with @xmath217 = i\ell^{2}$ ] .
the coefficient matrix @xmath218 at valley @xmath219 is constructed from the eigenvectors @xmath220 , @xmath221 and has the property @xmath222 . here
@xmath223 for @xmath224 , and @xmath225 ; @xmath226 ; it is understood that @xmath227 for @xmath228 or @xmath49 . within the pzm sector , @xmath229 at valley @xmath18 , and @xmath230 at valley @xmath169 , with @xmath231 and @xmath162 .
one can further show that , with only @xmath232 kept , these @xmath233 are only corrected to @xmath234 or smaller . in view of eqs .
( [ vn ] ) and ( [ hequi ] ) , the form factors @xmath235 enjoy the property @xmath236 for general @xmath237 , where it is understood that one sets @xmath238 for the pzm level @xmath239 .
this property plays a key role in our analysis later .
equations ( [ gpzmk ] ) and ( [ gpzmkp ] ) are expressions valid to zero@xmath240 order in perturbations @xmath241 , but they actually know the nature of perturbations through the mixing angles @xmath242 that depend on the relative strengths @xmath243 .
they indeed satisfy eq .
( [ gmnproperty ] ) .
the form factors @xmath233 generally differ between the two valleys .
interestingly , they happen to coincide for @xmath87 : indeed , for @xmath87 , one finds @xmath244 at valley @xmath18 , and analogous @xmath169-valley expressions with @xmath245 replaced by @xmath246 in the above .
this fact tells us that the charge @xmath247 takes a manifestly valley - symmetric form for zero bias @xmath188 while the one - body spectra @xmath248 inevitably break valley symmetry .
in addition , eq . ( [ gknu0 ] ) implies that , for @xmath87 , @xmath249 is isolated from @xmath250 , and similarly , @xmath251 from @xmath252 . from now on we frequently suppress summations over levels @xmath1 , spins @xmath253 and valleys @xmath254 , with the convention that the sum is taken over repeated indices .
the hamiltonian @xmath112 projected to the pzm sector is thereby written as @xmath255 with @xmath256 . here
the zeeman term @xmath257 $ ] mev is introduced via the spin matrix @xmath258 .
actually , the zeeman energy @xmath259 is only about 3mev even at @xmath260 t and is generally smaller than energy splitting due to valley breaking . accordingly , in what follows , we mostly suppose that the spin is practically unresolved and focus on energy gaps due to valley and orbital breaking .
in this section we examine the effect of coulombic quantum fluctuations on the pzm multiplet .
the coulomb interaction is written as @xmath261 where @xmath262 with @xmath263 and the substrate dielectric constant @xmath264 .
for simplicity we ignore the difference between the intralayer and interlayer coulomb potentials . in this paper
we generally study many - body ground states @xmath265 with a homogeneous density , realized at integer filling factor @xmath266 $ ] .
we set the expectation values @xmath267 with @xmath268 , so that the filling factor @xmath269 for a filled @xmath270 level .
let us define the dirac sea @xmath271 as the valence band with levels below the pzm sector ( i.e. , levels with @xmath272 , @xmath273 , ... ) all filled .
we construct the hartree - fock hamiltonian @xmath274 out of @xmath275 as the effective hamiltonian that governs the electron states over @xmath271 . as usual , the direct interaction @xmath276 is removed if one takes into account the neutralizing positive background .
we thus focus on the exchange interaction @xmath277 where we sum over filled levels @xmath237 and retain the pzm sector @xmath278 .
let us first extract the contribution from the dirac sea , @xmath279 where the sum is understood over spin @xmath253 and over @xmath280 for @xmath281 and @xmath282 for @xmath283 .
actually , the sum over infinitely many filled levels with @xmath284 gives rise to an ultraviolet divergence .
fortunately this infinite sum is evaluated exactly to zero@xmath240 order in perturbations @xmath241 , as done earlier , @xcite if one notes eq .
( [ gmnproperty ] ) and the completeness relation @xcite @xmath285 the result is @xmath286 for @xmath287 ; analogously for valley @xmath169 .
the @xmath288 term leads to a divergence upon integration over @xmath289 ; it , however , shifts all levels @xmath239 uniformly and is safely omitted .
the regularized dirac - sea contribution then reads @xmath290 analogously for @xmath291 .
integration over @xmath289 , with the formula @xmath292
= [ 1,1,3]\ , \tilde{v}_{c } $ ] , yields @xmath293\ , \tilde{v}_{c } , \nonumber\\ \epsilon^{\rm v}_{0_{\pm}}\!\!\ ! & = & \textstyle{1\over{2}}\ , \big [ 1 + \textstyle{1\over{2 } } c_{1}^2 ( 1 \mp \sin \phi ) \big]\ , \tilde{v}_{c } , \nonumber\\ \epsilon^{\rm v}_{1 } & = & \textstyle{1\over{2}}\ , ( 1 - c_{1}^2 + 3\ , c_{1}^4 ) \ , \tilde{v}_{c } , \label{vzeroreg}\end{aligned}\ ] ] where @xmath294 , etc .
, @xmath295 and @xmath296}\ , { \rm mev}.\ ] ] note that eqs .
( 34 ) @xmath297 ( 36 ) are the zeroth - order expressions , which depend on bias @xmath38 through the zero@xmath240-order ratios @xmath298 and @xmath299 .
numerically , for @xmath188 and @xmath192 t , and with @xmath300 taken as a typical value , @xmath301 in this way , the pzm levels are @xmath2lamb - shifted " due to vacuum fluctuations and the splitting among @xmath302 reflects the difference in their spatial ( or @xmath289 ) distributions . plotted in units of @xmath303 as a function of bias @xmath38 at @xmath304 t .
( b ) @xmath299 and @xmath298 as a function of @xmath38 for @xmath304 t . ] in fig .
2 the lamb shifts @xmath302 are plotted , in units of @xmath303 , as a function of bias @xmath38 .
they are ordered , e.g. , as @xmath305 at one valley , and are valley - symmetric for @xmath188 ( this reflects the manifest valley symmetry of the charge @xmath247 noted in sec .
ii ) , with valley asymmetry developing gradually with increasing bias @xmath38 . and
@xmath195 mev .
the upper and lower halves of each parabolic spectrum refer to the empty ( @xmath306 ) and filled ( @xmath307 ) case , respectively .
( b ) @xmath308 and @xmath195 mev , with a weaker coulomb potential . ]
the spectra ( [ vzeroreg ] ) or ( [ spectrum ] ) refer to those of empty levels .
actually the spectra vary ( i.e. , generally go down due to the exchange interaction ) with filling of the pzm levels . in particular , eq . ( [ vxds ] ) tells us that , when the pzm sector is filled up , @xmath302 change sign @xmath309 . in this sense , the lamb - shift corrections @xmath302 preserve @xmath199 symmetry .
thus , as the pzm sector is filled from @xmath310 ( empty ) to @xmath311 ( full ) , the spectrum of the @xmath239-th level varies from @xmath312 to @xmath313 .
see fig .
3(a ) , which depicts the ( empty / filled ) spectra @xmath314 as a function of @xmath63 , for @xmath188 and @xmath315 ; the upper and lower halves of each spectrum refer to empty and filled levels , respectively .
note that bias @xmath38 works to enhance the splitting of the @xmath316 spectra . for reference ,
3(b ) shows the level spectra for a weaker potential with @xmath317 . at valley @xmath18 ,
the one - body spectra @xmath248 are ordered so that @xmath318 while @xmath319 are ordered so that @xmath305 .
the lamb - shift contributions @xmath319 therefore enhance splitting among filled levels @xmath320 with spectra @xmath321 . for empty levels ( with @xmath322 ) they work oppositely and even reverse the ordering of the @xmath323 and @xmath324 spectra when the coulomb interaction @xmath303 is strong enough ( i.e. , for smaller @xmath325 and higher @xmath63 ) ; compare figs .
3(a ) and 3(b ) .
replacing @xmath326 also allows one to find essentially the same features for valley @xmath169 . in this way the lamb - shift contributions , though @xmath199 symmetric by themselves , work to enhance @xmath199 asymmetry in the full pzm spectra @xmath314 . to see how each level evolves with filling
, one has to examine the coulomb interaction acting within the lll ; we study this in the next section .
the coulomb exchange interaction acting within the pzm sector is written as @xmath327 with @xmath328 , where @xmath329 and @xmath237 are summed over @xmath330 .
for definiteness we focus on the @xmath87 case , where @xmath331 and @xmath332 considerably simplify . indeed , as noted in eq .
( [ gknu0 ] ) , for @xmath188 , the charge @xmath247 takes a valley - symmetric form and , in addition , one has @xmath333 and @xmath334 .
this structure suggests that , at valley @xmath18 , @xmath335 tends to be isolated from @xmath336 which may potentially get mixed ; similarly , @xmath337 tends to be isolated from @xmath338 at valley @xmath169 .
actually , for @xmath188 one finds that @xmath339 at valley @xmath18 , with obvious spin indices @xmath340 suppressed . replacing @xmath341 in the above yields expressions @xmath342 for valley @xmath169 and , in an analogous way , mixed - valley components @xmath343 as well .
one can further use new orbital labels @xmath344 and rename @xmath345 as @xmath346 with valley @xmath281 , and @xmath347 as @xmath346 with valley @xmath283 , so that , e.g. , @xmath348 , @xmath349 , etc
. then the exchange interaction @xmath350 itself is cast into a valley- ( and spin-)symmetric form composed of terms like @xmath351 .
the pzm levels are now governed by the effective hamiltonian @xmath352 .
note first that the interaction @xmath353 is symmetric in spin and valley ( for @xmath188 ) .
thus @xmath354 is made diagonal in valley and spin if one takes the valley basis @xmath355 and the spin basis @xmath356 of the one - body part @xmath357 .
accordingly , one can treat each valley and spin separately , and diagonalize @xmath354 with respect to the orbital modes @xmath358 and @xmath359 for each spin .
let us now discuss how the pzm levels evolve with filling . for definiteness , we first suppose filling the empty pzm sector with electrons gradually under a fixed magnetic field @xmath360 t and @xmath188 .
we consider 6 levels , @xmath361 per spin , and use @xmath362 to denote the filling factor for this subsector ; with electron spins supposed to be unresolved , the pzm sector thereby has the filling factor @xmath363 .
( we refer to the case of resolved spins later . )
our focus is on uniform ground states at integer filling . to follow their evolution , we choose to diagonalize the hamiltonian @xmath354 with uniform states at intermediate filling factors ; this serves to visualize how level mixing and crossing take place , as we shall see .
one can read from fig .
3(a ) the level spectra of the empty / filled pzm sector , @xmath364 where @xmath365 and @xmath366 .
it is seen from these spectra that the @xmath367 and @xmath368 levels potentially have crossing , so do @xmath369 and @xmath370 .
this signals level mixing within each pair .
it is the lowest - lying @xmath367 level that starts to be filled first .
as it is filled , it cooperates with the @xmath368 level , paired via the exchange interaction . to clarify how they evolve
, one can now try to diagonalize @xmath371 for the three lowlying levels @xmath372 , and subsequently for those at valley @xmath18 .
see the appendix for an analysis .
t and @xmath373 as the pzm sector is filled from @xmath374 ( empty ) to @xmath375 ( full ) ; the filling factor @xmath376 $ ] , with the electron spin supposed to be unresolved .
orbital mixing takes place over the interval @xmath377 and @xmath378 , indicated by colored stars .
thin dotted curves represent evolution of level spectra when no orbital rotation were allowed .
( a ) evolution of level spectra when the pzm sector is emptied from @xmath375 to @xmath374 .
( b ) and ( c ) level spectra for @xmath379 and @xmath373 .
( d ) level mixing is absent for a weaker coulomb potential with @xmath380 and at @xmath381 t . ]
figure 4(a ) summarizes the resulting evolution of level spectra . for @xmath382 with @xmath383 only the @xmath367 level is filled and gets lower in energy along with the ( paired ) empty @xmath368 level . beyond @xmath384 , @xmath367 is mixed with @xmath368 and turns into @xmath368 at @xmath385 ; at the same time , @xmath368 turns into @xmath367 .
both @xmath368 and @xmath367 levels are eventually filled up at @xmath386 . at integer filling @xmath387 , the @xmath388-mixed levels consist of a filled level of energy @xmath389 mev and an empty level of energy @xmath390 mev . a close look into fig .
4(a ) reveals that an @xmath2orbital " rotation takes place so as to avoid level crossing .
the remaining @xmath251 level evolves individually , and is filled over the interval @xmath391 $ ] . for @xmath392
an analogous process is repeated for @xmath393 levels at another valley . there
mixing of @xmath370 and @xmath369 takes place over the interval @xmath394 with @xmath395 , and avoids level crossing . from fig .
4(a ) one can read off the spectra of the pzm sector at each integer filling factor @xmath396 $ ] .
the spectra are @xmath199- and valley - asymmetric .
let us now note that , due to this @xmath199 asymmetry , the level spectra may evolve in a different pattern when one empties the pzm sector rather than filling it .
indeed , such a difference is clearly seen from fig .
4(a ) , which shows the evolution of level spectra when the filled pzm sector is gradually emptied ( i.e. , @xmath397 ) under the same @xmath360 t .
actually , fig .
4(a ) is a result of direct calculation , but it will be clear how to draw it by a glance at fig .
for comparison , see also figs .
4(b ) and 4(c ) , which show the evolution of level spectra under @xmath398 t . there
the pattern of evolution is uniquely fixed , independent of whether one fills or empties the pzm sector .
lastly , fig .
4(d ) illustrates the case of a weaker coulomb potential with @xmath317 and at @xmath381 t , corresponding to the level spectra in fig .
3(b ) . here
again the pattern of level spectra is uniquely fixed , but , unlike in the above cases , there is no level mixing . in general , the level spectra @xmath314 of the empty / filled pzm sector ( in fig . 3 ) are fixed in advance by specifying the value of magnetic field @xmath63 at @xmath310 and @xmath311 , respectively .
how the spectra evolve at intermediate filling factors , as we have seen , depends on whether one fills or empties the pzm sector and how one does it , e.g. , under fixed @xmath63 or fixed density @xmath399
. it will be clear from the model calculations above that the @xmath249 and @xmath251 levels evolve individually without mixing with others while @xmath370 and @xmath369 move in pairs , so do @xmath400 .
actually , with this experience , a close look into the empty / filled spectra in fig . 3
allows one to draw a general idea about how the level spectra evolve under fixed @xmath63 and @xmath188 ( or even for small @xmath38 as well ) .
for example , the presence or absence of level mixing is inferred from fig .
level mixing takes place so as to avoid crossing of paired levels @xmath401 or @xmath252 .
as noted in sec .
ii , the ordering of these paired levels , i.e. , @xmath402 and @xmath403 for @xmath404 , is reversed for the full spectra @xmath405 when the coulomb potential @xmath406 is strong enough .
it is thus this inversion of ( empty ) paired levels that drives level mixing .
accordingly , with @xmath407 , mixing of paired levels is necessarily present for almost all values of @xmath63 in fig .
3(a ) , as indeed seen from figs .
4(a ) and 4(b ) . for fig .
3(b ) , i.e. , for a weaker potential with @xmath317 , mixing is present only at low @xmath408 t and is absent at higher @xmath63 , as is the case with fig .
when bias @xmath38 is turned on , mixing of @xmath400 disappears rapidly with increasing @xmath38 , but mixing of @xmath401 tends to persist at low @xmath63 , as verified easily . in the level spectra of fig . 3
, the @xmath251 level is relatively isolated upward from the paired levels @xmath252 , so is @xmath249 from @xmath409 .
it is therefore the lowest - lying @xmath400 pair that is filled first as @xmath410 ( or emptied last as @xmath411 ) .
this leads to a unique @xmath412 ground state [ consisting of filled @xmath252 levels ] with a relatively large @xmath412 level gap , as is evident from fig .
4 . likewise
, an isolated @xmath249 level leads to a unique @xmath413 state with a relatively large gap .
the ground states at other filling factors , in contrast , vary in composition case by case .
in particular , one notices an equally large @xmath414 gap and a relatively small @xmath415 gap in the spectra of figs .
4(b ) and 4(d ) , which show essentially the same low-@xmath63 characteristics of the pzm sector
[ at @xmath416 in fig .
3(a ) or @xmath417 in fig . 3(b ) ] .
interestingly , in those low-@xmath63 cases , the @xmath415 state ( essentially ) consists of filled @xmath418 levels , which is the same in composition as the @xmath415 state one naively expects from the one - body spectra @xmath191 in fig . 1 alone .
we have so far supposed unresolved electron spins .
note that the exchange interaction acts on pairs of the same spin and valley .
accordingly , if , e.g. , in fig .
4(b ) , there were two @xmath400 pairs of spin up and down resolved against possible disorder , each pair would repeat the @xmath410 evolution in the figure over the interval @xmath419 , yielding a @xmath420 gap comparable to the @xmath412 gap . in this way ,
small spin gaps , if resolved , are equally well enhanced by the interaction , and will modify the evolution patterns in fig . 4 accordingly .
the transport properties of graphene trilayers have been studied in a number of experiments , @xcite and some nontrivial features of the lll of @xmath0-stacked trilayers have been observed .
evidence for the opening of the @xmath415 gap comes from early observations @xcite of an insulating @xmath415 state in both @xmath0 and @xmath3 trilayers .
recent experiments on substrate - supported @xmath0 trilayer graphene by henriksen _ et al . _
@xcite observed a robust @xmath412 hall plateau and a possible incipient @xmath414 or @xmath413 plateau under zero bias ( @xmath421 , and @xmath422 plateaus in biased samples .
subsequent measurements on dual - gated suspended devices by lee _
et al . _
@xcite observed @xmath423 plateaus at low magnetic field @xmath424 and also resolved , in high magnetic fields , additional plateaus at @xmath425 , and @xmath426 , indicating almost complete lifting of the 12-fold degeneracy of the lll .
common to these observations , in particular , is @xmath199 asymmetry in the sequence of plateaus , with a prominent @xmath412 plateau .
the @xmath379 and @xmath366 case of fig .
4(b ) appears to capture these features seen in experiment at low @xmath63 . for resolved spins
, this case will lead to large gaps at @xmath427 , 0 , 3 and 5 , and relatively small gaps at @xmath428 , -3 and -5 . in our picture ,
appreciable @xmath199 asymmetry is a result of coulombic enhancement of the @xmath199 asymmetry in @xmath429 and a large @xmath412 gap is triggered by the valley asymmetry of @xmath429 such that @xmath430^{k ' } < { \rm min}[\epsilon_{0}^{\rm h } , \epsilon_{1_{-}}^{\rm h } ] ^{k}$ ] , i.e. , the @xmath169 valley is relatively lower in spectrum . in general ,
large level gaps are associated with evolution of orbital modes @xmath401 and @xmath400 of basic filling step @xmath431 per spin and evolution of rather independent modes @xmath249 and @xmath251 of step @xmath432 .
this is in sharp contrast to the case of @xmath3 trilayers , where large gaps within the lll are associated with evolution ( actually , mixing ) of orbital modes of basic step @xcite @xmath433 per spin , leading to visible @xmath434 plateaus .
in a magnetic field graphene trilayers develop , as the lll , a multiplet of twelve nearly - zero - energy levels with a three - fold orbital degeneracy . in this paper
we have examined the quantum characteristics of this pzm multiplet in @xmath0 trilayers , with the coulomb interaction and the orbital lamb shift taken into account .
it turned out that @xmath0 trilayers are distinct in zero - mode characteristics from @xmath3 trilayers examined earlier .
@xcite we have , in particular , seen that both valley and @xmath199 symmetries are markedly broken in the lll of @xmath0 trilayers .
these asymmetries appear in the one - body spectra @xmath435 already to first order in nonleading hopping parameters ( such as @xmath203 , @xmath436 and @xmath28 ) , and are enhanced via the lamb - shift contributions @xmath437 and the coulomb interaction acting within the lll .
in contrast , for @xmath3 trilayers the one - body pzm spectra @xmath435 involve , for zero bias @xmath188 , only a tiny @xmath199 asymmetry of @xmath438 and no valley breaking @xcite linear in nonleading parameters @xmath439 . the lamb - shift corrections @xmath437 add no further breaking ( to the leading order ) .
accordingly , in @xmath3 trilayers the lll is far less afflicted by @xmath199 and valley asymmetries .
the pzm levels differ in structure between the two types of trilayers . in @xmath0 trilayers
they are composed of the @xmath172 and @xmath173 orbital modes distributed in a distinct way at each valley , as noted in sec .
ii , and in this sense the associated valley asymmetry is intrinsic .
in contrast , in @xmath3 trilayers these levels are characterized by the @xmath172 , @xmath173 and @xmath440 orbital modes residing predominantly on one of the outer layers , @xcite with the two valleys related symmetrically [ by layer and site interchange @xmath441 $ ] .
the two types of trilayers substantially differ in the way the coulomb interaction acts within the lll .
they thus differ in the way large level gaps or the associated conductance plateaus appear within the lll , with @xmath0 trilayers having basic filling steps of @xmath442 and @xmath3 trilayers having a step of @xmath433 .
interlayer bias @xmath38 also acts quite differently on the two types of trilayers . for @xmath3 trilayers ,
@xmath38 acts oppositely at the two valleys and enhances valley gaps .
in contrast , for @xmath0 trilayers , it works to further split @xmath443 ( and @xmath444 ) , i.e. , enhance orbital breaking at each valley .
the orbital lamb shift is a many - body vacuum effect but is intimately correlated with the coulomb interaction acting within the multiplet .
this is clear if one notes that the filled pzm sector and the empty one , both subject to quantum fluctuations of the filled valence band , differ by the amount of this coulomb interaction .
it will be clear now why this vacuum effect , though it could easily be overlooked if one naively relies on the coulomb interaction projected to the lll alone , has to be properly taken into account in every attempt to explore the pzm sector in graphene few - layers .
@xcite the @xmath199 and valley asymmetries inherent to @xmath0 trilayers substantially modify the electron and hole spectra within the lll .
the sequence of broken - symmetry states , observable via the quantum hall effect , is thereby both @xmath199- and valley - asymmetric and can change in pattern , depending on how one fills or empties the lll .
we have presented some model calculations in sec .
iv , assuming a typical set ( [ triparameter ] ) of parameters and @xmath325 .
they are intended to illustrate what would generally happen when the orbital lamb shift and coulomb interactions are properly taken into account .
they will also be a good base point for a more elaborate analysis when more data on graphene trilayers become available via future experiments .
this work was supported in part by a grant - in - aid for scientific research from the ministry of education , science , sports and culture of japan ( grant no .
24540270 ) .
in this appendix we diagonalize the effective hamiltonian @xmath445 for the three lowlying levels @xmath446 , with the electron spin kept frozen .
let us write @xmath447 and @xmath448 , with @xmath449 for @xmath450 .
note that @xmath451 are real for real filling factors @xmath452 , which we take .
it therefore suffices to use a real @xmath453 rotation to diagonalize the @xmath454 real symmetric matrix @xmath455 .
we thus rotate @xmath456 in orbital @xmath450 space , @xmath457 , with three euler angles @xmath458 parameterizing @xmath459 where the spin-1 generators @xmath460 in terms of the totally antisymmetric tensor @xmath461 with @xmath462 . note that @xmath463 mixes @xmath464 , @xmath465 mixes @xmath466 , etc .
we make @xmath467 diagonal , using the filling factors @xmath468 expressed in terms of the filling fraction @xmath469 of the diagonalized levels @xmath470 .
let us start filling the empty pzm sector at ( relative ) filling factor @xmath374 .
obviously , it is the lowest - lying @xmath367 level that starts to be filled . to follow how it evolves
let us suppose that it is filled with fraction @xmath471 and set @xmath472 .
@xmath473 is diagonalized if one can adjust @xmath474 so that @xmath475 .
the onset of possible rotations is seen from the behavior of these @xmath476 under small rotations .
to first order in @xmath474 , @xmath477 this structure reveals that @xmath478 for @xmath479 with @xmath480 while @xmath481 is possible for @xmath482 . solving for @xmath474
numerically shows that the energy eigenvalue @xmath483 is indeed lowered for @xmath482 with @xmath484 and @xmath485 .
a further analysis reveals that @xmath463 rises from 0 to @xmath486 for @xmath487 and then keeps @xmath486 up to @xmath386 .
( for @xmath488 we set @xmath489 and extend the solution across @xmath490 . ) thus , over the interval @xmath491 a rotation takes place in orbital space , and thereby the @xmath367 and @xmath368 levels are interchanged .
finally , the remaining @xmath337 level is filled individually for @xmath492 .
|
in a magnetic field graphene trilayers support a special multiplet of 12 zero(-energy)-mode landau levels with a threefold degeneracy in landau orbitals .
a close look is made into such zero - mode levels in @xmath0-stacked trilayers , with the coulomb interaction taken into account .
it turns out that the zero - mode landau levels of @xmath0 trilayers are greatly afflicted with electron - hole and valley asymmetries , which come from general hopping parameters and which are enhanced by the coulomb interaction and the associated vacuum effect , the orbital lamb shift , that lifts the zero - mode degeneracy .
these asymmetries substantially affect the way the zero - mode levels evolve , with filling , via coulomb interactions ; and its consequences are discussed in the light of experiments .
|
in recent years , power grids are experiencing a revolutionary technological transformation .
one significant feature is that electric appliances can receive realtime power price via communication networks and optimize its power consumption level according to the current power price .
then , the power utilization efficiency is significantly improved and the global energy consumption is reduced to combat the crisis of energy resource . in smart grid , a key challenge is how to adapt the communication network to the context of power price transmission .
obviously , the data flow of power price can not be elastic since it should be realtime ; otherwise , it may incur a significant loss if the expired power price is used .
therefore , the data transmission of power price must be equipped with quality of service ( qos ) guarantee .
this incurs two important questions unique to smart grid : * _ how to define the qos requirement in the context of smart grid ?
_ * _ how to ensure the qos requirement from the home appliance in the communication network ? _ in this paper , we _ answer the above two questions by proposing a qos system for smart grid . _
the proposed qos framework plays the role of interface between the power market and the communication networks .
once a set of reasonable qos metrics can be derived in the context of smart grid , many qos ensuing approaches can be applied to guarantee the performance gain introduced by the technology of smart grid . to answer the first question , we need to study the detailed mechanism of power price .
take video streaming for instance . to propose a qos requirement for video streaming
, the source coder must be aware of the impacts of different factors like delay or jitter on the video quality and then derive a suitable qos requirement .
therefore , we study the impact of qos parameters on the reward of home appliance . for simplicity
, we study only two qos parameters , namely the delay and outage probability .
the framework proposed in this paper also applies to many other qos metrics .
we first introduce the mechanism of power price based on the dynamics of load .
then , we build a reward system for the home appliance based on the power price and the utility function of the appliance , thus obtaining the impact of delay and outage on the reward of home appliance . finally , the qos requirement is derived by optimizing the reward . to answer the second question , we focus on routing methodologies meeting the derived qos requirement .
we focus on providing multiple qos - aware routing within multiple ( more than 2 ) constraints . given the heterogeneity of the smart grid , traditional schemes , such as fully polynomial - time approximations @xcite@xcite@xcite ,
can not be directly applied due to the requirements of high computing and storage capabilities .
an efficient , which can be implemented by both powerful and resource - limited devices , and effective , which provides provably - good performance , algorithm is needed for the qos routing in smart grid . in this part , we present a simple greedy algorithm for the multi - constrained qos routing .
moreover , we prove that our greedy algorithm is a @xmath0-approximation ( @xmath0 is the number of constraints ) .
in addition , our solution can be implemented in a distributed manner .
the remainder of this paper is organized as follows .
the system model is introduced in section [ sec : model ] .
then , the qos requirement is derived in section [ sec : derivation ] while the qos routing is discussed in section [ sec : routing ] .
numerical results and conclusions are provided in sections [ sec : numerical ] and [ sec : conclusion ] , respectively .
we consider a simplified model for smart grid by considering only the qos requirement in the power price inquiry .
we assume that a home appliance receives power price from the power market .
a qos requirement is sent from the home appliance to the control center of the communication network .
then , the control center assigns one or more route for the home appliance to guarantee the qos requirement .
smart devices , such as smart meter , and electricity generator can be viewed as the nodes throughout a network .
all the transmission medium , such as fiber , wireless , broadband over power line , wimax , gprs , ethernet , and radio , form the links in a network . as shown in fig .
[ fig : grid ] , the whole infrastructure of smart grid can be represented by a communication network structure , which is designed to optimize a smart grid investment .
it is worth noting that smart gird is a heterogeneous network .
various electric equipments , with dramatically different resource limits , such as computing power , storage capability , are integrated in the grid .
meanwhile , wireless network technology is utilized in combination with a utilities fiber or ethernet communications infrastructure . to provide qos - aware routing for smart grid
, we must consider the heterogeneity of the network and provide solutions that could be applied for all the devices in the networks .
a qos requirement usually includes specifications like average delay , jitter and connection outage probability . to derive the qos requirement ,
the following problems should be addressed in the study : * how to describe the probabilistic dynamics of the power system ? * how to evaluate the impact of different qos specifications on the smart grid system ?
for example , how does a long communication delay affect the system performance ? * how to derive qos requirement due to the corresponding impact ? in this section ,
we provide an approach to address the above three key questions and thus derive the qos requirements for delay and outage probability .
power price is typically determined by locational margin price ( lmp ) @xcite driven by the load which varies with time .
a constrained optimization problem can be used to derive the lmp from the load and other parameters , where the lagrange factors of the constraints are considered as prices @xcite . in practical systems , we can use a piecewise curve , as illustrated in fig .
[ fig : mapping ] , to accomplish the mapping between the load and the price .
note that , we have finite numbers of prices , denoted by @xmath1 , in fig .
[ fig : mapping ] .
therefore , we denote by @xmath2 , @xmath3 , ... , @xmath4 these prices .
the intervals of loads corresponding to the prices @xmath2 , ... , @xmath4 are denoted by @xmath5 , ... , @xmath6 , respectively .
we assumed that the load is uniformly distributed within the corresponding interval given the price .
the load is random due to many random factors like the power generation and consumption level .
we can model it as the positive part of a gaussian random variable , whose probability density function ( pdf ) is given by @xmath7 where @xmath8 is the load at time slot @xmath9 , @xmath10 is the maximal possible load , @xmath11 and @xmath12 are the expectation and variance .
then , we model the gaussian distribution parameters as functions of the elapsed time .
suppose that , at time slot 0 , the true value of the load is @xmath13 .
then , we assume that the load distribution at time slot @xmath9 satisfies that following laws : * the expectation @xmath11 of the gaussian distribution is equal to @xmath13 .
the rationale is that the prediction should be unbiased . *
the variance @xmath12 satisfies @xmath14 , where @xmath15 is a parameter and can be estimated from measurements , i.e. the variance increases linearly with the elapsed time , which is similar to a brownian motion . at time
slot @xmath9 , the power price and power consumption are denoted by @xmath16 and @xmath17 , respectively .
we assume a time - invariant utility function for the power consumption and denote it by @xmath18 .
the decision of power consumption is based on the known power price , which means that @xmath17 is a function of @xmath19 . for simplicity , we assume that the optimal power consumption level maximizes the following metric : @xmath20 where @xmath21 is the price adopted by the home appliance
. it could be different from the true value due to delay .
we assume that @xmath22 is an increasing , strictly concave and continuously differentiable function .
we also assume that the first order derivative of @xmath22 , denoted by @xmath23 , ranges from @xmath24 to @xmath25 .
based on these assumptions , the optimal power consumption level is thus given by @xmath26 which is derived from the first order condition of optimality , i.e. @xmath27 .
suppose that the communication delay is @xmath28 time slots .
then , at time slot @xmath9 , the price used for optimizing the power consumption level is @xmath29 .
hence , the cost incurred by the communication delay , as a function of the delay , is given by @xmath30,\end{aligned}\ ] ] where the expectation is over all realizations of the power price and can be computed using the probabilistic dynamics of the power price discussed in section [ sec : derivation ] .
a. it is also possible that the communication link experiences an outage such that the home appliance can not obtain the realtime power price .
in such a situation , the home appliance can only use a default power price , which is independent of the time .
we assume that the default power price equals the average power price , which is denoted by @xmath31 .
then , the expected loss incurred by the outage is given by @xmath32,\end{aligned}\ ] ] where @xmath33 is the outage probability .
if there is no constraint on the delay , the delay requirement of the home appliance should be as low as possible .
however , it is expensive for the network to achieve a very low communication delay .
therefore , the system can control the delay requirement using a delay dependent price , namely @xmath34 .
then , the delay requirement of the home appliance is to minimize the total cost , i.e. the average loss incurred by using the old price and the price taxed by the communication network .
the optimal delay requirement is then given by @xmath35 similar approach can be applied for deriving the requirement of outage probability .
suppose that there is a tax for the communication with outage probability @xmath33 , which is denoted by @xmath36 .
then , the optimal requirement of outage probability is given by @xmath37 when the qos specification includes both delay and outage probability , the optimal qos requirement is then given by @xmath38
after deriving the qos requirements , we will study how to deliver transmission in smart grid with multi - constrained qos routing problems with @xmath39 additive qos parameters .
a smart grid is modeled by an edge weighted directed graph @xmath40 ) , where @xmath41 is the set of _ n nodes _ , including end users , smart meters and other electric devices , _ e _ is the set of _ m edges _ , and @xmath42 ) is an _ edge weight vector _ so that @xmath43 _ 0 _ is the @xmath44 _ weight _ of edge @xmath45 . for a path @xmath21 in @xmath46 ,
the @xmath44 _ weight _ of @xmath21 , denoted by @xmath47 , is the sum of the @xmath44 weights over the edges on @xmath21 : @xmath48 .
[ multi - constrained routing ( mcr ) ] [ def : mcr ] given an edge weighted directed graph @xmath49 , with @xmath0 nonnegative real - valued edge weights @xmath50 associated with each edge @xmath45 , a constraint vector @xmath51 where each @xmath52 is a positive constant ; and a source - destination node pair @xmath53 .
the mcr problem seeks an @xmath54 @xmath55 @xmath9 path @xmath21 such that @xmath56 . @xmath57
the inequality @xmath58 is called _
the @xmath44 qos constraint_. a path @xmath21 satisfying all @xmath0 qos constraints is called a _ feasible path or a feasible solution _ of mcr problem .
an mcr problem is said to be _ feasible _ if it has a feasible path , and _
infeasible _ otherwise .
[ fig : mcr ] to see the incidences of this problem in smart grid , as shown in fig .
[ fig : mcr ] , one can consider that on each transmission line , there are different _ weights _ associated with it , representing _ the energy consumption for the transmission , edge delay , edge reliability _ , etc .
in smart grid , a transmission is required to satisfy several constraints , such as delay , energy consumption , and transmission reliability .
assume that electric generator ( @xmath59 ) needs to provide qos transmissions to the user ( @xmath60 ) . on each link ,
two different qos metrics : _ cost _ and _ delay _ , are considered . if the constraint vector @xmath61 is ( 3 , 5 ) , in other words , if users aim to find a path such that _ cost _ @xmath62 , _ delay _
@xmath63 , path ( 1 , 2 , 5 ) , marked by dotted red links , is a feasible path . for the constraint vector ( 4 , 4.5 ) , path ( 1 , 3 , 4 , 5 ) , marked by solid blue links , is a feasible solution . however , there is no feasible solution in this network for constraint vector ( 3 , 4 ) .
the mcr problem is known to be np - hard @xcite , even for the case of @xmath64 .
although qos routing in networks has become an active area in recent years , little work , particularly on _ performance - guaranteed multi - constrained _ qos routing , has been done in smart grid . given the characteristics of smart grid ,
there are several unique challenges for providing multi - constrained qos routing . among them
, one of the biggest concerns is routing for a heterogenous system like smart grid .
various devices with different resources and capabilities , from powerful large electrical generator to resource - limited sensors , are collaborated together .
most previous performance - guarantee qos routing schemes requires strong computing capability @xcite@xcite@xcite@xcite .
however , these sophisticated schemes can not be directly applied in smart grid due to the stringent requirements on the memory and computing capability .
second , a smart grid is a large distributed system .
most of the time , qos routing decision must be made locally by each device based on its local information .
for example , a smart meter needs to decide whether to accept a qos request based on its local reading and expectations .
most previous work , especially with performance guarantee , requires the globe knowledge of the network , and could not be directly applied to smart grid . to build a scheme for diverse heterogeneous system , simple and efficient qos routing scheme , which could be implemented by various devices ,
is needed .
our goal is to find a simple and effective routing solution for smart grid . to find simple and efficient solution that can be implement in a distributed manner , we target the problem from a different perspective . instead of studying the mcr problem directly ,
let us formulate an _ optimization _ version multi - constrained qos routing problem .
[ def : mcr ] given an undirected network @xmath46=@xmath65 , with @xmath0 _ nonnegative real - valued _ edge weights @xmath66 , associate with each edge @xmath67 ; a positive vector @xmath68 ; and a source - destination node pair @xmath69 , mcr seeks an @xmath70 path @xmath71 such that @xmath72 , where @xmath73 is the smallest real number @xmath74 such that there exists an @xmath70 path @xmath21 satisfying @xmath75 .
@xmath57 we call @xmath73 the _ optimal value _ of mcr and @xmath71 ad _ optimal path _ of mcr . note that @xmath76 * if and only if * mcr problem is feasible .
since @xmath73 could be smaller than 1 , the optimization problem omcr also introduces a metric to compare two feasible solutions to mcr @xmath77 the one with the smaller corresponding @xmath78 value is regarded as a better solution .
a very simple @xmath0-approximation algorithm , named omcr , is presented in algorithm [ alg : greedy ] .
the algorithm computes an auxiliary edge weight @xmath79 as the maximum of all edge weights @xmath80 divided by @xmath81 , respectively .
it then computes a shortest path @xmath82 using this auxiliary edge weight . the path @xmath83 is guaranteed to be a @xmath0-approximation of omcr .
note that the auxiliary edge weights can be computed locally at each node , and the shortest path can be computed using bellman - ford algorithm .
therefore , our @xmath0-approximation algorithm can be implemented as a distributed algorithm , and can be used by existing routing protocols such as ospf @xcite .
compute an auxiliary edge weight @xmath84 ; compute a shortest path @xmath82 from @xmath54 to @xmath9 with the auxiliary edge weight function @xmath85 the path @xmath86 found by algorithm [ alg : greedy ] is a @xmath0-approximation to omcr . in other words , @xmath87 @xmath88 , where @xmath73 is the optimal value of omcr .
@xmath57 proof : since @xmath73 is the optimal value of omcr , there exists an path @xmath71 such that @xmath89 .
this means that @xmath90 ( [ eqn:1 ] ) can be presented as @xmath91 summing ( [ eqn:2 ] ) over @xmath0 constraints , we have @xmath92 since @xmath93 , we have @xmath94 since @xmath86 is the shortest path with respect to edge weight function @xmath85 , we have @xmath95 .
therefore , @xmath96 since @xmath97 , we have @xmath98 therefore , we know that @xmath99 , and consequently , that @xmath86 is a @xmath0-approximation to omcr .
in this section , we use numerical simulations to demonstrate the proposed mechanism and algorithm in this paper . the pjm five - bus system @xcite is used for simulations , as illustrated in fig . [
fig : pjm ] .
the mapping between lmp and load ( one curve for each bus ) is given in table [ tab : curve ] ( the first column shows the lower boundary of the corresponding load interval @xmath100 ) @xcite . .lmp
( $ /mwh ) versus load ( mw ) [ cols="^,^,^,^,^,^",options="header " , ] we assume that utility function is @xmath101 and the price for communication delay is @xmath102 .
note that these functions are chosen arbitrarily for illustrative purpose . for practical systems
, they can be estimated from historical data .
[ fig : delay ] shows the curves of cost versus different delays ( measured in time slots ) for homes served by the five buses , respectively .
we observe that , for some buses , the cost increases monotonically with delay while the minimal cost is not achieved by the minimal delay for other cases . comparing the results with table [ tab : curve ] , we observe that , the higher the lmp is , the more sensitive the cost is to the delay .
the curves of cost versus different outage probabilities are shown in fig .
[ fig : out ] .
again , we observe the non - monotonicity of the cost , which demonstrates the existence of the optimal requirement of outage probability
. the optimal requirements of delay and outage probabilities using the joint optimization in ( [ eq : joint ] ) are shown in figures [ fig : opt_delay ] and [ fig : opt_out ] , respectively , for various values of @xmath15 .
note that the range of the outage probability is confined between 0 and 0.1 .
we observe that there exists some fluctuation in the optimal values .
particularly , the optimal qos requirements of bus e are quite loose .
an explanation is that the power price changes marginally for bus e. therefore , home appliances served by bus e can degrade their qos requirements to avoid the cost for communication . in this section ,
we present some numerical results to show the performance of our simple greedy algorithm .
we implemented our greedy algorithm of this paper ( denoted by omcr in the figures ) , and compared it with previous sophisticated approximation algorithm fptas of @xcite ( denoted by fptas in the figures ) , which is the best approximation solution to the omcr problem .
our numerical results are presented in figs .
[ fig03a ] and [ fig03c ] , where each figure shows the average of @xmath103 runs . first , to compare the routing performance , we define the _ length _ of a found path @xmath21 is @xmath104 .
we say path @xmath105 is better than path @xmath106 is @xmath107 . in fig .
[ fig03a ] , we show the _ qualitative _ comparison of the performances of omcr and fptas using the metric of _ path length_. we set @xmath108 for fptas , which means that fptas returns a 1.5-approximation to the omcr problem . we observed that for all test cases , fptas generally provides better results . in among the 100 connections , in 30% to 43% of the test cases ( 30% for the tight scenario , 35% for the medium scenario , and 43% for the loose scenario ) , the path computed by fptas is better than the path computed by omcr
. meanwhile , in 20% to 35% of the test cases(20% for the tight scenario , 25% for the tight scenario , and 35% for the loose scenario ) , the path computed by omcr is better than the path computed by fptas .
we can conclude that omcr has similar performance as fptas .
next , we compare the running times between the omcr and fptas in fig . [
fig : runningtime ] .
as we expected , the running time of omcr is much shorter than the running time of fptas , while the two algorithms computed paths with comparable lengths .
to study the scalability of fptas and omcr with the network size , we used four more random network topologies with the following sizes : 80 nodes with 314 edges , 120 nodes with 474 edges , 140 nodes with 560 edges , 160 nodes with 634 edges , to test the computational scalability of the algorithms .
here we have used @xmath109 and medium scenario for these test cases .
the running times of these two algorithms are shown in fig .
[ fig03c ] .
we can see that the running time of fptas increased dramatically with the increased network size .
meanwhile , omcr requested much less amount of time and is not affect much by the size of the networks .
this proves that our solution will be more adaptable in fast - developing smart grid enviroment .
we have addressed the qos routing in smart grid . to derive the qos requirement ,
we have analyzed the dynamics of power market and the impact of communication metrics like delay and outage on the revenue of home appliances .
then , we model the qos derivation as an optimization problem that maximizes the total reward .
based on the derived qos requirement , a simple greedy routing algorithm has been applied to secure the qos and address the strict realtime requirement .
we have shown that the proposed algorithm is a @xmath0-approximation .
we have run numerical simulations which demonstrated the effectiveness of the proposed mechanism and algorithm .
the authors would like to thank prof .
fangxing li in the dept .
of eecs in the university of tennessee , knoxville , for the discussion on the load - price mapping in power market .
|
smart grid is an emerging technology which is able to control the power load via price signaling .
the communication between the power supplier and power customers is a key issue in smart grid .
performance degradation like delay or outage may cause significant impact on the stability of the pricing based control and thus the reward of smart grid .
therefore , a qos mechanism is proposed for the communication system in smart grid , which incorporates the derivation of qos requirement and applies qos routing in the communication network . for deriving the qos requirement ,
the dynamics of power load and the load - price mapping are studied .
the corresponding impacts of different qos metrics like delay are analyzed .
then , the qos is derived via an optimization problem that maximizes the total revenue . based on the derived qos requirement ,
a simple greedy qos routing algorithm is proposed for the requirement of high speed routing in smart grid .
it is also proven that the proposed greedy algorithm is a @xmath0-approximation .
numerical simulation shows that the proposed mechanism and algorithm can effectively derive and secure the communication qos in smart grid .
|
learning in a single layer of feed forward neural network with binary synapses has been studied either based on statistical mechanics analysis @xcite or in algorithmic aspects @xcite .
this network can learn an extensive number @xmath0 of random patterns , where @xmath1 is the number of synapses and @xmath2 denotes the constraint density .
the critical @xmath2 ( also called the capacity ) separating the learnable phase from unlearnable phase is predicted to be @xmath3 where the entropy vanishes @xcite .
a solution is defined as a configuration of synaptic weights to implement the correct classification of @xmath4 random input patterns . above @xmath5
, no solutions can be found with high probability ( converging to @xmath6 in the thermodynamic limit ) . the replica symmetric solution presented in ref .
@xcite has been shown to be stable up to the capacity , which is in accordance with the convexity of the solution space @xcite .
note that the solutions disappear at the threshold @xmath5 still maintaining a typical finite value of hamming distance between them , which is quite distinct from the case in the continuous perceptron with real - valued synapses . in the continuous perceptron
, this distance tends to zero when the solutions disappear at the corresponding threshold @xcite . on the other hand , many local search algorithms @xcite were proposed to find solutions of the perceptron learning problem , however , the search process slows down with increasing @xmath2 , and the critical @xmath2 for the local search algorithm @xcite decreases when the number of synapses increases .
this typical behavior of the stochastic local search algorithm is conjectured to be related to the geometrical organization of the solution space @xcite . in order to acquire a better understanding for the failure of the local search strategy
, we compute the entropy landscape both from a reference configuration and for solution - pairs with a given distance in the solution space .
both distance landscapes contain rich information about the detailed structure of the solution space and then can help us understand the observed glassy behavior of the local search algorithms . throughout the paper
, the term distance refers to the hamming distance .
the distance landscape has been well studied in random constraint satisfaction problems defined on diluted or sparse random graphs @xcite .
learning in the binary perceptron can be mapped onto a bipartite graph where variable node represents synaptic weight and function node represents the input random pattern to be learned ( see figure [ perc ] ( b ) ) .
this graph is also called graphical model or factor graph @xcite .
the efficient message passing learning algorithm for the binary perceptronal learning problem has been derived using the cavity method and this factor graph representation @xcite . in this paper
, we focus on the typical property of the solution space in random ensembles of the binary perceptronal learning problem .
we apply the replica trick widely used to study disordered systems @xcite to compute the statistical properties in the thermodynamic limit . to confirm the mean field result computed using the replica approach , we derive the message passing equations in the cavity context which can be applied on single random instances of the current problem . in this context
, we apply the decorrelation assumption as well as the central limit theorem to derive the formula at the replica symmetric level .
this assumption arises from the weak correlation among synaptic weights ( within one pure state @xcite ) and among input patterns @xcite .
the efficiency of the inspired message passing algorithms in loopy systems has been observed in refs .
@xcite while the underlying mechanism still needs to be fully understood .
however , our cavity method focuses on the physical content and yields the same result as that obtained using replica approach @xcite .
the remainder of this paper is organized as follows .
the random classification by the binary perceptron is defined in sec .
[ sec_bperc ] . in sec .
[ sec_sdr ] , we derive the self - consistent equations to compute the distance landscape ( entropy landscape ) from a reference configuration , i.e. , to count the number of solutions at a distance from the reference configuration .
both the annealed and replica symmetric ( rs ) computations of this entropy landscape are presented .
we also derive the message passing equations for single instances in this section using the cavity method and factor graph representation . in sec .
[ sec : sdp ] , the landscape of hamming distances between pairs of solutions is evaluated at both annealed approximation and rs ansatz , and the associated message passing equations are proposed as well .
discussion and conclusion are given in sec .
[ sec_con ] .
input units ( open circles ) feed directly to a single output unit ( solid circle ) .
a binary input pattern @xmath7 of length @xmath1 is mapped through a sign function to a binary output @xmath8 , i.e. , @xmath9 .
the set of @xmath1 binary synaptic weights @xmath10 is regarded as a solution of the perceptron problem if the output @xmath11 for each of the @xmath12 input patterns @xmath13 $ ] , where @xmath14 is a preset binary value .
( b ) each circle denotes the variable node whose value takes @xmath15 with @xmath16 its index .
the square is the function node denoting a random binary pattern to be learned . if the pattern is learned by the synaptic vector @xmath17 , the value of the corresponding function node takes zero .
the dotted line represents other @xmath18 function nodes while the dashed line for variable node @xmath16 means @xmath16 is connected to other @xmath18 function nodes and that for function node means the function node ( e.g. , @xmath19 ) is connected to other @xmath20 variable nodes .
, title="fig : " ] .5 cm input units ( open circles ) feed directly to a single output unit ( solid circle ) .
a binary input pattern @xmath7 of length @xmath1 is mapped through a sign function to a binary output @xmath8 , i.e. , @xmath9 .
the set of @xmath1 binary synaptic weights @xmath10 is regarded as a solution of the perceptron problem if the output @xmath11 for each of the @xmath12 input patterns @xmath13 $ ] , where @xmath14 is a preset binary value .
( b ) each circle denotes the variable node whose value takes @xmath15 with @xmath16 its index .
the square is the function node denoting a random binary pattern to be learned .
if the pattern is learned by the synaptic vector @xmath17 , the value of the corresponding function node takes zero .
the dotted line represents other @xmath18 function nodes while the dashed line for variable node @xmath16 means @xmath16 is connected to other @xmath18 function nodes and that for function node means the function node ( e.g. , @xmath19 ) is connected to other @xmath20 variable nodes .
, title="fig:"].2 cm
the binary perceptron realizes a random classification of @xmath4 random input patterns ( see figure [ perc](a ) ) . to be more precise
, the learning task is to find an optimal set of binary synaptic weights ( solution ) @xmath21 that could map correctly each of random input patterns @xmath22 to the desired output @xmath23 which is assigned a value @xmath24 at random .
@xmath4 is proportional to @xmath1 with the coefficient @xmath2 defining the constraint density ( each input pattern serves as a constraint for all synaptic weights , see figure [ perc ] ( b ) ) .
the critical value is @xmath3 below which the solution space is non - empty @xcite . given the input pattern @xmath25 , the actual output @xmath8 of the perceptron is @xmath26 where @xmath15 takes @xmath24 and @xmath27 takes @xmath24 with equal probabilities . if @xmath28 , we say that the synaptic weight vector @xmath17 has learned the @xmath29-th pattern .
therefore we define the number of patterns mapped incorrectly as the energy cost @xmath30 where @xmath31 is a step function with the convention that @xmath32 if @xmath33 and @xmath34 otherwise . the prefactor @xmath35 is introduced to ensure that the argument of the step function remains at the order of unity , for the sake of the following statistical mechanical analysis in the thermodynamic limit . in the current setting , both @xmath36 and the desired output @xmath37 are generated randomly independently . without loss of generality ,
we assume @xmath38 for any input pattern in the remaining part of this paper , since one can perform a gauge transformation @xmath39 to each input pattern without affecting the result .
in this section , we consider the entropy landscape from a reference configuration ( which is not a solution ) . this entropy counts the number of solutions at a distance @xmath40 from the reference configuration @xmath41 .
the behavior of this entropy landscape reflects the geometrical organization of the solution space .
since we concentrate on the ground state ( @xmath42 ) , we take the inverse temperature @xmath43 and introduce a coupling field @xmath44 to control the distance between solutions and the reference configuration . the partition function for this setting
is @xmath45\ ] ] where the sum @xmath46 goes over all possible synaptic weight vectors and @xmath47 means the sum over all variable nodes . under the definition of the overlap @xmath48
, the partition function can be written as @xmath49\ ] ] where @xmath50 is the number of solutions with the overlap @xmath51 . in the thermodynamic limit @xmath52
, the saddle point analysis leads to @xmath53 $ ] where @xmath54 is defined as the free energy density .
therefore , we can determine the entropy @xmath55 by a legendre transform @xcite @xmath56,\\ \tilde{q}(x)&=&\frac{df(x)}{dx}\label{legendre02}\end{aligned}\ ] ] where @xmath51 is related to @xmath57 through @xmath58 and then the entropy density can be expressed as a function of the distance @xmath57 which can be understood as the probability that a synaptic weight takes different values in @xmath17 and @xmath41 .
one recovers the total number of solutions by setting @xmath59 in eq .
( [ ptf ] ) .
we first calculate the annealed entropy density @xmath61 which serves as the upper bound ( jensen s inequality @xcite ) for the true value of the entropy density .
actually , the free energy @xmath62 should be averaged over the random input patterns .
however , the annealed approximation alternatively performs the average of the partition function first and then takes the logarithmic operation as @xmath63 where the average is taken over the distribution of the random input patterns .
this can be computed as @xmath64\right > \nonumber\\ & = & \int d\tilde{q}\int\frac{d\hat{q}}{2\pi \mathrm{i}/n}\exp\left[n\left(-\hat{q}\tilde{q}+x\tilde{q}-\alpha\log2+\log(2\cosh\hat{q})\right)\right]\end{aligned}\ ] ] where the integral representation of @xmath65 is used and the conjugated counterpart @xmath66 of the overlap @xmath51 is introduced as a dirac delta function @xmath67 is inserted @xcite .
a saddle point analysis results in @xmath68 where the saddle point equation reads @xmath69 . using eq .
( [ legendre01 ] ) and the saddle point equation , we get the annealed entropy density @xmath70 the free energy density @xmath54 is a self - averaging quantity whose value concentrates in probability around its expectation in the thermodynamic limit @xcite , and its average over the random input patterns is very difficult to compute because the logarithm appears inside the average .
the replica trick bypasses this difficulty by using the identity @xmath71 .
then the disorder averaged free energy density can be computed by first averaging an integer power of the partition function and then letting @xmath72 as @xmath73 although the replica method is not generally rigorous , the obtained theoretical result can be checked by numerical simulations .
to compute @xmath74 , we introduce @xmath75 replicated synaptic weight vectors @xmath76(@xmath77 ) as follows .
@xmath78\right>\nonumber\\ \fl=\int\prod_{a < b}\frac{dq^{ab}d\hat{q}^{ab}}{2\pi \mathrm{i}/n}\exp\left[-n\sum_{a < b}q^{ab}\hat{q}^{ab}+n\alpha\log g_{0}(\{q^{ab}\})+ng_{1}(\{\hat{q}^{ab}\})\right],\end{aligned}\ ] ] where @xmath79 and @xmath80 are expressed respectively as @xmath81e^{\mathrm{i}\sum_{a}\lambda^{a}t^{a}-\sum_{a < b}\lambda^{a}\lambda^{b}q^{ab}-\frac{1}{2}\sum_{a}(\lambda^{a})^{2}},\\ g_{1}(\{\hat{q}^{ab}\})&=&\log\sum_{j^{a}:a=1,\ldots , n}e^{\sum_{a < b}\hat{q}^{ab}j^{a}j^{b}+x\sum_{a}j^{a}j^{*}},\end{aligned}\ ] ] where we have introduced the replica overlap @xmath82 and its associated conjugated counterpart @xmath83 .
the replica symmetric ansatz assumes @xmath84 for @xmath85 . now using the saddle point analysis
, we finally arrive at the formula of the free energy density and the corresponding saddle point equations , @xmath86,\\ q&=&\int dz\tanh^{2}(\sqrt{\hat{q}}z+x),\label{repf03a}\\ \hat{q}&=&\frac{\alpha}{1-q}\int dz\left[g\left(\sqrt{\frac{q}{1-q}}z\right)/h\left(\sqrt{\frac{q}{1-q}}z\right)\right]^{2},\label{repf03b}\end{aligned}\ ] ] where @xmath87 and @xmath88 with the gaussian measure @xmath89 .
after the fixed point of the self - consistent equations ( [ repf03a ] ) and ( [ repf03b ] ) is obtained , the entropy landscape @xmath60 is computed as @xmath90 note that the final expression of @xmath91 does not depend on the reference configuration and the integral in the second term of eq .
( [ sdrep ] ) is @xmath92 defined in eq .
( [ legendre02 ] ) . in this section ,
we derive the message passing equations to compute the entropy landscape for single instances under the replica symmetric ansatz . to derive the self - consistent equation
, we apply the cavity method @xcite and first define two kinds of cavity probabilities .
one is the probability @xmath93 that variable node @xmath16 in figure [ perc ] ( b ) takes value @xmath15 in the absence of constraint @xmath94 .
the other is @xmath95 staying for the probability that constraint @xmath19 is satisfied ( pattern @xmath96 is learned ) if synaptic weight @xmath16 takes @xmath15 .
according to the above definitions , the self - consistent equation for these two kinds of probabilities is readily obtained as @xmath97 where @xmath98 is a normalization constant , @xmath99 denotes the neighbors of node @xmath16 except constraint @xmath94 and @xmath100 denotes the neighbors of constraint @xmath19 except variable node @xmath16 .
( [ bp0 ] ) and ( [ bp1 ] ) are actually the belief propagation equations @xcite . for the binary perceptron , directly solving the belief propagation equations is impossible . to reduce the computational complexity , we define @xmath101 .
note that the sum involves @xmath102 independent random terms , as a result , the central limit theorem implies that @xmath103 follows a gaussian distribution with mean @xmath104 and variance @xmath105 where @xmath106 and @xmath107 . within the rs ansatz , the clustering property @xmath108 for @xmath109 in the thermodynamic limit
is used to get the variance @xcite .
@xmath110 is the magnetization in statistical physics language . by separating the term @xmath111 from the sum in the @xmath65 of eq .
( [ bp1 ] ) , and approximating the sum @xmath112 by an integral over @xmath103 , we get finally @xmath113 where @xmath114 and @xmath115 in which the cavity magnetization @xmath116 . using eqs .
( [ bp0 ] ) and ( [ phat ] ) , the cavity field @xmath117 can be obtained in the log - likelihood representation @xmath118.\label{cavh02}\end{aligned}\ ] ] notice that the cavity bias @xmath119 can be approximated by @xmath120 in the large @xmath1 limit .
( [ cavh01 ] ) and ( [ cavh02 ] ) constitute the recursive equations to compute the free energy density in the bethe approximation @xcite @xmath121,\\ \delta f_{a}&=&\log h\left(-\frac{\hat{w}_{a}}{\sqrt{\hat{\sigma}_{a}}}\right),\end{aligned}\ ] ] where @xmath122 and @xmath123 are the free energy shifts due to variable node ( @xmath16 ) addition ( and all its function nodes ) and function node ( @xmath94 ) addition @xcite respectively .
actually @xmath124 is the normalization constant of the full probability @xmath125 and @xmath126 the normalization constant of @xmath127 @xcite . @xmath128 and @xmath129 .
equations ( [ cavh01 ] ) and ( [ cavh02 ] ) can be solved by an iterative procedure with a random initialization of the corresponding messages .
after the iteration converges , the entropy landscape @xmath60 from the fixed reference configuration @xmath41 can be computed according to the legendre transform eqs .
( [ legendre01 ] ) and ( [ legendre02 ] ) .
the computational complexity is of the order @xmath130 for this densely connected graphical model .
@xmath54 computed based on eq .
( [ frs01 ] ) does not depend on the reference configuration since the change of @xmath131 does not affect the final result , consistent with eq .
( [ repf03 ] ) . ) ) for @xmath132 ( from the top to the bottom ) respectively .
the horizontal dashed line indicates the zero entropy value . the line connecting symbols is a guide to the eye .
the empty symbols stay for the numerical simulation results on systems with @xmath133 ( from the top to the bottom ) using message passing algorithms .
the result is the average over @xmath134 random instances .
solid symbols are the replica symmetric results computed numerically by solving the saddle point equations .
, scaledwidth=80.0% ] the distance landscape from a reference configuration is reported in figure [ sdr ] .
we choose the reference configuration @xmath135 for simplicity .
other choices of the reference configuration still yield the same behavior of the landscape .
note that the annealed entropy provides an upper bound for the rs one , and it roughly coincides with the rs one at low @xmath2 ( around the maximal point ) while the large deviation is observed when @xmath2 further increases .
it is clear that most of the solutions concentrate around the dominant point where the maximum of the entropy is reached .
when the given distance is larger or less than certain values ( @xmath136 or @xmath137 ) , the number of solutions at those distances becomes exponentially small in @xmath1 . in the intermediate range ( @xmath138 $ ] ) , as the distance increases , the number of solutions separated by the distance from the reference point in the solution space increases first and then reaches the maximum which dominates the typical value of the entropy in the original systems ( by setting @xmath59 , see figure [ sd ] ) .
the maximum is then followed by a decreasing trend as the distance is further increased . this mean field behavior is confirmed by the numerical simulations on large - size single random instances using the message passing algorithms derived in sec . [ subsec : mpa01 ] .
the consistency between the mean field result obtained by replica approach and the simulation result obtained on single random instances is clearly shown in figure [ sdr ] .
the bell shape in figure [ sdr ] is similar to that observed in calculating the growth rate of expected distance enumerator for a random code ensemble @xcite .
note that as the constraint density increases , the distance range where solutions exist shrinks , which illustrates clearly how the solution space changes as more patterns are presented to the binary perceptron .
we also compute the typical value of the entropy in the original system ( by setting @xmath59 ) and of the distance between any two solutions as a function of the constraint density using the replica method .
the result is reported in figure [ sd ] . here
we define the typical value of the distance between any two solutions as @xmath139 where @xmath140 is obtained from the stationary value of eq .
( [ repf03a ] ) .
the entropy vanishes at @xmath3 with a finite typical value of distance @xcite .
this typical distance is also in accordance with that computed on single instances by sampling a finite number of solutions @xcite .
note that this distance is evaluated here based on the rs ansatz .
one can further check its stability by the population dynamics @xcite on the one - step replica symmetry breaking ( @xmath6rsb ) solution @xcite where we define two typical distances : one is inter - cluster distance @xmath141}{2}$ ] where @xmath142 means the average over clusters and @xmath143 $ ] over the disorder , and the other is intra - cluster distance defined by @xmath144}{2}$ ] @xcite where @xmath145 is the local field defined in eq .
( [ cavh01 ] ) by including all contributions of patterns around the synaptic weight ( @xmath59 ) . in general , solutions within a single cluster
are separated by a sub - extensive number of synaptic weights while any two clusters are separated by an extensive number of synaptic weights .
our numerical simulations confirmed that @xmath146 and @xmath147 will turn out to be identical ( equal to @xmath148 ) after sufficient iterations implying that the rs ansatz is unbroken below the capacity .
however , for constraint density close to the capacity , one needs a much larger sampling interval ( by way of the metropolis importance sampling method ) @xcite in the population dynamics algorithm . to probe the fine structure of the connection pattern in the solution space
, we study the distance landscape of solution - pairs in the following section .
this is similar to the study of the spherical @xmath149-spin model in the presence of an attractive coupling with quenched configuration in ref .
@xcite , however , the rich information about the solution space structure of the current problem can also be attained by calculating the distance landscape of solution - pairs .
the geometrical property of the solution space can also be studied by counting the number of solution - pairs with a predefined distance @xmath57 , equivalently an overlap of value @xmath51 .
actually this entropy value may be much larger than the entropy density of the original problem ( which is obtained by setting @xmath59 in eq .
( [ frs01 ] ) ) .
as we shall present later , this case becomes more involved for the binary perceptron with an increasing computational cost .
considering distance between solutions , we write the partition function as @xmath150\ ] ] where the coupling field @xmath44 is used to control the distance between a pair of solutions @xmath151 and the associated overlap @xmath152 .
this partition function has been used to predict optimal coupling field for a multiple random walking strategy to find a solution for the perceptronal learning problem @xcite . in the following sections , we present an annealed computation as well as rs computation of the distance landscape @xmath60 . note that in this setting , eqs .
( [ ptf02 ] ) , ( [ legendre01 ] ) and ( [ legendre02 ] ) can also be used but here @xmath57 should be understood as the distance separating two solutions in the weight space . following the same techniques used in sec .
[ subsec : ann01 ] , we obtain the annealed free energy density as ( see also ref .
@xcite ) @xmath153 the maximization with respect to @xmath51 and @xmath66 leads to the following saddle point equation @xmath154 where the identity @xmath155 has been used . using eq .
( [ legendre01 ] ) and the above saddle point equation , we get the final expression for @xmath156 : @xmath157,\label{anns02}\end{aligned}\ ] ] where @xmath158 which is actually the annealed entropy density of the original system @xcite . if @xmath159 , then @xmath160 which is in accord with the fact that the number of solution - pairs with a distance @xmath161 should be the total number of solutions @xmath162 if no constraints are present .
in this section , we derive the free energy density @xmath54 for the landscape of solution - pairs under the replica symmetric approximation , using the replica trick introduced in sec .
[ subsec : rs01 ] .
since the partition function in this case involves a sum of all possible configurations of two synaptic weight vectors , the computation becomes a bit complicated .
the disorder average of the integer power of the partition function can be evaluated as @xmath163 where @xmath164 and @xmath165 . under the replica symmetric ansatz ,
the disorder average is carried out as @xmath166^{n},\ ] ] where @xmath167 , @xmath168 and we have used @xmath169 under the rs ansatz . after the computation of the summation in eq .
( [ repair01 ] ) by using @xmath170 and the hubbard - stratonovich transform , we get the replica symmetric free energy density @xmath171 where @xmath172 , @xmath173 and @xmath174 where @xmath175 .
the rs order parameters @xmath176 are determined by the following self - consistent equations @xmath177^{2}},\\ \fl\hat{q}&=\frac{\hat{r}}{2}+\frac{\alpha}{2(1-q - r+r)}\int d\boldsymbol{z}\left[\frac{\int dy\left[g(y_{1})h(y_{2})- g(y_{2})h(y_{1})\right]}{\int dy h(y_{1})h(y_{2})}\right]^{2},\label{pairb}\\ \fl r&=\int d\boldsymbol{\hat{z}}\frac{\tanh a_{3}+\tanh a_{1}\tanh a_{2}}{1+\tanh a_{3}\tanh a_{1}\tanh a_{2}},\label{pairr}\\ \fl r&=\int d\boldsymbol{\hat{z}}\frac{\tanh a_{3}(\tanh^{2}a_{1}+\tanh^{2}a_{2})+\tanh a_{1}\tanh a_{2}(1+\tanh^{2}a_{3})}{(1+\tanh a_{3}\tanh a_{1}\tanh a_{2})^{2}},\\ \fl q&=r+\int d\boldsymbol{\hat{z}}\frac{(\tanh a_{3}-1)^{2}(\tanh a_{1}-\tanh a_{2})^{2}}{2(1+\tanh a_{3}\tanh a_{1}\tanh a_{2})^{2}}.\end{aligned}\ ] ] in the derivation of the above saddle point equations , we have used a useful property of the gaussian measure @xmath178 where @xmath179 is the derivative of the function @xmath180 with respect to @xmath181 .
after the fixed point of the above saddle point equations is obtained , one can compute the entropy density @xmath182 with @xmath183 .
note that @xmath184 may become negative , in this case we replace @xmath185 and @xmath186 by @xmath187 and @xmath188 respectively , @xmath189 by @xmath190 only for @xmath191 in eqs .
( [ paira ] ) to ( [ pairb ] ) . by analogy with definitions in sec .
[ subsec : mpa01 ] , we define by @xmath192 the probability that the synaptic weight @xmath16 takes a two - component vector state @xmath193 in the absence of constraint @xmath94 and by @xmath194 the probability that constraint @xmath19 is satisfied given the vector state @xmath193 of weight @xmath16 .
these two cavity probabilities obey the following recursive equations @xmath195 where @xmath98 is a normalization constant .
in fact the belief propagation equations ( [ bp2 ] ) and ( [ bp3 ] ) correspond to the stationary point of the bethe free energy function of the current system @xcite .
the exchange of @xmath196 and @xmath197 does not change the partition function eq .
( [ ptf03 ] ) , thus the cavity probabilities have the property that @xmath198 and @xmath199 .
this symmetry property will simplify the following analysis a lot . to simplify eqs .
( [ bp2 ] ) and ( [ bp3 ] ) , we need the joint distribution of @xmath200 and @xmath201 where @xmath202 and @xmath203 . since we impose a distance constraint upon two solutions @xmath196 and @xmath197 in eq .
( [ ptf03 ] ) , there exists correlation between these two normally distributed random numbers and this correlation is characterized by the correlation coefficient @xmath204 due to the symmetry property . based on eq .
( [ bp2 ] ) , messages @xmath205 and @xmath206 are determined respectively by the following equations , @xmath207 - 2e^{-x}\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{+1,-1}}{e^{x}\left[\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{+1,+1}+\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{-1,-1}\right]+2e^{-x}\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{+1,-1}},\label{q}\\ m_{j\rightarrow b}&=&\frac{e^{x}\left[\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{+1,+1}-\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{-1,-1}\right]}{e^{x}\left[\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{+1,+1}+\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{-1,-1}\right]+2e^{-x}\prod_{a\in\partial j\backslash b}\hat{p}_{a\rightarrow j}^{+1,-1}}.\label{cavh03}\end{aligned}\ ] ] therefore , both @xmath200 and @xmath201 obey a bivariate normal distribution and @xmath194 is reduced to be @xmath208 where @xmath114 and @xmath115 .
the overlap @xmath51 is determined by @xmath209 where @xmath210 is given by @xmath211 - 2e^{-x}\prod_{b\in\partial i}\hat{p}_{b\rightarrow i}^{+1,-1}}{e^{x}\left[\prod_{b\in\partial i}\hat{p}_{b\rightarrow i}^{+1,+1}+\prod_{b\in\partial i}\hat{p}_{b\rightarrow i}^{-1,-1}\right]+2e^{-x}\prod_{b\in\partial i}\hat{p}_{b\rightarrow i}^{+1,-1}}.\ ] ] eq .
( [ phat02 ] ) is more computationally demanding than eq .
( [ phat ] ) since an additional numerical integral is required to compute @xmath212 here .
however , the integral in eq .
( [ phat02 ] ) can be approximated by @xmath213 if we write the right hand side of eq .
( [ phat02 ] ) as @xmath214 and expand @xmath215 up to the second order in @xmath216 .
the constants @xmath217 , @xmath218 and @xmath219 can be determined as a function of @xmath94 and @xmath19 .
therefore , this approximation is accurate only when large @xmath216 has vanishing contribution to the integral
. the free energy shift due to variable node addition ( and all its adjacent constraints ) can be obtained as @xmath122 and the free energy shift due to constraint addition @xmath220 where @xmath221 + 2e^{-x}\prod_{b\in\partial i}\hat{p}_{b\rightarrow i}^{+1,-1},\\ z_{a}&=&\int_{-\frac{\hat{w}_{a}}{\sqrt{\hat{\sigma}_{a}}}}^{\infty}dth\left(-\frac{\hat{w}_{a}}{\sqrt{(1-\hat{\rho}_{a}^{2})\hat{\sigma}_{a}}}-\frac{\hat{\rho}_{a}t}{\sqrt{1-\hat{\rho}_{a}^{2}}}\right),\end{aligned}\ ] ] where @xmath222 , @xmath223 and @xmath224 .
the free energy density can then be obtained using eq .
( [ frs01 ] ) and the entropy landscape can be obtained correspondingly .
the recursive equations eqs .
( [ rho ] ) , ( [ q ] ) , ( [ cavh03 ] ) and ( [ phat02 ] ) can be solved by an iterative procedure similar to that used in sec .
[ subsec : mpa01 ] . .
the solid lines are the analytic annealed approximation ( eq .
( [ anns02 ] ) ) for @xmath132 ( from the top to the bottom ) respectively .
the line connecting symbols is a guide to the eye .
the empty symbols stay for the numerical simulation results on systems with @xmath225 ( from the top to the bottom ) using message passing algorithms .
the result is the average over @xmath134 random instances .
solid symbols are the replica symmetric results computed numerically by solving the saddle point equations . , scaledwidth=80.0% ] as a function of the constraint density .
the data points are computed by solving numerically the saddle point equations in sec .
[ subsec : rs02 ] .
the error bars characterize the numerical fluctuations from ten different random initializations .
the upper inset shows the corresponding entropy values .
the lower inset shows _ schematically _ the typical concave and non - concave behavior of the entropy landscape , where the horizontal dashed line indicates @xmath226 and the vertical dashed line denotes @xmath227 , and the black point @xmath228 marks the first change of the concavity , i.e. , @xmath229 .
the vertical dotted line marks the first order thermodynamic transition point @xmath230 , where the dash - dotted line going through @xmath231 touches the concave part of @xmath60 and has the slope @xmath232 .
note that @xmath227 corresponds to the spinodal point @xmath233 ( @xmath234 ) , i.e. , @xmath235 .
, scaledwidth=80.0% ] for fixed @xmath57 .
the data points are computed by solving numerically the saddle point equations in sec .
[ subsec : rs02 ] .
the error bars characterize the numerical fluctuations from ten different random initializations .
inset : the corresponding entropy curve as a function of @xmath57 .
( a ) @xmath236 .
( b ) @xmath237 . , title="fig:",scaledwidth=80.0% ] .05 cm for fixed @xmath57 .
the data points are computed by solving numerically the saddle point equations in sec .
[ subsec : rs02 ] .
the error bars characterize the numerical fluctuations from ten different random initializations .
inset : the corresponding entropy curve as a function of @xmath57 .
( a ) @xmath236 .
( b ) @xmath237 .
, title="fig:",scaledwidth=80.0%].05 cm as is seen from figure [ sdp ] , the entropy density increases smoothly until a maximum is reached for @xmath238 and then decreases as the distance further grows .
interestingly , this behavior observed in figure [ sdp ] can be well fitted by the annealed approximation keeping the concavity of the entropy function .
however , as @xmath2 increases , large deviation from the annealed approximation occurs .
the mean field calculations are supported by the numerical simulations on single instances using the proposed message passing algorithms , as shown in figure [ sdp ] .
the distance corresponding to the maximum of the entropy landscape curve in figure [ sdp ] is actually the typical distance @xmath148 calculated in figure [ sd ] , and @xmath239 recovers the typical entropy density of the original problem . by taking the limit @xmath240 in eq .
( [ repair02 ] ) , one can show that @xmath241 where @xmath54 is given by eq .
( [ repf03 ] ) . as the constraint density increases ,
the maximal point of the entropy curve moves to the left , however , solution - pairs still maintain a relatively broad distribution in the solution space when @xmath2 approaches @xmath5 , e.g. , @xmath242 at @xmath243 ( @xmath244 ) , which may be responsible for the algorithmic hardness in this region .
as @xmath2 increases , the message passing algorithm requires a large number of iteration steps to converge ( especially at small distances ) and additionally a computationally expensive monte carlo integral involved in eq .
( [ phat02 ] ) can not be avoided . on the other hand ,
when @xmath2 is large enough , one can easily observe a rapid growth of the order parameter @xmath185 to unity , i.e. , at some critical coupling field @xmath245 , @xmath185 changes sharply from a value smaller than one to one .
this implies that at @xmath245 , @xmath246 becomes a globally stable solution of the saddle point equations in sec .
[ subsec : rs02 ] .
the first order thermodynamic transition is signalled by the change of the concavity at @xmath247 .
we define @xmath227 as the minimal distance before @xmath246 becomes a unique stable solution .
figure [ gap ] shows the entropy gap and @xmath227 as a function of the constraint density . the corresponding coupling field
@xmath233 marks the point where the concavity starts to change , i.e. , @xmath248 , as shown in the lower inset of figure [ gap ] ( @xmath249 ) .
after @xmath233 , @xmath246 becomes the unique stable solution of the saddle point equations .
note that in the entropy gap , there exists a non - concave part of the entropy curve ( for small distances ) , which can only be obtained by fixing @xmath57 instead of @xmath44 and searching for a compatible @xmath44 ( by the secant method ) .
the result is shown in figure [ nonconcav ] for @xmath236 and @xmath250 .
the compatible @xmath44 for small distances ( the left branch ) is smaller than @xmath233 .
when @xmath251 , the right branch is no longer globally stable solution but becomes metastable solution of the saddle point equations until @xmath252 , i.e. , the spinodal point is reached . by fixing @xmath253
, one typically observes the right branch or @xmath246 , which describes the equilibrium properties of the boltzmann measure in eq .
( [ ptf03 ] ) .
thus , the non - concave behavior observed in @xmath254 is thermodynamically non - dominant and unstable , suggesting that the solution space is made of isolated solutions instead of separated clusters of exponentially many close - by solutions , and this behavior becomes much more evident as @xmath2 increases .
this explains why the multiple random walking strategy is extremely difficult to find a solution by tuning the coupling field at high @xmath2 and large @xmath1 @xcite .
we argue that for any finite @xmath2 , the slope of the entropy curve @xmath60 near to @xmath161 ( @xmath246 ) tends to negative infinity .
letting @xmath255 , we can obtain @xmath256 , where @xmath257 is a positive constant independent of @xmath2 and @xmath258 , and @xmath259 is a constant as well .
the derivation is given in [ sec : appendix - a ] .
thus , as long as @xmath260 , the non - concave part exists in the entropy curve for the interval @xmath261 , implying that such solution - pairs are exponentially less numerous than the typical ones .
furthermore , @xmath246 is always a stable solution , and in this case @xmath262 . as shown in figure [ gap]@xmath263 increases as @xmath2 grows , making a uniform sampling of solutions extremely hard .
in addition , @xmath227 seems to grow continuously , being the order of @xmath264 or less for @xmath265 .
the isolations of solutions can be explained by the nature of the hard constraints @xcite for the binary perceptron .
unlike the random @xmath266-sat and graph - coloring problems @xcite , the hard constraint in the binary perceptron problem implies that the synaptic weight on one node in the factor graph is completely determined by the values of other nodes .
but for finite @xmath1 , solutions may not be strictly isolated .
this explains why some local search heuristics can find a solution when @xmath1 and @xmath2 is not large enough .
as @xmath2 increases , some frozen solutions are more prone to disappear , thus solutions become much more far apart from each other , as shown by increasing @xmath227 in figure [ gap ] .
however , the thermodynamic properties can still be derived from the rs solution before @xmath5 .
we conjecture that clustering and freezing coexist for @xmath267 , which is consistent with the computation in ref .
@xcite in the sense that the total entropy ( displayed in figure [ sd ] ) @xmath268 where @xmath269 is the complexity of clusters of entropy density @xmath270 and @xmath271 for the current problem .
the structure of the solution space is described by the dynamical one - step replica symmetry breaking scenario ( at the parisi parameter @xmath272 ) @xcite . by contrast , in the random @xmath266-xorsat , there exists a phase where going from a solution to another one involves a large but finite hamming distance @xcite and in locked constraint satisfaction problem , there exists logarithmic hamming distance separation between solutions in the liquid phase @xcite . for the binary perceptron problem
, we can say that the solution space is simple in the sense that it is made of isolated solutions instead of separated clusters of exponentially many solutions ; however , it becomes rather difficult to find a solution via stochastic local search algorithm .
below @xmath5 , @xmath273 , meaning that there exist exponentially many solutions , but they are widely dispersed ( much more apparent at large @xmath2 ) .
in other words , solution - pairs maintain a relatively broad distribution .
the typical property of the distance landscape either from a reference configuration or for pairs of solutions is studied . for the first distance landscape , as the distance increases ,
the number of associated solutions grows first and then reaches its maximum ( dominating the typical value of the entropy in the original system ) followed by a decreasing behavior .
this typical trend is confirmed by the numerical simulations on single instances using the proposed message passing algorithms .
this behavior suggests that most of the solutions concentrate around the dominant point ( the maximum in the distance landscape ) in the @xmath1-dimensional weight space .
it is clear that as the constraint density increases , the distance landscape shows larger and larger deviation from the analytic annealed approximation .
we also calculate the second distance landscape characterizing the number of solution - pairs separated by a given distance . in this case
, the replica symmetric result is in good agreement with the annealed computation at low @xmath2 , while the large deviation is observed between the replica symmetric approximation and annealed computation for high @xmath2 .
both landscapes are evaluated in the thermodynamic limit and confirmed by message passing simulations on large - size single instances . in this paper
, we calculate the whole picture of the distance ( entropy ) landscape and show that the entropy value rises to a maximum before declining at higher values of distance at certain range of distances . from the first landscape ( a random configuration as a reference ) , we see clearly how the solution space shrinks as more constraints are added . from the second landscape of solution - pairs , we deduce a picture in which each global minimum ( referred to as a canyon ) is occupied by a single solution with zero ground state energy , and is surrounded by local minima ( referred to as valleys ) with positive energy @xcite .
this is also known as the valleys - dominated energy landscape @xcite .
the isolation of solutions implies that one can not expect to satisfy all constraints by flipping a few synapses .
the necessary number of synapses to be flipped should be proportional to @xmath1 .
the distance between the isolated solutions increases as the constraint density grows .
this is the very reason why some simple local search heuristics can not find a solution at high @xmath2 or large @xmath1 @xcite and the critical @xmath2 for the local search algorithm decreases when the number of synapses increases .
simulated annealing process used in refs .
@xcite suffers from a critical slowing down when approaching a certain temperature , therefore , it would be interesting to study this picture within a finite temperature framework by focusing on the metastable states around the isolated solutions .
the structure of these states should also be responsible for the algorithmic hardness .
this issue will be addressed in our future work .
the distance landscape evaluated here is very similar to the weight enumerator in coding theory @xcite and the method can be extended to consider the landscape analysis for low - density parity - check codes or code - division multiple access multiuser detection problems @xcite , which will help to clarify what role the distance landscape plays with respect to the decoding performance .
we are grateful to haijun zhou for helpful comments on earlier versions of this paper .
this work was partially supported by the research council of hong kong ( grant nos .
hkust @xmath274 , @xmath275)(mw , hh ) and the jsps fellowship for foreign researchers grant no .
@xmath276 ( hh ) and jsps / mext kakenhi nos .
@xmath277 , @xmath278 , and @xmath279 ( yk ) .
in this section , we give a derivation of @xmath282 in the limit of @xmath281 . by noting that @xmath283 , one can write the derivative as @xmath284 where @xmath285 and @xmath286 is determined by eq .
( [ pairr ] ) .
we know that @xmath281 implies that @xmath287 , and let @xmath288 in eq .
( [ pairr ] ) , we can then obtain @xmath289 to derive eq .
( [ derr01 ] ) , we have used @xmath290 @xmath291 . when @xmath281 , @xmath292 and @xmath293 , thus the first term in eq .
( [ derr ] ) can be reexpressed in the limit @xmath281 as @xmath294\times \epsilon^{-1/2}.\ ] ] therefore , @xmath295 where the constants can be obtained from the above equations .
note that @xmath257 is positive , and the first term in the derivative dominates the divergent behavior when @xmath296 .
following the same line , one can also prove that @xmath297 reduces to @xmath298 in the limit of @xmath240 .
|
the statistical picture of the solution space for a binary perceptron is studied .
the binary perceptron learns a random classification of input random patterns by a set of binary synaptic weights .
the learning of this network is difficult especially when the pattern ( constraint ) density is close to the capacity , which is supposed to be intimately related to the structure of the solution space .
the geometrical organization is elucidated by the entropy landscape from a reference configuration and of solution - pairs separated by a given hamming distance in the solution space .
we evaluate the entropy at the annealed level as well as replica symmetric level and the mean field result is confirmed by the numerical simulations on single instances using the proposed message passing algorithms . from the first landscape ( a random configuration as a reference ) , we see clearly how the solution space shrinks as more constraints are added . from the second landscape of solution - pairs , we deduce the coexistence of clustering and freezing in the solution space .
|
the dynamics of multimode lasers has occupied a prime place in the investigation of laser behaviour since , starting from their first realizations @xcite , lasers operated on several longitudinal ( and at times transverse ) cavity modes .
indeed , only with further work did it become possible to restrict the operation to one longitudinal and transverse mode oscillation .
thus , questions on the correct modeling of the interaction between several ( longitudinal ) modes and the active medium were vigorously debated very early on .
one of the first and most successful proposals for a satisfactory description of laser dynamics came from tang , statz and de mars @xcite who introduced a model
widely used up until today which treats the individual cavity mode intensities as directly coupled to the energy reservoir ( population inversion ) , independently of one another .
the question of mutual coupling among modes was effectively solved in a later seminal paper @xcite . since the early work , multimode dynamics has generally been extensively studied either because of its intrinsic features , for the regular or irregular temporal oscillations which accompany it , or because of the disruption that it may cause whenever high spectral purity is required .
one particular regime has , however , attracted particular attention because of its peculiar features : the so - called antiphase regime . under this mode of operation ,
several longitudinal laser modes oscillate , but are mutually coupled in such a way as to phase - lock , thereby maintaining a constant total output power . in the simplest configuration
such a behaviour was observed in two - mode lasers @xcite , but its full impact was appreciated only with the very first attempts at building efficient frequency - doubled lasers @xcite , followed by detailed investigations of the dynamical properties of antiphase states @xcite .
the modal picture is somewhat different when one examines the operation of semiconductor lasers .
devices which are capable of multi - longitudinal mode operation ( i.e. , in - plane geometries , whether bulk , guided , or quantum well junctions ) are extremely sensitive to any form of external feedback . aside from large levels of re - injection , introduced for stabilization purposes , practically any amount of reinjection
no matter how small causes a destabilization of the laser output which manifests itself in the form of low - frequency - fluctuations @xcite , mode - hopping @xcite , coherent collapse @xcite , and fast phase jumps @xcite with antiphase oscillations during the fast transient which wash out the details of the dynamics @xcite .
an early overview @xcite of the different regimes , which predates their detailed investigations , and a summary @xcite provide a useful global picture .
noise dominates the multimode semiconductor laser dynamics and , as such , models have concentrated on the most appropriate ways of introducing its influence on the dynamical evolution .
a thorough analysis of the influence of noise on the operation of a semiconductor laser and on the associated fluctuations was reported by henry @xcite , while a review of the different physical sources of noise in a semiconductor laser has been given by osiski and buus @xcite .
phenomenological multimode rate equations have been devised @xcite which account for mode partition @xcite ( a regime where the instantaneous laser power is randomly distributed among several simultaneous modes ) or mode hopping @xcite ( instantaneous oscillation on an individual mode , with random jumps among modes ) . in 2004 new experimental results @xcite demonstrated that the general picture , put together over the course of two decades , where multimode semiconductor laser dynamics is driven and controlled by noise , was incomplete .
indeed , measurements performed on a multiple quantum well ( mqw ) device conclusively showed that a new kind of dynamics exists , at least for some devices , where the laser emission is composed of a regular , cyclic alternation of modes phase - locked in such a way as to maintain constant the power output by the device @xcite . among various other oscillatory regimes with mixed features ( e.g. , partial modulation of the total laser power )
, the most unusual one corresponds to a modal switching dynamics nearly devoid of a modulation of the total output , but with full modulation in the modal , frequency - resolved emission , with a sequence starting from the bluest mode and evolving towards the reddest mode ; once the last mode in the active range is reached , the sequence restarts from the bluest mode following the same regular features ( with emission times and mode transitions which depend on the pairs of modes ) @xcite .
another experiment followed @xcite which showed a dynamical regime with substantially the same features , although much more complex , in an entirely different device : a multilongitudinal mode quantum dot laser ( qdl ) .
these observations proved that the deterministic regime is not exclusive to mqw devices and that its origin must be sought outside the physical characteristics of mqws .
the regular , deterministic dynamics entirely contrasted with the stochastic picture of all previous observations and demanded a new kind of physical description for its interpretation . to this end
, new phenomenological multimode rate equation models were constructed , which included the interaction among modes through four - wave - mixing ( fwm ) @xcite , cross - saturation ( cs ) and self - saturation ( ss ) @xcite . in this approach , coupling among modes is phenomenologically introduced on the basis of physical considerations , and a satisfactory reproduction of the main observations is achieved without the need for introducing noise .
thus , determinism is recovered through the introduction of mode - coupling and excluding noise from the mathematical description of the system .
an interesting alternative approach @xcite , cited also in the first experimental report @xcite , makes a choice opposite to the usual one @xcite by introducing noise at the level of the global electric field ( i.e. , the total field contributed by all modes ) .
this choice automatically introduces a form of intermodal coupling , through noise , which is thereby partitioned onto each modal intensity . in this approach , while the _ global _ noise is a random process , the _ individual noise component _ , i.e. the instantaeous noise projected onto each individual mode , introduces a correlation between the fluctuations of different modes .
this choice is unusual and shows that determinism can be recovered through correlations , even though the underlying source of correlation is a random process .
summarizing , some kind of laser devices are capable of showing deterministic modal dynamics , accompanied , in some regimes , by a constant power output , which indicates modal phase - locking .
modeling has been achieved either through multimode deterministic rate equations modified to phenomenologically include mode coupling , or though equivalent multimode equations ( without mode coupling ) including noise introduced for the global field ( partitioned among the individual modes ) .
following the publication of the experimental results @xcite , a more advanced form of modelling was introduced where the medium s susceptibility , which can be expressed in an approximate analytical form @xcite , was explicitely used to obtain coupled partial differential equations ( pdes ) which describe the evolution of the electric field and of the carrier density @xcite .
this modeling choice represents the most sophisticated approach developed up until our contribution @xcite and contains several similarities .
we are therefore going to closely analyse this model , interpret its different components , comment on its predictions and compare them to those we obtain @xcite .
the model derived by serrat and masoller @xcite describes the evolution of the laser field , in interaction with the carriers of a semiconducting material , in a bidirectional ring cavity .
their starting point is the same as the one we take @xcite , i.e. , the wave equation for the electromagnetic ( e.m . )
field coupled to a formally described polarization for the source term , together with the evolution equation for the carrier density .
however , the adiabatic elimination ( ae ) of the polarization variable and the slowly varying envelope approximation ( svea ) , together with the rotating wave approximation ( rwa ) , are assumed to hold @xcite .
thus , the equations for the slowly varying field amplitudes , in the two counterpropagating directions , are directly written out .
in addition , the frequency dependence of the susceptibility is described phenomenologically in the parabolic approximation @xcite thereby fixing all the physical functional dependencies .
the parabolic approximation , commonly used for semiconductor models @xcite , forces a symmetric lineshape around the lasing frequency and thereby misses a crucial point ( cf .
( [ betadef ] ) ) in the description of the phase instability ( cf .
( [ eigenv]-[alphabeta ] ) ) .
periodic boundary conditions are introduced @xcite , together with finite mirror reflectivity , to take into account the role of the ( ring ) cavity .
rewritten in a notation more similar to ours , the model @xcite reads : @xmath0f \ , , \\
% \cr \label{serratequations2 } { { \partial b}\over{\partial t } } & = & + { { \partial b}\over{\partial z}}+ \mathcal{g } \left [ -i \alpha n+\left(n-1\right)\left(1+g_{d}{{\partial^2}\over{\partial z^2}}\right)-\gamma_{int}\right]b \ , ,
\\ % \cr % \cr \label{serratequations3 } { { \partial n}\over{\partial t } } & = & j-\gamma_{n}n-\left(n-1\right)\left(\vert f \vert^2+\vert b \vert^2\right)+d{{\partial^2n}\over{\partial z^2 } } \ , , % \end{array } % \label{serratequations } % \end{equation}\end{aligned}\ ] ] where @xmath1 and @xmath2 represent the slowly - varying amplitudes of the forward and backward propagating e.m .
field , @xmath3 stands for normalized linear gain , @xmath4 is the linewidth enhancement factor , @xmath5 is the contribution of the susceptibility in the parabolic approximation , @xmath6 represents the non - radiative losses for the field , @xmath7 stands for the injected current density ( in normalized units ) and @xmath8 plays the role of a normalized diffusion constant .
it is important to notice a fundamental difference between the results that we show below and those previously reported @xcite , where the authors performed a reduction of the predicted field evolution by projecting the total intensity onto a set of cavity modes .
as specifically stated in the original paper @xcite in the multimode regime the output power exhibits fast oscillations due to mode beating " removed by convolving the predicted ( total ) field evolution with a reference field centered at each modal frequency .
it is the sum of the resulting set of modal intensities
labelled @xcite _ incoherent sum _ which was shown to display a behaviour similar to that of the experiments @xcite ( mode alternation with phase locking " ) .
this choice of information reduction was justified @xcite by the limited bandwidth encountered in experiments ( at least at the time when they were conducted ) which prevented the detection of the intermode beats .
the system of equations ( [ serratequations1]-[serratequations3 ] ) possesses a 4-continuous - parameters ( @xmath9 , @xmath10 , @xmath11 and @xmath12 ) family of stationary solutions @xmath13 where @xmath9 and @xmath10 represent the phases of the solutions , @xmath11 the phase angle of the vector @xmath14 which represents the amplitude of the e.m .
field ( decomposed into the forward , @xmath1 , and backwards , @xmath2 , directions ) , and @xmath12 the wavevector mismatch , with the following values obtained from the formal solution : @xmath15 @xmath16 being the carrier stationary amplitude and @xmath17 the angular frequency mismatch relative to that of the ( rotating ) reference frame .
we are now interested in the linear stability of these solutions . for the sake of clarity ,
we restrict ourselves to the case where @xmath18@xmath19@xmath9@xmath19@xmath10@xmath19@xmath12 , but situations with @xmath12@xmath20@xmath18 are not fundamentally different . among the linear eigenvalues , three are critical and take the form @xmath21 where @xmath22 is the perturbation wavenumber and the dots stand for higher - order powers of @xmath22 .
@xmath23 and @xmath24 are real numbers defined as @xmath25 for the parameter values used in the simulations @xcite these coefficients are positive , thus the solutions ( eqs .
( [ pos1]-[solutionserrat3 ] ) ) are stable with respect to the phase . in a numerical box of length @xmath26 , the phase gradients
are then expected to decrease with a characteristic ( normalized ) time @xmath27 given by @xmath28 expressed in the units of the original paper @xcite . without the support of the previous theoretical analysis ( as was the case @xcite )
, this long characteristic time can lead to confusing the transient evolution with an asymptotic state . with the integration time step used in the original paper @xcite , @xmath29 ,
the characteristic time @xmath27 is reached only after @xmath30@xmath31 iterations !
hence , it is not surprising that especially with the standard computing means of the mid2000 s the very long transient may have been assumed to be a steady state and may have been interpreted as a phase instability . [
cols="^ " , ]
in the exact form in which they are presently expressed , it is not straightforward to compare eqs .
( [ serratequations1]-[serratequations3 ] ) with eqs .
( [ modela]-[modelb ] ) because the former is expected to describe the dynamics for all regimes of parameters while the latter is valid only in the neighborhood of the laser bifurcation threshold .
hence @xmath32 stands for the true carrier density in eqs .
( [ serratequations1]-[serratequations3 ] ) while @xmath33 is the deviation from the critical transparency value in eqs .
( [ modela]-[modelb ] ) . nevertheless , some straightforward algebraic manipulations help drawing the comparison .
let us first assume @xmath34 in eqs .
( [ serratequations1]-[serratequations3 ] ) and rewrite @xmath7@xmath19@xmath35 .
then the straightforward solution @xmath32@xmath19@xmath36 and @xmath1@xmath19@xmath18 is linearly stable provided that @xmath36 @xmath37 @xmath38 @xmath19 @xmath39 . substituting @xmath40 in eqs .
( [ serratequations1]-[serratequations3 ] ) leads to @xmath41 where we have grouped on the first line of each of the previous equations the terms which are common to both eqs .
( [ serratequations1]-[serratequations3 ] ) and eq .
( [ modela]-[modelb ] ) .
@xmath42 represents the distance from threshold .
comparing the two descriptions we remark that : 1 . the terms @xmath43 and @xmath44 , which are present in eqs . ( [ smrewritten1]-[smrewritten2 ] ) and not in eqs . ( [ modela]-[modelb ] ) ,
do appear in the course of the derivation of eqs .
( [ modela]-[modelb ] ) but have been neglected because of their smallness in the neighborhood of the laser threshold @xcite .
the term @xmath45 is present in eq .
( [ modela ] ) but not in eq .
( [ smrewritten1 ] ) .
this term describes how the selected wave vector is shifted with the carrier density .
note that this term does not enter into the benjamin - feir phase instability criterium .
the coupling between phase and amplitude through the @xmath4 factor is the same for both description .
4 . in eqs .
( [ modela]-[modelb ] ) , the coefficient in front of @xmath46 is complex , while it is purely real in eq .
( [ serratequations1]-[serratequations3 ] ) . in other terms , @xmath47 is strictly zero ( by construction ) in the sm model , which definitively excludes the possibility of appearence of a phase instability regime .
we conclude that , although the two models ( sm and cd2 ) are structurally identical
i.e. , contain the same leading terms , the nature of one of their parameters is different . while the coefficient in front of the diffraction term ( second line of eq .
( [ smrewritten1 ] ) ) is complex in the cd2 model , in the sm model it is strictly real .
this is exactly the reason why a phase instability can not occur in the sm model , but can appear in the cd2 one . as a general remark , one should notice that for a coefficient ( or one of its complex parts ) to be exactly zero , an underlying profound physical selection process must exist ( e.g. , a symmetry of the problem or the existence of a conservation law ) . in general , coefficients should possess both components : their values will determine whether the corresponding process is sizeable or neglibible ( here the phase instability , elsewhere , e.g. , the dependence of the laser frequency on the electric field strength @xcite ) . up until here
we have discussed the analytical comparison between the two models using the restriction , for the sm one , to a single direction of propagation .
this assumption bears no weight on our conclusions , since the two counterpropagating waves are coupled through amplitude and not phase terms . thus , considering the full bidirectional nature of the sm model does not change the results of this analysis .
this is confirmed by the analysis of section [ smmodel ] which proves that the stability properties are generic , independently of the value of @xmath11 ( eqs .
( [ coeff12]-[coeff3 ] ) ) , thus of the relative amplitude of the two counterpropagating e.m .
field components .
the description of the semiconductor laser based on the interaction between the full e.m . field and
the lasing medium , introduced by serrat and masoller @xcite , has represented a innovative step beyond the modal decomposition , traditionally used in the description of the longitudinal multimode operation of semiconductor lasers . however , the approximations introduced to obtain a manageable model @xcite remove all possibilities for observing a true phase instability . the very slow convergence of the computer codes , owing to the presence of very long time scales , led to the false conclusion that long - term phase dynamics could be reproduced in this way @xcite .
the full mathematical description that our model provides , based on a complete codimension 2 expansion , offers the advantage of a proper and global handling of the various physical processes ( fwm , cs , ss ) in the neighbourhood of threshold , without ressorting to _ ad hoc _ choices to introduce a physical coupling mechanism .
the _ ab initio _ derivation of the equations , avoiding all the usual approximations as already done @xcite , has allowed us to discover the existence of a phase - instability which arises through a _ new fundamental parameter _ , named @xmath47-parameter , which in conjunction with the traditional @xmath4-parameter of semiconductor lasers , controls the growth of the phase - eigenvalue and maps the semiconductor laser problem into the benjamin - feir instability @xcite of fluid dynamics : a problem whose parameter space has been carefully explored and characterized @xcite .
the existence of the @xmath47-parameter and its joint - control , with @xmath4 , of the instability is the probable reason why semiconductor lasers as different as mqws and qdls may present similar dynamics .
this point is left for further investigation .
we have shown that the phase instability operates in a way which is similar to the random excitation due to noise , but in an entirely deterministic kind of description .
this comes from the width of the unstable interval of the lyapunov exponents in the spectrum of the eigenvalues @xcite .
thus , our description reconciles the phenomenological deterministic description of the problem @xcite with the stochastic one which couples the mode through the projection of noise on the mode ensemble @xcite . at the same time , however , our results open new questions on the relative role of the phase instability ( as _ noise source _ ) compared to the physical , unavoidable , sources of noise @xcite . answering these questions requires further work ; at the present stage we can only point out their existence .
relevant references can be found in a comprehensive review of polarization isotropic ( i.e. , two - mode ) lasers : lenstra d. , on the theory of polarization effects in gas lasers " , phys
. repts . *
59 * , 299 - 373 ( 1980 ) .
bidirectional ring lasers may present the same kind of behaviour .
for an extensive literature cf . the introduction and bibliography of : zeghlache h. , mandel paul , abraham n.b . , hoffer l.m .
, lippi g.l . , and mello t. , bidirectional ring laser : stability analysis and time - dependent solutions " , phys . rev .
a * 37 * , 470 - 497 ( 1988 ) .
huyet g. , balle s. , giudici m. , green c. , giacomelli g. , and tredicce j.r .
, low frequency fluctuations and multimode behaviour of a semiconductor laser with optical feedback " , opt . commun . *
149 * , 341 - 347 ( 1998 ) .
wallace i. , yu dejin , harrison r.g . , and gavrielides a. , experimental observations of multi - mode dynamics in an external cavity semiconductor laser " , j. opt .
b. : quantum semiclass . opt .
* 2 * , 447 - 450 ( 2000 ) .
yacomotti a.m. , furfaro l. , hachair x. , pedaci f. , giudici m. , tredicce j.r . , javaloyes j. , balle s. , viktorov e.a . , and mandel paul ,
dynamics of multimode semiconductor lasers " , phys .
a * 69 * , 053816 ( 1 - 9 ) ( 2004 ) .
manneville p. , lyapunov exponents for the kuramoto - sivashinsky model " , in [ _ macroscopic modelling of turbulent flows _ ] , u. frisch , j. keller , g. papanicolau , and o. pironneau eds .
, lecture notes in physics * 230 * ( springer , berlin , 1985 ) , pp .
319 - 326 .
|
following an overview of modeling of ( longitudinal ) multimode semiconductor laser dynamics , we analyze in detail a model proposed in 2006 to explain deterministic , phase - locked modal alternation , experimentally observed a decade ago . through a stability analysis , we prove that the numerically obtained electromagnetic field evolution , interpreted as an explanation of the experiments , is nothing more than an extremely long transient , so long as to be hardly identifiable in an entirely numerical approach . comparison with a model
we have recently derived , which predicts a phase instability ( benjamin - feir - like ) compatible with the experimental observations , highlights the crucial ingredient for the dynamics .
the wide spectrum of unstable eigenvalues accompanying the phase instability plays the role of an equivalent noise in a fully deterministic description , thus reconciling the heuristic models which could qualitatively reproduce the experimental observation either with deterministic equations in the presence of mode - coupling , or through stochastically driven modal decompositions .
|
in @xcite it was found that the distances on a one dimensional lattice in a noncommutative geometry differ from their classical analog . in noncommutative geometry the distances on a one dimensional lattice are given by @xmath0 where @xmath1 , @xmath2 , and @xmath3 is the lattice constant , with units of length . as can be seen directly , the distance to an odd @xmath4 point has a unit anomaly , whereas the distance to an even @xmath4 point is the geometric mean of the distances to the two nearby odd points .
moreover , we note that translation invariance is broken , since @xmath5 . in the large @xmath4 limit , however , translation invariance is restored asymptotically .
+ the fact that the distances given by ( eq.[dist ] ) thus break the lattice s translation invariance , represents a new phenomenon of a distance s dependence on reference frames .
it follows directly from the basic fact that all the distances in ( eq.[dist ] ) were computed from some zero point , which the lattice sites labeling is been referred to .
the selection of one of the lattice point as the zero point , was done without loss of generality .
therefore , _ if several spectators are present on the lattice sites , with each spectator defining his location as the zero point in his own world , then due to the different sites labeling by different spectators , the distance between two lattice sites becomes reference frames dependent .
_ we may thus define the distance between the @xmath6-th and @xmath7-th sites , as been labeled in some reference frame as the absolute value of the difference between the distances i.e. @xmath8 we remind the reader of the reason for the introduction of a new definition of distance in noncommutative geometry .
it follows from the fact that the classical definition uses the concept of a _ path _ , i.e. @xmath9 where @xmath10 is the length of the path @xmath11 connecting points @xmath3 and @xmath12 .
the infimum is taken over all such path lengths .
however , the concept of a path is only well defined on a smooth manifold .
one is thus forced to apply the noncommutative geometry definition , which is more fundamental than the classical , and more general , since it fits all topological spaces .
this definition of a distance in noncommutative geometry is given by@xcite : @xmath13 \;\right\| \leq 1\right\}\ ] ] where @xmath14 @xmath15 is the algebra of functions on @xmath16 @xmath17 is a dirac operator ( which is a self adjoint operator with compact resolvent ) acting in the hilbert space @xmath18 and the norm on the r.h.s is the norm of operators in @xmath19 .
both definitions give the same result when the base space is a riemannian manifold .
however the n.c.g definition has the advantage of being applicable to discrete spaces too .
+ noncommutative geometry , from its foundation was constructed in such a way that it fits the concepts used in quantum mechanics - mainly the passage from variables to operators .
more details about the connection to q.m . can be found in @xcite .
+ in that context the reader should not be surprised that we were led to the concept of reference - frames - dependent distances . not with standing its novelty , what can be more natural in q.m .
than an influence of the observer on the results ?
we believe that a similar phenomenon will be revealed in a full theory of quantum gravity when this will be found . in the following ,
we analyze some of the possible physical implications of the above result .
assume a one dimensional lattice with lattice constant @xmath3 .
on the lattice there are a _ spectator _ and a _ test particle ( t.p.)_. we assume that the spectator sticks to one site , which serves as the zero reference point in the spectator s system , from which distances are measured on the lattice .
the t.p . has a mass @xmath20 and is situated at a site @xmath6 .
the dynamics of the t.p . on the lattice are characterized by a hopping motion from site to site on the lattice , where for each jump to occur there is some probability .
such dynamics are particularly appropriate in a scenario in which the system is a quantum system , in analogy with the motion of a particle between discrete energy levels .
note that the dynamics on the lattice can not be `` smooth '' , i.e. the t.p .
can jump between two lattice sites without `` passing '' through the intermediate sites , as in jumps between bohr s orbits for atomic electrons .
essentially , such hopping dynamics are known from solid state physics ( e.g. mott hopping conductance , also connected to percolation theory @xcite ) .
quantum behavior is very much in the spirit of dynamics in a noncommutative geometry .
two questions now arise .
first , how do we treat _ time _ ?
the second question relates to the _ probabilities _ for the test particle s hops how do we go about evaluating them ?
as we shall see , there are three possible answers ( and perhaps even more ) for the first question .
the first possible answer would claim that there might be no need at all for an answer , since one can deal with probabilities without involving time dependence , in the same way as one deals with the probabilities in dice , namely , without caring about the rate at which the dice are thrown .
thus , in this solution , the jumps rate does not affect the probabilities .
however , having in mind physical models and not just mathematical ones , we believe that the question can not be eliminated .
the second possible answer is that time is an extra dimension , as is usually done .
thus the total spacetime is @xmath21 or @xmath22 , depending on whether the time is continuous or discrete , respectively .
but , since the results for the distances on a lattice which we use here are taken only for the one- dimensional case , and the distances on those two - dimensional spaces could be different from what we use , we prefer not to adopt this answer .
thus , although the second answer might be the correct one physically , we select for our purposes the following third answer , which enables us also to apply our result ( [ dist ] ) for the distances .
this third answer consists in using our quantum hops to define and measure time intervals , in the spectator system . _ to measure time , we count the t.p .
jumps_. in this context , time is not an extra dimension , but rather a parameterization defined by counting , similar to the counting of the vibrations of the quartz in a digital watch .
although some vibrations might have longer periods ( as measured by counting the vibrations of a more accurate watch ) time can still be defined as the result of an exact counting of the vibrations . to determine the probabilities for the t.p.s hops on the lattice
, we first examine the propagator of a free boson on a translation - invariant space : @xmath23 .
the fourier transform of this propagator is essentially the correlation function between different locations in the coordinate space .
it seems reasonable to take the correlation function between two different sites as proportional to the probability of a jump between them : the more they are correlated , the stronger the chances for the t.p . to make that jump . working with a one - dimensional lattice , we take the one - dimensional fourier transform of the propagator
. the probabilities will thereby be of the form : @xmath24 where @xmath25(@xmath20 - is the t.p .
mass , and @xmath3 is the lattice constant ) , @xmath6 and @xmath26 are the initial and the final sites of the t.p .
relative to the spectator , @xmath27 , and @xmath28 is the normalization of the probabilities . remembering that the lattice translation invariance is broken and only reached asymptotically - we take the above expression for the probabilities only as a first guess , a form that they should roughly follow .
the normalization is given by the following formula : @xmath29 note that for @xmath30 the t.p .
is initially situated at the same site as the spectator . as can be immediately seen , if @xmath31 is a constant , the hopping probabilities for the t.p . as seen in the spectator s frame
differ from those evaluated in the t.p .
this is a direct outcome of the fact that @xmath32 where the l.h.s is the distance as observed in the spectator frame and the r.h.s is the distance as observed in the t.p . frame . in the following section ,
we discuss the meaning of this incompatibility between the probabilities in the two systems , and ways of resolving this difficulty .
a closer look at the distances involved will reveal that for a t.p .
initially situated at an even site @xmath33 , the most probable jump , as evaluated in the spectator frame is to the site @xmath34 ( i.e. @xmath35 ) . if the t.p .
was initially located at the site @xmath36 , the most probable jump , as evaluated in the spectator frame is to @xmath37 ( i.e. @xmath38 ) . however in the t.p .
own frame , there is no difference in the probabilities for jumps to the left or to the right .
another anomaly which is an outcome of the above incompatibility can be seen when inspecting distances between sites whose indices differ by two units . though the distance between two neighboring odd points is always @xmath39 ( except for @xmath40 ) , the distances between two neighboring even points form a decreasing series , tending to @xmath39 , when @xmath41 . in the spectator frame , the effective probability for the test particle s motion will thus be larger in an outward direction ( relative to the spectator ) , i.e. the t.p .
_ drifts away_. in the t.p .
frame , however , as there is no difference between its two sides , the t.p . should just stay in its initial position ( on the average ) .
note , moreover , that different spectators - located each at a different lattice site ( i.e. with different origins in their respective frames ) - will observe different probability distributions .
though the probability space in the t.p .
frame is a commutative space ( in the sense that when defining @xmath42 as the probability of jumping @xmath7 sites to the right , then @xmath43 ) , it is noncommutative in the spectator frame , where @xmath44 .
this point summarizes the inconsistencies between the two frames . since the probabilities are thus not the same , whether between spectator and t.p .
frames , or between two spectator frames , applying our `` third option '' as defined in section 2.1 , we observe that _ the time intervals are not synchronized_. we thus set upon removing the anomalies by an appropriate corrective mechanism .
all above anomalies were obtained under the assumption that the mass - @xmath20 , of the t.p .
is the same in both the spectator and the t.p . frames .
our solution is to abandon the assumption that the t.p .
mass is a global parameter and assume instead that it is a local variable . in the following we demonstrate that this is a complete resolution of our paradoxes .
we require the probabilities @xmath45 and @xmath46 to be equal in both spectator and t.p .
frames , i.e. using localized masses , we restore translation invariance for the probability of hopping @xmath47sites . where @xmath49 is the mass of the t.p . as observed in the spectator frame , while hopping from the @xmath6 site to the @xmath26 site , and @xmath50 is the mass in the t.p .
the normalization constant @xmath51 is defined as : @xmath52 where we are using ( [ dist])[see figure 1 in the appendix ] , thus normalizing the probabilities ( i.e. @xmath53 ) .
as one can see , equation([mass2 ] ) can not be analytically solved for @xmath55 , but if one assumes for simplicity that : @xmath56 one finds from ( [ prob2 ] ) that the mass as seen by the spectator is given by @xmath57 } \approx \mu _ 0\left [ 1-a\mu _ 0\left ( \left| x_{i+j}-x_i\right| -\left| x_j\right| \right ) \right ] \label{mass}\ ] ] we emphasize that if one assigns planck - length value @xmath58 to the lattice constant @xmath3 ( i.e. @xmath59 ) , then the limit ( [ limit ] ) should be interpreted as the low mass limit @xmath60 , where @xmath61 is a planck - mass ( i.e. @xmath62 ) .
thus , the mass formula ( [ mass ] ) should be seen as semi - classical . by setting the probabilities to be equal in the two frames
, we are essentially synchronizing the relevant times - applying our definition of time intervals , based on counting jumps .
the price , however , has been paid by the transformation of the mass @xmath63 into a local and dynamical variable - @xmath64 ( local and dynamical because of the @xmath6 and the @xmath7 dependence respectively ) .
there is a very rough analogy with the emergence of the doppler effect in special relativity .
however , whereas in the doppler effect the frequency of a wave emitted by a system moving towards the spectator is blue - shifted , i.e. always increases in the spectator frame , in our case , the t.p .
mass as observed in the spectator frame @xmath64 can be smaller than @xmath50 , even when the t.p .
jump is directed towards the spectator .
the fact that the physical mass @xmath64 depends on the lattice constant @xmath3 , aside from its dependence on @xmath63 and on the location , suggests that @xmath64 depends on the scale .
this behavior allows us to think of the mass as a `` running mass '' , as happens in the context of the renormalization procedure in field theory ( see figure 2 in the appendix ) . adopting the above concept of a local mass , we have thus achieved full agreement between the spectator and t.p .
probability values .
there is thus also a synchronization of time measurements between the frames , and there is no preferred direction on the lattice . instead of modifying the kinematics , by constraining the mass to be a local variable , in our effort at restoring covariant dynamics for the t.p .
, we could use an alternative , `` dynamical '' approach , by introducing potentials . with the distances on the lattice
not being translation invariant , the propagator of a boson particle situated at the @xmath65 site ( in the spectator frame ) could be written in terms of a potential @xmath66 , , @xmath67 ^{-1 } \label{propagator}\ ] ] where @xmath68 is induced by the metric defined in ( [ dist ] ) and ( [ disij ] ) . as we demonstrated , this is , on the average , a repulsive force , _ an induced attractive potential @xmath70 _ which will exactly cancel _ @xmath68 .
this is somewhat reminiscent of einstein s 1917 introduction of the cosmological constant , to stabilize the universal geometry .
the propagator of the t.p . in all spectator frames are now the same as for a free particle @xmath71 ^{-1}$ ] .
the existence of the induced potential , in some aspects , is also similar to the lentz law in classical electromagnetism .
_ one can identify the modified mass in ( [ prob2 ] ) with the following effective mass , defined as : @xmath72 in other words , one can deduce from @xmath49 the value of the induced potential between the t.p . and the spectator ( in the low mass limit ( [ limit ] ) one can use ( [ mass ] ) and ( [ effmass ] ) ) . summarizing this approach , we note that though the distances break translation invariance and while the mass preserves its global definition , the hopping probabilities and the related beats of the clock have become translation invariant . + summary + we have shown that the induced metric of a one - dimensional lattice , ( from a noncommutative geometry calculation ) breaks the lattice translation invariance and therefore also becomes reference - frame dependent , possibly leading to some paradoxes concerning the fundamental issue of general covariance .
we have suggested a solution of the paradoxes , essentially at the probabilistic level - which in our opinion is the most important aspect , at least from the measurement aspects of quantum mechanics .
it was done by either allowing an effective locally dependent mass which thus becomes a `` running mass '' , or by introducing a compensating potential .
both ways lead to a commutative probabilistic space which is general covariant .
+ acknowledgment + i am most greatful to prof .
yuval neeman for the enlightening comments .
i would like also to thank the anonymous referee for his constructive suggestions .
appendix + a. the test particle as a wave function + up to this point , we have treated the t.p . as a classical point particle . in the following ,
we go over to quantum mechanics - i.e. the t.p .
is now a quantum particle , represented by a wave function - @xmath73 which is defined over the entire lattice .
let @xmath74@xmath75 ( i.e. a square integrable function ) be normalized so that @xmath76 .
let @xmath77be a complete set of orthogonal states which span the hilbert space - @xmath78which is defined over the lattice .
the hilbert space basis @xmath79 is defined as follows : @xmath80 where the @xmath7 -index stands for a state in the hilbert space , while the @xmath81 -index stands for the lattice site .
any @xmath82 can now be represented by the following sum : @xmath83 where the @xmath84 -s are complex numbers . since @xmath85 it follows that @xmath86 . as in q.m .
the @xmath87 represent the probability for the particle to be found in a measurement at the @xmath7-site of the lattice .
one can thus treat the @xmath88 as a distribution , for which the @xmath87 -s form a probability space .
let @xmath89 be a distribution centered at the @xmath6 site of the lattice .
we define @xmath90 as the coefficients in ( [ cigma ] ) for the @xmath91 .
we observe the similarity between @xmath92 and @xmath45 ; both define the probability for the t.p . to be initially at the @xmath6 - site and after a measurement , in the @xmath93 - site .
we have thus essentially showed that the t.p .
can also be represented by a wave function , with the effective mass ( or the induced potential ) having the same features as before .
+ + b. in the following we exemplify some of the formulae found in the article .
figures 2a - 2e reproduce some examples of the running mass in ( [ mass2 ] ) .
the @xmath6 and the @xmath7 indices represent the initial and the hopped - to sites of the test particle , respectively .
@xmath63 is the t.p .
mass in its own frame , and @xmath3 is the lattice constant .
|
in a one - dimensional lattice , the induced metric ( from a noncommutative geometry calculation ) breaks translation invariance .
this leads to some inconsistencies among different spectator frames , in the observation of the hoppings of a test particle between lattice sites . to resolve the inconsistencies between the different spectator frames , we replace the test particle s bare mass by an effective locally dependent mass
this effective mass also depends on the lattice constant - i.e. it is a scale dependent variable ( a `` running '' mass ) .
we also develop an alternative approach based on a compensating potential .
the induced potential between a spectator frame and the test particle is attractive on the average .
we then show that the entire formalism holds for a quantum particle represented by a wave function , just as it applies to the mechanics of a classical point particle .
taup-2335 - 96 -1.8 cm -2.2 cm
|
the work of amari and others @xcite on the use of methods of riemannian geometry to analyze statistical estimation problems is of increasing interest to researchers in signal processing .
this methodology , known as _ information geometry _ , provides a rigorous framework for measuring the power of data to discriminate values of parameters .
these ideas date back to rao @xcite , who showed that the fisher information of a likelihood used in an estimation problem can be seen as a riemannian metric on the parameter manifold .
this paper brings an information - geometric perspective to a class of sensor management problems by casting the objective of sensor management as parameter estimation and describing how this leads to the role of sensor management as selecting a riemannian metric for the parameter manifold .
established results in riemannian geometry @xcite , outside the context of information geometry , show that the collection of all riemannian metrics on a riemannian manifold is itself an ( infinite - dimensional ) riemannian manifold . in problems where the collection of possible sensor actions is suitably modeled by a smooth finite - dimensional manifold , the space of interest is a finite - dimensional sub - manifold of this infinite - dimensional riemannian manifold .
a perspective is developed in which the best sensor management action to take , in terms of gathering the most information relevant to the estimation objective , is locally characterised in terms of geodesic curves in this space . much of the development in subsequent sections of this paper is rather abstract and draws upon mathematical machinery that is unfamiliar to many researchers in sensor management area . to provide a more concrete context in which to illustrate some of the concepts that arise in later sections , is it helpful to begin by setting forth an example problem .
suppose two mobile sensor platforms and one stationary target ( emitter ) are located in the plane @xmath0 , as depicted in fig . 1 .
the goal is to estimate the position of the target from bearings - only measurements taken at the sensors . since the sensors are mobile , the sensor management problem is to identify the trajectories of sensor motion that will yield the best estimate of the target position . of a single stationary emitter from bearings - only measurements received at two mobile sensing platforms located at respective coordinates @xmath1 and @xmath2 in the plane .
] more specifically , the target position is @xmath3 and the sensor positions are @xmath4 for @xmath5 . denoting @xmath6 and @xmath7 ,
the bearing of the target from sensor @xmath8 is @xmath9 .
the sensor measurements are independent and von mises distributed , each with common inverse dispersion parameter @xmath10 and with the circular mean of the measurement at sensor @xmath8 having circular mean @xmath11 .
the following sections proceed to describe the nature of a sensor model from an information - geometric viewpoint , to define the parameter manifold , the sensor manifold , its metric structure , and to derive a differential equation that characterizes geodesic curves on the sensor manifold .
the development departs from the purely geometric treatment in @xcite in that it allows for informative prior distributions on the parameter manifold rather than restricting attention to a volume form corresponding to the jeffreys prior .
further , the riemannian metric with respect to which geodesics maximize `` energy integrals '' on the sensor manifold is shown to arise from both kullback - leibler and mutual information perspectives . throughout this process
, the example just introduced will be used to illustrate these concepts in a concrete fashion .
consider the problem of estimating a parameter @xmath12 from data @xmath13 collected by sensors up to time @xmath14 . beginning with a prior probability distribution for @xmath12 , which may reflect what is known from previous measurements or side information , the effect of taking measurement at time
@xmath14 is to provide a posterior probability distribution , which will be assumed to be represented by a posterior probability density @xmath15 . if the option exists to use one of a parametrized set of sensors or sensor configurations , each of these will produce its own posterior density . when these posteriors are known , selecting a sensor configuration to use for a measurement amounts to choosing one of them from which to estimate @xmath12 .
the parameter @xmath12 will henceforth be assumed to be an element of an @xmath16-dimensional smooth manifold @xmath17 ( @xmath18 will be assumed , although @xmath19 is sufficient for most of the discussion here ) , which will be called the `` parameter manifold . '' in the example problem , @xmath17 is @xmath0 because the parameter @xmath20 is a physical location in the @xmath21 plane . denoting by @xmath22 the log - likelihood for this problem , the fisher information is @xmath23 , \ ] ] where @xmath24 denotes the derivative of @xmath25 with respect to the parameter @xmath12 .
this is well known to be equivalently expressed as @xmath26 .\ ] ] in this expression , @xmath27 represents the covariant derivative along any connection in @xmath17 ; all choices of connection give the same quantity as .
@xmath28 is always a non - negative definite @xmath29 matrix , and in what follows it will be assumed to be non - singular , thus giving rise to a continuous family of inner products over the tangent spaces of the manifold .
direct calculation in the example problem shows that , in the coordinate system described above and depicted in fig . 1 , @xmath28 has the form @xmath30 where @xmath31 . through this mechanism
, the choice of a particular sensor leads to the association of a positive definite matrix with each @xmath32 , thereby imbuing @xmath17 with a _
riemannian metric _ that measures the ability of that sensor s data , at least locally , to discriminate between parameter values .
it is possible to calculate the shortest distance , in terms of this metric , between two values @xmath12 .
as discussed in @xcite , the kullback - leibler divergence between @xmath33 and @xmath34 is approximately half of the square of the distance between @xmath12 and @xmath35 when @xmath12 and @xmath35 are close .
it has been shown @xcite that the collection @xmath36 of all riemannian metrics on the manifold @xmath17 is an infinite - dimensional ( weak ) riemannian manifold .
the structure of its tangent space is described in @xcite .
a point in @xmath37 is a riemannian metric on @xmath17 ; i.e. , it associates a positive definite form @xmath38 with each @xmath32 . under suitable assumptions ,
a metric on @xmath37 is defined by @xmath39 where @xmath40 .
specific assumptions guaranteeing finiteness of this integral are beyond the scope of this discussion , and it will suffice for the purposes here to assume directly that it is finite .
although the nature of @xmath37 appears formidable , realistic sensor management problems do not require one to work with this entire space , rather with a finite - dimensional sub - manifold that inherits the metric from @xmath37 .
the assumption that leads to this situation is that the collection of all possible sensor configurations is parametrized by a smooth manifold @xmath41 , which will be called the `` sensor manifold . '' in the example problem , the sensor configuration is completely specified by the positions of the two sensor platforms in the plane ; i.e. , by @xmath42 . in this case
, the sensor manifold is @xmath43 and the only elements of @xmath37 of relevance are those metrics on @xmath17 that arise from a sensor configuration @xmath44 in this four - dimensional manifold .
beginning with a sensor configuration @xmath45 gives rise first to a likelihood @xmath46 and consequently to a riemannian metric @xmath47 on the parameter manifold @xmath17 , as described in section [ sec : model ] . as a riemannian metric on @xmath17
, @xmath47 is an element of @xmath36 .
this mapping @xmath48 taking @xmath44 to @xmath47 will be called the `` sensor geometry . '' in what follows , @xmath49 will be assumed to be smooth and one - to - one and @xmath50 a sub - manifold of @xmath37 .
weaker assumptions are possible ( e.g. , @xmath49 is an immersion ) , but full generality is is not needed here to adequately illustrate the method . through the sensor geometry map , the finite - dimensional manifold @xmath41 inherits the riemannian structure of @xmath37 ; i.e. , the distance between two sensor configurations in @xmath51 and @xmath52 in @xmath41 is taken to be the distance between @xmath53 and @xmath54 in @xmath37 .
this construction endows the sensor manifold @xmath41 with its own riemannian metric which captures , in information - theoretic terms , the `` complementariness '' of sensor configurations . allows the finite - dimension manifold @xmath41 of sensor configurations to inherit a riemannian metric from the infinite - dimensional manifold @xmath37 of all riemannian metrics on the parameter manifold @xmath17 . ]
the objective of determining good trajectories in the sensor manifold @xmath41 will be addressed by relating these to geodesic curves in @xmath37 .
following @xcite , consider a smooth curve @xmath55\to{\mathcal{m}}$ ] .
for each @xmath56 $ ] , @xmath57 is a riemannian metric on the parameter manifold @xmath17 and thus associates a positive definite matrix @xmath58 with each point in @xmath32 .
the energy integral along the curve @xmath59 is @xmath60 in this expression , @xmath61 is a probability density on @xmath17 , @xmath59 means @xmath58 , and @xmath62 is the derivative of @xmath59 with respect to @xmath14 .
geodesics in @xmath37 minimize @xmath63 , and a variational approach is used in @xcite to obtain the differential equation @xmath64 for @xmath57 , which implies @xmath65 the right - hand side of this differential equation is observed to be a christoffel symbol .
the induced metric at a point @xmath45 is @xmath66 where @xmath67 and @xmath68 are in the tangent space @xmath69 of @xmath41 at @xmath44 and @xmath70 is the push - forward of @xmath71 . for a smooth curve @xmath55\rightarrow s$ ] ,
the energy integral restricts to @xmath72 the geodesics , which are the extremal curves of @xmath63 , satisfy @xmath73 where @xmath74 denotes the christoffel symbol for the levi - civita connection on @xmath41 . in terms of local coordinates in @xmath41 , geodesic equation in @xmath41
may be obtained by solving a variational problem on the path @xmath67 . to set this up , it is convenient to abuse notation and define a smooth function @xmath75 ^ 2\to s$ ] such that @xmath76 . with this notation , and
coordinatizing @xmath77 as @xmath78 , @xmath79 if this expression is set to zero , further algebraic simplification leads to a differential equation ( in coordinates ) that characterizes geodesics in @xmath41 : @xmath80 returning to the example pictured in fig . 1 , the local coordinates in @xmath43
are @xmath81 , @xmath82 , @xmath83 , and @xmath84 .
the positive definite matrix @xmath85 corresponds to the fisher information matrix @xmath86 given in .
the inverses and derivatives needed are calculable , with @xmath87 and @xmath88 where @xmath89 fig .
[ fig : example ] shows trajectories obtained for a particular case of the example scenario .
the target is stationary at ( 1,1 ) , and the sensors prior distribution on the target location is normal with mean ( 1,1 ) and covariance @xmath90 .
sensor 1 starts at ( 0,1 ) and sensor 2 starts at ( 1,0 ) , and initial directions of motion are defined by the geodesic for this configuration .
the sensors move in this direction for a fixed period of time , a new set of directions is determined from geodesic calculations based on the new configuration , the sensors move again , and so forth .
the dotted trajectories are extrapolations ; they indicate the directions defined by the geodesic computation at the last iteration computed .
[ fig : example ]
the proposed scheme for sensor management involves following , at least locally , geodesic curves in @xmath41 defined by the riemannian metric @xmath41 inherits from @xmath37 . geodesics maximize energy integrals of the form , so it is desirable to understand how optimization in this sense relates to the amount of information gathered by the sensor .
consider first the kullback - leibler divergence @xmath91 for two multivariate normal distributions with equal means and respective non - singular covariance matrices @xmath49 and @xmath92 .
this is given by @xmath93 where @xmath94 denotes determinant and @xmath95 is the identity matrix .
a divergence on @xmath37 may be defined by @xmath96 \;df(\theta ) .\end{gathered}\ ] ] here , the two positive definite matrices @xmath49 and @xmath92 are regarded as arising at each point of @xmath17 from two riemannian metrics .
it is evident that @xmath97 and @xmath98 .
the corresponding riemannian metric on @xmath37 is @xmath99 as appears in .
similarly , one can define a divergence on @xmath37 motivated by mutual information by @xmath100 this `` symmetrized '' mutual information expression is equivalent to @xmath101 as with @xmath102 , it is clear that @xmath103 .
calculation reveals that @xmath104 and that the corresponding riemannian metric on @xmath37 is @xmath105 thus , despite arising from different concepts of information ( i.e. , @xmath102 from fisher and @xmath106 from shannon ) , both of these divergences give rise to exactly the riemannian metric on @xmath37 used in the geodesic computations of section [ sec : geodesics ] .
in this short paper , we have built upon results in differential geometry , outside the context of information geometry , to introduce an information - geometric approach to sensor management .
the approach begins with the observation that , when the goal of sensing is parameter estimation , the effect of selecting a particular sensor configuration amounts to imparting a riemannian metric on the parameter manifold @xmath17 via the fisher information .
the collection of all such metrics is the riemannian manifold @xmath36 , for which the metric , geodesic equations , and other differential geometric aspects are known . with the assumption that our choices of sensor configuration are parametrized by a smooth sensor manifold " @xmath41
, we observed that @xmath41 inherits a riemannian structure from @xmath37 and used this to obtain a differential equation characterizing geodesic curves in @xmath41 . in the purely geometrical work on which we have built , the measure on @xmath17
is a volume form that corresponds to the statistical notion of a ( minimally informative ) jeffreys prior .
we observe that this may be replaced by an informative prior , as would typically be desirable in sensor management applications .
navigation along geodesic curves in a riemannian manifold maximizes an energy integral involving the metric .
we have constructed two distinct divergences on @xmath37 corresponding to familiar information - theoretic quantities ( kullback - leibler divergence and mutual information ) that have been used by various authors as criteria in designing sensor scheduling algorithms .
both of these are shown to lead to the same riemannian metric on @xmath37 , suggesting the information gathering merit of sensor scheduling based on following geodesic curves defined with respect to this metric . while the work presented here
is mostly conceptual , we have shown enough specifics of how the proposed method manifests in a concrete example to indicate its feasibility .
we are continuing to develop complete application examples while simultaneously working out rigorous specifics of some of the mathematical foundations .
the authors are grateful to sofia suvorova who supplied the numerical results presented in sec . [
sec : geodesics ] in response to reviewer remarks on our original manuscript .
we regret that icassp policy prevents us from including her as an author on this revised version of the paper .
this work was supported in part by the university of michigan and the u.s .
army research office under muri award no .
w911nf-11 - 1 - 0391 and by the u.s
. air force office of scientific research under grant no .
fa9550 - 09 - 1 - 0561 .
|
an information - geometric approach to sensor management is introduced that is based on following geodesic curves in a manifold of possible sensor configurations .
this perspective arises by observing that , given a parameter estimation problem to be addressed through management of sensor assets , any particular sensor configuration corresponds to a riemannian metric on the parameter manifold . with this perspective ,
managing sensors involves navigation on the space of all riemannian metrics on the parameter manifold , which is itself a riemannian manifold .
existing work assumes the metric on the parameter manifold is one that , in statistical terms , corresponds to a jeffreys prior on the parameter to be estimated .
it is observed that informative priors , as arise in sensor management , can also be accommodated . given an initial sensor configuration , the trajectory along which to move in sensor configuration space to gather most information is seen to be locally defined by the geodesic structure of this manifold .
further , divergences based on fisher and shannon information lead to the same riemannian metric and geodesics . information geometry ; sensor management
|
the discovery of several peculiar type ia supernovae ( sne ia ) has drawn the attention both to the photometric and spectroscopic diversity among this class of otherwise homogeneous transients .
the dispersion of the luminosity decline rate relation @xcite can be explained by an additional correlation between the decline rate and the colour at maximum light @xcite .
hence sne ia can be arranged into a photometric sequence extending from luminous , blue , slowly declining sn ia , like sn 1991 t , to normal events @xcite , and finally to sub - luminous , red , quickly declining objects , like sn 1991bg @xcite .
sne ia also appear to form a spectroscopic sequence based on the systematic variations in the flux ratios of several spectral features near maximum light ( e.g. si ii @xmath65972 , 6355 , see * ? ? ?
the common view is , despite their diversity , peculiar events such as the luminous 1991t - like and sub - luminous 1991bg - like sne ia , just like the normal population of sne ia , originate from the thermonuclear explosion of a c / o white dwarf ( wd ) that exceeds the chandrasekhar mass after accreting mass from a companion star in a binary system . however , there is a group of peculiar sne ia that challenges the canonical chandrasekhar - mass explosion channel .
the prototype of this class is sn 2002cx @xcite , which shows a peak luminosity significantly lower than that of normal sne ia , even though its light curve decline - rate parameter is comparable to normal events .
spectra obtained near maximum light resemble those of over - luminous 1991t - like objects ( with a blue continuum and absorption from higher - ionisation species ) , even if a low ejecta velocity ( @xmath7 km s@xmath3 at the epoch of @xmath8-band maximum light ) points towards a moderate kinetic energy from the explosion @xcite .
the late - time spectra show narrow iron and cobalt lines @xcite , in stark contrast to normal sne ia at similar epochs .
after the pioneering studies by @xcite and @xcite on sn 2002cx , both new and old sn discoveries have been classified or reclassified as 2002cx - like events , and it has become clear these transients are not so rare .
this class was labelled type iax supernovae ( sne iax ) by @xcite , who presented a review on the entire group and defined clear observational criteria to classify a type iax event .
a variety of explosion scenarios and potential progenitors or progenitor systems have been proposed to explain each event ( see * ? ? ? * and references therein for a recent review ) .
although the leading models for sne iax are thermonuclear explosion of a c / o wd @xcite , a core - collapse scenario has been proposed at least for sn 2008 ha @xcite , which is the most extreme member of sn iax class to date .
the latter however is controversial because of the detection of c / o burning products in the maximum - light spectrum of sn 2008 ha @xcite , providing a link to thermonuclear explosions . in principle , the best way to shed light on this issue would be the detection of a progenitor in pre - explosion images .
recently , @xcite reported the detection of a luminous blue source coincident ( at the 0.8@xmath9 level ) with the location of type iax sn 2012z in _ hubble space telescope _ ( _ hst _ ) pre - explosion images .
although the photometric properties of this object suggest a c / o wd primary plus a he - star companion progenitor system , the explosion of a single massive star has not definitely been ruled out . in this case , post - explosion imaging , obtained after the sn fades away , should help to distinguish between the two models .
for two other sne iax , no sources were detected in pre - explosion images , but limits were obtained that exclude massive stars as potential progenitors ( sne 2008ge and 2014dt , see * ? ? ? * ; * ? ? ? * ) . given the diversity of this sn class
, one may consider the possibility that multiple progenitor channels may lead to the production of sne iax .
in fact , the forty - some objects classified as sne iax have a number of similarities , but also noteworthy differences .
in particular , they show a large range in luminosity at maximum , from @xmath10 @xmath11 @xmath1214.2 mag of the faint sn 2008 ha @xcite to @xmath10 @xmath11 @xmath1218.5 mag of sne 2009ku @xcite and 2012z @xcite .
the ejecta velocities near maximum brightness also exhibit a large spread , ranging from @xmath13 to @xmath14 km s@xmath3 .
for the majority of sne iax , there appears to be a correlation between ejecta velocity and peak luminosity , with the higher - velocity objects being also the brighter ones @xcite .
however , sn 2009ku , a low - velocity , high - luminosity sn studied by @xcite , does not follow the trend ( note however that the first spectrum was taken long after maximum and so the inferred ejecta velocity is uncertain , see * ? ? ?
* ) . in this paper
, we present the results of a comprehensive observational campaign of sn 2014ck , which started well before maximum light .
it turns out that sn 2014ck is an outlier among sne iax , as it mirrors sn 2002cx from a photometric point of view , while the early spectra exhibit extremely narrow spectral lines , indicating very low expansion velocities of the ejecta .
this paper is organised as follows : in section 2 , we give some basic information about the sn discovery and the host galaxy , and we describe the follow - up campaign . in section 3 ,
we analyse _ hst _ pre - discovery images .
we discuss data reduction and present the photometric evolution and visual and near - infrared spectroscopic sequences of sn 2014ck in section 4 . in section 5 the galactic and host galaxy reddening is estimated .
descriptions of the photometric and spectroscopic properties of sn 2014ck are reported in sections 6 and 7 , respectively .
expansion velocities of the ejecta , along with the photospheric temperatures , are deduced from the spectra .
spectral modelling with the synow code is used to assist in line identification .
a final discussion of the available data in the context of the explosion models follows in section 8 .
sn 2014ck was discovered by the lick observatory supernova search ( loss ; * ? ? ?
* ) , on 2014 june 29.47 ut , at an apparent magnitude of 16.4 mag using the katzman automatic imaging telescope ( kait ; * ? ? ?
a marginal detection on 2014 june 24.5 ut was also reported by loss with an approximate @xmath15-band magnitude of 17.0 mag . however , a subsequent analysis of kait images on 2014 june 13 , 23 , 24 , 25 , 28 and 29 , performed independently by the loss team ( zheng 2016 , private communication ) and by us , postpones this marginal detection by approximately one day ( 2014 june 25.5 ut , with an approximate @xmath16-band magnitude of [email protected] mag , as reported in table [ sloan ] ) .
the sn is located 43 e and 05 s from the centre of the spiral galaxy ugc 12182 ( figure 1 ) .
an heliocentric recessional velocity of 1490 km s@xmath3 for ugc 12182 is listed in the nasa / ipac extragalactic database ( ned ) , as taken from `` the updated zwicky catalogue '' @xcite .
the distance and distance modulus ( adopting @xmath18 km s@xmath3 mpc@xmath3 ) , corrected for the virgo , great attractor and shapley infall , are @xmath19 mpc and @xmath20 mag , respectively @xcite .
we note that the correction for virgo infall ( see appendix a in * ? ? ?
* ) includes two components : the correction for the infall to virgo plus the vector contribution due to the local group s peculiar velocity with respect to virgo .
we also note that in the local group the radial peculiar velocity dispersion is estimated to be @xmath21 km s@xmath3 ( see for example * ? ? ?
* ) , which accounts for about 25% of the total error budget on @xmath22 .
soon after discovery , spectroscopic classifications of sn 2014ck were obtained independently at the lick observatory and under the asiago classification program @xcite .
the earliest spectrum indicated it was a sn iax on the rise @xcite , resembling sn 2005hk @xcite , sn 2008 ha @xcite and sn 2010ae @xcite .
given the relatively small number of well - observed sne iax in the literature and the early detection and classification , we initiated a follow - up campaign aimed to collect detailed optical and near - infrared ( nir ) observations using several telescopes available to our collaboration .
basic information of sn 2014ck and its host galaxy are summarised in table [ info ] . following the discussion in section 6 ,
we adopt a @xmath23-band maximum estimate of @xmath24 , which we use as a reference epoch throughout this work .
[ info ] .basic information on sn 2014ck and its host galaxy , ugc 12182 .
[ cols= " < , < , < , < " , ] km s@xmath3 . the blending of lines can produce a systematic shift of features and increase the uncertainty at least to @xmath25 km s@xmath3 . ] one of the main properties of sne iax is the low expansion velocity of the ejecta , which suggests low explosion energies compared to normal sne ia ( see section [ bolometric ] ) . the ejecta velocity of sn 2014ck was estimated from the location of the absorption minima of spectral lines with little line blending , based on line identifications from synow .
the results are listed in table [ tbb_vel ] and plotted in figure [ vel_lines ] .
early spectra can be used to probe the velocity distribution of various elements , i.e. unburned ( c+o ) material , intermediate mass elements ( imes ) and completely burned elements close to nuclear statistical equilibrium ( nse ) , namely iron , cobalt and nickel . in principle , the velocity evolution can provide solid constraints on the explosion physics , as a layered structure might be revealed ( a signature of detonations , * ? ? ?
* ; * ? ? ?
* ) unless extensive mixing has destroyed the original stratification ( a signature of deflagrations , * ? ? ?
however , for sne iax there is severe blending of lines over the full optical and nir spectral range that prevent secure line identifications and plague our velocity estimates @xcite . before maximum light ,
expansion velocities are measured from the minima of ca ii h&k and @xmath268498 ; c ii @xmath66580 , 7234 ; s ii @xmath65454 , 5640 ; and si ii @xmath266355 absorption features and are found to lie between 2800 and 4100 km s@xmath3 .
mg ii @xmath264481 soon becomes blended with fe iii and is not easily distinguished . on the contrary ,
the o i line at @xmath267774 is clearly detected redward of the telluric @xmath27-band at 7590 - 7650 and the calculated velocities are similar to the ones inferred for si ii . at these early phases ,
iron features fe ii @xmath266149 , 6247 and 4555 exhibit consistent line velocities that are @xmath25 km s@xmath3 higher than those of imes and fe iii @xmath28
. however , we note that complex blending with emerging co ii and ti ii features could change the position of fe iii @xmath28 ( and also fe iii @xmath29 ) absorption minima . indeed , around maximum light , the blending with emerging fe ii , ti ii and co ii lines might broaden the observed profile and shift the middles of several lines .
in particular , even during pre - maximum phases , the feature around 6300 , mainly attributed to si ii @xmath266355 , could be a blend with fe ii @xmath30 ( or a more complex blending either with fe ii plus co ii or s ii , see * ? ? ?
* their figures 11 and 12 ) .
after maximum , the `` iron curtain '' prevents the secure identification of unburned ( c+o ) material or imes , forming at similar or higher velocities @xcite , and the velocity measurements might be ill constrained .
close to maximum , the absorption - minimum velocities of the co ii nir lines at 1.5759 , 1.6064 and 1.6361 are in good agreement either with si ii @xmath266355 or s ii @xmath65454 , 5640 , and are @xmath31 km s@xmath3 lower than for fe ii @xmath32 .
about twenty days after maximum , the velocities of the co ii nir features are systematically @xmath33 km s@xmath3 lower than those inferred from optical fe ii lines at the same phase .
a similar trend was noted by @xcite for sn 2010ae .
hereafter the line velocities evolve rather slowly .
overall , the velocity structure of sn 2014ck indicates outer layers rich in iron - group ions , while c+o elements , si ii and ca ii , identified at lower velocities , seem to be well mixed . in principle
, they should be present even at higher velocities ( i.e. in the outer layers ) if earlier spectra were available for an in - depth analysis .
consequently , we can not exclude either a mixed or a layered structure for sn 2014ck and , in turn , it is not easy to discriminate between the different explosion mechanisms ( see section [ discussion ] ) . in table
[ tbb_vel ] we also list our estimates of the photospheric temperature of sn 2014ck as derived from a blackbody fit to the spectral continuum ( the spectra were corrected for the redshift and extinction ) . at phases beyond @xmath34 d ,
emerging emission lines and line blanketing drive a flux deficit at the shorter wavelengths , and the fit becomes difficult .
the errors are estimated from the dispersion of measurements obtained with different choices for the spectral fitting regions .
the early photospheric temperature of sn 2014ck is above 8000 k , but it decreases quickly to @xmath35 k at maximum light and flattens to about 4000 k afterwards .
a late - phase spectrum of sn 2014ck was obtained at @xmath36 d and is plotted in figure [ fig_spectra4 ] . as already remarked for other sne iax @xcite , the late - phase spectrum does not appear to be truly nebular despite no clear indication of a continuum or absorption lines . at these late epochs
, sn 2014ck shows narrow permitted fe ii lines superimposed on a pseudo - continuum and several forbidden lines associated with of fe , co and ca .
the dominant feature is the ca ii nir triplet , but also comparable in strength is the forbidden [ ca ii ] @xmath67292 , 7324 doublet ( see figure [ fig_spectra4 ] ) .
both permitted and forbidden calcium lines are significantly more prominent in sn 2014ck than in sn 2002cx , and are comparable to those of the fainter sne 2008 ha and 2010ae @xcite .
the [ fe ii ] @xmath267155 line is the strongest iron feature in the late - time spectrum .
a relatively broad hump at 4700 is identified as [ fe iii ] and [ co ii ] .
a number of features blueward of 5800 are likely a blend of permitted fe ii lines .
the same lines are present at earlier epochs but at higher velocities .
this identification was suggested by @xcite for sn 2002cx . as a test
, we attempted to fit the late time spectrum with a synow model , including fe ii , ca ii , na i and o i ions ( see * ? ? ?
* their table 2 and their figures 3 and 4 ) .
although the observed spectrum is not fully reproduced by the photospheric model , the synthetic spectrum provides a good match to many of the absorption features blueward of 5800 , and possibly the p - cygni profiles of na i d at @xmath65890 , 5896 .
an alternative identification of this feature could be [ co iii ] @xmath265888 @xcite , also suggested for sn 2012z by @xcite .
this last interpretation is supported by the unambiguous presence of other [ co iii ] lines in the 6000 region .
we conclude that the late - time spectrum of sn 2014ck is a combination of p - cygni profiles of recombination lines and emission lines of forbidden fe , ca and co features , but no [ o i ] @xmath66300 , 6364 emission .
this feature is typically not present in late spectra of sne ia @xcite . in order to get [ o i ] emission
, we need a significant amount of o in a region where @xmath37-rays are being absorbed , and the o - emitting region can not be too contaminated by ca .
in fact , the [ ca ii ] @xmath67291 , 7324 feature can limit the strength of [ o i ] @xmath66300 , 6364 emission from a region in which both these ions co - exist @xcite .
the emission of [ o i ] @xmath66300 , 6364 is absent from relatively late spectra of sne 2014ck and 2008 ha @xcite , while in both cases o i @xmath267774 absorption is identified in photospheric phase spectra .
the fwhm values were estimated as 1900 , 1200 and 1600 km s@xmath3 for the [ ca ii ] , [ fe ii ] and ca ii nir lines , respectively .
the nebular lines have slightly diverse velocity shifts : about @xmath38 km s@xmath3 for [ fe ii ] @xmath67155 , 7453 , @xmath39 km s@xmath3 for [ ca ii ] @xmath67291 , 7324 and @xmath40 km s@xmath3 for [ ni ii ] @xmath267376 ( the lastter is difficult to measure as it is in the wing of [ ca ii ] ) .
as already noted by @xcite for sne 2002cx , 2005hk , 2008a and 2008ge , the [ fe ii ] and [ ca ii ] features have shifts in opposite directions , highlighting a quite complex velocity structure of sne iax .
d ( black ) and our best - fit synow synthetic spectrum ( red ) . the contribution of prevalent ions is also shown .
telluric regions are indicated with the @xmath41 symbol and vertical grey bands .
inset : close - up of the @xmath42-band spectral region , showing the ubiquitous signature of co ii .
( a colour version of this figure is available in the online journal ) . ]
the nir spectral evolution of sn 2014ck is presented in figure [ spec_nir ] . before maximum light ,
large parts of the spectra resemble the infrared tail of a hot blackbody continuum , with the exception of few spectral features around 1 and humps between 1.5 and 1.8 . from @xmath43 d to @xmath44 d , the most prominent features are attributed to fe ii , in particular the stronger line with a p cygni profile at @xmath45 is fe ii 0.9998 . mg ii 0.9227 and 1.0927 ( and possibly weaker lines around 2.4 due to mg ii transitions , i.e. 2.4041 , 2.4044 , 2.4125 , see * ? ? ?
* ) produce shallow notches partially blended with fe .
moreover , also around 1 , there might be traces of c i lines at 0.9093 , 0.9406 , 1.0693 and 1.1754 , which have been tentatively identified both in the sub - luminous type ia sn 1999by @xcite and in the type iax sn 2012z @xcite , in addition to normal sne ia ( hsiao et al .
2013 , 2015 ) .
however , no confident carbon detections can be made in sn 2014ck nir spectra .
o i 0.9264 should be blended with mg ii 0.9227 .
actually , in our optical spectra the o i @xmath267773 which is expected to be 320 times stronger than the 0.9264 line is already weak , so the o i 0.9264 is not expected to be a strong feature .
the 0.8446 o i line may contribute to the absorption , dominated in later phases by the ca ii nir triplet .
signatures of si ii might be present blueward of 1 ( 0.9413 ) and the 1.6930 si ii line may be part of the hump at these wavelengths , together with mg ii ( 1.6787 ) and emerging co ii lines . a synow fit ( adopting @xmath46 k and @xmath47 km s@xmath3 , see section [ bb ] ) of the @xmath48 d spectrum was used to assist for the above line identification , including an extended set of ions ( c i , c ii , o i , mg i , mg ii , si ii , si iii , ca ii , fe ii , fe iii and co ii , see figure [ spec_nir ] ) .
the inset of figure [ spec_nir ] shows the most prominent feature in the earliest spectrum , attributed to fe ii 0.9998 .
three weeks later , the nir spectrum radically changes , being strongly dominated by co ii lines , as previously documented in all other sne iax with similar data @xcite .
co ii clearly contributes with numerous lines between 1.6 and 1.8 soon after maximum light , as it is already present in the spectrum taken at @xmath44 d. spectra obtained at @xmath49 d or later show distinct absorption at the location of several co ii lines , most prominently at 1.5759 , 1.6064 , 1.6361 , 1.7772 , 1.7462 , 2.2205 , 2.4596 , 2.3613 .
the increasing strength of co ii with time is attributed both to a lower opacity and a higher abundance of @xmath5co in the external ejecta compared with sne ia @xcite .
the synow fit to the @xmath49 d spectrum of sn 2014ck is plotted in figure [ synow_nir ] .
we adopted @xmath50 k and @xmath51 km s@xmath3 ( see section [ bb ] ) and included a smaller subset of the above imes and fe - group ions ( co ii , fe ii , ca ii , si iii , o i and mg ii ) . while it is confirmed that co ii largely dominates the spectrum redward of 1.5 ( @xmath42- and @xmath52-bands ) , fe ii prevails blueward of that wavelength , as in the spectra of sne ia during the transition from the photospheric to the nebular phase @xcite .
the ca ii nir triplet is also a prevalent feature starting from the spectrum at @xmath49 d , and si iii might help to improve the fit of the features between 1.1 and 1.4 ( as in sn 2012z , see * ? ? ?
the nir spectral evolution of sn 2014ck is reminiscent of what is observed in sn 2010ae ( see * ? ? ?
* their figure 4 ) and , overall , in normal sne ia @xcite . for both sne ia and sne
iax the characteristic @xmath42- and @xmath52-band iron - peak complex rapidly emerges soon after @xmath8-band maximum and becomes the dominant feature in the nir .
similarly to sn 2008 ha @xcite , the pre - maximum spectra of sn 2014ck show the signatures of si ii @xmath66347 , 6371 , s ii @xmath65454 , 5640 ( together with ca , o , na and fe ) and no sign of hydrogen or helium . as in sn
2008 ha , unburned ( c+o ) material , specifically c ii @xmath66580 , 7234 is also detected , as is c iii @xmath264647 .
such analogies in the early phases might suggest they share analogous physical conditions and composition .
however , unlike sn 2008 ha , the late spectrum of sn 2014ck is dominated by iron - peak elements , with the presence of both permitted fe ii and forbidden [ fe ii ] , [ fe iii ] and [ co ii ] emission lines .
it also displays relatively strong features of both the ca ii nir triplet and [ ca ii ] @xmath67291 , 7324 , which have a comparable strength in sn 2008 ha but are weak or absent in sn 2002cx . in the nir spectral region , sn 2014ck shows features typical of sne iax , with the ubiquitous presence of co ii .
fe ii lines are present in sn 2014ck at higher expansion velocities than imes and unburned c+o elements . before maximum , ca ii lines velocities are bracketed by measures of fe ii , fe iii and s ii .
, @xmath53 and photospheric velocities @xmath54 derived around + 10 days for several well studied sne iax ( see narayan et al .
2011 , their figure 3 ) . ] mirroring the behaviour already observed for sn 2009ku by @xcite , the high luminosity and low ejecta velocity for sn 2014ck is contrary to the trend generally seen in the heterogeneous group of peculiar sne iax . in fact , from a photometric point of view sn 2014ck resembles sn 2002cx , the prototype of the class , showing similar peak luminosities ( @xmath0 mag for sn 2014ck vs. @xmath55 mag for sn 2002cx ) and decline rates ( @xmath56 mag vs. @xmath57 mag ) in all bands . given the almost identical bolometric luminosities at maximum ,
the synthesised mass of @xmath5ni is nearly the same for both objects , @xmath58 . yet despite the relatively high peak luminosity and slow light curve decline , the spectra of sn 2014ck are more similar to those of the three - magnitudes - fainter sne 2008 ha @xcite , 2007qd @xcite and 2010ae @xcite .
all these objects exhibit narrow spectral lines indicating low expansion velocities of the ejecta ( @xmath54 from 2500 to 3000 km s@xmath3 at maximum from the absorption minima of si ii ) . @xcite and @xcite suggested the existence of a relation between peak luminosity , ejecta velocity and light - curve shape , with higher - velocity sne iax being more luminous and more slowly declining . on this basis , they argued that the full class of sne iax might originate from a single explosion mechanism . both sne 2009ku and 2014ck , with their low velocity and relatively high luminosity , are outliers in these relations , and the results of this work confirm and reinforce the results of @xcite .
this is illustrated in figure [ conf ] , which shows the positions of a sample of sn iax in a three - variable phase space composed of @xmath59 , @xmath53 and @xmath54 measured 10 days after maximum ( see * ? ? ?
* their figure 3 ) .
in fact , this sample of sne iax shows : ( i ) there is not a monotonic correlation between @xmath54 and @xmath60 , as at low velocities the spread in the decline rate is wide ( top - left panel ) ; ( ii ) there is evidence of a linear correlation between @xmath60 and @xmath53 ( bottom - left ) ; and ( iii ) sn 2014ck ( and also sn 2009ku ) are at odd with the suggestion of @xcite that higher velocity sne iax also have higher peak luminosities ( bottom - right panel ) .
we note that these are still small - number statistics , and therefore these findings need further confirmation .
however , at present we can not conclude sne iax are a one - dimensional sequence from fainter to brighter events .
comprehensive explosion models are required to describe both the diversity and homogeneity among sne iax . on the whole ,
the observations seem to indicate that they may originate from a homogeneous population @xcite
. however , specifically , there appears to be a diverse range of rise times , peak luminosities , expansion velocities , etc . , and , as underlined above , there is not a clear correlation among these physical parameters .
thus , there may be multiple progenitor paths and explosion mechanisms to create a sn iax .
to date , the progenitors of sne iax are still a subject of debate @xcite .
the search and detection of progenitor stars or systems in pre - explosion images is , in principle , a promising technique to test different progenitor models @xcite . for sn 2014ck ,
archival _ hst _ images were obtained and no progenitor was detected at the sn position .
the available images are not very deep and provide a @xmath61 limit of @xmath62 mag .
this limit allows us to rule out only the most - luminous wolf - rayet stars as a potential progenitor .
overall , the optical and nir observational features summarised above favour a thermonuclear explosion of a c / o wd for sn 2014ck . a failed deflagration model
@xcite might explain most of its observed properties , from the low peak luminosity and energetics ( see section [ bolometric ] ) to the spectral characteristics ( see section [ spec_evol ] ) .
in particular , the moderate @xmath63 production derived for this event and the low expansion velocities are well matched in this scenario .
the failed deflagration is too weak to unbind the wd , leaving behind a gravitationally bound , compact remnant around @xmath64 .
the low ejecta mass of sn 2014ck is partially consistent with that prediction .
the rise time predicted by these models ( see table 4 in * ? ? ?
* ) is too fast compared to the rise time derived for sn 2014ck ( @xmath65 days , see section 6.2 ) . above all , the modelled burning of a c / o wd via a turbulent deflagration flame produces a homogeneous mixing of elements in the ejecta , with unburned material , partially burned material and fully burned ( to the iron peak ) material throughout the ejecta .
significant mixing in the ejecta was suggested for sn 2005hk by @xcite .
the failed deflagration model was also considered the most probable scenario for sn 2012z by @xcite . on the contrary , for the same event , @xcite pointed out evidence of a layered structure for calcium , silicon and magnesium , which is instead the signature of a detonation .
thus , they suggested that these elements are produced in a detonation phase after the mixing has already occurred and the majority of the iron peak elements have been produced in a previous deflagration phase , i.e. a pulsational delayed detonation ( pdd ; * ? ? ?
* ; * ? ? ?
looking at the velocity distribution of elements in sn 2014ck ( figure 12 ) , the presence of fe ii features up to high velocities , as reported by @xcite for sn 2012z ( and also for few normal sne ia ) , seems to suggest a layered structure in the ejecta , arguing in favour of a detonation phase , which might have followed a deflagration . on the other hand , c+o elements and
si ii are moderately mixed , and one could argue that they would be identified at higher velocities if earlier spectra were available .
moreover , explaining iron - group material in the outer layers is still a challenge for explosion models , even if some results have been obtained within a delayed - detonation scenario ( see for example * ? ? ?
* and references therein ) .
so far , several questions are still open , among them the mechanism of transition from deflagration to detonation . to be thorough , it is difficult to explain within pdd models the extremely low photospheric velocity at maximum of sn 2014ck .
we also underline , once again , that the severe blending across the spectra of sne iax , might prevent secure line identifications , as was detailed by @xcite .
consequently , the derived expansion velocities of the elements is ill constrained , as well as the distinction between a layered or a mixed structure of the ejecta .
the analysis of our extended data set of sn 2014ck can not exclude either the failed deflagration or the pdd models for this event .
in addition , the comparison with other sne iax highlights that the diversity within this class of transients can not be reduced to a one - parameter description .
this may also imply that distinct progenitors and/or explosion mechanisms are involved , despite the overall similarity of the main observables .
this paper is based on observations made with : the lcogt 2.0 m ( haleakala , hawaii , usa ) and 1.0 m ( mcdonald observatory , texas , usa ) telescopes ; the 2.56 m nordic optical telescope ( la palma , spain ) ; the inaf osservatorio astronomico di padova copernico 1.82 m telescope ( mt . ekar , asiago , italy ) ; the 3.58 m telescopio nazionale galileo ( la palma , spain ) ; the 8.1 m gemini - n telescope ( hilo , hawaii , usa ) ; and the 10.4 m gran telescopio canarias ( la palma , spain ) .
we acknowledge the staff at inaf oapd in asiago and at lcogt for their support . based on observations obtained at the gemini observatory under program i d gn-2014a - q-8 and gn-2014b - q-41 .
gemini is operated by the association of universities for research in astronomy , inc .
, under a cooperative agreement with the nsf on behalf of the gemini partnership : the national science foundation ( united states ) , the national research council ( canada ) , conicyt ( chile ) , the australian research council ( australia ) , ministrio da cincia , tecnologia e inovao ( brazil ) and ministerio de ciencia , tecnologa e innovacin productiva ( argentina ) . also based on observations made with the nasa / esa hubble space telescope , obtained from the data archive at the space telescope science institute .
stsci is operated by the association of universities for research in astronomy , inc . under nasa contract
nas 5 - 26555 .
we thank weikang zheng and alex filippenko for sending us pre - discovery and discovery lick / kait images of ugc 12182 from which non - detection upper limits of sn 2014ck were determined and the rise time estimated .
we also thank the anonymous referee for the thorough review of the paper .
ap , sb , ner , ltar , gt and mt are partially supported by the prin - inaf 2014 with the project
transient universe : unveiling new types of stellar explosions with pessto " .
ner acknowledge the support from the european union seventh framework programme ( fp7/2007 - 2013 ) under grant agreement n. 267251 `` astronomy fellowships in italy '' ( astrofit ) .
amg acknowledges financial support by the spanish _ ministerio de economa y competitividad _ ( mineco ) , grant esp2013 - 41268-r .
ms and eyh acknowledge support provided by the danish agency for science and technology and innovation through a sapere aude level 2 grant . 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 ; iraf packages , distributed by the national optical astronomy observatory , operated by the associated universities for research in astronomy , inc .
, under cooperative agreement with the national science fundation .
|
we present a comprehensive set of optical and near - infrared photometric and spectroscopic observations for sn 2014ck , extending from pre - maximum to six months later .
these data indicate that sn 2014ck is photometrically nearly identical to sn 2002cx , which is the prototype of the class of peculiar transients named sne iax .
similar to sn 2002cx , sn 2014ck reached a peak brightness @xmath0 mag , with a post - maximum decline - rate @xmath1 mag .
however , the spectroscopic sequence shows similarities with sn 2008 ha , which was three magnitudes fainter and faster declining . in particular , sn 2014ck exhibits extremely low ejecta velocities , @xmath2 km s@xmath3 at maximum , which are close to the value measured for sn 2008 ha and half the value inferred for sn 2002cx .
the bolometric light curve of sn 2014ck is consistent with the production of @xmath4 of @xmath5ni .
the spectral identification of several iron - peak features , in particular co ii lines in the nir , provides a clear link to sne ia .
also , the detection of narrow si , s and c features in the pre - maximum spectra suggests a thermonuclear explosion mechanism .
the late - phase spectra show a complex overlap of both permitted and forbidden fe , ca and co lines .
the appearance of strong [ ca ii ] @xmath67292 , 7324 again mirrors the late - time spectra of sn 2008 ha and sn 2002cx .
the photometric resemblance to sn 2002cx and the spectral similarities to sn 2008 ha highlight the peculiarity of sn 2014ck , and the complexity and heterogeneity of the sne iax class .
[ firstpage ] supernovae : general supernovae : individual : sn 2014ck , sn 2006fp galaxies : individual : ugc 12182
|
in the @xmath5 limit the behaviour of quantum systems can be expressed in terms of the classical phase space trajectories .
for one dimensional scalar problems the wkb method @xcite yields semi - classical wavefunctions and energy levels , an approach that can be generalized to both integrable multi - dimensional systems ( ekb torus quantization @xcite ) , as well as to systems with multi - component wave functions@xcite . in the latter case , however , the multi - component wavefunction matching inherent in the wkb approach rapidly becomes prohibitively cumbersome as the number of components increases .
the gutzwiller trace formula circumvents this matching problem by giving a direct expression for the semi - classical density of states .
it is , furthermore , valid for systems with a non - integrable ( i.e. , chaotic ) classical limit .
while generalizations of the trace formula for the multi - component case have been presented for specific matrix hamiltonians such as the relativistic dirac hamiltonian@xcite and the dirac - weyl hamiltonian of graphene@xcite , a general multi - component version of the trace formula has not been considered . in the solid state theory context , however , a semi - classical method applicable to _ arbitrary matrix hamiltonians _ would be extremely useful .
this is so for two reasons .
firstly , in many systems , for instance few layer graphenes , topological insulators , and semiconductors , one encounters multi - band effective hamiltonians and hence multicomponent wavefunctions .
secondly , and perhaps most importantly , in the emerging class of low dimensional materials one very often encounters structural deformations occurring on length scales far in excess of the lattice constant .
such deformations are very difficult to treat fully quantum mechanically ( due to the huge unit cell sizes involved ) while at the same time presenting a natural case for a semi - classical treatment ( due to the slowly varying spatial deformation ) .
examples include flexural ripples in graphene@xcite , rotational stacking faults in few layer graphene systems@xcite , the recently discovered partial dislocation networks in few layer graphenes@xcite , and graphene nanostructures for which interesting semi - classical work already exists@xcite .
a general semi - classical approach for treating such systems thus has the potential of providing a very useful investigative tool .
the purpose of the present paper is therefore twofold : ( i ) to generalize the gutzwiller trace formula to the case of arbitrary matrix - valued hamiltonians and ( ii ) to demonstrate that the semi - classical approach yields an _ accurate and tractable _ scheme for the treatment of deformations in graphene based systems . to that end
, we will first focus on fundamental theory and some simple applications , and in the final part of the paper consider application of the theory to a realistically complex example of a deformation in bilayer graphene , a one dimensional strain moir .
let us briefly outline the differences between the matrix - valued case and the scalar case within a semi - classical treatment
. at @xmath6 a scalar schrdinger equation reduces to the hamilton - jacobi equation of classical mechanics . for a @xmath0 hamiltonian ,
however , there are two important differences in the @xmath6 classical structures .
firstly , we obtain @xmath2 hamilton - jacobi equations , some of which may be identical .
this situation arises , for instance , in the @xmath7 limit of the dirac equation@xcite , where each of the two hamilton - jacobi equations ( for particles and antiparticles ) is twice degenerate - in the limit @xmath7 there is no spin and hence one obtains two pairs of degenerate equations . as for the case of ordinary perturbation theory , such degenerate cases require a special treatment .
secondly , the hamiltonian phase space is found to be endowed with a classical berry curvature arising from anholonomy in the transport of the eigenvectors @xmath8 of @xmath3 around the classical orbits .
in contrast to the scalar case , for multi - band quantum hamiltonians the semi - classical particle types have `` internal structure '' ( pseudospin structure in the case of graphene , for example ) and it is this that is encoded in the classical eigenvectors of @xmath9 ; the classical berry curvature is thus a direct consequence of the internal degrees of freedom of the underlying quantum system . at @xmath10
we obtain an equation for the amplitude that is transported along the classical orbits described by the hamilton - jacobi equations of motion .
in contrast to the scalar case , for a matrix hamiltonian this amplitude acquires an addition phase resulting from the @xmath6 berry curvature which , for non - degenerate systems , consists of ( i ) a geometric phase and ( ii ) a dynamical phase acquired by transport of the classical particle through this berry curvature .
the geometric part of this phase depends only on individual hamilton - jacobi orbits , while the dynamical phase couples all the @xmath2 hamilton - jacobi systems . for degenerate systems one finds the non - abelian analogue of the berry curvature and phase .
the existence of a dynamical phase related to the berry curvature ( in addition to the usual geometric phase ) is somewhat unusual , and reflects the fact that we have a transport equation for the amplitude function and not the semi - classical vectors @xmath8 .
our approach differs from that taken by carmier and ullmo@xcite in their semi - classical study of graphene and 2-band bilayer graphene in that their description begins with the energy dependent greens function , which then enforces a complex matching procedure for the boundary condition of this function ; as in the case of the wkb method , this becomes prohibitively cumbersome for a general matrix valued hamiltonian .
following bolte and keppeler@xcite we implement the boundary conditions for the greens functions at the level of the time dependent greens functions , which allows an elegant solution that circumvents the boundary matching problem .
the procedure leading from an arbitrary matrix valued hamiltonian to the density of states is presented as an explicit set of steps , and we apply it to a number of cases where the exact solution is known : the landau spectra of graphene , a 4-band model of bilayer graphene , and silicene .
as one would expect , the agreement between the exact and the semi - classical results becomes considerably degraded at low energies , and in particular for bilayer graphene the zero mode found in the exact solution is not captured within the semi - classical approximation ( and is , of course , also not captured in the 2-band approximation to this problem@xcite ) .
it is therefore by no means obvious that a semi - classical approach is suitable for graphene based systems with slowly spatial deformations , as one is always interested in the low energy behaviour .
to explore this situation more closely we consider a realistic example of such a deformation : a one dimensional strain moir in bilayer graphene , which serves as an instructive analogue of the graphene twist bilayer@xcite , two mutually rotated layers of graphene .
the twist bilayer displays extraordinarily rich electronic structure in small angle limit @xcite ( i.e. , as the moir length becomes large ) , and in particular a strong electron localization on the aa stacked regions of the lattice .
we present an analytical semi - classical analysis of the strain moir , finding that : ( i ) at the dirac point the action is orders of magnitude larger than @xmath11 when the moir length becomes large compared to the lattice constant - thus validating the semi - classical approach - and ( ii ) that the electron localization is driven by the existence of a _ semi - classical potential well _ centered at the aa spots , that arises from the stacking potential in the quantum hamiltonian . we should stress that this `` potential well '' picture , which provides a very natural description of charge localization , is fundamentally semi - classical : no such `` potential well '' could localize quantum mechanically due to the klien paradox which prevents localization of electrons in graphene by a scalar potential .
we furthermore find an analytical form for the semi - classical density of states arising from electrons trapped in this potential well , which we show to be in almost perfect agreement with exact quantum mechanical calculations .
this demonstrates both that the semi - classical approach provides a valid tool for investigating slow deformations in few layer systems , as well as the insight it can bring into the physics of these rather complex systems .
it should be stressed that treating such a system on the basis of either the standard wkb approach ( in principle possible as we have an effective one - dimensional system ) , or that of ref .
, could not be contemplated due to the extraordinarily complexity of the matching procedure that would be involved .
we consider schrdinger s equation for the greens function @xmath12 with a _ matrix - valued _
hamiltonian @xmath13 @xmath14 and the boundary condition @xmath15 we assume that @xmath13 is analytic in @xmath16 , or more precisely that for all smooth test functions @xmath17 the matrix @xmath18 is analytical in @xmath19 .
we further assume that the hamiltonian may be decomposed as @xmath20 , in which @xmath21 is a matrix containing only derivatives ( such as momentum operators ) and @xmath22 contains only coordinate functions such as potentials .
this restriction is easily lifted without qualitatively changing what follows , however it covers most physical situations and simplifies the analysis so we retain it .
finally , we introduce the following convention : the _ primed _ coordinates are _ initial _ coordinates with the _ non - primed _ ones the _ final _ coordinates on a classical trajectory .
+ let us first derive an expression for the time dependent greens function using a generalization of the ansatz provided by bolte and keppeler @xcite .
if @xmath13 is a @xmath0 matrix we search for the _ time - dependent greens function _ of the form @xmath23 this ansatz is completely general and could describe the exact solution to eq .
( [ sg1 ] ) , however the @xmath24 term allows a wkb - like expansion in orders of @xmath11 and , due to the integral in eq .
( [ bkan ] ) , one can implement the boundary condition eq . by requiring @xmath25
we now assume that @xmath26 is analytical in @xmath11 and insert eq . into the schrdinger equation eq . and collect terms of @xmath6 .
this procedure leads to the _ zero eigenvalue condition _ @xmath27 ( we may drop the sum and integral in eq . as the equation is linear . )
there are evidently @xmath2 solutions to this equation which occur when the @xmath28 are equal to the @xmath2 orthonormal eigenvectors of @xmath29 . denoting the corresponding eigenvalue by @xmath30
, we find that the zero eigenvalue equation implies the existence of @xmath2 hamilton - jacobi equations @xmath31 from which we then immediately identify for system @xmath32 the following objects : @xmath30 is a classical scalar hamiltonian , the real scalar function @xmath33 a classical action , and @xmath34 the corresponding canonical momentum .
thus a general @xmath0 quantum hamiltonian leads to @xmath2 separate classical systems at order @xmath6 .
the full classical information is contained in the @xmath2 hamilton - jacobi equations ( [ hjeq ] ) and the @xmath2 vectors @xmath28 . the situation becomes more complicated if the matrix @xmath35 has degenerate eigenvalues . in this circumstance recourse to degenerate perturbation theory may be circumvented following the scheme of bolte and keppeler , in which the @xmath36 orthogonalized eigenvectors corresponding to the degenerate eigenfunction @xmath30 are combined into an @xmath37 `` full degenerate eigenvector '' that evidently satisfies @xmath38 . with the initial conditions
@xmath39 the boundary condition of the green s function equation is satisfied and the calculations thereafter proceed in a formally identical way for both the degenerate and non - degenerate cases , with the difference of course that the rank of the various objects in the theory differs between the two cases .
+ we now consider the @xmath10 corrections to the @xmath2 hamilton - jacobi equations ; sewing `` quantum flesh on classical bones '' to use the evocative phrase of berry and mount @xcite .
the derivative operators @xmath16 in @xmath21 are substituted by the canonical momenta @xmath40 upon acting on @xmath41 . to generate the @xmath10 terms from @xmath21
we should allow every matrix element @xmath42_{ij}$ ] to act on the exponential until only first order operator terms remain .
for example , the term @xmath43 would generate @xmath44 .
the resulting matrix operator @xmath45 is therefore linear in first order derivatives . considering @xmath45 as a perturbation , we obtain in first order perturbation theory the @xmath10 equation @xmath46\hat v_\alpha^{(0)}\hat f_\alpha({{\bf r}})=0 , \label{teq1}\ ] ] which , as we shall now show , is a _ transport equation _ for @xmath47 along the classical trajectories governed by the @xmath30 hamiltonian .
( note that substitution of eq . into the schrdinger equation eq .
, collecting terms of @xmath10 , and pre - multiplying by @xmath48 also leads to eq .
[ teq1 ] . ) in the case of degenerate hamilton - jacobi systems @xmath47 will be an @xmath49 matrix ; in the non - degenerate case @xmath47 is a scalar function .
it is not apparent from eq .
( [ teq1 ] ) that the @xmath2 formally independent classical systems found at @xmath10 are coupled at @xmath10 although , as we shall see , different particle types are indeed coupled by the transport equation . to simplify the notation
we now define @xmath50 such that eq .
and eq . take the form @xmath51 and @xmath52 respectively .
we now demonstrate that eq .
may be written in the form of a transport equation familiar from the scalar case .
we first note a rather obvious relation connecting @xmath53 and @xmath54 namely : @xmath55_{lm}=\sum_{k=1}^d\partial_{p_k}\left[h_{cl}\right]_{lm}\partial_k . \label{hdd}\ ] ] this identity follows from deployment of the leibnitz rule on the right hand side that generates all terms from @xmath35 with one momentum variable removed and replaced by the operator @xmath56 , exactly the definition of @xmath57 on the left hand side .
the identity eq . generalizes a similar relation used by carmier and ullmo@xcite in their semi - classical treatment of single layer graphene and the two band model of bilayer graphene . in accordance with convention we set @xmath58 , @xmath59 and make use of the sum convention letting the sum run from @xmath60 , where @xmath61 is the space dimension . we may then compactly write @xmath62 and using this result can write eq .
( [ teq1 ] ) as @xmath63\partial_\mu\hat v_\alpha^{(0 ) } + \hat v_\alpha^{(0)\dagger}\left[\partial_{p_\mu}h_{cl}\right]\hat v_\alpha^{(0)}\partial_\mu\right]\hat f_\alpha({{\bf r}})=0 . \label{eq212}\ ] ] since we do not intend to calculate the higher order corrections to @xmath28 from now on we shall drop the index @xmath64 at all vectors @xmath65 . now , following carmier and ullmo @xcite we split @xmath66 into a hermitian and an anti - hermitian part .
we denote the _ antihermitian _
part @xmath67 : @xmath68 ( note that in the case of an @xmath36-fold degeneracy @xmath69 is a @xmath70 matrix . )
this term includes a semi - classical analogue of a berry phase , as explained in detail later .
by repeatedly applying the leibnitz rule and using the fact that @xmath71 is hermitian and @xmath72 ( recall our assumption of no coordinate functions in front of derivatives in the hamiltonian ) we see that the hermitian part of @xmath66 takes the following simple form @xmath73\partial_\mu \hat v_\alpha ) = \frac{1}{2}\partial_\mu(\partial_{p^\alpha_\mu}\lambda_\alpha ) .
\label{transpeqrealpart}\ ] ] where we have also used the hellmann - feynman theorem @xmath74 . applying the hellmann - feynman theorem to the second part of eq .
we find that the equation simplifies to @xmath75\hat f_\alpha({{\bf r}})=0 . \label{eq2124}\ ] ] from hamilton s equations we have @xmath76 with @xmath77 and so eq .
reduces to the _ transport equation _
@xmath78\hat f_\alpha({{\bf r}})=0 , \label{eq2123}\ ] ] which is similar to the scalar case but _ with an additional matrix @xmath69_. it should be stressed that @xmath69 appears only for matrix - valued hamiltonians .
all terms in eq . are expressed in terms of the classical orbits obtained from the order @xmath6 approximation .
equation must now be integrated along the classical orbits such that @xmath47 is `` transported '' along these orbits .
the first two terms of this expression are scalars , and hence the ansatz @xmath79 , where @xmath80 is a scalar function that solves an auxiliary equation @xmath81g_\alpha({{\bf r}})=0 $ ] , is sensible .
the term @xmath82 is an @xmath70 matrix in case of an @xmath36-fold degeneracy for @xmath30 .
the expression for @xmath83 with initial condition @xmath84 is known from the scalar case ( see , for example , refs . ):
@xmath85 inserting @xmath79 into eq .
yields @xmath86\hat h_\alpha({{\bf r}})=0 , \label{insertion}\ ] ] which is then solved by a _ time - ordered exponential _ ( matrices @xmath69 at different times will in general not commute ) .
we now have all ingredients required to calculate the time - dependent greens function eq . .
the remaining steps of the derivation now follow closely the scalar case , and we will , therefore , be brief in the presentation . from the initial condition eq . for the classical actions @xmath87 it follows that @xmath88@xcite , where @xmath89 are as yet unknown functions .
we evaluate eq . with the stationary phase approximation to find @xmath90 with @xmath91 here the stationary phase applied to the @xmath92 integral has enforced the sum over @xmath93 to be over all classical paths leading from @xmath94 to @xmath95 within the time interval @xmath96 .
the actions @xmath97 are the classical actions that solve the hamilton - jacobi equations with this requirement .
additionally , @xmath98 is the path specific time - dependent maslov index ( or morse index ) that arises from taking the absolute value in the expression for @xmath99 .
the morse index accounts for the sign changes of the determinant under the square root , which occurs when the lagrangian manifold corresponding to the dynamics of the hamiltonian @xmath30 develops a fold ( see , for example , the book of cvitanovic@xcite ) . in a one dimensional problem
this simply corresponds to the turning points of the classical path at which the greens function picks up a phase @xmath100 .
we now determine the retarded energy - dependent greens function @xmath101 within the stationary phase approximation .
the calculation is again almost identical to the scalar case and we thus , for brevity of presentation , refer the reader to the standard literature for the scalar gutzwiller formula ( see , for example , ref . ) and quote the result of the stationary phase approximation for the matrix valued case : @xmath102 where @xmath103 and where the stationary phase condition now sets the summation to be on the _ classical energy shell_. in this expression @xmath104 is now the energy - dependent action , and @xmath105 are all classical paths connecting @xmath95 and @xmath94 at energy @xmath106 . as is customary within a semi - classical formulation we have chosen a coordinate system with one axis parallel ( @xmath107 ) to the trajectory and the other coordinates perpendicular ( @xmath108 ) to the trajectory .
similar to eq .
, @xmath109 is the _ maslov index _ and counts the sign changes of the expression under the absolute value in @xmath110 .
we can now calculate the density of states simply by taking a trace over positions and over the matrix structure @xmath111\right\}. \label{densityofstates}\ ] ] the calculation of the position trace is , once again , identical to the scalar case which , following the standard procedure , evaluated in a stationary phase approximation for perpendicular coordinates and without any approximation ( but assuming isolated orbits ) for the parallel coordinates .
this yields for the density of states @xmath112 where the position trace restricts the summation to the closed orbits and the stationary phase condition picks up only periodic orbits @xmath113 , and where the monodromy matrix @xmath114 is given as @xmath115 and which describes the stability of an orbit with respect to small deviations of the initial positions and momenta .
it is evaluated for the orbit @xmath113 and is independent of the point on the orbit which we choose for its evaluation@xcite . the _ maslov index _ is @xmath116 and it is important to note that , in general , it is not the same as the maslov index in eq . .
for one dimensional systems , however , these indices coincide .
lastly @xmath117 is the _ time _ needed to traverse the _ primitive orbit _ ( primitive means traversing the orbit only once ) .
note that this result is valid only for the extended orbit contributions . for the short orbits
the exponential is not a fast oscillating function and the stationary phase approximation is inadmissible .
equation provides the so - called oscillatory part of the density of states .
for the non - degenerate case the @xmath118 phase is a scalar , and thus eq .
( [ insertion ] ) may be immediately solved to yield @xmath119 , leading to a simpler gutzwiller formula @xmath120,\nonumber\end{aligned}\ ] ] an expression that , in fact , differs from the scalar formula only by the addition of the @xmath69 phase .
the semi - classical @xmath69 phase appearing in the matrix generalization of the gutzwiller equation has been discussed by carmier and ullmo@xcite in the context of graphene and a 2-band model of bilayer graphene .
these authors concluded that for quantum hamiltonians with no mass term , but arbitrary field @xmath121 , this phase was exactly the classical analogue of the adiabatic berry phase . here
we wish to show that this semi - classical phase , that would seem to appear only at @xmath10 , is in fact a manifestation of a deeper @xmath6 classical structure in the semi - classical theory of matrix hamiltonians .
the full form of this phase for the @xmath32 hj system is given by @xmath122\partial_\mu v_\alpha \label{scph}\ ] ] where for simplicity of notation we consider the case of non - degenerate hamilton - jacobi equations ( the generalization is straightforward ) . switching from 4-vector notation to separate space and time derivatives we have for @xmath69 @xmath123\partial_\mu v_\alpha = v^\dagger_\alpha \left[\partial_{p_i } h\right]\partial_i v_\alpha + v^\dagger_\alpha \partial_t v_\alpha\ ] ]
if we insert into the right hand side of this equation @xmath124 we then find for the integrand @xmath125 the first two terms of which are evidently the total time derivative @xmath126 i.e. , represent a berry phase that depends only on the geometry of the classical path ( as the total time derivative allows us to eliminate time from the integral @xmath127 by a change of variables ) .
we write this , for a general @xmath128 parameter space ( @xmath61 is the dimension of space ) and using the notation @xmath129 , as @xmath130 where @xmath131 is the berry curvature tensor , @xmath132 the berry connection for the @xmath32 hamilton - jacobi system , and @xmath133 a hyper - surface in the @xmath128 hamiltonian phase space .
note that the greek indices @xmath134 , @xmath135 now run over the @xmath128 dimensions of phase space in the vector @xmath136 .
what is the third term of eq . ?
evaluation of the derivative @xmath137 and insertion of the identity operator yields @xmath138 ( note that the @xmath139 term is identically zero in this sum ) .
using @xmath140 we may write this as @xmath141 \label{eqph1}\end{aligned}\ ] ] which has some resemblance to the antisymmetric structure of the berry curvature tensor , @xmath142 eq .
( [ eqph1 ] ) , however , contains only terms diagonal in dimension and that have mixed @xmath143 and @xmath144 derivatives , while the curvature tensor contains all possible combinations of these indices .
however , by contracting the curvature tensor with the symplectic matrix @xmath145 we find @xmath146\end{aligned}\ ] ] which is , apart from the weight factor @xmath147 , identical to eq .
( [ eqph1 ] ) .
we may therefore express the semi - classical phase as @xmath148 where we have introduced the @xmath149 state part of the curvature tensor : @xmath150\ ] ] and used the fact that on the @xmath32 orbit we have @xmath151 .
. has a simple visual interpretation .
the first term is the geometric berry phase while the second term is clearly dynamical , and represents the time integral of the classical particle moving through a hamiltonian phase space endowed with a berry curvature .
contraction of the berry curvature with the symplectic matrix yields a scalar , and in the line integral this is , for each semi - classical vector @xmath149 , weighted by the energy separation of the @xmath149 manifold and the @xmath32 particle orbit : @xmath152 .
this term evidently encodes the coupling between the @xmath2 non - degenerate hamilton - jacobi systems
. one might wonder how important is the dynamical part of the semi - classical phase . as we will subsequently show , in the case of silicene ( which has an explicit mass term ) this term in the semi - classical phase considerably improves the accuracy of the theory .
_ structure of the dynamical semi - classical phase _ : the structure of the berry curvature depends , as usual , on the degeneracy structure , in this case of the semi - classical energy manifolds .
it is worthwhile exploring this point , and to that end we insert the formula for the matrix element of the derivative of an eigenvector @xmath153 into @xmath154 to find @xmath155 the structure of @xmath154 is thus dominated by degeneracies amongst the classical eigenvalues .
the geometric part of the semi - classical phase , of course , depends on the global structure of the berry curvature @xmath156 .
this structure is , however , equally important for understanding the dynamical phase . in particular ,
if the berry curvature is a @xmath157-function - the case for which the berry phase is `` topological '' , i.e. , depends only on the winding number around the pole - then any classical trajectory that does not pass through such a source will have zero for the line integral in eq .
( [ scp ] ) . in other words , _ if the semi - classical berry phase is topological then the semi - classical dynamical phase is zero_. a special case of this is the graphene dirac - weyl hamiltonian in the presence of arbitrary @xmath121 but with no mass term @xmath158 .
in such a situation the degeneracy at the dirac point is preserved , the curvature retains the delta function structure , and hence the semi - classical phase will coincide with the berry phase , as stated by carmier and ullmo@xcite . if the hamiltonian has one or more cyclic coordinates , then the stationary phase approximation for these coordinates can not be applied . furthermore ,
if orbits are not isolated the trace over starting positions is not given by the integral @xmath159 as this clearly assumes a single closed orbit ( here @xmath160 , as before , denotes the time period of a primitive orbit i.e. the time to travel once around a closed orbit ) .
we consider here the case where only one coordinate @xmath161 is non - cyclic and indicate how this result may be ( straightforwardly ) generalized .
the method presented here differs from that of carmier and ullmo@xcite in that we derive a solution constructively , beginning at the level of the transport equation , whereas in ref .
[ ] the treatment of orbit degeneracies is performed post a general solution of the transport equation .
the two methods are , however , equivalent .
for the case of a single non - cyclic coordinate the transport equation takes on the much simpler form @xmath162 with the solution @xmath163 this result for @xmath164 gives a new expression for the time - dependent greens function @xmath165 where @xmath166 while similarly for the energy dependent greens function we find @xmath167 where @xmath168 from the energy dependent greens function may then be found the oscillatory part of the density of states : @xmath169 for problems with only one non - cyclic coordinate the 0 length orbit contribution to the density of states has a very similar form @xmath170 where objects with the index @xmath113 label different solutions to the hamilton - jacobi equations .
consistent with the notion that we are treating very small action orbits we have expanded the argument of the dirac @xmath157-function about zero position and momentum , deployed a @xmath171 length approximation @xmath172 , and subsequently used hamilton s equation to arrive at the final result .
it is straightforward to generalize the procedure to problems with arbitrary combinations of cyclic and non - cyclic coordinates with the only change for more than one non - cyclic coordinate is the reappearance of a monodromy matrix in the density of states which is , however , then restricted to the space of the non - cyclic coordinates . in the case that the action , the semi - classical phase , and the maslov index do not depend on initial positions ( and the hamilton - jacobi equations are non - degenerate ) then @xmath173 and @xmath174 may be straightforwardly combined and the dirac comb identity used to yield an expression for density of states in terms of delta functions : @xmath175 where @xmath176 denotes once more a primitive orbit . from the @xmath157-function
one can read off a generalization of the bohr - sommerfeld quantization condition which is given as @xmath177 we briefly present a summary of the steps required to obtain the oscillatory density of states for a generic matrix hamiltonian @xmath178 : _ reverse quantization _ : in the hamiltonian @xmath178 replace @xmath179 .
one thus finds the matrix @xmath35 . _ introducing a set of classical particle types _ : determine the eigenvalues @xmath30 of @xmath35 and corresponding normalized eigenvectors @xmath180 for @xmath181 .
in the case of degenerate eigenvalues @xmath30 orthonormalize the corresponding eigenvectors and write them next to each other as columns giving the `` full eigenvector '' @xmath180 ( an @xmath37 matrix with @xmath36 the degeneracy number of the @xmath32th set of distinct eigenvalues ) . _ solving the classical problems _ : the eigenvalues @xmath30 correspond to hamilton - jacobi equations @xmath182 , which must be solved for the actions @xmath183 of all periodic orbits @xmath184 at energy @xmath106 . _
determining the maslov indices _ : calculate the maslov index @xmath116 for each orbit , which is given by the sum of all sign changes of ( i ) @xmath185 ( the velocity along the orbit ) and ( ii ) @xmath186 ( see eq . for the definition of @xmath187 ) . in a the case of a one dimensional problem
the maslov index is just the number of classical `` wall - reflections '' ( i.e. sign changes of @xmath188 ) along the orbit in phase space . _
calculating the semi - classical phase _ : express the @xmath189 in terms of @xmath190 and use eq . to calculate @xmath69 .
_ the density of states _ : the expressions resulting from the previous steps must then be entered into eq . or , in the case of non - isolated orbits and cyclic coordinates , into eq .
and the integrals over the cyclic coordinates performed .
the procedure for obtaining the semi - classical greens functions is almost the same , however , it includes non - closed orbits and , as is well known ( see for example ref . ) , classification of all possible such orbits is a difficult problem , and this procedure is rarely used to explicitly evaluate the greens function .
the above steps present a systematic `` recipe '' for calculating the semi - classical density of states of an arbitrary matrix valued hamiltonian . in the next two sections we will apply this procedure first to a number of systems for which the exact quantum mechanical result is known ( section iii ) , as well as subsequently ( in section iv ) to a problem , the one - dimension strain moir in bilayer graphene , for which the quantum result may only be obtained numerically ( the semi - classical result , however , remains of simple analytical form ) .
as a first test of the semi - classical procedure outlined in the previous sections we consider a number of cases for which the exact analytical solution is known .
_ single layer graphene _ : we first consider a single layer of graphene in a uniform out - of - plane magnetic field .
the hamiltonian is thus simply the dirac - weyl operator with minimal substitution : @xmath191 and where we have employed the landau gauge so that @xmath192 is a good quantum number of the problem .
this system has been treated by carmier and ullmo@xcite , and we thus omit details of the derivation . the final result , which agrees with that given in ref . , is given by @xmath193 _ 4-band model of bilayer graphene _ : a much more difficult system to treat in any method that employs a matching procedure is the full four - band model of ab stacked bilayer graphene ( in ref .
only the two band down - folded version of the full hamiltonian was treated ) .
we take the simplest model of this material in which the interlayer coupling matrix @xmath194 is independent of momentum ( although lifting this condition would not significantly complicate the analysis ) @xmath195 where @xmath196 describes the interlayer hopping . with the method outlined in the previous section ,
this system yields straightforwardly to a semi - classical analysis as we now show .
we first send @xmath179 and diagonalize the resulting hamiltonian to find the eigenvalues @xmath197^{(1/2 ) } , \label{be}\ ] ] where @xmath198 $ ] and @xmath199 label the four bands ; the low energy chiral bands have @xmath200 and the high energy bonding and anti - bonding bands @xmath201 .
the corresponding eigenvectors are @xmath202 where @xmath203 where @xmath204 .
these are , of course , are formally identical to the standard eigenvalues and eigenvectors of the bernal bilayer .
. [ be ] may be straightforwardly solved for @xmath205 and then the action found from @xmath206 giving @xmath207 from the eigenvectors the semi - classical @xmath69 phase is immediately found to be zero ( the berry phase is also zero in this system ) . note this contrasts with the case of single layer graphene where the berry phase is @xmath208 , and the semi - classical phase equals the berry phase ,
i.e. the dynamical phase is zero .
this is a consequence of the @xmath157-function structure of the berry curvature for single layer graphene .
as the maslov phase is evidently simply 2 , the number of turning points , and as none of these depend on the initial position of the classical orbit , we can deploy eq .
[ bs ] to immediately find the semi - classical spectrum @xmath209 where @xmath210 and @xmath211 .
the exact quantum solution is given as @xmath212\\&\pm\sqrt{\left[c\left(n+\frac{1}{2}\right)+\frac{1}{2}\right]^2-c^2n(n+1)}.\\ \end{split } \label{exactab_stack_levels}\ ] ] for large @xmath2 , the limit in which the semi - classical approximation must hold , we find @xmath213 for our approximation , while from the exact result we find @xmath214 thus the semi - classical result agrees up to @xmath215 with the exact result , and the agreement for smaller magnetic fields is better as can be seen by the fact that the higher order terms depend on @xmath216 , which is essentially the magnetic field . for both of these example problems ( i )
the hamilton - jacobi equations were non - degenerate and ( ii ) the dynamical part of the semi - classical phase vanished ( and thus the semi - classical phase was identical to the berry phase ) .
we now turn to a problem which is both non - degenerate and , as we shall see , one in which the dynamical phase is non - zero .
_ silicene _ : we consider silicene , a two - dimensional allotrope of silicon with a hexagonal honeycomb structure similar to that of graphene .
spin - orbit coupling is more important in this material than in graphene ( where it can generally be neglected ) , and therefore in the treatment that follows the spin - orbit coupling term is included .
the hamiltonian for this system is given as@xcite @xmath217 where @xmath218 is the tensor product for matrices and we use the basis @xmath219 ( see ref . for details of this model hamiltonian ) .
the term proportional to the fermi velocity , @xmath220 , is just the graphene hamiltonian with a mass term @xmath221 , while the term proportional to @xmath222 describes spin - orbit coupling .
we will consider silicene in an uniform out - of - plane magnetic field and therefore introduce into eq .
the minimal substitution @xmath223 . replacing momentum operators by momentum functions @xmath224 in the resulting hamiltonian , yields the matrix @xmath35 , diagonalization of which results in two pairs of twice degenerate eigenvalues and thus twice degenerate hamilton - jacobi equations .
these are given as @xmath225 and evidently describe a relativistic particle with a `` speed of light '' @xmath226 and mass energy @xmath227 .
the corresponding `` full eigenvectors '' are @xmath228 and @xmath229 the action is easily found from these hamilton - jacobi equations , eq .
, to be @xmath230 , and the maslov index to be @xmath231 with @xmath232 the number of circuits of one primitive orbit . from the `` full eigenvectors '' we find the ( matrix valued ) semi - classical phase to be @xmath233 this expression is diagonal ( yet this is generally not the case , see , for example ref . ) . on the other hand the berry phase can also be calculated directly from the `` full eigenvectors '' and is found to be @xmath234 which clearly does not coincide with the full semi - classical phase eq . .
it is interesting to note that both phases coincide in the limit @xmath235 , where they limit to the energy independent constant @xmath236 , but that the next order of the semi - classical phase is quadratic in @xmath237 while it is linear for the berry phase . as both the semi - classical phase and the maslov index
do _ not _ depend on initial positions we may use directly the generalized bohr - sommerfeld quantization condition , eq . .
to this end we require the maximum initial momentum in the cyclic direction and this provides the bounds for the cyclic momentum integral , which can straightforwardly be found from the hamilton - jacobi equations as @xmath238 .
we hence encounter the integral @xmath239 and hence the density of states is given by @xmath240 the exact energy levels are given as @xmath241 , where @xmath242 , and we thus see that if we retain only the energy independent part of the semi - classical phase we obtain the exact result .
this implies that , as the semi - classical phase is @xmath243 and the berry phase @xmath244 , that the extra dynamical term in the semi - classical phase is important . in figure [ f1 ]
we present a graphical comparison of the exact and semi - classical energy levels and , as may be seen , while the disagreement is pronounced at low energies , for higher energies the agreement is , as expected , very good .
we now consider the semi - classical analysis of a complex graphene system that is an analogue of the well studied graphene twist bilayer , the one - dimensional strain moir . as may be seen in fig .
[ m0 ] this consists of a uniform strain applied to ( without loss of generality ) layer 1 of an initially ab stacked bilayer that leads to a moir lattice in which ( as in the case of the twist bilayer ) all possible stacking types occur over one moir period .
we choose to apply the strain in the armchair direction ( @xmath245 in the coordinate system displayed in fig .
[ m0 ] ) as this makes the semi - classical analysis somewhat more tractable .
the deformation field of the problem is therefore , using the coordinate system indicated in fig .
[ m0 ] , given by @xmath246 , and it is convenient to express this as @xmath247 with @xmath248 some function encoding the particular strain , and the constant shift term introduced for a more symmetrical hamiltonian . with these definitions the interlayer potential @xmath249 can be obtained via the general theory outlined in ref . with the result @xmath250 where @xmath251 with @xmath252ev the interlayer hopping .
the hamiltonian of the strain moir system is then given by @xmath253 note that this interlayer field is rather similar to that deployed for a the one - dimensional moir treated in ref .
; however in that work the moir was created by shear and not by strain .
uniform strain requires @xmath254 , with @xmath255 the moir period as may be deduced from eq . .
in particular we have aa stacking at @xmath256 , ab stacking at @xmath257 , and ac stacking at @xmath258 , as may readily be seen by substitution of these @xmath259 values into eq . .
solving the classical @xmath6 problem we find 4 non - degenerate hamilton - jacobi systems .
evidently @xmath260 is a good quantum number of the effectively one - dimensional problem , and for simplicity we will consider here the case @xmath261 ; this is not a singular limit and as such the small @xmath260 behaviour is very similar to @xmath261 . for larger @xmath260
the system develops a much richer and interesting structure , which we will not investigate here .
the 4 distinct hamilton - jacobi equations have 4 distinct momenta of which two differ only by a minus sign : @xmath262 where the index @xmath263 indicates the sign in front of the square - root . for simplicity of discussion
we will now adopt the habit of referring to the solution with index 1 and 2 as particles 1 and 2 . in eq .
we have expressed the momenta in terms of an effective mass @xmath264 given by @xmath265 which is energy dependent , and an effective potential @xmath266 that takes the form @xmath267 the cosine form of this potential reveals immediately that we have a quantum well structure to the problem with the maxima ( particle type 1 ) or minima ( particle type 2 ) centred at the aa spots @xmath268 . for particle type 1
the effective mass is negative at all @xmath106 where @xmath269 is defined and hence the usual regions of classically allowed and forbidden motion are inverted , and thus a maxima of the effective potential at the aa spot indicates bound orbits centred on this region of the moir .
particle type 2 with a minimum of the potential well on the aa spot and a positive mass well , evidently , also describes orbits centred at the aa spot .
this is illustrated in fig .
[ m2 ] in which the shaded areas represent the regions of allowed particle motion and one can see that the situation is symmetric if we send particle type 1 to 2 and @xmath270 provided the sign of the mass changes as well .
the effective potentials of particle types 1 and 2 are shown as the blue and green lines respectively , with the shaded areas illustrating the classically allowed regions ; note that this is inverted for particle type 2 due to the negative effective mass of that particle type , see eq . .
for an energy of @xmath271ev the two orbits of the two particle types are indicated , along with the corresponding local density of states .
this latter quantity displays , as expected , pronounced peaks at the turning points of the classical particle .
the aa spots of the moir correspond to @xmath268 with @xmath272.,scaledwidth=45.0% ] to examine this situation more closely we determine the turning points of the orbits which for particle type 1 are given by @xmath273 and for particle type 2 by @xmath274 the orbit length for particle 2 is given by @xmath275 and ( as may also be seen from fig .
[ m2 ] ) decreases with increasing @xmath106 until we find a zero length orbit at the band edge @xmath276 , after which the particle trajectory abruptly jumps from a zero length orbit to non - localized behavior .
in contrast there are no bound states of this particle type at the other band edge of @xmath277 as at @xmath278 the bound orbits merge together , and the behavior for energies lower than this is again non - localized . for particle type 1
the situation is the same but with @xmath270 . interestingly , and contrary to what one might expect given the results of the previous section ,
the semi - classical approximation for the strain moir is therefore _
better for lower energies than for large energies _ , in particular close to band edges where the approximation is guaranteed to fail as the orbit length approaches zero . on the other hand
, one should stress that for slowly varying structural perturbations - such as considered here - the semi - classical approximation is expected to be good . to calculate the semi - classical spectrum via the gutzwiller formula
we require the semi - classical phase , the actions @xmath279 , and the orbit times @xmath194 .
we now proceed to calculate each of these in turn .
nm , @xmath280 , and @xmath281.,scaledwidth=40.0% ] _ the semi - classical phase _
: the integral of the @xmath118 phase @xmath282 may be conveniently found by the change of variables @xmath283 and is found to be @xmath284 where we also find , as we must , that @xmath285 .
in fact only the result for @xmath286 could be obtained fully analytically ; for @xmath287 and @xmath288 the integral was taken numerically with the values @xmath289 and @xmath171 obtained to @xmath290 accuracy . _ the action @xmath279 _ : the actions may be obtained analytically from the hamilton - jacobi equations with the result that @xmath291 for particle type 1 , where @xmath292 is the elliptic integral of the second kind . for particle type 2
we simply have @xmath293 .
the action for particle type 2 is shown in fig .
[ actionplot ] , and evidently is larger for smaller energies and large compared to @xmath11 , thus justifying the semi - classical approximation at low energies .
nm ( a ) , @xmath294 nm ( b ) , and @xmath295 nm ( c ) showing the increasingly good agreement at low energies between the exact result and semi - classical approximation .
note that by construction the semi - classical approximation only treats bounds orbits , and hence it is only near the dirac point where bound orbits exist in the strain moir ( but at which the dos is zero for pristine graphene ) that the agreement between the semi - classical and exact solution is very good . , scaledwidth=45.0% ] _ orbit times _ : the time for an orbit @xmath296 can be found as @xmath297 , where @xmath298 which follows from the hamilton s equations and which we calculated using the hellman - feynman theorem @xmath299 .
we thus arrive at the following expression for the period @xmath296 @xmath300 where @xmath301 is the elliptic integral of the first kind and @xmath292 the elliptic integral of the second kind .
we now have all the ingredients required to obtain the semi - classical spectrum via the gutzwiller trace formula .
as all of these quantities evidently do not depend on the initial position of the orbit we may use the generalized bohr - sommerfeld quantization condition , eq .
we note that in a sample of length @xmath302 the orbits are , evidently , @xmath303-times degenerate and the resulting density of states is then @xmath304 in fig .
[ m3 ] we plot this semi - classical density of states against the exact result , obtained simply by solving the quantum problem of eq . in a basis of single layer graphene states , for three moir lengths , @xmath305 nm , @xmath294 nm , and @xmath295 nm . as may be seen as the moir length increases an increasing number of bound states are trapped in the aa centred quantum wells .
interestingly , even for a rather small moir length of @xmath305 nm the 4 bound states that are trapped in the well are reasonably well described in the semi - classical approximation .
as the moir length increases the agreement between the semi - classical and exact results becomes increasingly good , with at @xmath295 nm the low energy agreement between the two results almost perfect .
, plotted in the energy position plane . the high intensity regions in the ldos
mark the turning points of the classical orbits described by eqs .
-.,scaledwidth=50.0% ] we may also calculate the local density of states ( ldos ) , obtained simply by leaving out the position trace @xmath306 in the general formula eq . : @xmath307 the ldos at an energy of @xmath271ev is shown in fig . [ m2 ] along with the classically allowed orbits indicated by the arrowed lines .
evidently the ldos shows a pronounced intensity peak at the turning points of the classical orbit , exactly as one would expect .
this can also be seen in a density plot of the ldos as a function of energy and @xmath259 , shown in fig .
[ m33 ] , where we have smoothed the result via a gaussian convolution for ease of plotting .
the bright high intensity regions of ldos follow the turning point structure of the classical orbits and result in regions of high ldos surrounding the aa spots of the strain moir .
for a general @xmath0 matrix hamiltonian the classical dynamics is governed by @xmath2 hamilton - jacobi ( hj ) equations , corresponding to @xmath308 semi - classical particle types , in a hamiltonian phase space that is endowed with a berry curvature . as an @xmath6 object ,
i.e. a classical object , this curvature is of corresponding importance to the hamilton - jacobi equations in the semi - classical theory of matrix hamiltonians , and encodes anholonomy in the transport of the eigenvectors of the object @xmath3 around the classical orbits .
while for some systems the semi - classical particle types have an intuitive interpretation , e.g. in the case of the dirac equation these correspond to electrons and positrons@xcite , in a more general solid state context these simply represent the emergence of the internal semi - classical structure of the underlying quasiparticles . in particular , the number of semi - classical particle types may be less than @xmath2 , as was the case for the strain moir treated in section iv . for each particle type @xmath32
we find at @xmath4 a _ transport equation _ for the wavefunction amplitude , which differs from a similar equation found in the case of a scalar hamiltonian by the presence of a semi - classical phase @xmath69 . in the case where all hj equations are distinct , the non - degenerate case ,
@xmath69 may be expressed as the sum of a geometric ( berry ) phase that encodes global features of the @xmath6 berry curvature , as well as a dynamical phase that arises from the motion of the semi - classical particle `` through the berry curvature '' .
this latter phase is responsible for coupling the @xmath2 hj equations . in the case of degeneracies amongst the hj systems
, @xmath69 becomes matrix valued and the geometric phase is then the analogue of the non - abelian berry phase , although the theory remains formally very similar to the non - degenerate case .
this scheme leads to expressions for the semi - classical greens functions and the semi - classical density of states , and we have provided such expressions for both the degenerate and non - degenerate case , as well as for further more specialized situations such as the presence of cyclic coordinates .
in particular for effectively one degree of freedom systems ( @xmath309 cyclic coordinates ) we have presented a generalization of the bohr - sommerfeld quantization rule for closed orbits .
we have applied this formalism to a number of low dimensional systems : graphene , bernal stacked bilayer graphene , silicene , and a one - dimension strain moir in bilayer graphene . in the latter case
we find almost perfect agreement between the exact and semi - classical density of states arising from localized states that the moir induces near the dirac point , while for the former cases good agreement is observed between the exact and semi - classical landau spectra at high energies .
using the formalism here , even the strain moir yielded rather easily to a semi - classical treatment which , we stress , would be almost impossibly complex to treat with the wkb method , or indeed any approach utilizing a matching condition between wavefunctions or green s functions@xcite .
the semi - classical treatment of the strain moir reveals the existence of two particle types ( that dominate the negative and positive energy regions ) that for energies near the dirac point are trapped in a semi - classical potential well centred on the aa stacked regions of the moir lattice .
it should be stressed that such physics is in an essential way semi - classical : at the quantum level a potential well in graphene will not form bound states due to the klein tunneling effect .
the existence of such `` moir potential wells '' has been discussed on the basis of numerical tight - binding calculations in which localization seen on the aa regions of the graphene twist bilayer@xcite , a two dimensional analogue of the one - dimensional system considered in this work , but never rigorously shown to exist .
our work shows that this `` moir potential well '' concept is in fact most naturally a semi - classical concept .
this illustrates the conceptual usefulness of the semi - classical approach in providing insight into the physics of these complex systems .
as much of the structural complexity low dimensional systems is due to deformations that are spatially slow on the scale of the lattice constant , for example partial dislocation networks in bernal stacked bilayer graphene@xcite , the formalism presented in this work may provide not only a numerically tractable scheme for such complex systems , but also one which may yield transparent access to the underlying physics of this new class of materials .
in addition , the results of this paper may facilitate a treatment of a semi - classical transport in systems with internal degrees of freedom where , in analogy to the scalar case@xcite , one should be able to find semi - classical expressions for the kubo conductivity formula .
jonathan s. alden , adam w. tsen , pinshane y. huang , robert hovden , lola brown , jiwoong park , david a. muller , and paul l. mceuen .
strain solitons and topological defects in bilayer graphene . , 110(28):1125611260 , 2013 .
ferdinand kisslinger , christian ott , christian heide , erik kampert , benjamin butz , erdmann spiecker , sam shallcross , and heiko b. weber .
linear magnetoresistance in mosaic - like bilayer graphene . , 2015 .
|
we have extended the semi - classical theory to include a general account of matrix valued hamiltonians , i.e. those that describe quantum systems with internal degrees of freedoms , based on a generalization of the gutzwiller trace formula for a @xmath0 dimensional hamiltonian @xmath1 .
the classical dynamics is governed by @xmath2 hamilton - jacobi ( hj ) equations , that act in a phase space endowed with a classical berry curvature encoding anholonomy in the parallel transport of the eigenvectors of @xmath3 ; which describe the internal structure of the semi - classical particles .
this berry curvature is a fully classical object and is , in that sense , as fundamental to the semi - classical theory of matrix hamiltonians as the hamilton - jacobi equations . at the @xmath4 level
, it results in an additional semi - classical phase composed of ( i ) a berry phase and ( ii ) a dynamical phase resulting from the classical particles `` moving through the berry curvature '' .
we show that the dynamical part of this semi - classical phase will , generally , only be zero only for the case in which the berry phase is topological ( i.e. depends only on the winding number ) .
we illustrate the method by calculating the landau spectrum for monolayer graphene , the four - band model of ab bilayer graphene , and for a more complicated matrix hamiltonian describing the silicene band structure .
finally we apply our method to an inhomogeneous system consisting of a strain engineered one dimensional moir in bilayer graphene , finding localized states near the dirac point that arise from electron trapping in a semi - classical moir potential
. the semi - classical density of states of these localized states we show to be in perfect agreement with an exact quantum mechanical calculation of the density of states .
|
it is a well known fact that solutions of equations determining the motion of a homogeneous inviscid fluid possess a remarkable property the lines of the vorticity field are frozen - in @xcite .
mathematical reason for this is the so called relabeling symmetry of fluids that provides necessary conditions for the noether theorem applicability and results in infinite number of the conservation laws @xcite . due to this basic property , in the framework of ideal hydrodynamics
such flows are possible where during sufficiently long time interval the vorticity is concentrated in quasi - one - dimensional structures , vortex filaments , that fill a small part of entire bulk of the fluid .
the motion of vortex filaments is very interesting problem both from theoretical and practical viewpoints and is a classical subject of hydrodynamics ( see , for instance , @xcite and references therein for various analytical and numerical approaches to this problem ) . in general case analytical study in this field is highly complicated because of several reasons , the main of them being non - locality and nonlinearity of governing equations of motion .
a less significant trouble seems to be the necessity of some regularization procedure for the hamiltonian functional ( the total energy ) of the system in the limit of `` infinitely thin '' vortex filaments , since a logarithmic divergency takes place in some observable physical quantities ( for instance , in the velocity of displacement of curved pieces of the filament ) as the thickness decreases .
however , in few limit cases the dynamics of a single vortex filament can turn out to be effectively integrable .
a known and very interesting example of such integrable system is a slender non - stretched vortex filament ( in the boundless three - dimensional ( 3d ) space filled by an ideal fluid ) in the so called localized induction approximation ( lia ) , when in the energy of the filament only logarithmically large contributions from interaction of adjacent pieces are taken into account . in this approximation
the hamiltonian is simply proportional to the length of the filament , resulting in conservation this quantity , thus application of the so called hasimoto transformation @xcite is appropriate and reduces the problem to ( 1 + 1)-dimensional nonlinear schroedinger equation that is known to be integrable by the inverse scattering method @xcite . in present work
another integrable case in vortex dynamics is recognized , the long - scale motion of a thin vortex filament near a flat fixed boundary .
mathematically the problem of a single filament in a half - space is equivalent to the problem of two symmetric filaments in the boundless space , that allows us simplify some further calculations .
our immediate purpose will be to consider the configurations of the vortex filament that satisfy the following conditions : \a ) the angle is everywhere small between the tangent vector on the curve determining the shape of the filament and the boundary plane ; \b ) the distance from an arbitrary point of the curve to the plane is small comparatively to the curvature radius at the given point but large in comparison with the thickness of the filament ; \c ) the filament projection on the boundary plane does not have self - intersections or closely approaching one - to - another different pieces . in these conditions the system dynamics
is known to be unstable ( the so called crow instability @xcite ) , with the instability increment directly proportional to the wave number of ( some small ) long - scale perturbation of the filament shape .
it is a well known fact that such dependence of the increment is usual for a class of local @xmath1 partial differential systems that can be exactly linearized by so called hodograph transformation @xcite exchanging dependent and independent variables .
this observation has served as a weighty reason to look for a natural local nonlinear approximation in description of the long - scale dynamics of a vortex filament near a flat boundary and to examine that approximation for the hodograph method applicability . as the result
, a consistent derivation of the corresponding local approximate equations of motion has been performed , and also the fact has been demonstrated that the nonlinear partial differential system for two functions @xmath2 and @xmath3 determining the shape of the filament and depending on the time moment @xmath4 and on the cartesian coordinate @xmath5 is reduced by hodograph transformation to a linear equation .
moreover , it is possible to choose the pair of new independent variables in such a manner that in the linear partial differential equation for the function @xmath6 the coefficients will not depend on @xmath3-variable . for this purpose
it is convenient to define @xmath2-variable as double distance from a filament point to the boundary plane @xmath7 , while @xmath3-variable will be the angle between @xmath5-direction and the tangent to the filament projection on @xmath8-plane .
obviously , an explicit dependence of the coefficients on @xmath3 will be absent due to the symmetry of the system with respect to rotations in @xmath8-plane .
therefore separation of the variables will be possible and most simple solutions will have the form @xmath9 where @xmath0 is an arbitrary complex parameter , and the complex function @xmath10 satisfies the simple equation @xmath11 to find the complex function @xmath12 , it will be necessary to solve some ordinary linear differential equation of the second order with variable coefficients that will be considered in appropriate section of this paper .
the corresponding geometrical configurations of the vortex filament strongly depend on @xmath0 .
in particular , it will be shown the solutions ( [ t_lambda ] ) with real @xmath13 describe such a shape of the ( moving ) vortex filament that its projection on @xmath8-plane has two asymptotes with the angle between them @xmath14 , while in the case @xmath15 the projection has the shape of a two - branch spiral ( see the figures ) .
this article is organized as follows . in section ii a necessary brief review is given concerning the hamiltonian formalism adopted to the problem of frozen - in vorticity , since this approach is the most clear and compact way to treat ideal flows
. then approximate local equations of motion for a vortex filament near a flat boundary are derived . in section iii
we demonstrate applicability of the hodograph method and introduce variables that are most convenient for the particular problem .
section iv is devoted to investigation of simple solutions obtained by the variables separation in the governing linear equation .
existence itself of the ideal - hydrodynamic solutions in the form of quasi - one - dimensional vortex structures ( vortex filaments ) filling just a small part of the total fluid bulk is closely connected with the freezing - in property of the vortex lines @xcite .
mathematically this property is expressed by the special form of the equation of motion for the divergence - free vorticity field @xmath16 , @xmath17,\ ] ] where @xmath18 is the velocity field . since in this article
we consider incompressible flows , we may write @xmath19 where @xmath20 is the 3d laplace operator . as it is known , the action of the inverse nonlocal operator @xmath21 on an arbitrary function @xmath22 is given by the formula @xmath23 where @xmath24 is the green function of the @xmath25 operator in the boundless space
. hamiltonian noncanonical structure @xcite of the equations of ideal incompressible hydrodynamics is based on the relation @xmath26 where the hamiltonian functional @xmath27 is the kinetic energy of a homogeneous incompressible fluid ( with unit density ) expressed through the vorticity , @xmath28 our approach to investigation of the vortex filament motion is based on representation of the ideal homogeneous fluid flows in terms of the frozen - in vortex lines , as described , for instance , in @xcite .
the special form ( [ frozen - in_omega ] ) of the equation of motion allows us express the vorticity field @xmath29 in a self - consistent manner through the shape of the vortex lines ( the so called formalism of vortex lines ) , @xmath30 where @xmath31 is the 3d delta - function , @xmath32 is some 2d manifold of labels enumerating the vortex lines ( @xmath32 is determined by topological properties of a particular flow ) , @xmath33 is a label of an individual vortex line , @xmath34 is an arbitrary longitudinal parameter along the line .
what is important , the dynamics of the line shape @xmath35 is determined by the variational principle @xmath36/\delta{\mbox{\boldmath $ { r}$}}(\nu,\xi , t)=0,\ ] ] with the lagrangian of the form @xmath37\cdot{\mbox{\boldmath $ { d}$}}({\mbox{\boldmath $ { r}$}}))d\xi \,\ , -{\cal h}\{{\mbox{\boldmath $ { \omega}$}}\{{\mbox{\boldmath $ { r}$}}\}\},\ ] ] where the vector function @xmath38 in the case of incompressible flows must satisfy the condition @xmath39 below we choose @xmath40 .
since we are going to deal with a very thin vortex filament , we will neglect the @xmath41-dependence of the shapes of individual vortex lines constituting the filament . by doing this step
we exclude from further consideration all effects related to finite variable cross - section and longitudinal flows inside the filament .
thus , we consider an `` infinitely narrow '' vortex string with a shape @xmath42 and with a finite circulation @xmath43 .
however , the hamiltonian of such singular filament diverges logarithmically , @xmath44 in order to regularize this double integral , it is possible , as a variant , to modify the green function @xcite .
for example , instead of the singular function @xmath45 one can use a smooth function like @xmath46 or some other appropriate expression depending on a parameter @xmath47 .
it should be emphasized that relation @xmath48 is exactly satisfied only in the original non - regularized system , but in the case of a finite @xmath47 it is not valid on distances of order @xmath47 from the singular vortex string .
thus , the meaning of vorticity in regularized models is not so simple , but nevertheless , relation ( [ v_omega_hamiltonian ] ) remains valid in any case .
relatively small parameter @xmath47 in regularized models serves to imitate a finite width of vortex filament in the usual ( non - regularized ) hydrodynamics .
the energy of the string turns out to be logarithmically large , @xmath49 where @xmath50 is a typical scale depending on a given problem ( in particular , the usual lia uses @xmath51 const @xmath52 ) . in our case
we consider two symmetric vortex filaments in the long - scale limit , when direction of the tangent vector varies weakly on a length of order @xmath53 .
for such configurations , the energy concentrated in the half - space @xmath54 is approximately equal to the following expression @xmath55 this local hamiltonian is able to provide qualitatively correct dynamics of the filament down to longitudinal scales of order @xmath53 where perturbations become stable and where non - locality comes to play .
unfortunately , we do not have a simple method to treat the hamiltonian ( [ lia_generalized ] ) analytically , that is why we will consider only very large scales ( @xmath56 ) and thus suppose the slope of the tangent vector to the boundary plane to be negligibly small ( this means @xmath57 ) . then , choosing as a longitudinal parameter @xmath34 simply the cartesian coordinate @xmath5 , we have the following approximate lagrangian : @xmath58 where the functions @xmath53 and @xmath59 depend on @xmath5 and @xmath4 , @xmath60 , @xmath61 . having neglected the term @xmath62 under the square root , we sacrifice correct behaviour of perturbations with wave - lengths of order @xmath53 , but instead we obtain exactly solvable system , as it will be shown below .
let us say a few words about geometrical meaning of the second term in r.h.s . of the expression ( [ approx_lagrangian ] ) .
since we study the very long - scale limit , locally the flow under consideration looks almost like a two - dimensional flow with a small vortex at the distance @xmath53 from the straight boundary , and the expression @xmath63 is just the energy of such 2d flow per unit length in the third ( longitudinal ) direction , while the multiplier @xmath64 gives the arc - length element in the longitudinal direction .
now for simplicity we take new time and length scales to satisfy @xmath65 and @xmath66 . after
that we introduce new quantities @xmath67 and @xmath68 , and also the function @xmath69 , @xmath70 where @xmath71 the corresponding equations of motion then can be written in the following remarkable general form : @xmath72 more explicitly , the last two equations are @xmath73=0 , & & \label{rho_t}\\ { \partial_t}z+f'(\rho)\sqrt{1+z_x^2}=0.&&\label{z_t}\end{aligned}\ ] ] these equations have a simple geometrical treatment . indeed , eq.([z_t ] ) means if we consider the dynamics of the filament projection on @xmath8-plane , then we see an element of the projection moving in the normal to the projection tangent direction with the velocity depending only on @xmath2 and equal to @xmath74 .
simultaneously , in @xmath75-direction the element of the filament moves with the velocity proportional to the @xmath8-projection curvature multiplied by the function @xmath76 , as eq.([rho_t ] ) shows .
it is interesting to note that an analogous consideration can give us also long - scale hamiltonian equations of motion for a thin vortex filament in a slab of an ideal fluid between two parallel fixed boundaries at @xmath77 and @xmath78 .
one has just to define the @xmath2-variable by the formula @xmath79 and make in eq.([h_rm ] ) substitution @xmath80 where @xmath81 with a small dimensionless parameter @xmath82 .
it is known that any nonlinear system of the form ( [ mu_definition])-([bernoulli_general ] ) can be locally reduced to a linear equation if we take @xmath2 and @xmath83 as new independent variables ( this is the so called hodograph method ; see , for instance , @xcite where a particular case is discussed , the 1d gas dynamics , with @xmath84 , where @xmath2 , @xmath83 , @xmath85 are the gas density , gas velocity , and the internal energy density respectively ) .
indeed , as from equations ( [ mu_definition ] ) and ( [ bernoulli_general ] ) we see the relation @xmath86 it is convenient to introduce the auxiliary function @xmath87 , @xmath88 in order to obtain @xmath89 from the above expression we easily derive @xmath90 after that we rewrite eq.([contin_general ] ) in the form @xmath91 and multiply it by the jacobian @xmath92 , @xmath93 thus , now we have @xmath94 substitution of the relations ( [ tx ] ) into this equation and subsequent simplification give us the linear partial differential equation of the second order for the function @xmath87 , @xmath95 as the function @xmath69 has the special form ( [ h_rm ] ) , it is convenient to change variables , @xmath96 where @xmath3 is the angle in @xmath8-plane between @xmath5-direction and the tangent to the corresponding projection of the vortex filament . as the result , the relations ( [ chi_definition ] ) and ( [ tx ] ) will be rewritten in the form @xmath97 and the coefficients of the linear equation for the function @xmath98 will not depend on @xmath3-variable , @xmath99 the same is true for the coefficients of the equation determining the function @xmath100 , @xmath101 once some particular solution of eq.([equation_for_u_rho_theta ] ) is known , then further procedure consists in the following two steps : \i ) find in terms of some parameter @xmath34 the curves of constant values of the function @xmath102 .
it is this point where nonlinearity comes to play , since we need to solve nonlinear equation ; \ii ) substitute the obtained expressions @xmath103 and @xmath104 into eqs.([x_u]-[z_u ] ) and get complete description of the filament motion , @xmath105 , @xmath106 , @xmath107 .
thus , the long - scale local approximation ( [ approx_lagrangian ] ) turns out to be integrable in the sense it is reduced to the _ linear _ equation ( [ equation_for_u_rho_theta ] ) .
however , the function @xmath98 is multi - valued in the general case
. therefore statement of the cauchy problem becomes much more complicated .
besides , the functions @xmath76 and @xmath108 determined by expressions ( [ f_log ] ) and ( [ f_epsilon ] ) result in _ elliptic _ linear equations as against the usual 1d gas dynamics where the corresponding equations were _ hyperbolic_. generally speaking , the ellipticity makes the cauchy problem ill - posed in the mathematical sense if initial data are not very smooth .
however , in this article we will not discuss these questions , instead in the following section we will present simple particular solutions that during some time interval satisfy the applicability conditions for the long - scale approximation .
we are going to consider the simplest particular solutions of eq.([equation_for_u_rho_theta ] ) obtainable by separation of the variables @xmath109 where @xmath0 is an arbitrary complex spectral parameter , @xmath110 and the function @xmath10 contains two arbitrary complex constants @xmath111 and @xmath112 , @xmath113 + c_\lambda^{-}\exp[-(\varkappa+ik)\vartheta].\ ] ] the motion of the vortex filament will be described by the formulas @xmath114 the function @xmath115 must satisfy the ordinary differential equation of the second order @xmath116 let us turn a bit of attention to the special value @xmath117 of the spectral parameter , when the solution of eq.([u_lambda ] ) can be explicitly written for any function @xmath76 , @xmath118 where @xmath119 and @xmath120 are arbitrary complex constants . at @xmath121 it can be convenient to deal with the function @xmath122 that satisfies the equation @xmath123 in particular , eq.([t_lambda ] )
is simply solvable at @xmath124 ( this solution describes the motion of a perfect vortex ring ) , @xmath125 simple manipulations with formulas ( [ t_lambda_u]-[u_lambda ] ) allow us rewrite the solutions in the form @xmath126'\theta_\lambda'(\vartheta)\cos\vartheta \nonumber\\ & & \quad+[\lambda f'(\rho)t_\lambda(\rho)-f(\rho)t_\lambda'(\rho ) ] \theta_\lambda(\vartheta)\sin\vartheta \}\big\},\\ \label{z_lambda_t } & & z_\lambda=\mbox{re}\big\{(\lambda+1)^{-1}\ { -[f(\rho)t_\lambda(\rho)]'\theta_\lambda'(\vartheta)\sin\vartheta \nonumber\\ & & \quad-[\lambda f'(\rho)t_\lambda(\rho)-f(\rho)t_\lambda'(\rho ) ] \theta_\lambda(\vartheta)\cos\vartheta \}\big\}.\end{aligned}\ ] ] let us first consider real values of the spectral parameter , @xmath127 , and the corresponding real functions @xmath10 and @xmath115 . since @xmath128 , @xmath129 , we may expect the solutions @xmath115 with @xmath130 to have a number of oscillations , the more , the more @xmath0 is . in the opposite case ,
when @xmath131 , the solutions will be a linear combination of two functions , one of them being increasing , and other decreasing .
it is sufficient to know these general properties to get an impression concerning geometrical configurations of the vortex filament described by the formulas ( [ t_lambda_t]-[z_lambda_t ] ) .
let us take @xmath132 with @xmath133 and suppose the explicit dependence @xmath134 to be known and increasing at large @xmath2
. for simplicity we take @xmath135 and after that resolve the relation ( [ t_lambda_t ] ) with respect to @xmath3 , @xmath136.\ ] ] substitution of this expression into formulas ( [ x_lambda_t]-[z_lambda_t ] ) gives us final form of the solutions as dependences @xmath137 and @xmath138 , @xmath139\ , \sin\left(\frac{1}{k}\arccos\left[\frac{t}{t_{-k^2}(\rho)}\right]\right)\nonumber \\ & & \mp\frac{k[f(\rho)t_{-k^2}(\rho)]'}{k^2 - 1 } \left[1-\frac{t^2}{t_{-k^2}^2(\rho)}\right]^{1/2 } \cos\left(\frac{1}{k}\arccos\left[\frac{t}{t_{-k^2}(\rho)}\right]\right),\end{aligned}\ ] ] @xmath140\ , \cos\left(\frac{1}{k}\arccos\left[\frac{t}{t_{-k^2}(\rho)}\right]\right)\nonumber \\ & & \mp\frac{k[f(\rho)t_{-k^2}(\rho)]'}{k^2 - 1 } \left[1-\frac{t^2}{t_{-k^2}^2(\rho)}\right]^{1/2 } \sin\left(\frac{1}{k}\arccos\left[\frac{t}{t_{-k^2}(\rho)}\right]\right).\end{aligned}\ ] ] the @xmath2-variable in the above expressions varies in the limits from @xmath141 such that @xmath142 , to @xmath143 .
the corresponding curve in @xmath8-plane is a smoothed angle @xmath144 [ see figs.([k_sqrt9]-[k_sqrt10 ] ) where for the case @xmath145 the filament shape is shown at several time moments , @xmath146 const ] .
completely different form is obtained at @xmath147 , two - branch spirals [ see fig.([spirals ] ) ] .
let us take @xmath148 .
then @xmath149,\ ] ] @xmath150\right)\nonumber \\ & & \qquad\qquad\qquad\qquad -\frac{\varkappa[f(\rho)t_{\varkappa^2}(\rho)]'}{\varkappa^2 + 1 }
\cos\left(\frac{1}{\varkappa } \log\left[\frac{t}{t_{\varkappa^2}(\rho)}\right]\right)\biggr\},\end{aligned}\ ] ] @xmath151\right)\nonumber \\ & & \qquad\qquad\qquad\qquad -\frac{\varkappa[f(\rho)t_{\varkappa^2}(\rho)]'}{\varkappa^2 + 1 } \sin\left(\frac{1}{\varkappa } \log\left[\frac{t}{t_{\varkappa^2}(\rho)}\right]\right)\biggr\}.\end{aligned}\ ] ] the variable @xmath2 runs here between two neighbour zeros of the function @xmath152 and approaches these values at two logarithmic branches of the spiral , @xmath153 . for further investigation
let us substitute @xmath154 into the equations ( [ t_lambda_u]-[u_lambda ] ) and change the variable @xmath155 as the result , we will obtain @xmath156 \theta_\lambda(\vartheta)\sin\vartheta \},\\ z_\lambda&=&\mbox{re}\ { u_\lambda(q)\theta_\lambda'(\vartheta)\sin\vartheta \nonumber \\ \label{z_lambda_uq } & & \qquad-\left[u_\lambda(q)+u_\lambda'(q)\right ] \theta_\lambda(\vartheta)\cos\vartheta \ } .\end{aligned}\ ] ] @xmath157 general solution of eq.([uq ] ) is representable by the laplace method @xcite as arbitrary linear combination @xmath158 of two contour integrals in a complex plane , @xmath159 here the first closed contour @xmath160 goes around the points @xmath161 and @xmath162 . the second contour @xmath163 is not closed , at positive @xmath164 it starts at @xmath165 .
if @xmath166 , then @xmath163 ends at @xmath167 , but if @xmath168 , then its end point is @xmath169 . in both cases at the end point of the contour @xmath163 the integrand multiplied by @xmath170 tends to zero .
it is interesting to note that at the integer values of the parameter @xmath0 the integral @xmath171 can be expressed in terms of polynomials : @xmath172 these expressions have been used to prepare figs.([k_sqrt9])-([spirals ] ) where the vortex filament shape corresponding to @xmath173 is drawn for several moments of time .
it is easily to see , at sufficiently large times the spirals satisfy the conditions a ) , b ) , c ) that have been formulated in the introduction . as to the angle - shaped configurations ,
the condition @xmath57 , generally speaking , is not satisfied at @xmath174 , since at very large @xmath164 ( on the asymptotes of the angle ) the growth of @xmath175 is faster than growth of @xmath176 .
therefore , if we take a particular solution @xmath177 separately , not as a term in a more complex linear combination , then we have to deal only with large @xmath178 , and consider only the pieces of the filament where @xmath179 . in ref.@xcite we investigated another regularization of the hamiltonian functional that corresponds to @xmath180 , with some small positive parameter @xmath181 .
that time we did not see applicability of the hodograph method and therefore we were able to find only few particular solutions .
now it has been clear that in this case a simple substitution exists that reduces the problem to 2d equation @xmath182 .
thus , it becomes possible to present a very wide class of solutions of the equation @xmath183 as linear combinations of singular fundamental solutions ( that are expressed through the mcdonald function @xmath184 ) and regular exponential or polynomial solutions .
indeed , by the substitutions @xmath185 eq.([equation_for_t_alpha ] ) is reduced to the equation with constant coefficients , @xmath186 as it is well known , the fundamental solutions of this equation have the form @xmath187 where @xmath188 and @xmath189 are arbitrary parameters
. therefore eq.([equation_for_t_alpha ] ) has particular solutions @xmath190 ^ 2 + \frac{(\vartheta-\vartheta_0)^2}{\alpha(1-\alpha)}}\right).\ ] ] it is interesting to note that at @xmath191 the system possesses conformal symmetry
. a deep reason of this symmetry is not clear yet . as concerning separation of the variables , the function @xmath12 in eqs.([t_lambda_t])-([z_lambda_t ] )
is given by the expression @xmath192 where @xmath193 are arbitrary constants .
the complex exponents @xmath194 are the roots of the quadratic equation @xmath195 thus , @xmath196 it should be mentioned the solutions presented in @xcite correspond to the particular case @xmath197 .
in this article an approximate exactly solvable nonlinear model has been derived to describe unstable locally - quasi-2d ideal flows with a thin vortex filament near a flat boundary .
the hodograph method has been applied and some particular solutions have been analytically found by variables separation in the governing linear partial differential equation for auxiliary function @xmath198 . more general solutions @xmath98 can be obtained as linear combinations of the terms ( [ u_lambda ] ) with different @xmath0 , but only in few cases it will be possible to resolve analytically the dependence @xmath102 . however , this procedure can be performed numerically . though we derived the exactly solvable model under several restrictive simplifications , the solutions obtained in this work promise benefit in many aspects .
for instance , they may serve as basic approximations in future more advanced analytical studies that will take into account effects of non - locality and/or finite variable cross - section of the filament , as well as surface waves in the case of free boundary .
|
hamiltonian dynamics of a thin vortex filament in ideal incompressible fluid near a flat fixed boundary is considered at the conditions that at any point of the curve determining shape of the filament the angle between tangent vector and the boundary plane is small , also the distance from a point on the curve to the plane is small in comparison with the curvature radius .
the dynamics is shown to be effectively described by a nonlinear system of two ( 1 + 1)-dimensional partial differential equations .
the hodograph transformation reduces that system to a single linear differential equation of the second order with separable variables .
simple solutions of the linear equation are investigated at real values of spectral parameter @xmath0 when the filament projection on the boundary plane has shape of a two - branch spiral or a smoothed angle , depending on the sign of @xmath0 .
|
the anderson localization of light has been one of the most fascinating phenomena in condensed matter physics since the mid - eighties @xcite . like electron systems this phenomenon finds its origin in coherent multiple scattering which slows down diffusion of photons and eventually brings them to stop .
parallel to studies of disordered conductors the subject in this field ranges from light localization near or far below the mobility edge in bulk ( infinite ) systems @xcite ( where low - energy photon motion enjoys the translational symmetry ) to their detection such as transmission measurements in the slab geometry ( e.g. , refs . ) .
a unique subject of localization in optical ( and other classical wave ) systems is the enhanced coherent backscattering ( cbs ) phenomenon @xcite . in this subject
the issue of semi - infinite geometry is heavily addressed because the cbs line shape is responsible for by optical paths near the vacuum - medium interface .
although it is well known that in the weak disorder region , i.e. , @xmath0 ( @xmath1 the mean free path and @xmath2 the wavelength ) incident photons enjoy diffusion as in bulk media @xcite , in the strong disorder region @xmath3 the role played by the leakage at the interface has been of long term interests @xcite and , still , remains in the central position of cbs studies , particularly to forecast or observe the cbs line shape @xcite .
pressingly , in the latter region strong localization emerges in the bulk and a new scale namely the localization length @xmath4 appears . on the experimental side ,
there has been increasing evidence indicating that inside the boundary layer of thickness @xmath5 the photon leakage at the vacuum - medium interface strongly interplays with strong localization @xcite . on the theoretical side ,
some time ago exact solution of semi - infinite one - dimensional disordered chains shed the light on the existence of so - called radiative localization states in the boundary layer @xcite , which lead to anomalous slowing decay of reflected ( backscattered ) incident light pulses @xcite .
recently , in an insightful theoretical work @xcite it was realized that ( in three - dimensional disordered media ) inside the boundary layer the translational symmetry of low - energy ( hydrodynamic ) photon motion is strongly destroyed resulting in the so - called `` local diffusion '' .
remarkably , constructive wave interference renders the static diffusion coefficient depending on the distance from the interface .
consequently , the local diffusion was found to lead to a rounded cbs line shape resembling that observed experimentally @xcite and , thus , might overcome the conceptual difficulty of earlier theoretical proposal @xcite .
surprisingly , the dynamic generalization of the local diffusion equation @xcite provides an explanation of some key phenomena observed in quasi - one - dimensional microwave experiment @xcite , and well captures anomalous slowing decay of reflected incident pluses in both quasi - one - dimensional @xcite and three - dimensional disordered media @xcite .
moreover , such novel prediction the position dependence of diffusion coefficient seems to have been confirmed by a very recent experiment @xcite . despite of these progress theoretical investigations on local diffusion are , however , restricted on the self - consistent diagrammatical method @xcite .
thus , an intellectual challenge is to seek for the genuine microscopic origin underlying this novel concept .
this is , indeed , the purpose of this work . the last few decades have witnessed spectacular success of applications of supersymmetric field theory to various disordered systems in the absence of interactions ( such condition is perfectly satisfied by optical systems . ) @xcite . among them there are few exact nonperturbative results for quasi - one - dimensional disordered wires such as density - density correlation function ( in the infinite geometry ) @xcite and transmission statistics @xcite .
they allow one to make important insight on the strong localization .
most importantly , for periodic disordered media by using the supersymmetric field - theoretic method a local light diffusion equation , similar to that proposed in ref .
, recently has been derived at the microscopic level @xcite . in view of these
it is natural and inevitably necessary to proceed along the same line to explore the concept of local light diffusion and its effects for more general fully disordered media , which differ drastically from the former @xcite from both physical and technical view .
the main results of this paper are as follows .
( i ) we present a field - theoretic proof showing that , in contrary to the conjecture of ref .
, no scaling behavior exists inside a layer of thickness @xmath6 extrapolating into the vacuum .
( ii ) we justify at the perturbative level the local diffusion equation proposed in refs . .
( iii ) we analyze signatures of the static local diffusion in the cbs line shape .
it should be stressed that in this paper the suspersymmetric field theory is treated perturbatively , and the nonperturbative treatise will be reported in the forthcoming paper .
the rest of this paper is organized as follows . in the next section
we produce nonlinear supersymmetric @xmath7 model in the context of optical systems .
most importantly , we derive the boundary constraint satisfied by the supersymmetric matrix field .
the supersymmetric field theory is then applied to the two - dimensional medium bar .
[ renormal ] is devoted to exploring states residing deeply inside the semi - infinite medium bar ( namely far away from the interface ) by investigating renormalization effects of infinite medium bar . in sec .
[ wlsemiinfinite ] weak localization in the semi - infinite medium bar is studied , where the general dynamic local diffusion equation is justified .
the static limit of the local diffusion equation is studied in sec .
[ localdiffueq ] . in particular , the weak localization correction to the bare diffusion constant is explicitly calculated , and its signatures in the cbs line shape are analyzed .
we conclude in sec . [ conclusion ] and give some technical details in appendix [ derivationeffgreen]-[izdiscussion ] .
in this section a supersymmetric field - theoretic formalism is presented for light scattering in semi - infinite disordered medium .
we first show that as interactionless electron systems low - energy photon motion in bulk disordered media is well described by the nonlinear @xmath7 model .
the derivation is rather standard @xcite .
here we only outline the scheme with an emphasis on the main difference , while refer the reader to ref . for the details .
in the present work for simplicity the scalar wave will be considered .
the wave propagation in a bulk disordered medium is described by the helmholtz equation : @xmath8\right\}e({\bf r } ) = j({\bf r } ) \ , , \label{helmholtz}\ ] ] where the field @xmath9 has the radiation frequency @xmath10 ( velocity @xmath11 set to be unity ) , and @xmath12 is the source . here
the fluctuating dielectric field @xmath13 has zero mean and is distributed according to the gaussian @xmath14-correlated law : @xmath15 the helmholtz equation resembles the schrdinger equation with the hamiltonian now read out as @xmath16 . as usual
we may introduce the retarded / advanced green function @xmath17 defined as @xmath18 where @xmath19 . the electric field and the source are related via @xmath20 .
we may further introduce the diffuson @xmath21 and the cooperon @xmath22 propagator defined as @xmath23 with @xmath24 and @xmath25 , where the overline stands for the average over random dielectric field .
these two propagators describe elegantly the light propagation over large scales .
the propagators above are represented in terms of superintegrals . for this purpose
we define a supervector field @xmath26 : @xmath27 with @xmath28 s ( @xmath29 s ) complex commuting ( anticommunting ) variables , where the superscript @xmath30 ( @xmath31 ) refers to retarded ( advanced ) green function , and its charge conjugate @xmath32 . here
@xmath33 is an @xmath34 supermatrix : @xmath35 hereafter supermatrices are defined on the retarded / advanced ( ar ) , bosonic / fermonic ( bf ) and time - reversal ( tr ) sector .
then @xmath36\ , \psi^1_\alpha({\bf r}){\bar \psi}^1_\alpha({\bf r } ' ) \psi^2_\beta({\bf r}'){\bar \psi}^2_\beta({\bf r})\ , \overline{e^{-\mathfrak{l}[\psi,{\bar \psi}]}}\ , , \nonumber\\ & & { \cal y}^{\rm { c } } ( { \bf r},{\bf r}';\omega ) \nonumber\\ & = & -4 \int d[\psi]\ , \psi^1_\alpha({\bf r}){\bar \psi}^1_\alpha({\bf r } ' ) \psi^2_\beta({\bf r}){\bar \psi}^2_\beta({\bf r}')\ , \overline { e^{-\mathfrak{l}[\psi,{\bar \psi}]}}\ , .
\label{dcsuper}\end{aligned}\ ] ] here @xmath37\psi({\bf r } ) d{\bf r}\ , .
\label{lagrange}\ ] ] with @xmath38 , where the @xmath39 term is omitted since @xmath25 .
performing the average we arrive at @xmath40 } } = e^{-i\int { \bar \psi } \left[-{\cal h}_0 - \omega\ , \omega^+\lambda\right]\psi \
, d{\bf r}-\frac{\delta } { 2}\,\int \left({\bar \psi}\psi\right)^2 d{\bf r } } \ , . \label{lagrangeaverage}\ ] ] the quartic term is decoupled by the standard hubbard - stratonovich transformation .
introduce an @xmath34 supermatrix field @xmath41 conjugate to @xmath42 . here
@xmath43 is related to the photon density of states per unit volume @xmath44 by @xmath45 .
then , @xmath46 \label{hstransformation } \\ & = & \!\!\int\!\ ! \exp\left[-\pi\delta n(\omega^2)\!\!\int \!\ !
\left(\ ! { \bar \psi}q\psi+\frac{\pi n(\omega^2)}{4}\ , q^2 \ !
\right)d{\bf r}\right]\!\ !
d[q ] \ , .
\nonumber\end{aligned}\ ] ] substituting it into eqs .
( [ dcsuper ] ) and ( [ lagrangeaverage ] ) and integrating out the @xmath26 fields using the wick theorem , we obtain : @xmath47 ^ 2 \times
\label{dc } \\ & & \langle{\rm str } [ k(1+\lambda)(1- \tau_3 ) q({\bf r } ) ( 1-\lambda)(1\mp \tau_3 ) kq({\bf r } ' ) ] \rangle \nonumber\end{aligned}\ ] ] with @xmath48 and @xmath49 the pauli matrices defined on the time - reversal sector . in eq .
( [ dc ] ) the following average is introduced : @xmath50\rangle = \int d[q]\ , p[q]\ , e^{-f[q ] } \ , , \label{average}\ ] ] where the action @xmath51 $ ] is @xmath52 & = & \int d{\bf r}\ , { \rm str}\bigg\{\left(\frac{\pi n(\omega^2)}{2}\right)^2\delta q^2- \label{f}\\ & & \frac{1}{2}\ln\left[-i{\cal h}_0- i\omega\omega^+ \lambda+\pi\delta n(\omega^2)q({\bf r})\right ] \bigg\ } \ , . \nonumber\end{aligned}\ ] ] minimizing @xmath51 $ ] gives the saddle point equation : @xmath53 in the limit @xmath54 , eq .
( [ spa ] ) gives the saddle point as @xmath55 .
so far the derivation is exact .
fluctuations analysis around the saddle point may be performed for eqs .
( [ average ] ) and ( [ f ] ) .
yet , we could not proceed further and only give the results here , instead refer the reader to ref .
for all the details .
first , after standard procedure the mean field approximation namely @xmath55 gives the averaged retarded / advanced green function as @xmath56 the imaginary part of the self - energy gives the elastic mean free path which is @xmath57 and has the rayleigh form , i.e. , @xmath58 .
then , with @xmath59 taken into account the action is simplified to be @xmath51=\int d{\bf r}\ , { \cal l}[q]$ ] , where ( from now on we set @xmath60 to shorten formula . )
@xmath61=\frac{\pi\nu}{8 } \ , { \rm str}\ , [ d_0(\partial q)^2 + 2i\omega^+\lambda q ] \label{action}\ ] ] with the bare diffusion constant @xmath62 .
here @xmath63 describes goldstone modes with @xmath64 a matrix field taking the value in the coset space @xmath65 reflecting the orthogonal symmetry .
an explicit parametrization of @xmath66 will be given in the next section .
the action @xmath51 $ ] obtained above is invariant under the translational symmetry , which is broken in the presence of the vacuum - medium interface .
the broken translational symmetry may profoundly affect light propagation .
experience in mesoscopic physics shows that to take into account the vacuum - medium interface effect one may impose some appropriate boundary condition on the @xmath67-field in the field - theoretic formalism .
however this is a nontrivial task and in meseoscopic physics investigations so far have been restricted on interface structures of quasi - one - dimensional disordered wires and small quantum dots @xcite . in this part
we switch to optical systems and study the vacuum - medium coupling where the interface may be infinite and bear arbitrary geometry .
though the derivation below may be generalized to arbitrary dimension to simplify discussions we will focus on the two - dimensional case . let us suppose an arbitrary curve @xmath68 which divides the space @xmath69 into two disconnected subspaces @xmath70 and @xmath71 , i.e. , @xmath72 and @xmath73 .
we are interested in light propagation in some subspace say @xmath71 described by an effective green function @xmath74 , which is identical to @xmath75 for @xmath76 . to study such green functions for @xmath77 we introduce auxiliary green functions @xmath78 satisfying @xmath79 then our starting point is the following theorem due to zirnbauer @xcite and refined by efetov @xcite , which was originally established for description of coupling between leads and mesoscopic devices .
the theorem is stated as follows : ( for the self - contained purpose the proof tailored to the present context is given in appendix [ derivationeffgreen ] . )
+ _ for @xmath80 the green function @xmath74 solves @xmath81 where the normal unit vector @xmath82 at @xmath83 points to @xmath71 . here _
@xmath84 f({\bf r}')\ , , \nonumber\\ b({\bf r},{\bf r } ' ) & = & \partial_{{\bf n}({\bf r})}\partial_{{\bf n}({\bf r } ' ) } g^r_{\omega^2}({\bf r},{\bf r } ' ) \ , , for\ , { \bf r}\ , , { \bf r}'\in c\ , . \label{effgf6}\end{aligned}\ ] ] the effective hamiltonian for the retarded ( advanced ) green function is @xmath85 .
remarkably , it is non - hermitian due to the escape from @xmath71 into @xmath70 through @xmath68 . in the present case the vacuum - medium interface namely the curve @xmath68 is a straight line . to proceed we choose the coordinate system @xmath86 with the @xmath87 ( @xmath88)-direction perpendicular ( parallel ) to the vacuum - medium interface .
the vacuum fills the space @xmath89 where no dielectric scatterers are available . for technical reasons we assume that the dielectric scatterers located at @xmath90 ( @xmath91 ) are uniformly ( in the statistical sense ) distributed with the distance between nearest scatterers @xmath92 order of @xmath1 .
@xmath68 is located at @xmath93 and @xmath71 ( @xmath70 ) is set to be the medium ( vacuum ) .
taking into account the boundary condition specified in eq .
( [ auxiliarygreenfunction ] ) we find that the green function @xmath94 is @xmath95 upon the substitution of eq .
( [ gfinterface ] ) into eq .
( [ effgf6 ] ) the operator @xmath96 is simplified to be @xmath97 repeating the derivation of sec . [ nsm ] with the effective hamiltonian @xmath85 we arrive again at eq .
( [ f ] ) except that the action is modified according to @xmath51 \rightarrow f[q ] + f_{\rm inter}[q]$ ] with @xmath98 = -\frac{1}{2}\ , { \rm str}\ln\ , \left\{1-{\hat b } \lambda { \cal g}_0\right\ } \ , .
\label{interf}\end{aligned}\ ] ] expanding the logarithm and substituting eq .
( [ bdefinition ] ) into it we obtain @xmath98 = -\frac{1}{2 } \ , { \rm str } \left\{\sum_{n=1}^\infty\ , \frac{(-1)^{n+1}}{n } \left[{\hat b } \lambda { \cal g}_0\right]^n\right\ } \
, , \label{interf1}\end{aligned}\ ] ] where the supertrace @xmath99 includes the integration over @xmath88 , and @xmath100 exponentially decays for @xmath101 according to eq .
( [ spa ] ) . to calculate eq .
( [ interf1 ] ) we introduce , for arbitrary @xmath102 , the auxiliary variable @xmath103 $ ] . following ref . in the layer @xmath104 the green function @xmath105
may be approximated by @xmath106 where the longitudinal wave function is determined by the boundary condition of eq .
( [ theorem ] ) . here
@xmath107 and @xmath108 stand for the @xmath67- and the dielectric field , respectively in the regime : @xmath109\times [ 0\ , , l]$ ] .
they are considered to be a constant ( matrix ) since both @xmath67 and @xmath110 varies over the scale @xmath1 .
moreover , the transverse component @xmath111 is defined as @xmath112 substituting eq .
( [ g0 ] ) into eq .
( [ interf1 ] ) , with the help of the following identity : @xmath113\qquad \label{orthogonalrelation}\end{aligned}\ ] ] we obtain : @xmath98 = -\frac{1}{2 } \ , \sum_i \sum_{0
< \frac{k_\perp}{\omega } \leq 1 } { \rm str } \ln \left\{1 + \alpha^i_{k_\perp}\ , \lambda q_i\right\ } \label{interf2}\end{aligned}\ ] ] with @xmath114 taking the advantage of large channel number @xmath115 we may simplify it to be ( the details are given in appendix [ simplification ] . ) @xmath98 = -\frac{1 } { 4 } \ , \sum_i \frac{\omega l_i } { \pi}\ , t_0(i ) \ , { \rm str } \ , ( \lambda q_i ) \label{interfresult}\end{aligned}\ ] ] by assuming that @xmath116 does not depend on @xmath117 , i.e. , @xmath118 . here
@xmath119 is the well known transmission coefficient of electromagnetic wave @xcite . passing to the continuum limit : @xmath120 we rewrite the action @xmath121 $ ] as @xmath98 = -\frac{\omega}{4\pi}\ , \int dr_\perp \ ,
t_0(r_\perp ) \ , { \rm str } \ , [ \lambda q(r_\perp , z=0)]\ , . \label{interfresult1}\end{aligned}\ ] ] in the quasi - one - dimensional geometry the summation over @xmath102 is suppressed , and the coupling action @xmath121 $ ] namely eq .
( [ interfresult ] ) recovers the one obtained previously @xcite . for @xmath122
although to generalize the derivation above is straightforward , the coupling action may be obtained by simple physical arguments below .
notice that the coefficient of eq .
( [ interfresult ] ) allows a simple physical explanation @xcite : according to eq .
( [ greenfunctionaverage ] ) the ( single ) photon green function decays over the scale @xmath1 .
suppose that the medium is partitioned into boxes of volume @xmath123 , then the states ( denoted as @xmath124 ) in different boxes are uncorrelated .
the box states neighboring to the interface may be translated into the vacuum state ( denoted as @xmath125)so - called lead channel in the terminology of mesoscopic physics .
the coupling strength is @xmath126 with @xmath127 the scattering matrix element , which scales as @xmath128 with @xmath129 the interface area .
thus , although the total channel number is @xmath130 , the number of channels to which the interface box state is transmitted ( denoted as @xmath131 ) is much smaller @xmath132 .
more precisely , @xmath131 may be found to be @xmath133 for @xmath134 this gives @xmath135 namely the coefficient of eq .
( [ interfresult ] ) . for arbitrary @xmath136 with the replacement of @xmath137 ( and @xmath138 ) in eq .
( [ interfresult ] ) the vacuum - medium action becomes @xmath139 & = & -\frac{{\tilde n}_d } { 4 } \
, \sum_i \frac{(\omega l_i)^{d-1 } } { \pi}\ , t_0(i ) \ , { \rm str } \ , ( \lambda q_i ) \nonumber\\ & = & -\frac{{\tilde n}_d \omega^{d-1}}{4 } \ , \int d{\bf r}\delta_c\ , t_0({\bf r})\ , { \rm str } [ q({\bf r})\lambda ] \label{interfresult2}\end{aligned}\ ] ] with @xmath140 in the last equality of eq .
( [ interfresult2 ] ) we again pass to the continuum limit , and the operator @xmath141 is defined as @xmath142 . we then come to derive the boundary condition satisfied by @xmath67 . for this purpose
we employ the so - called boundary ward identity @xcite .
it states that an arbitrary local observable say @xmath143 ( with @xmath83 inside the medium ) , which is expressed in terms of the average of the functional @xmath144 $ ] namely @xmath145 { \cal p}[q({\bf r})]\ , e^{-\int_{z\in \mathbb{r}^+}d{\bf r}{\cal l}[q({\bf r})]-{\tilde f}_{\rm inter}[q]}\ , , \label{p}\ ] ] must be invariant under an infinitesimal boundary rotation below : @xmath146 \,,\label{rotation}\\ r & = & \left ( \begin{array}{cc } 0 & { \cal r}({\bf r}_\perp , z=0 ) \\ { \bar { \cal r}}({\bf r}_\perp , z=0 ) & 0 \\ \end{array } \right)^{ar}\otimes \mathbf{1}^{bf } \otimes \mathbf{1}^{tr}\ , .
\nonumber\end{aligned}\ ] ] notice that the boundary rotation alters neither @xmath147 $ ] nor @xmath148 $ ] for @xmath83 inside the medium .
the boundary ward identity then demands @xmath149 , i.e. , @xmath150\right)\right\ } \nonumber\\ & = & 0 \ , .
\nonumber\end{aligned}\ ] ] as @xmath151 are arbitrary this requires @xmath152\right)_\perp\big|_{z=0}=0 \label{bc}\ ] ] to be met , where the subscript @xmath153 stands for the offdiagonal component in the retarded / advanced sector ( thereby anticommunting with @xmath33 ) and @xmath154 .
( [ bc ] ) is the field - theoretic version of the radiative boundary condition @xcite . as we will show in sec .
[ wlsemiinfinite ] , it describes that the low - energy coherent dynamics penetrates into the vacuum with the extrapolation length @xmath155 notice that it is proportional to the inverse transmission coefficient in agreement with ref . .
for @xmath156 in the case of perfect transmission , i.e. , @xmath157 the extrapolation length is @xmath158 in agreement with ref . , and is closed to the one that obtained by solving milne equation @xcite which gives @xmath159 . traditionally the radiative boundary condition is imposed to diffusion equation to mimics the leakage at the interface @xcite and is justified for one - dimensional discrete random walk @xcite .
in the rest of this paper we will apply the supersymmetric field - theoretic formalism to the semi - infinite two - dimensional medium bar ( with the width @xmath160 ) , where in the transverse ( @xmath161- ) direction the photon motion is confined .
the purpose of this section is two - fold : on the physical side , we wish to explore how a finite width affects localization in the bulk which differs inessentially from localization in an infinite bar . accordingly , through out this section the action reads out as @xmath51=\int_{-\infty}^\infty dz\int_0^a d\rho\ , { \cal l}[q]$ ] . on the technical side , by presenting some details we wish to address the difference of calculations between semi - infinite and infinite bar , which originates at the fact that in the former system the translational symmetry of low - energy modes , i.e. , the @xmath67-field is broken . following the standard strategy we factorize the @xmath66-field into the slow and fast mode in terms of @xmath162 , where rotations @xmath163 ( @xmath164 ) involve spatial fluctuations on short ( large ) scales . substituting it into the action
we then obtain : @xmath52 & = & \frac{\pi\nu}{8}\int_{-\infty}^\infty \!\!\!\!dz \!\!\int_0^a \!\!\!\ !
d\rho\ , { \rm str } \ , \big\{d_0\big((\partial q_>)^2 + \label{action1 } \\ & & 4q_>\partial q_>\cdot \phi + [ \phi , q_>]^2\big ) + 2i\omega^+ q _
> t_<^{-1}\lambda t _
< \big\}\ , , \nonumber\end{aligned}\ ] ] where @xmath165 and @xmath166 . integrating out @xmath167 results in an effective action of @xmath168 . to work out the strategy outlined above we set @xmath169 with @xmath170 parameterized by @xmath171 since photons are confined the current vanishes at @xmath172 , i.e. , @xmath173 .
we may thus introduce the fourier transformation @xmath174 where @xmath175 and @xmath176 are ultraviolet cut - off of longitudinal and transverse wave number , respectively . in eq .
( [ parametrization ] ) the matrix @xmath177 has the structure as @xmath178 with @xmath179 where @xmath125 s , @xmath180 s ( @xmath7 s , @xmath181 s ) are complex bosonic ( grassmann ) numbers , and the charge conjugation transformation of a matrix @xmath182 is defined as @xmath183 straightforward calculations justify the useful identity : @xmath184 .
importantly , @xmath170 satisfies the following relation : @xmath185 which , as shown in appendix [ wsymmetry ] , enforces the invariance of @xmath170 under the charge conjugation , i.e. , @xmath186 .
now we study the one - loop renormalization . in doing
so we expand @xmath167 to quadratic order in @xmath170 .
consequently the action separates into three contributions : @xmath187 where the slow mode action is @xmath188 \label{saction}\end{aligned}\ ] ] with @xmath189 , the fast mode action is @xmath190 and the slow - fast mode coupling is described by the action : @xmath191\cdot \phi -(\phi\lambda w_>)^2 - ( \phi \lambda ) ^2 w_>^2 -i \partial w_>\cdot \phi - i(\phi\lambda)^2 w _
> \right ) - \nonumber\\ & & \frac{i\omega^+}{2 } \left(i w_>\lambda t_<^{-1}\lambda t _ < + w_>^2 \lambda t_<^{-1}\lambda t _
< \right)\bigg\ } \ , . \label{sfaction}\end{aligned}\ ] ] in order to calculate the general average with respect to the fast mode action : @xmath192\ , ( \cdots)e^{-f_f}$ ] we employ the contraction rule below : @xmath193\ , { \rm str } [ w_>({\bf r } ' ) m_2 ] \rangle_f & = & { \cal d}_f({\bf r}\ , , { \bf r}')\ , { \rm str}[(m_1+{\bar m}_1 ) m_2 ] \ , , \qquad m_{1,2}=\left ( \begin{array}{cc } 0 & m^{12}_{1,2 } \\ m^{21}_{1,2 } & 0 \\ \end{array } \right)^{ar } \label{wcontraction}\end{aligned}\ ] ] with the fast mode propagator @xmath194 } \nonumber\end{aligned}\ ] ] and the wick theorem .
notice that we assume that the fast mode propagator does not depend on the transverse center - of - mass coordinate , i.e. , @xmath195 and the self - average over this variable has been performed . performing the average of @xmath196 we obtain an effective action @xmath197=
f_s+ \langle f_{sf } \rangle_f$ ] , where @xmath198 with @xmath199 eqs .
( [ sfactionresult ] ) and ( [ i0 ] ) show that the one - loop renormalization results in the weak localization correction to the bare diffusion constant @xmath200 : @xmath201 & = & \frac{\pi\nu}{8}\ , \int_{-\infty}^{\infty } \!\!\!\!dz \!\!\int_0^a \!\!\!\ !
d\rho\ , { \rm str } \{[d_0+\delta d^{(2)}]\ , ( \partial q _ < ) ^2 + \nonumber\\ & & 2i\omega^+\lambda q_<\ } \label{effaction}\end{aligned}\ ] ] with the two - dimensional weak localization correction @xmath202 .
in the high - frequency region , i.e. , @xmath203 , the condition : @xmath204 is met .
the photon motion is thus two - dimensional described by the action , eq .
( [ effaction ] ) .
furthermore , since @xmath59 the two - dimensional weak localization correction @xmath205 is much smaller than @xmath200 , the photon motion thereby is diffusive . in the low - frequency region ,
i.e. , @xmath206 one may further enforce @xmath207 and @xmath208 and thereby obtain a quasi-@xmath30d effective action of @xmath209 which is homogeneous in the transverse direction : @xmath210 & = & \int_{-\infty}^{\infty } dz { \cal l}_{\rm eff}[\breve{q}]\ , , \nonumber\\ { \cal l}_{\rm eff}[\breve{q } ] & = & \frac{\pi\nu a}{8}\ , { \rm str}\ , [ d_{\rm eff}\ , ( \partial_z \breve{q})^2 + 2i\omega^+\lambda \breve{q } ] \ , , \label{action2dwl}\end{aligned}\ ] ] where the renormalized diffusion constant is @xmath211 note that in the above the second term suffers logarithmic divergence which , as usual , may be regularized by introducing the upper cut - off @xmath212 which is order of @xmath213 . as a result , @xmath214 thus , in the low - frequency region : @xmath215 the system is quasi - one - dimensional provided that the bar width satisfies @xmath216 . for
wider bar the system display two - dimensional strong localization which is beyond the scope of present perturbative analysis .
the discussions of sec . [ renormal ] break down in the semi - infinite geometry due to the absence of the translational symmetry . in this section and the next we turn to study the vacuum - medium interface effect on wave interference . in order to explore the physics implied by the boundary constraint eq .
( [ bc ] ) let us parameterize @xmath66 in the same way as eqs .
( [ parametrization ] ) , ( [ b ] ) and ( [ absigmaeta ] ) .
( to distinct notations from those of sec . [ renormal ]
we eliminate all the subscript @xmath217 . ) with the substitution of the parametrization and keeping eq .
( [ bc ] ) up to the first order in @xmath218 we obtain : @xmath219 implying @xmath220 with @xmath221 .
hence the low - energy goldenstone modes penetrate into the vacuum of a depth @xmath222 then exponentially decays .
that is , the optical paths underlying coherent multiple scattering do not cross the line located at @xmath223 . from now on
we assume that the interface is almost transparent namely @xmath224 closed to @xmath30 . in this case @xmath158 .
since the mean free path @xmath1 is much smaller than any other macroscopic scale we may safely assume that the crossing line where @xmath218 vanishes coincides with @xmath68 .
consequently the boundary constraint eq .
( [ bc ] ) is simplified as @xmath225 it is the action : @xmath226=\int_0^{\infty } dz\int_0^a d\rho \ , { \cal l}[q ] $ ] with the @xmath67-field subject to this boundary constraint that we will use in the rest of this paper .
expanding @xmath67 in terms of @xmath218 gives @xmath52=f_{2}[w ] + f_{4}[w ] + \cdots\ , , \label{actionexpansion}\end{aligned}\ ] ] where the gaussian action : @xmath227 = \frac{\pi\nu}{2 } \!\!\int_{0}^{\infty}\!\!\!\ ! dz \!\!\int_0^a \!\!\!\ !
d\rho \ , { \rm str } \ , [ d_0 ( \partial w )
^2 -i\omega w^2 ] \label{effectiveaction}\end{aligned}\ ] ] and @xmath228 & = & \frac{\pi\nu}{2 } \int_0^{\infty}\!\!\!\ !
dz \!\!\int_0^a \!\!\!\ !
d\rho\,\big\ { -2 d_0\ , { \rm str}\ , \left[(\partial w)^2 w^2\right ] \nonumber\\ & & \qquad\qquad\qquad \quad \,\,+ i\omega\ , { \rm str}\ , w^4 \big\ } \ , .
\label{f4}\end{aligned}\ ] ] from eq .
( [ effectiveaction ] ) immediately we obtain the same contraction rules as eq .
( [ wcontraction ] ) except making the replacement : @xmath229 where the propagator solves the diffusion equation : @xmath230 the boundary condition above is inherent from the constraint eq .
( [ bcsimplify2 ] ) which imposes @xmath231 . keeping the prefactor of eq .
( [ dc ] ) up to quadratic term we obtain the leading cooperon propagator : @xmath232 ^ 2 \ , \big\langle { \rm str } \big\{k ( 1+\lambda)(1- \tau_3 ) w({\bf r } ) \nonumber\\ & & \qquad\qquad \quad \times k ( 1-\lambda)(1 + \tau_3 ) w({\bf r } ' ) \big\ } \big\rangle_{f_2 } \nonumber\\ & = & \frac{2\pi\nu}{\omega^2 } \ , { \cal d } ( { \bf r},{\bf r}';\omega)\ , . \label{cooperon}\end{aligned}\ ] ] it is easy to see that the propagator above preserves the symmetry @xmath233 inherent from eq .
( [ dc ] ) .
( [ cooperon ] ) is traditionally obtained by summing up all the ladder diagrams and imposing appropriate boundary condition @xcite .
we proceed to calculate the one - loop correction to the bare propagator @xmath234 . for this purpose
we keep the @xmath218-expansion up to the quartic terms for both the prefactor and the action which , after straightforward calculations , gives the cooperon as @xmath235 with @xmath236 ^ 2 \ , \big\langle { \rm str } \left\{k ( 1+\lambda)(1- \tau_3 ) w^3({\bf r } ) k ( 1-\lambda)(1 + \tau_3 ) w ( { \bf r}')\right\ } \nonumber\\ & & + \ , { \rm str } \left\{k ( 1+\lambda)(1- \tau_3 ) w({\bf r } ) k ( 1-\lambda)(1 + \tau_3 ) w^3 ( { \bf r}')\right\ } \nonumber\\ & & - \ , { \rm str } \left\{k ( 1+\lambda)(1- \tau_3 ) w^2({\bf r } ) k ( 1-\lambda)(1 + \tau_3 ) w^2 ( { \bf r}')\right\ } \nonumber\\ & & + \ , { \rm str } \left\{k ( 1+\lambda)(1- \tau_3 ) w({\bf r } ) k
( 1-\lambda)(1 + \tau_3 ) w ( { \bf r}')\right\ } f_{4}[w ] \big\rangle_{f_2}\ , . \label{cooperonsl1}\end{aligned}\ ] ] first , it is easy to show that the third term in the right hand side of eq .
( [ cooperonsl1 ] ) vanishes .
second , as shown in appendix [ cancelation ] the first two terms partly cancel the last term .
eventually eq .
( [ cooperonsl1 ] ) is reduced into @xmath237 ^ 2 \ , ( \pi\nu d_0 ) \int_0^{\infty}\!\!\!\ !
dz_1 \!\!\int_0^a \!\!\!\ !
d\rho_1\ , \big\langle { \rm str } \left\{k ( 1+\lambda)(1- \tau_3 ) w({\bf r } ) k
( 1-\lambda)(1 + \tau_3 ) w ( { \bf r}')\right\ } \times \nonumber\\ & & { \rm str}\ , \big[\partial^2 \overbrace{w \,({\bf r}_1 ) w({\bf r}_1)w}\ , ( { \bf r}_1 ) w({\bf r}_1)+\left(\partial w ( { \bf r}_1)\ , w({\bf r}_1)\right)^2+\left(\partial w({\bf r}_1)\right)^2 w^2({\bf r}_1)\big ] \big\rangle_{f_2 } \ , , \label{cooperonsl2}\end{aligned}\ ] ] where the overbrace fixes the contraction and the derivative acts only on the nearest @xmath218 .
we remark that @xmath238 vanishes when @xmath239 is spatially homogeneous , which is a reflection of the flux conservation law or ward identity at the one - loop level .
notice that @xmath238 vanishes if either @xmath83 or @xmath240 belongs to the interface @xmath68 .
such property is inherent from eq .
( [ dc ] ) which vanishes upon sending either @xmath241 or @xmath242 to @xmath33 . using the contraction rules and integral by parts
we further reduce eq .
( [ cooperonsl2 ] ) into @xmath243 after tedious but straightforward calculations .
notice that @xmath244 preserves the symmetry : @xmath245 .
( [ cooperonsl3 ] ) justifies that @xmath246 solves the following local diffusion equation : @xmath247 at the one - loop level . here
@xmath248 with the weak localization correction @xmath249 the local diffusion equation differs from the traditional one in that the diffusion coefficient is position - dependent .
it may be amounted to incompletely developed constructive interference between two counter - propagating optical paths which leads to the weak localization near the boundary . indeed ,
although deep inside the medium @xmath250 saturates recovering the bulk weak localization , at the interface it vanishes , i.e. , @xmath251 importantly , this is contrary to the theoretical proposal of ref . which claims that wave interference democratically renormalizes the diffusion constant appearing in both the diffusion equation in the bulk and the radiative boundary condition at the interface . here
several remarks are in order : ( i ) higher order loop corrections preserve eq .
( [ localdiffusion ] ) .
they affect the local diffusion equation by introducing higher order weak localization corrections which are also position - dependent .
this peculiar property reflects the photon number conservation law and is protected by ward identity .
( ii ) in the presence of internal reflection namely @xmath252 ( far ) below @xmath30 , ( i ) is no longer applicable because the simplification namely eq .
( [ bcsimplify2 ] ) breaks down due to large extrapolation length .
in fact , ref . falls into this case .
( iii ) the concept of local diffusion originally introduced in ref . at the static limit , i.e. , @xmath253 together with its dynamic generalization @xcite
is now justified at the perturbative level .
in this section we study the static limit : @xmath253 ( for this reason below we suppress the argument @xmath254 in all the formulae . ) of the local diffusion equation namely eq .
( [ localdiffusion ] ) for a bar with the width satisfying @xmath255 .
in particular we will explicitly calculate the weak localization correction eq .
( [ wl ] ) , and study its effects on the coherent backscattering phenomenon . in the static limit the weak localization correction eq .
( [ wl ] ) becomes self - averaged over the center - of - mass @xmath161 and thereby is @xmath161-independent .
that is , @xmath256 substituting eq .
( [ wlselfaverage ] ) into eq .
( [ localdiffusion ] ) we find that @xmath257 depends on @xmath258 , but not on the center - of - mass @xmath195 .
therefore , we may introduce the fourier transform : @xmath259 and insert it into eq .
( [ localdiffusion ] ) to obtain ( @xmath260 ) @xmath261 where @xmath262 is infinitesimal positive constant .
( [ localdiffusion1d ] ) may be considered to be a quasi - one - dimensional local diffusion equation with a mass @xmath263 .
the weak localization correction eq .
( [ wlselfaverage ] ) then becomes @xmath264\ , .
\label{wlselfaverage1}\ ] ] here @xmath265 satisfies @xmath266-i0^+\right\ } { \cal d}_{q_\perp}(z , z ' ) & = & \delta(z - z')\,,\nonumber\\ { \cal d}_{q_\perp}(z=0\ , , z ' ) & = & 0 \ , .
\label{bareproagator}\end{aligned}\ ] ] it is solved by ( with the introduction of @xmath267 ) @xmath268 where @xmath269 . substituting it into eq .
( [ wlselfaverage1 ] ) gives @xmath270 where the first term is the quasi - one - dimensional contribution , and the second term is the two - dimensional contribution with @xmath271 standing for the transverse hydrodynamic wave number .
as expected at @xmath272 the weak localization correction @xmath273 vanishes .
away from the interface i.e. , @xmath274 it may be approximated by @xmath275 as shown in appendix [ izdiscussion ] .
this suggests that in this region ( even in the static limit ) the two - dimensional low - energy motion dominates the weak localization .
the two - dimensional contribution saturates at @xmath276 : @xmath277 where the second term is none but the bulk weak localization correction ( see eq .
( [ deff ] ) ) renormalizing the bare diffusion constant @xmath200 . with this taken into account eq .
( [ i1approx1 ] ) may be rewritten as @xmath278 where @xmath279 stands for the quasi - one - dimensional weak localization correction , and @xmath280 is the exact localization length @xcite .
( [ i1approx2 ] ) agrees with the leading @xmath281-expansion of the local diffusion coefficient given in ref . .
it thereby justifies that for @xmath282 the medium bar displays the quasi - one - dimensional ( interface ) weak localization .
indeed , at early times @xmath283 incident photons explore a region of size @xmath284 neighboring to the interface , approaching a uniform distribution in the transverse direction . at later times they diffuse as in a quasi - one - dimensional medium .
technically , starting from quasi - one - dimensional @xmath7 model by performing the one - loop calculation one finds eq .
( [ i1approx2 ] ) .
importantly , from eqs .
( [ i1approx ] ) and ( [ i1approx2 ] ) we find that within the boundary layer @xmath285 weak _ rather than strong _ localization occurs even in the static limit .
in this part we turn to investigate effects of local diffusion on the cbs line shape .
we will consider a medium illuminated by light of frequency @xmath10 parallel to the bar , and calculate the angular resolution of the backscattered light intensity @xmath286 near the inverse incident direction . since the bar is wide enough so that @xmath287 a large parametric region : @xmath288 is opened . below
we pay particular attention to the line shape at @xmath289 .
( notice that the line shape is symmetric with respect to @xmath290 . )
it is well known that the backscattered light intensity may be decomposed into the background @xmath291 and the coherent part @xmath292 according to @xmath293 here @xmath294\ , { \cal y}^{\rm { c } } ( { \bf r},{\bf r } ' ) \ , , \nonumber\end{aligned}\ ] ] where @xmath295 because of @xmath296 , and the overall normalization factor is omitted .
first of all , it is easy to show that @xmath297 and therefore only the coherent part @xmath292 which determines the line shape will be studied below . inserting the fourier transform , namely eq .
( [ cooperonfourier ] ) into @xmath292 we arrive at @xmath298 where the propagator @xmath299 solves eq .
( [ localdiffusion1d ] ) .
the interfering optical paths penetrate into the medium of a depth @xmath300 .
let us first study the cbs line shape in the region : @xmath301 . because of the condition : @xmath302 the cbs line shape is mainly responsible for by photons which diffuse around the interface , i.e. , @xmath303 and thereby undergo two - dimensional weak localization .
indeed , the first term of eq .
( [ wlresult ] ) is much smaller due to the condition @xmath304 . setting the ultraviolet cutoff @xmath305 ( as eq .
( [ deffapprox ] ) ) we may approximate eq .
( [ wlresult ] ) by @xmath306 with the substitution of such weak localization correction into eq .
( [ localdiffusion1d ] ) we find @xmath307 \ , , \nonumber\\ & & \pi /a
\lesssim l^{-1 } \ , .
\label{lineshaperesult1}\end{aligned}\ ] ] according to eq .
( [ lineshaperesult1 ] ) the conventional triangular peak described by the factor @xmath308 is enhanced by a logarithmic factor ( dashed line in fig .
[ fig ] ) .
( [ wlresultapprox ] ) indicates that in the region : @xmath309 the local diffusion is of minor importance .
it is the two - dimensional bulk weak localization that is responsible for such a logarithmic enhancement .
thus in such region the scaling theory , still , is applicable .
at the exact backscattering direction , i.e. , @xmath290 interfering optical paths penetrate into the medium bulk : @xmath310 where quasi - one - dimensional ( bulk ) strong localization states are formed .
in contrast to @xmath301 for @xmath311 the local diffusion plays crucial roles and thus strongly affects the cbs line shape . indeed , despite of the nonperturbative nature of strong localization the backscattering light intensity at @xmath311 may be easily found provided that in the region @xmath312 the local diffusion equation is still valid @xcite . for @xmath311 from eq .
( [ localdiffusion1d ] ) one may find @xmath313 where in the second equality the substitution of eq .
( [ i1approx1 ] ) is made .
this immediately gives @xmath314 eq . ( [ albedoexactdirection ] ) shows that although due to two - dimensional bulk weak localization the cbs line shape develops a logarithmic singularity at @xmath315 , the singularity is cut off at @xmath311 ( fig .
[ fig ] ) where quasi - one - dimensional strong localization occurs in the bulk .
it is in order to remark that formally there is a region @xmath316 where the rounding due to the local diffusion occurs ( yet , the detailed rounding form depends on @xmath317 in the region @xmath318 , to find which is beyond the present perturbative treatise . ) , however , it is unobservable because the finite bar width renders @xmath319 .
finally we anticipate that the predicted cbs line shape is qualitatively correct for @xmath320 , though the analytical result here is obtained for @xmath0 .
for light propagation in fully disordered media a supersymmetric field theory is presented . the supersymmetric @xmath7 model described by eq .
( [ action ] ) may be applied to bulk ( infinite ) media for studies of optical localization transition . in this direction
it may serve as an alternative to the replica field theory @xcite . however , the supersymmetric field - theoretic formalism turns out to be far more powerful as propagation of incident light in semi - infinite media @xcite concerned , which is the subject of this paper and is closely related to the coherent backscattering phenomenon . differing from infinite medium in the presence of the vacuum - medium interface @xmath68 (
the interface may bear arbitrary geometry but must be smooth over the scale of the mean free path . ) the supermatrix field ( locally ) satisfies the radiative boundary condition eq .
( [ bc ] ) .
accordingly , the bare diffusion constant acquires a position - dependent wave interference correction namely @xmath321 which roots in the incomplete constructive interference ( weak localization ) near the interface .
thus , we justify the ( static ) local diffusion equation , originally proposed in ref . , as well as its dynamic ( i.e. , @xmath322 ) generalization @xcite .
most importantly , for ( almost ) transparent interface , i.e. , @xmath323 , the weak localization correction @xmath324 vanishes at the interface .
this immediately shows that the radiative boundary condition is protected against wave interference effects , and constitutes an explicit proof that no scaling hypothesis might exist in the extrapolation layer of thickness @xmath6 .
therefore , the present work supports the criticism of refs . on the earlier theoretical proposal @xcite . in the present work the static limit of the wave interference ( weak localization )
correction namely @xmath325 is explicitly calculated for the two - dimensional semi - infinite medium with a finite width @xmath125 ( the bar geometry ) , where @xmath326 solely depends on the distance from the interface @xmath87 . for @xmath327 a dimensional crossover of the wave interference correction @xmath328 is found .
indeed , @xmath329 displays two - dimensional weak localization at @xmath330 with a logarithmic dependence on @xmath87 , while displays one - dimensional weak localization at @xmath331 .
furthermore , for the latter region it is not difficult to generalize eq .
( [ i1approx2 ] ) to higher order loop corrections which reads out as @xmath332 this at the perturbative level formally confirms the result of ref . for one - dimensional geometry .
notice that the unimportant numerical expansion coefficients @xmath333 may vary depending on the strict / quasi- one - dimensional geometry . for wider medium bar such that
@xmath334 ( e.g. , infinite medium plane ) eq . ( [ i1approx ] ) , in fact , is still applicable except that the one - dimensional contribution vanishes .
that is , @xmath335 this suggests that in the medium there exists a boundary layer of thickness @xmath336 outside which two - dimensional strong localization occurs .
surprisingly , inside the layer the diffusion coefficient logarithmically depends on the distance from the interface .
how to reproduce this logarithmic dependence by the self - consistent diagrammatical method @xcite is unclear .
finally , it should be stressed that the present field - theoretic justification of local diffusion is perturbative .
therefore , the validity of such a concept in the nonperturbative strong localization region remains an important question .
this problem is far beyond the scope of this paper and will be addressed in a forthcoming paper , especially the issue how eq .
( [ i1approx3 ] ) is extended to the nonperturbative region : @xmath318 .
it is well known that in faraday - active medium the one - loop weak localization may be strongly suppressed @xcite .
therefore , to take into account such medium within the present field - theoretic formalism remains another important problem .
it is also interesting to generalize the present field - theoretic formalism to include medium gain @xcite .
these issues are left for future work .
i am deeply grateful to a. altland , t. micklitz for useful conversations , especially m. r. zirnbauer for several fruitful discussions .
work supported by transregio sfb 12 of the deutsche forschungsgemeinschaft .
assuming that @xmath337 and @xmath338 , from the helmholtz equation ( [ helmholtz ] ) we obtain : @xmath339\ } [ g^a_{\omega^2}({\bf r},{\bf r}'')]^ * \nonumber\\ & = & \delta ( { \bf r}-{\bf r}'')\ , g^r_{\omega^2}({\bf r},{\bf r } ' ) \label{effgf1}\end{aligned}\ ] ] and @xmath340^*\ , \{\nabla^2 + \omega_+^2[1+\epsilon({\bf r})]\ } g^r_{\omega^2}({\bf r},{\bf r } ' ) = 0 \ , . \label{effgf2}\end{aligned}\ ] ] subtracting eq .
( [ effgf2 ] ) from eq .
( [ effgf1 ] ) gives @xmath341 g^r_{\omega^2}({\bf r},{\bf r}')- [ g^{a*}_{\omega^2}({\bf r},{\bf r}'')]\nabla g^r_{\omega^2}({\bf r},{\bf r } ' ) \ } \nonumber\\ & = & \delta ( { \bf r}-{\bf r}'')\ , g^r_{\omega^2}({\bf r},{\bf r } ' ) \ , , \label{effgf3}\end{aligned}\ ] ] where @xmath342 acts only on @xmath83 .
noticing that @xmath343 and @xmath344^*=g^r_{\omega^2}({\bf r}'',{\bf r})$ ] , with @xmath345 integrated out we find @xmath346 here @xmath347 stands for the normal derivative at @xmath83 with @xmath82 pointing to @xmath71 . taking the derivative
we obtain : @xmath348 with @xmath349 following the definition of eq .
( [ effgf6 ] ) . now suppose that @xmath83 is shuffled to @xmath68 from the @xmath70 side .
let us integrate out eq .
( [ greenfunction ] ) over the line element along an infinitesimal piece of a curve passing from @xmath70 to @xmath71 . in doing
so we obtain : @xmath350 where @xmath351 ( @xmath352 ) stands for the curve infinitesimally closed to @xmath68 from the @xmath71 ( @xmath70 ) side .
taking eq .
( [ effgf5 ] ) into account we may rewrite eq .
( [ greenfunction ] ) as @xmath353 which is supplemented by the boundary condition @xmath354 .
@xmath355 ( @xmath356 ) consists of the real and imaginary part .
the former is small and may be absorbed into @xmath357 renormalizing @xmath13 .
it is thus ignored .
in contrast , the latter is important determining the analytical structure . taking this into account we prove eq .
( [ theorem ] ) for the retarded ( and similarly for the advanced ) green function .
for the moment let us suppress the indices @xmath102 and @xmath117 , and decompose @xmath67 according to @xmath358=0\ , , \quad \{q_\parallel\ , , \lambda\}=0 \ , .
\label{qdecomposition}\end{aligned}\ ] ] taking such decomposition into account we obtain : @xmath359 upon taylor expanding the second logarithm only the even order terms contribute .
thus , eq .
( [ simplification1 ] ) may be rewritten as @xmath360 where in deriving the third equality we use the identity @xmath361 , and in deriving the last two equalities we use the identity @xmath362 . restoring the index @xmath102 and substituting eq .
( [ simplification2 ] ) into eq .
( [ interf2 ] ) we obtain : @xmath98 & = & -\frac{\omega l}{4\pi}\,\sum_i { \rm str}\ln \left[2-t_0(i ) + t_0(i)\lambda q_{i\parallel}\right]\nonumber\\ & \approx & -\frac{\omega l}{4\pi}\,\sum_i t_0(i ) { \rm str}\ , \left(\lambda
q_{i\parallel}\right ) \nonumber\\ & = & -\frac{\omega
l}{4\pi}\,\sum_i t_0(i ) { \rm str}\ , \left(\lambda q_i\right)\ , , \label{simplification3}\end{aligned}\ ] ] where in the second line we take advantage of strong coupling , i.e. , @xmath59 , and in the third equality we use the identity @xmath363 .
the charge - conjugation symmetry is irrespective of fast - slow mode separation and therefore we ignore the subscript @xmath364 . substituting the parametrization of @xmath218 into eq .
( [ wrelation ] ) we obtain : @xmath365 giving @xmath366 and @xmath367 .
the first relation may be rewritten as @xmath368 substituting the second relation into eq .
( [ brelation ] ) we obtain : @xmath369 giving @xmath370 .
noticing the relation : @xmath371 one finds @xmath372 and thus justifies the charge - conjugation symmetry of @xmath218 .
using integral by parts we transform eq .
( [ f4 ] ) into ( noticing that @xmath373 ) @xmath374 = \frac{\pi\nu}{2 } \int_0^{\infty}\!\!\!\ ! dz \!\!\int_0^a \!\!\!\ ! d\rho\,\left\ { 2 d_0\ , { \rm str}\ , \left[\partial^2 w\ , w^3+\left(\partial w \ , w\right)^2+\left(\partial w\right)^2 w^2\right ] + i\omega \ , { \rm str}\ , w^4 \right\ } \ , .
\label{f4transformation}\ ] ] using the contraction rule we obtain : @xmath375\ , { \rm str}\ , [ \partial^2 w \,({\bf r}_1 ) w^3({\bf r}_1)]\right\rangle_{f_2 } \label{ia}\\ & = & i_a + \pi\nu d_0\ , \int_0^{\infty}\!\!\!\ ! dz_1 \!\!\int_0^a \!\!\!\ ! d\rho_1 \ , \big\langle{\rm str}\ , [ k ( 1+\lambda)(1- \tau_3 ) w({\bf r } ) k
( 1-\lambda)(1 + \tau_3 ) w ( { \bf r}')]\ , { \rm str}\ , \big[\partial^2 \overbrace{w \,({\bf r}_1 ) w({\bf r}_1)w}\ , ( { \bf r}_1)w({\bf r}_1)\big ] \big\rangle_{f_2 } \ , , \nonumber\end{aligned}\ ] ] where the overbrace fixes the contraction and @xmath376 \big\rangle_{f_2}+\nonumber\\ & & \qquad\qquad\qquad\qquad\,\,\,\,\partial_{{\bf r}_1}^2 { \cal d}({\bf r } , { \bf r}_1;\omega)\big\langle { \rm str}\ , [ k ( 1+\lambda)(1- \tau_3 ) w^3({\bf r}_1 ) k ( 1-\lambda)(1 + \tau_3 ) w({\bf r } ' ) ] \big\rangle_{f_2}\bigg\ } \ , . \label{iaresult}\end{aligned}\ ] ] likewise , we also obtain : @xmath377\ , { \rm str}\ , [ w^4({\bf r}_1 ) ] \right\rangle_{f_2}\nonumber\\ & = & i\omega\ , \int_0^{\infty}\!\!\!\ ! dz_1 \!\!\int_0^a \!\!\!\ ! d\rho_1 \
, \bigg\ { { \cal d}({\bf r } ' , { \bf r}_1;\omega ) \left\langle{\rm str}\ , [ k ( 1+\lambda)(1- \tau_3 ) w({\bf r } ) k ( 1-\lambda)(1 + \tau_3 ) w^3({\bf r}_1 ) ] \right\rangle_{f_2}+ \nonumber\\ & & \qquad\qquad\qquad\qquad{\cal d}({\bf r } , { \bf r}_1;\omega ) \left\langle{\rm str}\ , [ k ( 1+\lambda)(1- \tau_3 ) w^3({\bf r}_1 ) k ( 1-\lambda)(1 + \tau_3 ) w({\bf r}')]\right\rangle_{f_2}\bigg\}\ , . \label{ib}\end{aligned}\ ] ] noticing eq .
( [ diffusion ] ) we find that @xmath378 exactly cancels the first two terms of eq .
( [ cooperonsl1 ] ) .
let us introduce the function : @xmath379/(n\pi)$ ] . taking its derivative
we obtain : @xmath380 on the other hand , the low - energy diffusion occurs on the scale @xmath6 . over this scale the interface where @xmath381 vanishes is smeared .
therefore , without loss of any physics we may reformulate the boundary condition as @xmath382 .
taking it into account and integrating out eq .
( [ fderivative ] ) we obtain @xmath383 for @xmath384 , which gives @xmath385 for @xmath274 justifying eq .
( [ i1approx ] ) .
a. a. golubentsev , zh .
. fiz . * 86 * , 47 ( 1984 ) [ sov .
jetp * 59 * , 26 ( 1984 ) ] ; m. p. van albada and a. lagendijk , phys .
lett . * 55 * , 2692 ( 1985 ) ; p. e. wolf and g. maret , _ ibid . _ * 55 * , 2696 ( 1985 ) .
|
a supersymmetric field theory of light diffusion in semi - infinite disordered media is presented . with the help of this technique
we justify at the perturbative level the local light diffusion proposed by tiggelen , lagendijk , and wiersma [ phys .
rev .
lett . * 84 * , 4333 ( 2000 ) ] , and show that the coherent backscattering line shape of medium bar displays a crossover from two - dimensional weak to quasi - one - dimensional strong localization .
|
wall dynamics in thin magnetic strips or wires driven by either applied external magnetic fields or electric currents has been an active field of research during recent years , due to both the fundamental aspects of the underlying physics as well as due to the potential spintronics applications @xcite .
most of the theoretical studies of such phenomena have so far focused on `` perfect '' systems free of any imperfections in the sample that could affect the dynamics of the domain wall .
however , disorder is present in practically any realistic material , either in the form of edge roughness or various point - like defects in the bulk of the system .
while the effect of edge roughness @xcite or notches @xcite on domain wall motion has received much attention in the literature , bulk disorder is usually assumed to be insignificant in narrow nanostrips .
however , this might not be true in general : for instance , thickness fluctuations of the strip might give rise to bulk disorder similarly to the disorder due to the rough edges of the system . also
other forms of impurities might be present in the bulk of the system , affecting the domain wall dynamics .
in this paper we present a line - based model of a transverse domain wall in a narrow and thin disordered magnetic ( nano)strip .
such a coarse grained model is useful in understanding the effect of various kinds of disorder on the domain wall dynamics , because it does not suffer from the shortcomings of some other approaches such as micromagnetic simulations where inclusion of disorder is tricky due to mesh - related problems , or point - particle models in which the internal degrees of freedom of the domain wall are not considered .
we focus on the effect of randomly distributed point - like pinning centers in the bulk of the system on the domain wall dynamics in the sub - threshold thermally activated creep regime . in some recent studies
, it has been demonstrated that the experimentally relevant domain wall velocities often are within this regime @xcite .
the paper is organized as follows : in the next section , the line based model of a transverse domain wall is presented , and some numerical results on the creep motion of such a domain wall are presented in section iii .
section iv finishes the paper with discussion and conclusions .
we consider here strips of length @xmath0 , width @xmath1 and thickness @xmath2 , satisfying @xmath3 , made of a soft magnetic material such as permalloy .
the domain wall structure in such a strip depends on the balance between exchange and anisotropy energies .
the latter is here taken to be dominated by shape anisotropy , and thus the domains lie along the long axis of the strip .
we focus on the case in which the width @xmath1 and thickness @xmath2 of the strip are sufficiently small , so that the stable domain wall structure is the so called transverse wall @xcite . for an example of its micromagnetic structure , see the top panel of fig .
[ fig : tdw ] . for wider and/or thicker strips ( not considered here ) , a domain wall with vortex topology would have a lower energy .
due to the balance between exchange interactions and the effect of the demagnetizing fields @xmath4 arising from the `` magnetic charges '' within the transverse domain wall , the equilibrium shape of the wall resembles the letter v. these charges are associated to the discontinuities of the normal component of the magnetization .
a simplified description of the wall structure is obtained by considering the wall to be composed of _ two _ @xmath5 domain walls , with the first one from the left in fig .
1 separating domains with spins pointing to the right and up , respectively , while the second @xmath5 wall separates the domain with spins up from from the one with spins to the left . clearly , such a domain description can not be completely accurate as the magnetization within the wall rotates smoothly from right to left as in fig .
[ fig : tdw ] , so the magnetic charges associated to this rotation are actually distributed across the transition region . nevertheless , by imposing the above approximation of two 90-degree walls with the magnetic charges concentrated along these walls yields a qualitatively correct equilibrium shape for the transverse wall , i.e. the two 90-degree walls form a v - shaped structure . the lower panel of fig .
[ fig : tdw ] shows a schematic of the model .
the two lines , labeled ( a ) and ( b ) in the lower panel of fig [ fig : tdw ] , have line tensions @xmath6 , and interact repulsively due to the exchange interaction . with the local width of the wall @xmath7 , exchange interactions along the strip axis lead to an energy @xmath8 , resulting in a repulsive force @xmath9 between any pair of line segments with the same @xmath10-coordinates . these forces are balanced by the forces due to the demagnetizing fields arising from the magnetic charges . the top edge ( with @xmath11 ) has a positive charge density @xmath12 , where @xmath13 is the saturation magnetization . similarly , the bottom edge ( where @xmath14 ) has @xmath15 . as described above , also the lines are charged : whenever the local angle between the line and the @xmath16-axis is different from 45 degrees , the lines have a non - zero charge density .
notice that the total charge of the domain wall is counterbalanced by the negative charges at the two ends of the strip ( with @xmath17 and @xmath18 ) .
furthermore , the domain wall can be driven towards the positive @xmath16 direction either by applying an external magnetic field @xmath19 or an electric current @xmath20 ( with the direction chosen in such a way that the electrons move along the @xmath21 direction ) . here
we restrict ourselves to the case of low fields and currents : for stronger driving one should also include a `` phase '' angle of the domain wall magnetization as a variable to the model . for the moment we assume that all magnetic moments remain within the plane of the strip . in other words , the fields and currents
are restricted to be well below the so called walker threshold , where precession of the domain wall magnetization about the long axis of the strip would take place ( something that in the strip geometry considered here would in fact proceed via nucleation and propagation of an _ antivortex _ ) . to see how the effect of a small current can be included in the present model , we point out that the landau - lifshitz - gilbert ( llg ) equation describing the time evolution of the magnetization @xmath22 , with the additional terms describing the effect of current can be written @xcite @xmath23 where @xmath24 is the total effective magnetic field , @xmath25 is the gyromagnetic ratio and the second term is the gilbert damping term .
the effects due to an electric current are taken into account by the last two terms , usually referred to as the _ adiabatic _ and _ non - adiabatic _ terms , respectively .
the essential point here is that the effect of current enters in both terms through the gradient of @xmath26 : the non - adiabatic term enters like a magnetic field that is proportional to @xmath27 , while the adiabatic term exerts a torque on the spins of the domain wall , which is again proportional to @xmath27 .
thus , the simplest way to include the effects of the current is to add a local magnetic field which is proportional to the rate of change of the magnetization across the domain wall , i.e. to @xmath28 .
thus , the difference between the two forms of driving is that whereas the effect of the field is independent of the local domain wall width , the same is not true for the current drive .
, @xmath29 and @xmath30 . *
bottom : * a schematic of the line - based model , where two lines ( labeled ( a ) and ( b ) , respectively ) separating three different magnetic domains ( with the magnetization direction indicated by the thick solid arrows ) form a model of the transverse wall .
the dashed arrows indicate the directions @xmath31 ( @xmath32 ) of the components of the local field which drive the lines to the positive @xmath16 direction , as well as the line normals @xmath33 .
, title="fig:",width=288 ] + , @xmath29 and @xmath30 . *
bottom : * a schematic of the line - based model , where two lines ( labeled ( a ) and ( b ) , respectively ) separating three different magnetic domains ( with the magnetization direction indicated by the thick solid arrows ) form a model of the transverse wall .
the dashed arrows indicate the directions @xmath31 ( @xmath32 ) of the components of the local field which drive the lines to the positive @xmath16 direction , as well as the line normals @xmath33 .
, title="fig:",width=297 ] the total magnetic field due to all these contributions is then evaluated at each segment of the discretized lines . while the domain wall as a whole is driven towards the @xmath21 direction by a field along the positive @xmath16 direction ,
the same is not true for the two `` sub - walls '' separately : due to the different spin orientations on each side of the two sub - walls , the relevant field components driving the sub - walls are also different from each other . for the line ( a ) ,
the relevant component of the local magnetic field driving the line towards the @xmath21 direction is the component along @xmath34 , while the relevant component for line ( b ) is along @xmath35 .
notice that the _ total _ field acting on the domain wall as a whole is along the positive @xmath16 direction ( @xmath36 ) . a force with a magnitude of these components of the local magnetic field
is then taken to act on each segment of the lines along the local normal @xmath37 .
this leads to a local normal force along each segment of the two lines , which is then used to calculate the dynamics of the system . to account for the effect of the finite damping coefficient in eq .
( 3 ) , we introduce a domain wall mass @xmath38 , along with a friction force , so that the equation of motion for the line segment @xmath39 reads @xmath40\vec{n}_j,\end{aligned}\ ] ] where @xmath41 is the effective friction coefficient , @xmath6 is the line tension , @xmath42 is the local curvature of the line and @xmath37 is the local normal vector of the line .
@xmath43 ^ 2 $ ] and @xmath44 ^ 2 $ ] are the repulsive forces between the two line segments of the different lines with the same @xmath10 coordinate due to exchange interactions .
@xmath45 is a quenched random potential describing the interaction of the line segments with various defects localized at random positions @xmath46 within the strip ( with @xmath47 the range of interaction ) . @xmath48 and @xmath49 are the demagnetizing and external fields acting on the line segment @xmath39 , respectively . as discussed earlier , we include the effects due to an electric current as a local field with a magnitude inversely proportional to @xmath7 , by setting @xmath50\hat{x}$ ] , where @xmath51 is the magnitude of the applied external field and @xmath52 is the applied current density .
finally thermal effects are included by a random force @xmath53 acting on each segment @xmath39 of the line , satifying @xmath54 and @xmath55 .
the model is implemented by using the software package surface evolver @xcite , within which the line is taken to be formed by a set of vertices connected to each other by edges .
the program allows automatic remeshing of the discretized line , with the line moving in continuous space , thus avoiding spurious lattice effects .
the equations of motion are integrated numerically with the euler algorithm . in the absense of disorder ,
thermal effects and external drive , the initial configuration of two parallel lines relax to reach a v - shaped stable structure , with the angle of the v - shape depending on the relative magnitude of the demagnetizing fields and effects arising from the exchange interactions ( the line tension of the lines , the repulsive interaction between the two ) . after this initial relaxation ,
the external drive , the interaction with disorder and thermal effects are turned on and the evolution of the system is monitored . in the spirit of the model ( where we imposed a restriction that the external driving is weak ) , we focus on the thermally activated subthreshold creep regime
. the parameters of the simulations in dimensionless units of eq .
( [ eq : eom ] ) are as follows : @xmath56 , @xmath57 , @xmath58 , @xmath59 , @xmath60 and @xmath61 . the demagnetizing field due to a charge @xmath62 ( with @xmath63 measuring the charge density in units of the saturation magnetization @xmath13 ) within a line segment of length @xmath64 positioned at the origin
is taken have a magnitude @xmath65 , where the saturation magnetization is taken to have a value @xmath66 , and @xmath67 is the distance from the origin .
the unit of length is set by the sample dimensions , @xmath68 and @xmath69 ( the very small thickness @xmath2 of the strip is neglected by treating the system as two - dimensional ) .
[ fig : snapshots ] shows an example of a sequence of configurations of the creeping domain wall in the field driven case .
( from left to right , @xmath70 , @xmath71 and @xmath72 in dimensionless units ) of the domain wall moving in a disordered nanostrip ( with the dots representing the pinning centers ) under the influence of a subthreshold external field @xmath51 and thermal fluctuations.,title="fig:",width=72 ] ( from left to right , @xmath70 , @xmath71 and @xmath72 in dimensionless units ) of the domain wall moving in a disordered nanostrip ( with the dots representing the pinning centers ) under the influence of a subthreshold external field @xmath51 and thermal fluctuations.,title="fig:",width=72 ] ( from left to right , @xmath70 , @xmath71 and @xmath72 in dimensionless units ) of the domain wall moving in a disordered nanostrip ( with the dots representing the pinning centers ) under the influence of a subthreshold external field @xmath51 and thermal fluctuations.,title="fig:",width=72 ] the dynamics of the domain wall in this regime displays the typical characteristics of creep motion : the domain wall moves in discrete jumps , separated by periods of pinning , see fig .
[ fig : position ] for examples of the domain wall position as a function of time . due to the thermally activated nature of the motion ,
the dynamics is inherently stochastic , and the relative fluctuations of the dynamics become increasingly pronounced as the driving force is deacreased .
the _ average _ creep velocity is found to exhibit non - linear dependence on both external field and current . in general , for creep motion of elastic manifolds subject to an applied force @xmath73 and temperature @xmath74 , one expects the creep velocity to obey @xmath75 where @xmath76 is the creep exponent , characterizing the divergence of the height of the energy barriers as the driving force tends towards zero .
our results presented in fig .
[ fig : velocity ] indicate that for both external field and current ( i.e. @xmath77 and @xmath78 ) , the data is consistent with the above form with @xmath79 .
a more detailed study of these issues will be published elsewhere .
in this paper we have presented a line - based model of a transverse domain wall in a disordered magnetic nanostrip , and studied its dynamics within the thermally activated subthreshold creep regime .
the creep motion of the domain wall interacting with point - like pinning centers randomly positioned within the bulk of the strip is found to exhibit typical features of creep of elastic manifolds , i.e. stochastic velocity fluctuations and a non - linear relation between the average creep velocity and the external drive .
the finding that the creep exponents for the field and current driven cases are similar indicates a universality of the two forms of driving in the present case .
this is in contrast with experiments on a different system with a different wall configuration where different creep exponents were found in the two cases @xcite .
thus , a detailed experimental study of the creep dynamics of transverse domain walls in nanostrips would be interesting , to see if the simplified model presented here captures the statistical properties of the true dynamics . notice also that the non - linear behaviour of the domain wall velocity versus the applied external drive is out of scope of any point particle based models , where one would expect @xmath80 .
stochastic velocity fluctuations of domain wall motion in nanostrips have been observed also in experiments @xcite .
this might suggest that these experiments are typically probing the creep regime , with the velocity distribution arising from the random sequence of pinning and depinning events governed by thermal fluctuations . from the point of view of practical spintronics - based applications ,
this presents a fundamental problem : how to displace a domain wall within a nanostrip in a controllable fashion ? from the point of view of low power consumptions of the devices , it might be desireable to use low driving forces , but we have seen that within this regime the relative fluctuations in the domain wall dynamics are large . using stronger external drives leads to different problems , as the wall magnetization starts to precess around the long axis of strip , something that in the presence of disorder could also happen in a stochastic manner . as the translational and phase degrees of freedom of the wall
are coupled , this would presumably affect also the translational motion of the domain wall .
ll wishes to thank academy of finland for financial support . 1 g. s. d. beach , m. tsoi , and j. l. erskine , j. magn .
magn . mater . *
320 * 1272 ( 2008 ) .
s. s. p. parkin , m. hayashi , and l. thomas , science * 320 * , 190 ( 2008 ) .
g. s. .d .
beach _ et al .
_ , nature mater . * 4 * , 741 ( 2005 ) ; r. cowburn and d. petit , nature mat . * 4 * , 721 ( 2005 ) .
d. a. allwood _ et al .
_ , science * 309 * , 1688 ( 2005 ) .
e. martinez _ et al .
b * 75 * , 174409 ( 2007 ) .
y. nakatani , a. thiaville , and j. miltat , nature mat .
* 2 * 521 ( 2003 ) .
m. t. bryan , d. atkinson and r. p. cowburn , j. phys : conf .
* 17 * , 40 ( 2005 ) .
e. martinez _ et al .
98 * , 267202 ( 2007 ) .
r. d. mcmichael _ et al .
_ , ieee trans . magn . * 33 * 4167 ( 1997 )
. y. nakatani , a. thiaville , j. magn .
* 290 - 291 * , 750 ( 2005 ) . m. laufenberg _ et al . _ ,
. lett . * 88 * , 052507 ( 2006 ) .
s. zhang and z. li , phys .
93 * , 127204 ( 2004 ) .
m. j. donahue and d. g. porter , oommf user s guide , version 1.0 , interagency report nistir 6376 , national institute of standards and technology , gaithersburg , md ( sept 1999 ) .
m. yamanouchi _ et al .
_ , science * 317 * , 1726 ( 2007 ) . g. meier _ et al .
* 98 * , 187202 ( 2007 ) .
p. mhrke _ et al .
_ , j. phys .
d * 41 * , 164009 ( 2008 ) .
a. vanhaverbeke , a. bischof , and r. allenspach , phys .
* 101 * , 107202 ( 2008 ) .
|
we present a line - based model of transverse domain walls in thin magnetic strips , to study the effect of bulk disorder on the domain wall dynamics within the thermally activated creep regime .
the creep velocity is found to exhibit a non - linear dependence on both applied magnetic fields and electric currents , characterized by similar creep exponents for both forms of the external drive .
we discuss briefly the significance of the inherently stochastic thermally activated domain wall motion from the point of view of spintronics applications , where it generally is essential to be able to control the domain wall displacement in a deterministic manner .
|
the field of extrasolar planet research is growing rapidly with more than 1000 planets found ( as of november , 2013 ) .
many of these objects are large and/or close to their host star .
these large extrasolar planets have come to be known as extra - solar giant planets ( egps ) and those in close proximity to their host stars are often referred to as `` hot jupiters '' .
many of these planets are observed to transit their host stars as viewed from earth . combined with radial velocity data ,
this allows calculation of the planetary mass , radius and surface gravity . and
because they are likely synchronously rotating , planetary rotation rate can also be inferred . because these close - in planets are likely tidally locked they have permanent day and night sides .
such asymmetric heating drives strong winds in the planetary atmosphere which have been modeled by several groups @xcite .
while these models vary in their approach and complexity , they have several common features , such as transonic wind speeds and eastward equatorial jets .
the eastward equatorial jet implies that the hottest point would be advected eastward of the substellar point , as was first predicted by @xcite , and which was subsequently observed on hd 189733b by @xcite .
although discrepancies exist , it appears that atmospheric circulation models with varying forms of radiation transport are doing well at describing the large scale dynamics of these planets atmospheres . however , there are still several unexplained observations .
first , it was observed early on that many egp s , particularly those that are the most strongly irradiated , have radii much larger than that expected from standard evolutionary models @xcite . the most likely solution to this observation
is that there is an additional heat source in the planetary interior that slows gravitational contraction causing the radius to appear larger than expected from standard evolutionary models .
many heating mechanisms have been proposed such as tidal heating @xcite , downward transport of mechanical energy by atmospheric circulation via waves or advection @xcite , atmospheric turbulence @xcite and ohmic dissipation @xcite .
second , observations of infrared and visible light curves have been able to constrain how efficiently the planetary atmosphere transfers heat from the irradiated dayside to the cooler nightside by measuring the phase variation over the planetary orbital period .
the picture emerging is that these planets likely have a variety of atmospheric circulation patterns , with varying recirculation efficiency @xcite .
observations of hd 209458b and 51 peg @xcite , hd189733b @xcite and hd 149026b @xcite indicate that these planets have efficient recirculation from their day to night side and hence , show small ( @xmath1k ) temperature variations .
however , observations of ups and b @xcite , hd 179949 @xcite , hat - p-7b @xcite , wasp-12b @xcite and wasp-18b @xcite indicate weak recirculation and hence , large day - night temperature variations . it has been argued that a strong magnetic field could prevent efficient circulation from the day to night side by opposing such flow .
the effects of the magnetic field are therefore , speculated to be ( at least ) twofold .
first a planetary magnetic field could slow day - night winds causing reduced circulation efficiencies and hence , larger day - night temperature variations @xcite .
second , ohmic dissipation associated with the magnetic field could lead to the heating required to explain the inflated radii of many hot jupiters @xcite .
these pioneering studies demonstrate the possible influence of magnetic effects on the atmospheric winds and quantify the possible behavior of the ohmic dissipation .
however , these early studies did not solve the full mhd equations .
rather , they adopt a kinematic approach where the lorentz force is estimated from the winds using a prescribed dipole magnetic field and assuming the induced fields generated by electrical currents can be neglected . this approach is a natural first step , but is only valid at magnetic reynolds numbers @xmath2 ( @xmath3 , where @xmath4 is a typical speed and @xmath5 is a typical length scale and @xmath6 is the magnetic diffusivity )
. however , the actual @xmath7 in these planetary atmospheres often exceed unity , particularly for the hotter planets .
this indicates the need for a fully self - consistent mhd treatment .
an exception to this prescription is the work by @xcite , which solves the mhd equations , but in the boussinesq approximation , so they are effectively two - dimensional and they omit ohmic heating . here ,
we present fully 3d mhd simulations including the effects of compressibility and ohmic heating .
we use a 3d , mhd model in the anelastic approximation .
the model is based on the glatzmaier dynamo code @xcite , but has some differences in the actual equations solved , the discretization and the implementation .
the model decomposes the magnetic and mass flux into a poloidal - toroidal decomposition and it is discretized using spherical harmonics in the horizontal and a fourth - order finite difference scheme in the vertical .
a description of the axisymmetric version of this code can be found in @xcite .
the 3d version has been benchmarked with several other dynamo codes @xcite .
the model solves the following equations : @xmath8\\ \frac{\partial \bvec}{\partial t } & = & \curl ( \vvec \times \bvec ) - \curl ( \eta \curl \bvec ) \\ { { \frac{\partial{t}}{\partial{t}}}}+(\bf{v}\cdot\bf{\nabla}){t } & = & \nonumber -v_{r}\left[{{\frac{\partial{\overline{t}}}{\partial{r}}}}-(\gamma-1)\overline{t}h_{\rho}\right]+ \\\nonumber & & \hspace{-25 mm } \mbox{}+(\gamma-1)th_{\rho}v_{r } + \gamma\overline{\kappa}\left[\nabla^{2}t+(h_{\rho}+h_{\kappa}){{\frac{\partial{t}}{\partial{r}}}}\right]+ \\ & & \mbox{}+ \frac{t_{eq}-t}{\tau_{rad}}+\frac{\eta}{\mu_{o}\rho c_{p}}|\nabla \times { \bf b}|^{2}\end{aligned}\ ] ] equation ( 1 ) represents the continuity equation in the anelastic approximation @xcite .
this approximation allows some level of compressibility by allowing variation of the reference state density , @xmath9 , which varies in this model by five orders of magnitude .
equation ( 2 ) represents the conservation of magnetic flux .
equation ( 3 ) represents conservation of momentum including coriolis and lorentz forces , equation ( 4 ) represents the magnetic induction equation and equation ( 5 ) represents the energy equation .
all variables take their usual meaning and details can be found in @xcite , with the following few exceptions .
the magnetic diffusivity , @xmath6 , is a function of temperature , as in @xcite , but , in this first approach , we take it to be only a function of the reference state temperature ( and therefore , only a function of radius ) .
strong stellar irradiation is treated using a newtonian radiative scheme as in @xcite .
this prescription is indicated by the fourth term on the right hand side ( rhs ) of equation ( 5 ) .
the last term of equation ( 5 ) represents ohmic heating .
we note that these equations are non - hydrostatic and the diffusion coefficients in equation ( 3 ) and ( 5 ) are explicit with no hyper - diffusion employed .
these values are set as low as possible while maintaining numerical stability ] .
we resolve the radial extent from @xmath10bar .
the resolution of the models presented is 64 x 128 x 100 ( @xmath11 ( latitude ) x @xmath12 ( longitude ) x @xmath13 , or t42 with 100 radial levels ) and the code is parallelized using message passing interface ( mpi ) .
we initially run purely hydrodynamic models for @xmath14 rotation periods ( @xmath15 ) , then add a magnetic field .
we ran two models varying the reference state ( and equilibrium ) temperature : 1 ) the model for hd209458b from @xcite and 2 ) the @xcite model with 300k added at every radial level .
these represent our `` cool '' and `` hot '' models , respectively ( in our hot model temperatures vary from 2100k at depth to 1300k aloft ) and we adopt a planetary radius , rotation rate and gravity of @xmath16 , 3 days and @xmath17 .
we also present two different magnetic field strengths : a ) a model with a 5 g dipolar field at the base of the atmosphere and 3.5 g at the top of the atmosphere initially and b ) a higher magnetic field strength of 15 g at the bottom and 10 g at the top initially .
these values correspond to the radial field strength at the pole or the latitudinal field at the equator , and the radial structure is consistent with a dipolar field which falls off as @xmath18 .
the magnetic field strength at the base of the atmosphere is held fixed and matched to a potential field at the top of the domain .
the magnetised models are then run for @xmath19 and are in a steady state .
we note at the outset , that the @xcite model is too cool to show any appreciable magnetic effects and so the results displayed in section 3 are those for the hotter model .
figure 1 shows the temperature structure along with wind patterns within our purely hydrodynamic simulation .
we see that these models recover some of the typical features of radiatively forced atmospheric models : azimuthal flow marked by a prograde equatorial jet at most longitudes ( and on zonal average ) , predominantly retrograde flow at higher latitudes with meridional flow towards the poles near the dayside and returning toward the equator at the nightside .
the chevron shaped patterns are typical of the kelvin / rossby wave forced flows described in @xcite . however , our flows have lower speeds in general , peaking around a couple km / s and we note that at depth , our pattern of temperature fluctuations deviates from previous models .
this is likely due to our treatment of thermal forcing and diffusion .
we are currently looking into this issue , but note that it does not affect our results ( the velocity patterns are similar to previous models and these temperature fluctuations are only a few k ) .
previous work has indicated that a planetary scale magnetic field of predominantly dipolar structure could slow winds and diminish day - night circulation @xcite . because previous works did not solve mhd equations this physical concept was introduced in the form of a drag term on the zonal flow equation which was proportional to the zonal flow speed divided by a magnetic timescale that represents the time required for the field to halt the zonal flow .
an order of magnitude estimate of this timescale is : @xmath20 the lorentz force is then calculated as @xmath21 and the ohmic dissipation rate is @xmath22 , where @xmath23 is the zonal flow speed , @xmath24 is the latitudinal current , b is the imposed dipolar field strength , @xmath25 is the latitude , and c is the speed of light . in figure 2
we show the azimuthal component of the lorentz force calculated in our simulation ( a , c ) , last term on rhs of equation 3 ) and the drag term using the prescription described in equation ( 6 ) with a constant field strength ( as done by previous authors b , d ) at 0.5bar and 100bar .
there are several important differences to note .
first , our mhd simulation predicts a lorentz force whose peak value is an order of magnitude smaller the peak value given by the prescription encompassed in equation ( 6 ) and second , the actual lorentz force is more localized spatially than the prescription .
the differences arise from spatial variation of the background dipole field ( not included in previous kinematic approaches ) as well as from field evolution - the toroidal field reaches @xmath2660 g at depth ( consistent with @xmath27 ) and the induced poloidal field reaches values three times larger than the imposed poloidal field , in contradiction of the simplifying assumptions made in previous models , see figure 3 .
overall , our 5 g model wind speeds are decelerated 10 - 40% compared to their non - magnetic counterparts ( depending on latitude and radius ) , our 15 g field shows similar slowing , except near the surface at high latitudes ( figure 2e shows the average of the absolute value of the zonal flow as a function of pressure for three different latitude ranges ) .
contrary to previous results @xcite , we find ohmic heating to be orders of magnitude too small to explain the inflated radius of hd209458b .
figure 4 shows the ohmic dissipation rate calculated from the mhd model at 0.5 bar and 100 bar ( a , c ) and using the prescription outlined in equation ( 6 ) , with velocities from these simulations and a constant 5 g field ( b , d ) .
similar to the lorentz force , the prescription overestimates the maximum value of the ohmic dissipation at a give height by more than an order of magnitude .
we also see that ohmic dissipation calculated self - consistently in these simulations is more localized than the prescription given by equation ( 6 )
. therefore , on integration over a given radius the ohmic dissipation will be substantially smaller than that given by the prescription .
this is true regardless of the windspeeds .
the integrated ohmic dissipation rate within a pressure scale height as a function of pressure for out hot model with two different magnetic field strengths is shown in ( e ) . using a stellar insolation of 3.3@xmath28w and the nominal required heating @xcite of @xmath26 0.1% deep within the interior ( @xmath26200 bar )
, one would require @xmath29w . in figure
4e we see that the peak ohmic dissipation occurs high in the atmosphere where it can not impact radius inflation .
deeper in the atmosphere the measured ohmic dissipation is two orders of magnitude too small to explain inflation .
@xcite suggest a lower value of @xmath30w at 90bar could explain the radius of hd209458b , our models still show dissipation rates an order of magnitude too small .
the above results are consistent with the results of @xcite who also found ohmic dissipation was 1 - 2 orders of magnitude too small to explain the inflated radius of hd209458b . their model , which includes the back reaction of the field on the flow in determining ohmic dissipation rates , is closer to these mhd simulations than those of @xcite or @xcite .
we have presented mhd simulations of the atmosphere of the hot jupiter hd209458b .
we have found that magnetic fields slow winds by 10 - 40% and contribute to ohmic heating of @xmath0w at depth .
therefore , we find that ohmic dissipation , at least in the atmosphere modeled here , is unable to explain the inflated radius of this planet .
we also do not find any evidence that pole - ward meridional flows are arrested due to field geometry , as suggested in @xcite .
this is because our hot model is not hot enough to be in a magnetically dominated regime as assumed by @xcite .
previous prescriptions treating magnetism in hot jupiter atmospheres use a parameterization for the electric current based on ohm s law .
however , that current must also satisfy ampere s law and faraday s law requires the time rate of change of the magnetic field be determined by the curl of the electric field .
that is , the solution to the full mhd equations is required to obtain a physically self - consistent representation of the appropriate fields and forces ; a parameterization of ohm s law alone is not sufficient .
stated simply , a kinematic approach to mhd in these hot atmospheres does not adequately represent magnetic effects . in the original work by @xcite
it was argued that the magnetic reynolds number was generally less than one , hence validating their kinematic approach .
however , in hotter atmospheres ( precisely those being investigated for ohmic inflation ) , the magnetic reynolds number can be substantially larger than one .
in the hot model presented here ( which is admittedly rather cool ) , the magnetic reynolds number reaches values @xmath31 , for hotter atmospheres of interest for ohmic heating , the reynolds number could easily exceed @xmath32 .
there are several shortcomings of the models presented which we address here .
first , our reference state model is not realistic . the cool model of @xcite ,
is known to be too cool and it shows virtually no magnetic effects .
our adhoc `` hot '' model is also not physical .
however , such variations would not bring our models into better alignment with previous estimates .
furthermore , our radiative transfer is crude , however , for these first mhd simulations , this heating is adequate and more realistic radiative transfer is not likely to alter our main results .
another shortcoming of these models is that they are anelastic , so they do not accurately represent supersonic flows .
the comparisons shown in figures 2 and 4 , use the flow speeds in these mhd models , and we show that it is the details of the magnetic field that are inadequate , this would be true regardless of the flow speed .
however , larger flow speeds at depth could increase the value of the ohmic dissipation there , and therefore deep jets may be required to inflate planets . in order to close the @xmath262 order of magnitude gap in ohmic dissipation one would need mhd wind speeds an order of magnitude larger at depths greater than 10bar . while such wind speeds may be possible in purely hydrodynamic models , it is not clear whether they are likely in the presence of magnetic fields .
additionally , our artificial boundary conditions could affect the profile of the radial current and therefore , the heating rate at depth .
finally , probably the most problematic of the approximations we have made is that the magnetic diffusivity is only a function of the reference state temperature , instead of being a function of all space .
this is important because the stellar insolation changes the surface temperature substantially leading to dayside - nightside temperature variations which would cause large horizontal variations in the magnetic diffusivity .
this is likely to change the surface wind structure .
orders of magnitude variation in magnetic reynolds number could lead to magnetic instabilities which could lead to turbulence and mixing , which may affect horizontal temperature variations . at the moment
, it is unclear how these effects would alter wind structure .
we are currently implementing a spatially dependent magnetic diffusivity into our mhd model and results will be forthcoming . despite its effect on wind structure at the surface
, we do not expect a horizontally varying magnetic diffusivity to significantly alter our results on ohmic dissipation .
the stellar insolation does not effect the deeper layers much and therefore , the magnetic diffusivity will vary little from their reference state values . since it is those deeper layers that require heating in order to arrest contraction and lead to inflated radii , and those radii
are affected the least by a horizontally dependent magnetic diffusivity we expect our result for ohmic dissipation to be robust ( although see the caveats above with regard to wind speeds at depth )
. hotter internal temperatures may increase ohmic heating but those models will have higher @xmath7 and require substantially more computational resources .
such models are currently being investigated .
we are grateful to g. glatzmaier , d. lin , a.cumming and k. batygin for helpful discussion and suggestions on improving the manuscript .
support for this research was provided by a nasa grant nng06gd44 g .
t. rogers is supported by an nsf atm faculty position in solar physics under award number 0457631 .
showman is supported by nasa origins grant nnx12a179 g .
|
we present the first three - dimensional ( 3d ) magnetohydrodynamic ( mhd ) simulations of the atmosphere of hd209458b which self - consistently include reduction of winds due to the lorentz force and ohmic heating .
we find overall wind structures similar to that seen in previous models of hot jupiter atmospheres , with strong equatorial jets and meridional flows poleward near the dayside and equatorward near the night .
inclusion of magnetic fields slows those winds and leads to ohmic dissipation .
we find wind slowing ranging from 10 - 40% for reasonable field strengths .
we find ohmic dissipation rates @xmath0w at 100 bar , orders of magnitude too small to explain the inflated radius of this planet .
faster wind speeds , not achievable in these anelastic calculations , may be able to increase this value somewhat , but likely wo nt be able to close the gap necessary to explain the inflated radius .
we demonstrate that the discrepancy between the simulations presented here and previous models is due to inadequate treatment of magnetic field geometry and evolution
. induced poloidal fields become much larger than those imposed , highlighting the need for a self - consistent mhd treatment of these hot atmospheres .
|
the nature of halo gas ( a.k.a .
circumgalactic medium ; cgm hereafter ) , on galactocentric distances of @xmath1 kpc , is a problem of significant ongoing interest in galaxy formation .
halo gas links galaxies from the intergalactic medium and is the conduit for exchange of matter energy density , momentum , angular momentum and metals between star formation and active galactic nucleus induced outflow and gravitational inflow of gas .
thus , understanding halo gas is imperative before a satisfactory theory of galaxy formation and evolution may be constructed .
there has been recent rapid advances on the observational front to address this important issue , made possible primarily by hst observations ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
in @xcite we investigate the overall gas composition with respect to the density , temperature and metallicity and find that , on average , for @xmath2 ( red , blue ) galaxies cold ( @xmath3k ) gas is the primary component in the inner regions , with its mass comprising 50% of all gas within @xmath4 kpc . at @xmath5 kpc for ( red , blue ) galaxies ,
the hot ( @xmath6k ) component becomes the majority component .
the warm ( @xmath7k ) component is , on average , a perpetual minority in both red and blue galaxies , with its contribution peaking at @xmath8 at @xmath9 kpc in blue galaxies and never exceeding 5% in red galaxies .
these findings are in agreement with recent observations in many aspects , in particular with respect to the amount of warm gas in star forming galaxies and the amount of cold gas in early type galaxies at low redshift , both of which are physically intriguing and at first instant less than intuitive . in light of a new observational development with respect to the nv to ovi absorption line ratio and in particular the apparent need of seemingly complicated , perhaps contrived , models to explain the data
, we here perform a detailed analysis of our high resolution cosmological hydrodynamic simulations to assess whether _ ab initio _ cosmological simulations are capable of accounting for this particular observation , in the larger context of the success of the model able to match the overall composition of halo gas , among others .
it is particularly relevant to note that the good agreement between our simulations and observations with respect ovi @xmath101032 , 1038 absorption lines , presented earlier in @xcite , suggests that the statistical description of the properties of the warm component in the simulations - mass , spatial distribution , density , temperature , metallicity , and their environmental dependences - has now been firmly validated and provides a critical anchor point for our model .
consequently , this additional , independent analysis with respect to nv / ovi ratio and other ratios becomes very powerful to further strengthen or falsify our model or our simulations .
our findings here are both encouraging and intriguing .
if one uses a fixed , solar n / o ratio regardless of the o / h ratio , our model is acceptable , with all 4 ks ( kolmogorov - smirnov ) test p - values greater than 0.28 for either haardt - madau ( * ? ? ?
* hm hereafter ) or hm+local radiation field , where the local radiation field is due to hot gas in the host galaxy .
if one allows for a dependence of the n / o ratio on the o / h ratio , both measured by independent observations and motivated by theoretical considerations of two different sources of n , then our model is able to account for the observations highly successfully , with all ks test p - values exceeding 0.9 .
we additionally examine the following absorption line column density ratios where comparisons to observations may be made in a reasonable statistical fashion , siiv / ovi , nii / ovi and niii / ovi , and find that the ratios from our simulations are fully consistent with observations .
we also investigate the model where uv radiation from local shock heated gas in concerned galaxies are added to the hm background , which is found to also agree with observations with comparable p - values for all line ratios examined .
however , these good agreements come about because observational data points are dominated by upper and lower limits instead of actual detections .
we discuss how some moderate improvment in observational sensitivity may provide much stronger tests of models .
we perform cosmological simulations with the amr eulerian hydro code , enzo @xcite .
we use the following cosmological parameters that are consistent with the wmap7-normalized @xcite lcdm model : @xmath11 , @xmath12 , @xmath13 , @xmath14 , @xmath15 and @xmath16 .
these parameters are also consistent with the latest planck results @xcite , if one adopts the hubble constant that is the average between planck value and those derived based on sne ia and hst key program @xcite .
we use the power spectrum transfer functions for cold dark matter particles and baryons using fitting formulae from @xcite .
we use the enzo inits program to generate initial conditions .
first we ran a low resolution simulation with a periodic box of @xmath17mpc on a side .
we identified two regions separately , one centered on a cluster of mass of @xmath18 and the other centered on a void region at @xmath19 .
we then re - simulate each of the two regions separately with high resolution , but embedded in the outer @xmath17mpc box to properly take into account large - scale tidal field and appropriate boundary conditions at the surface of the refined region .
we name the simulation centered on the cluster
c " run and the one centered on the void v " run .
the refined region for c " run has a size of @xmath20mpc@xmath21 and that for v " run is @xmath22mpc@xmath21 . at their respective volumes , they represent @xmath23 and @xmath24 fluctuations .
the root grid has a size of @xmath25 with @xmath25 dark matter particles .
the initial static grids in the two refined boxes correspond to a @xmath26 grid on the outer box . the initial number of dark matter particles in the two refined boxes correspond to @xmath26 particles on the outer box .
this translates to initial condition in the refined region having a mean interparticle - separation of @xmath27 kpc comoving and dark matter particle mass of @xmath28 .
the refined region is surrounded by two layers ( each of @xmath29mpc ) of buffer zones with particle masses successively larger by a factor of @xmath30 for each layer , which then connects with the outer root grid that has a dark matter particle mass @xmath31 times that in the refined region .
the initial density fluctuations are included up to the nyquist frequency in the refined region .
the surrounding volume outside the refined region is also followed hydrodynamically , which is important in order to properly capture matter and energy exchanges at the boundaries of the refined region . because we still can not run a very large volume simulation with adequate resolution and physics , we choose these two runs of moderate volumes to represent two opposite environments that possibly bracket the universal average .
we choose a varying mesh refinement criterion scheme such that the resolution is always better than @xmath32/h proper parsecs within the refined region , corresponding to a maximum mesh refinement level of @xmath33 above @xmath34 , of @xmath35 at @xmath36 and @xmath37 at @xmath38 .
the simulations include a metagalactic uv background @xcite , and a model for shielding of uv radiation @xcite .
the simulations also include metallicity - dependent radiative cooling and heating @xcite .
the enzo version used includes metallicity - dependent radiative cooling extended down to @xmath35k , molecular formation on dust grains , photoelectric heating and other features that are different from or not in the public version of enzo code .
we clarify that our group has included metal cooling and metal heating ( due to photoionization of metals ) in all our studies since @xcite for the avoidance of doubt ( e.g. , * ? ? ?
* ; * ? ? ?
star particles are created in cells that satisfy a set of criteria for star formation proposed by @xcite .
each star particle is tagged with its initial mass , creation time , and metallicity ; star particles typically have masses of @xmath39@xmath40 .
supernova feedback from star formation is modeled following @xcite .
feedback energy and ejected metal - enriched mass are distributed into 27 local gas cells centered at the star particle in question , weighted by the specific volume of each cell ( i.e. , weighting is equal to the inverse of density ) , which is to mimic the physical process of supernova blastwave propagation that tends to channel energy , momentum and mass into the least dense regions ( with the least resistance and cooling ) .
we allow the whole feedback processes to be hydrodynamically coupled to surroundings and subject to relevant physical processes , such as cooling and heating , as in nature .
the extremely inhomogeneous metal enrichment process demands that both metals and energy ( and momentum ) are correctly modeled so that they are transported into right directions in a physically sound ( albeit still approximate at the current resolution ) way , at least in a statistical sense . in our simulations metals
are followed hydrodynamically by solving the metal density continuity equation with sources ( from star formation feedback ) and sinks ( due to subsequent star formation ) .
thus , metal mixing and diffusion through advection , turbulence and other hydrodynamic processes are properly treated in our simulations .
the primary advantages of this supernova energy based feedback mechanism are three - fold .
first , nature does drive winds in this way and energy input is realistic .
second , it has only one free parameter @xmath41 , namely , the fraction of the rest mass energy of stars formed that is deposited as thermal energy on the cell scale at the location of supernovae .
third , the processes are treated physically , obeying their respective conservation laws ( where they apply ) , allowing transport of metals , mass , energy and momentum to be treated self - consistently and taking into account relevant heating / cooling processes at all times .
we use @xmath42 in these simulations .
the total amount of explosion kinetic energy from type ii supernovae with a chabrier imf translates to @xmath43 .
observations of local starburst galaxies indicate that nearly all of the star formation produced kinetic energy ( due to type ii supernovae ) is used to power galactic superwinds ( e.g. , * ? ? ?
given the uncertainties on the evolution of imf with redshift ( i.e. , possibly more top heavy at higher redshift ) and the fact that newly discovered prompt type i supernovae contribute a comparable amount of energy compared to type ii supernovae , it seems that our adopted value for @xmath41 is consistent with observations and physically realistic .
the validity of this thermal energy - based feedback approach comes empirically . in @xcite the metal distribution in and around galaxies over a wide range of redshift ( @xmath44 )
is shown to be in excellent agreement with respect to the properties of observed damped @xmath45 systems @xcite , whereas in @xcite we further show that the properties of ovi absorption lines at low redshift , including their abundance , doppler - column density distribution , temperature range , metallicity and coincidence between ovii and ovi lines , are all in good agreement with observations @xcite .
this is non - trivial by any means , because they require that the transport of metals and energy from galaxies to star formation sites to megaparsec scale be correctly modeled as a function of distance over the entire cosmic timeline , at least in a statistical sense .
we identify galaxies at each redshift in the simulations using the hop algorithm @xcite operating on the stellar particles , which is tested to be robust and insensitive to specific choices of concerned parameters within reasonable ranges .
satellites within a galaxy down to mass of @xmath46 are clearly identified separately in most cases .
the luminosity of each stellar particle in each of the sloan digital sky survey ( sdss ) five bands is computed using the gissel stellar synthesis code @xcite , by supplying the formation time , metallicity and stellar mass .
collecting luminosity and other quantities of member stellar particles , gas cells and dark matter particles yields the following physical parameters for each galaxy : position , velocity , total mass , stellar mass , gas mass , mean formation time , mean stellar metallicity , mean gas metallicity , sfr , luminosities in five sdss bands ( and various colors ) and others .
we show , among others , that the simulated luminosity functions of galaxies at @xmath19 are reasonably matched to observations @xcite .
in the analysis presented here we choose randomly ten galaxies from our simulation that have properties that are similar to observed galaxies in the cos - halo program @xcite with respect to the star formation rate ( sfr ) and stellar mass .
some relevant properties of the ten simulated galaxies are tabulated in table 1 .
a central galaxy is defined to be one that is not within the virial radius of a larger ( halo - mass - wise ) galaxy .
.properties of 10 simulated galaxies used in this study . [ cols="^,^,^,^,^,^ " , ] [ table : ks ] to gain a more quantitative statistical test between simulations and observations , we perform two - sample ks tests between simulated and observed column density ratios for four pairs of lines , nv / ovi , nii / nv , niii / ovi and siiv / ovi .
since most of the observational data are upper and lower limits , instead of actual detections , our analysis is performed as follows . for the case where the numerator line is an upper limit and
the denominator line is a detection , we allow the ratio to be drawn from the simulation distribution with value upper - bounded by the upper limit . the same is done for the case where the numerator line is an upper limit and the denominator line is a lower limit .
then , in conjunction with detections , where both lines are detected , we perform a two - sample ks test for each of the four line pairs , nv / ovi , nii / nv , niii / ovi and siiv / ovi , between simulations and observations . needless to say , our presently adopted procedure to treat the upper and lower limits cases favors agreement with observations and simulations . nevertheless , the procedure is consistent with the current data .
the results are tabulated in table [ table : ks ] . clearly , no major disagreements can be claimed as to reject the simulation results in all four cases , ( constant n / o , varying
n / o ) times ( hm , hm+local )
. however , there are hints , at face value , that the constant n / o cases are less preferred than the varying n / o cases .
nonetheless , it is premature to make any firm statistical conclusion on that at this juncture .
thus , the only robust conclusion we can reach at this time is that our simulation predictions are fully consistent with extant observational data with respect to the four line ratios , nv / ovi , nii / nv , niii / ovi and siiv / ovi .
what would exponentially increase the statistical power of testing the models is to turn these current upper limits into real detections .
we have performed the following exercise to demonstrate this point .
let us assume that all current upper limits of the column density ratios become detections and the detection values are lower than current upper limits uniformly by a factor of @xmath47 dex .
we find that , if @xmath48 , the ks p - values for the nv / ovi line ratio become @xmath49 for the ( hm , hm+local ) cases with constant n / o ; with @xmath50 , the ks p - values for the nv / ovi line ratio become @xmath51 for the ( hm , hm+local ) cases with constant n / o . for the varying n / o cases , the situation is non - monotonic in the following sense : the ks p - values for the nv / ovi line ratio are @xmath52 for the ( hm , hm+local ) with @xmath53 , increasing to a maximum of @xmath54 with @xmath55 , then downturning to @xmath51 with @xmath56 . obviously , a uniform shift is an oversimplification .
nevertheless , this shows clearly an urgent need to increase observational sensitivity in order to place significantly stronger constraints on models than currently possible .
when all line pairs are deployed , the statistical power will be still , likely much , greater .
-0.0 cm -1.0 cm -0.0 cm -0.0 cm -1.0 cm -0.0 cm -0.0
cm -0.0 cm -1.0 cm -0.0 cm -0.0 cm -1.0 cm -0.0 cm -0.0 cm finally , we turn to a closer examination of the physical conditions that give rise to the various absorption lines in our simulations shown above . in figure
[ fig : contour_zn ] we show the number density of lines in the column density - metallicity plane for ovi ( top - left panel ) , nv ( top - right panel ) , niii ( bottom - right panel ) and siiv ( bottom - left panel ) .
overall , while there is a significant span in metallicity , with as low as @xmath57=-3.5 $ ] at the low column density end for ovi , the vast majority of absorbers have metallicities falling into the range @xmath58=-2 $ ] to @xmath59 for ovi , @xmath58=-2 $ ] to @xmath59 for nv , @xmath58=-2 $ ] to @xmath59 for nv and @xmath58=-1 $ ] to @xmath59 for siiv . at the high column density end , we see @xmath58\sim -0.5 $ ] to @xmath60 for ovi , @xmath58\sim -0.5 $ ] to @xmath61 for nv , @xmath58\sim 0 $ ] to @xmath61 for niii and @xmath58\sim 0 $ ] to @xmath61 for siiv .
these trends and significant disparities between different lines are a results of complex multi - phase cgm with a very inhomogeneous metallicity distribution .
simplistic collisional excitation / ionization models are unlikely to be able to capture all of the key elements of the physical processes involved and may lead to conclusions that are not necessarily conformal to direct analyses of the simulations .
figure [ fig : contour_tn ] shows the number density of lines in the column density - temperature plane for ovi ( top - left panel ) , nv ( top - right panel ) , niii ( bottom - right panel ) and siiv ( bottom - left panel ) . to set the context of collisionally dominated ionization processes , we note that , under the assumption of collisional ionization equilibrium , as in cloudy , the peak temperature for the element in question with half - width - half - maximum is approximately @xmath62k for ovi , @xmath63k for nv , @xmath64k for niii and @xmath65k for siiv ( e.g. , * ? ? ?
a one - to - one comparison between each of these four peak temperature ( and its width ) and the contour levels indicates that for ovi and nv lines the collisional ionization dominates the process for creating ovi at @xmath66@xmath67 and nvi at @xmath68@xmath67 , respectively , manifested in the horizontal extension of the contours pointing to the right at the temperature ( with an appropriate width ) in question .
the same can be said about the siiv line at the high column end @xmath69@xmath67 ; however , at lower @xmath70 values ( @xmath71@xmath67 ) , the contours are no longer aligned horizontally , indicative of enhanced contribution of photoionization due to lower ionization potential of siiii ( @xmath72ev ) versus say ov ( @xmath73ev ) .
similar statements about niii lines to those for siiv can be made due to similar reasons .
overall , the similarity between ovi and nv lines suggests that collisional ionization processes are dominant and results with respect to these two lines are relatively immune to uncertainties in the radiation field used .
however , the apparent insensitivity of results on the radiation field with detailed calculation we have performed , in the way of comparing hm and hm+local results , indicates that the net effect due to an increase of radiation field is relatively small due to the large ranges of density and metallicity of gas involved , although the actual situation appears to be more intertwined because of nonlinear relationships between density , metallicity , ionization parameter and column density . as an example ,
as shown earlier in figure [ fig : cover ] , a stronger radiation field tend to increase the abundance of niii , although the difference between hm and hm+local cases for is apparently minor , seeming to suggest conflictingly that collisional processes are the primary powering source for producing niii
. a more thorough theoretical study and a more detailed comparison to observations will be desirable , when a larger observational sample with more sensitive column density detection limits becomes available .
as we have demonstrated , a fraction of a dex increase in sensitivity may warrant a revisit to a detailed comparison .
in light of a recent conclusion that the observed line ratios of uv absorbers in the cgm may pose a significant challenge for theoretical models @xcite , we study five uv absorption lines , ovi @xmath101032 , 1038 , siiv @xmath101394 , 1402 , nv @xmath101239 , 1243 , niii @xmath74990 , nii @xmath741084 , in the cgm of simulated galaxies , utilizing _ ab initio _ ultra - high resolution ( * laozi * ) hydrodynamical simulations .
our simulated galaxies are chosen to have stellar masses and star formation rates similar to their observed counterpart .
we examine uncertainties related to the radiation background by computing separately for two cases of radiation field , one with the hm radiation background and the other with both hm and local radiation due to hot gas in the host galaxy . separately and orthogonally ,
we examine two separate cases of nitrogen to oxygen ratio , in one case with constant n / o and in the other with varying n / o that is theoretically consistent with two different synthetic sources of nitrogen and observationally confirmed by independent observations .
@xcite find constant density photoionization models to be excluded by the data .
they find collisional ( both in and out of equilibrium ) ionization to be only broadly consistent with the data .
they suggest either collisionally - ionized gas cooling behind a fast shock or a highly structured gas photo - ionized by a local high energy source as plausible models to account of the observed ovi column density range and line ratios .
in contrast , we do not find significant difficulty in accounting for the same observational data in our cosmological simulations that capture the complex multi - phase structure of the cgm , as reflected by the acceptable ks test p - values for column density ratios of four pairs of lines nv / ovi , siiv / ovi , niii / ovi and nii / ovi examined .
inter - comparisons between results from different models employing different radiation fields in our simulations and comparisons between properties of the absorbers and expectations of collisional ionization indicate that collisional ionization play a major role in producing all the lines studied in realistic cgm produced in cosmological simulations .
photoionization process plays a significant role as well , to a varying degree , depending on the ion in question , although it seems clear that for nv and ovi lines photoionization effect is relatively minor .
the success of our largely collisional ionization model in all cases is , however , in a significant part , due to very accommodative observational line ratio data points that are dominated , in number , by upper limits rather than actual detections .
we find the apparent satisfactory agreement between simulations and extant observational data can be strongly tested and different cases ( different radiation fields and different n / o ratio assumptions ) differentiated , if most of the current upper limits in the observational data become detections . to demonstrate the power , we show , as an example , that if the upper limits of nv / ovi become detections with values that are lower by a mere @xmath75 dex than their respective current upper limits , the ks p - value for the nv / ovi line ratio becomes @xmath76 for the constant n / o case and @xmath77 for the varying n / o case .
if , on the other hand , the actual detection values turn out to be lower than current upper limits by @xmath78 dex , then the varying n / o case obtains a satisfactory p - value of @xmath54 , whereas the constant n / o case is endowed with a p - value equal to zero .
thus , it is highly desirable to increase the observational sensitivity and/or enlarge observational data sample size , in order to have a definitive test .
we are grateful to jessica werk for sharing observational data with us prior to publication and stimulating discussion .
we thank j. xavier prochaska for useful discussion .
we have used the very useful and versatile analysis software yt version 2.6 @xcite for some of our analysis .
computing resources were in part provided by the nasa high- end computing ( hec ) program through the nasa advanced supercomputing ( nas ) division at ames research center .
the research is supported in part by nasa grant nnx11ai23 g .
|
utilizing high - resolution cosmological hydrodynamic simulations we investigate various ultra - violet absorption lines in the circumgalactic medium of star forming galaxies at low redshift , in hopes of checking and alleviating the claimed observational conundrum of the ratio of nv to ovi absorbers , among others .
we find a satisfactory agreement between simulations and extant observational data with respect to the ratios of the following four line pairs examined , nv / ovi , siiv / ovi , niii / ovi and nii / ovi . for the pairs involving nitrogen lines , we examine two cases of nitrogen abundance , one with constant n / o ratio and the other with varying n / o ratio , with the latter motivated by theoretical considerations of two different synthetic sources of nitrogen that is empirically verified independently . along a separate vector , for all line pairs , we examine two cases of radiation field , one with the haardt - madau background radiation field and the other with an additional local radiation field sourced by hot gas in the host galaxy . in all cases ,
two - sample kolmogorov - smirnov tests indicate excellent agreements .
we find that the apparent agreements between simulations and observations will be strongly tested , if the bulk of current upper limits of various line ratios are turned into actual detections .
we show that an increase in observational sensitivity by @xmath0 dex will already start to significantly constrain the models .
= 1
|
given a graph @xmath0 and a natural number @xmath1 , there is a game called the _ c - coloring game _ of @xmath0 and it is known that there is a perfect deterministic strategy for this game if and only if @xmath0 has a c - coloring . thus , the _ chromatic number _ of @xmath0 , @xmath2 can be characterized as the least integer for which a perfect deterministic strategy exists for the c - coloring game .
this led researchers to consider various kinds of probabilistic strategies for games , especially cases where the probabilities arose from the random outcomes of quantum experiments that were in finite dimensional entangled states , called _
quantum strategies_. the least @xmath1 for which the c - coloring game has such a perfect strategy became known as the _ quantum chromatic _ number of a graph , denoted @xmath3 . for an introduction to this literature ,
see @xcite , @xcite .
there are some open questions about the proper model for the set of probabilities that arise from entangled quantum experiments and this has led to the study of several , potentially different , definitions of quantum chromatic numbers denoted @xmath4 and @xmath5 , see @xcite and @xcite .
however , while there are many graphs @xmath0 for which @xmath6 , it is not yet known whether or not @xmath7 , @xmath8 , and @xmath9 can assume different values .
furthermore , there are currently no general algorithms for computing these three quantum chromatic numbers .
there is also an alternative characterization of @xmath2 in terms of graph homomorphisms . if @xmath10 denotes the complete graph on @xmath1 vertices , then @xmath2 is the least integer @xmath1 for which a graph homomorphism exists from @xmath0 to @xmath10 .
again there is a corresponding _ graph homomorphism game _ and several , potentially different definitions of _ quantum graph homomorphisms _ defined in terms of the existence of perfect quantum strategies for the graph homomorphism game . in each case
, the least c for which there exists a perfect quantum strategy for the graph homomorphism game from @xmath0 to @xmath10 is the corresponding quantum chromatic number . for more on this literature
see , @xcite , @xcite , and @xcite .
given graphs @xmath0 and @xmath11 , c. ortiz and the third author @xcite affiliated a * -algebra to this pair , denoted @xmath12 , and characterized the existence of graph homomorphisms from @xmath0 to @xmath11 in terms of representations of this algebra : @xcite proves that @xmath12 has a non - zero homomorphism into the complex numbers if and only if there exists a classical graph homomorphism from @xmath0 to @xmath11 . because the problem of determining if @xmath13 is an np - complete problem , the results of @xcite allow us to deduce that the problem of determining if @xmath14 has a non - trivial homomorphism into the complex numbers is an np - complete problem .
similarly , @xcite also shows that there exists a _ quantum graph homomorphism _ , as defined by @xcite and @xcite , from @xmath0 to @xmath11 if and only if @xmath12 has a non - zero * -homomorphism into the matrices , that is , a non - zero finite dimensional representation .
a result of @xcite shows that the problem of determining if @xmath15 is an np - hard problem .
hence , the problem of determining if @xmath14 has any non - zero finite dimensional representations is np - hard .
currently , the computational complexity of these other variations of the quantum chromatic number is unknown .
the results of @xcite characterize the existence of these other types of quantum graph homomorphisms in terms of the existence of various types of traces on the algebra @xmath12 .
thus , questions about the existence or non - existence of these various types of graph homomorphisms , and consequently quantum colorings , is reduced to questions about these algebras .
this leads us to a more detailed study of these algebras , which is the main topic of this paper .
since the existence of these various types of representations can be characterized in terms of whether or not certain ideals in these algebras are proper , we are naturally led to study new chromatic numbers of a graph determined by the least integer @xmath1 so that the type of ideal in which we are interested is proper . thus , whenever a certain type of ideal is proper , we can use that as the definition of a new type of graph homomorphism and obtain a new _ algebraic _ chromatic number of a graph , which we then try to relate to prior chromatic numbers .
three of these new parameters are denoted @xmath16 , @xmath17 , and @xmath18 .
one goal of this paper is to study the properties of these new chromatic numbers . in general
, no algorithms are known for computing the quantum chromatic numbers of a graph .
one advantage to @xmath16 is that its computation reduces to a grbner basis problem in a non - commutative algebra .
however , using a machine - aided proof , we are able to show that @xmath19 for every graph , which is not the case for any of the earlier chromatic numbers .
on the other hand , we know of no graphs that separate @xmath20 from the other quantum chromatic numbers , while computing this parameter can be done approximately with a combination of algebra and optimization ( semidefinite programming ) . since all the earlier notions of quantum graph homomorphism and the corresponding quantum chromatic numbers , were defined in terms of the existence of _ perfect quantum strategies _ for certain games , we begin by identifying a certain family of games for which we can carry out the construction of @xcite and affiliate an algebra to the game such that the existence of various types of representations of the algebra determines whether or not various types of perfect quantum strategies exist for the game . for this family of games
it is possible that all of the notions of perfect quantum strategies coincide and are equivalent to a certain _ hereditary _ ideal in this algebra being proper . in section [ sec : def ]
, we provide some background on games and strategies , introduce the family of _ synchronous games _ which are the games for which we can extend the results of @xcite and construct an algebra . in section [ sec : alg - sync - game ]
, we introduce the * -algebra of a synchronous game , the chromatic numbers @xmath16 and @xmath17 and prove some of their properties .
in section [ sec:123colors ] , we focus on the case of 1 , 2 , and 3 colors . in section [ sec : alg ] , we show that removing the assumption that the algebra be a free * -algebra instead of just a free algebra , changes nothing essential . we also show that it is enough to study these algebras over @xmath21 instead of @xmath22 .
this allows us to use a grbner basis approach . in section [ sec:4colors ]
, we present the details of the machine - assisted proof that @xmath19 for all graphs . in sections [ sec : loc ] and [ sec : loc properties ] , we introduce and study the _ locally commuting _ chromatic number @xmath18 .
we lay out some definitions and a few basic properties of games and strategies .
we will primarily be concerned with the @xmath1-coloring game and the graph homomorphism game . by a * two - person finite input - output game *
we mean a tuple @xmath23 where @xmath24 are finite sets and @xmath25 is a function that represents the rules of the game , sometimes called the predicate .
the sets @xmath26 and @xmath27 represent the inputs that alice and bob can receive , and the sets @xmath28 and @xmath29 , represent the outputs that alice and bob can produce , respectively .
a referee selects a pair @xmath30 , gives alice @xmath31 and bob @xmath32 , and they then produce outputs ( answers ) , @xmath33 and @xmath34 , respectively .
they win the game if @xmath35 and lose otherwise .
alice and bob are allowed to know the sets and the function @xmath36 and cooperate before the game to produce a strategy for providing outputs , but while producing outputs , alice and bob only know their own inputs and are not allowed to know the other person s input . each time that they are given an input and produce an output
is referred to as a * round * of the game .
we call such a game * synchronous * provided that : ( i ) alice and bob have the same input sets and the same output sets , which we denote by @xmath37 and @xmath38 , respectively , and ( ii ) @xmath36 satisfies : @xmath39 that is , whenever alice and bob receive the same inputs then they must produce the same outputs . to simplify notation
we write a synchronous game as @xmath40 .
a _ graph _ @xmath0 is specified by a vertex set @xmath41 and an edge set @xmath42 , satisfying @xmath43 and @xmath44 .
the * c - coloring game * for @xmath0 has inputs @xmath45 and outputs @xmath46 where the outputs are thought of as different colors .
they win provided that whenever alice and bob receive adjacent vertices , i.e. , @xmath47 , their outputs are different colors and when they receive the same vertex they must output the same color .
thus , @xmath48 , @xmath49 and the rule function is equal to 1 for all other tuples .
it is easy to see that this is a synchronous game
. given two graphs @xmath0 and @xmath11 , a _ graph homomorphism from g to h _ is a function @xmath50 with the property that @xmath51 .
the * graph homomorphism game * from @xmath0 to @xmath11 has inputs @xmath45 and outputs @xmath52 they win provided that whenever alice and bob receive inputs that are an edge in @xmath0 , then their outputs are an edge in @xmath11 and that whenever alice and bob receive the same vertex in @xmath0 they produce the same vertex in @xmath11 .
this is also a synchronous game .
a * deterministic strategy * for a game is a pair of functions , @xmath53 and @xmath54 such that if alice and bob receive inputs @xmath55 then they produce outputs @xmath56 .
a deterministic strategy wins every round of the game if and only if @xmath57 such a strategy is called a * perfect deterministic strategy*. it is not hard to see that for a synchronous game , any perfect deterministic strategy must satisfy , @xmath58 . in particular , a perfect deterministic strategy for the c - coloring game for @xmath0 is a function @xmath59 such that @xmath60 .
thus , a perfect deterministic strategy is precisely a c - coloring of @xmath0 .
similarly , a perfect deterministic strategy for the graph homomorphism is precisely a graph homomorphism . finally , it is not difficult to see that if @xmath10 denotes the complete graph on @xmath1 vertices then a graph homomorphism exists from @xmath0 to @xmath10 if and only if @xmath0 has a c - coloring .
this is because any time @xmath61 then a graph homomorphism must send them to distinct vertices in @xmath10 . indeed
, the rule function for the c - coloring game is exactly the same as the rule function for the graph homomorphism game from @xmath0 to @xmath10 . a * random strategy * for such a game is a conditional probability density @xmath62 , which represents the probability that , given inputs @xmath30 , alice and bob produce outputs @xmath63 .
thus , @xmath64 and for each @xmath65 @xmath66 in this paper we identify random strategies with their conditional probability density , so that a random strategy will simply be a conditional probability density @xmath62 .
a random strategy is called * perfect * if @xmath67 thus , for each round , a perfect strategy gives a winning output with probability 1 .
we next discuss * local * random strategies , which are also sometimes called * classical * , meaning not quantum .
they are obtained as follows : alice and bob share a probability space @xmath68 , for each input @xmath69 , alice has a random variable , @xmath70 and for each input @xmath71 , bob has a random variable , @xmath72 such that for each round of the game , alice and bob will evaluate their random variables at a point @xmath73 via a formula that has been agreed upon in advance .
this yields conditional probabilities , @xmath74 this will be a perfect strategy if and only if @xmath75 or equivalently , @xmath76 if we have a perfect local strategy and set @xmath77 then @xmath78 since @xmath26 and @xmath27 are finite sets ; in particular , @xmath79 is non - empty . if we choose any @xmath80 and set @xmath81 and @xmath82 , then it is easily checked that this is a perfect deterministic strategy .
thus , a perfect classical random strategy exists if and only if a perfect deterministic strategy exists .
an advantage to using a perfect classical random strategy over a perfect deterministic strategy , is that it is difficult for an observer to find a deterministic strategy even after observing the outputs of many rounds .
the idea behind * nonlocal games * is to allow a larger set of conditional probabilities , namely , those that can be obtained by allowing alice and bob to run quantum experiments to obtain their outputs .
definitions of these various sets of probability densities , including loc , q , qa , qc , vect , nsb can be found in ( * ? ? ?
* section 6 ) or @xcite , so we will avoid repeating them here .
we only remark that if for @xmath83 we use @xmath84 to denote the corresponding set of conditional probabilities , then it is known that @xmath85 the sets @xmath86 , @xmath87 and @xmath88 represent three potentially different mathematical models for the set of all probabilities that can arise as outcomes from entangled quantum experiments .
the question of whether or not @xmath89 for any number of experiments and any number of outputs is known to be equivalent to connes embedding conjecture due to results of @xcite .
we say that @xmath62 is a * perfect t - strategy * for a game provided that it is a perfect strategy that belongs to the corresponding set of probability densities .
given a graph @xmath0 we set @xmath90 equal to the least @xmath1 for which there exists a perfect t - strategy for the c - coloring game for @xmath0 .
the above inclusions imply that @xmath91 currently , it is unknown if there are any graphs that separate @xmath92 and @xmath5 or whether these three parameters are always equal .
examples of graphs are known for which @xmath93 , for which @xmath94 and for which @xmath95 . for details , see @xcite , @xcite and @xcite .
other versions of quantum chromatic type graph parameters appear in @xcite and a comparison of those parameters with @xmath8 and @xmath96 can be found in ( * ? ? ?
* section 1 ) .
similarly , we say that there is a _
t - homomorphism _ from @xmath0 to @xmath11 if and only if there exists a perfect t - strategy for the graph homomorphism game from @xmath0 to @xmath11 and it is unknown if q - homomorphisms , qa - homomorphisms and qc - homomorphisms are distinct or coincide .
finally , we close this section by showing that it is enough to consider so - called symmetric games . note that in a synchronous game there is no requirement that @xmath97 .
that is , the rule function does not need to be * symmetric * in this sense .
the following shows that it is enough to consider synchronous games with this additional symmetry .
given @xmath98 a synchronous game , we define @xmath99 by setting @xmath100 and set @xmath101
. then it is easily seen that @xmath102 is a synchronous game with the property that @xmath103 . in the remainder of this section , we prove the following slight extension of @xcite .
let @xmath40 be a synchronous game and let @xmath104 be a perfect vect - strategy for @xmath105 , where the vectors @xmath106 and @xmath107 are as in the definition of a vector correlation(see @xcite )
. then @xmath108 .
by definition , for each @xmath109 the vectors @xmath110 are mutually orthogonal and @xmath111 are mutually orthogonal .
so , @xmath112 thus , the inequalities are equalities , which forces @xmath113 for all @xmath109 and all @xmath114 .
let @xmath115 be a synchronous game and let @xmath116 . if @xmath62 is a perfect t - strategy for @xmath105 , then @xmath117 for all @xmath118 and all @xmath119 .
if @xmath62 is a perfect t - strategy , then it is a perfect vect - strategy and hence there exist vectors as in the definition such that , @xmath120 where the middle equality follows since the inner products are assumed to be non - negative .
this corollary readily yields the following result . [
prop : reduce - to - synch - games ] let @xmath121 be a synchronous game and let @xmath122 . then @xmath62 is a perfect t - strategy for @xmath105 if and only if @xmath62 is a perfect t - strategy for @xmath102 .
we begin by constructing a * -algebra , defined by generators and relations , that is affiliated with a synchronous game .
the existence or non - existence of various types of perfect quantum strategies for the game then corresponds to the existence or non - existence of various types of representations of this algebra .
this leads us to examine various ideals in the algebra .
let @xmath123 be a synchronous game and assume that the cardinality of @xmath37 is @xmath124 while the cardinality of @xmath38 is @xmath125 .
we will often identify @xmath37 with @xmath126 and @xmath38 with @xmath127 .
we let @xmath128 denote the free product of @xmath129 copies of the cyclic group of order @xmath130 and let @xmath131 $ ] denote the complex * -algebra of the group .
we regard the group algebra as both a * -algebra , where for each group element @xmath132 we have @xmath133 , and as an ( incomplete ) inner product space , with the group elements forming an orthonormal set and the inner product is given by @xmath134 where @xmath135 is the trace functional . for each @xmath109
we have a unitary generator @xmath136 $ ] such that @xmath137 .
if we set @xmath138 then the eigenvalues of each @xmath139 is the set @xmath140 .
the orthogonal projection onto the eigenspace corresponding to @xmath141 is given by [ eq : efromu ] e_v , a = _ k=0^m-1 ( ^-a u_v ) ^k , and these satisfy @xmath142 the set @xmath143 is another set of generators for @xmath131 $ ] .
we let @xmath144 denote the 2-sided * -ideal in @xmath131 $ ] generated by the set @xmath145 and refer to it as * the ideal of the game @xmath105*. we define the * * -algebra of @xmath105 * to be the quotient @xmath146/{\mathcal{i}}({\mathcal{g } } ) .\ ] ] a familiar case occurs when we are given two graphs @xmath0 and @xmath11 and @xmath105 is the graph homomorphism game from @xmath0 to @xmath11 .
then @xmath147 , where the algebra on the right hand side is the algebra introduced in @xcite , so we shall continue that notation in this instance . recall that @xmath148 is then the algebra of the c - coloring game for @xmath0 .
we say that a game has a perfect algebraic strategy if @xmath149 is nontrivial .
given graphs g and h , we write @xmath150 if @xmath12 is nontirvial .
we define the algebraic chromatic number of g to be @xmath151 the following is a slight generalization of ( * ? ? ?
* theorem 4.7 ) .
let @xmath121 be a synchronous game . 1 .
@xmath105 has a perfect deterministic strategy if and only if there exists a unital * -homomorphism from @xmath149 to @xmath22 .
2 . @xmath105 has a perfect q - strategy if and only if there exists a unital * -homomorphism from @xmath149 to @xmath152 for some non - zero finite dimensional hilbert space .
3 . @xmath105 has a perfect qc - strategy if and only if there exists a unital c*-algebra @xmath153 with a faithful trace and a unital * -homomorphism @xmath154 .
hence , if @xmath105 has a perfect qc - strategy , then it has a perfect algebraic strategy and so @xmath155 for every graph @xmath0 .
we start with the third statement .
since the game is synchronous , any perfect strategy @xmath62 must also be synchronous . by (
* theorem 5.5 ) , any synchronous density is of the following form : @xmath156 , where @xmath157 is a tracial state for a a unital c*-algebra @xmath153 generated by projections @xmath158 satisfying @xmath159 for all @xmath31 .
if we take the gns representation @xcite of @xmath153 induced by @xmath135 , then the image of @xmath153 under this representation will be a quotient of @xmath153 with all the same properties and the additional property that @xmath135 is a faithful trace on the quotient . now if , in addition , @xmath62 belongs to the smaller family of perfect q - strategies ,
then by ( * ? ? ?
* theorem 5.3 ) the c*-algebra @xmath153 will be finite dimensional . hence , the second statement follows .
finally , if @xmath62 belongs to the smaller family of perfect loc - strategies , then the c*-algebra @xmath153 will be abelian , and hence , the first statement follows .
it is also possible to characterize the existence of perfect qa - strategies , but the proof is a bit long for here : @xmath105 has a perfect qa - strategy if and only if there exists a unital * -homomorphism of @xmath149 into the von neumann algebra @xmath160 .
the * -algebra @xmath131 $ ] also possesses an order defined as follows : let @xmath161 be the cone generated by all elements of the form @xmath162 for @xmath163.$ ] if @xmath164 $ ] are self - adjoint elements , we write @xmath165 if @xmath166 .
next , notice that @xmath161 induces a cone on @xmath149 , which we regard as the positive elements , by setting @xmath167 given two self - adjoint elements @xmath168 , we again write @xmath165 if and only if @xmath169 .
in the language of ozawa @xcite this makes @xmath149 into a * semi - pre - c*-algebra*. a self - adjoint vector subspace @xmath170 $ ] is called * hereditary * provided that @xmath171 and @xmath172 implies that @xmath173 .
let @xmath105 be a synchronous game .
find conditions on the game so that the 2-sided ideal @xmath144 is hereditary .
later we will see an example of a game such that @xmath144 is not hereditary .
the following result shows why the hereditary condition is important .
let @xmath105 be a synchronous game and let @xmath144 be the ideal of the game .
then @xmath144 is a hereditary subspace of @xmath131 $ ] if and only if @xmath174 .
let @xmath175 $ ] .
we begin by characterizing when the equivalence class @xmath176 is contained in @xmath177 . by definition
, this occurs if and only if there are elements @xmath178 in @xmath144 such that @xmath179 and @xmath180 .
this is equivalent to @xmath181 .
now suppose that @xmath182 and that the equivalence class @xmath183 is non - zero in @xmath149 .
if the class is contained in @xmath184 then choosing @xmath185 and @xmath186 as in the previous paragraph , the element @xmath187 demonstrates that @xmath144 is not hereditary .
conversely , if @xmath144 is not hereditary , then there exists @xmath188 and @xmath189 such that @xmath190 .
the inequality @xmath191 implies that @xmath192 , while @xmath193 implies that @xmath194 .
clearly , this element is non - zero . given a subspace @xmath195 we let @xmath196 denote the smallest hereditary subspace that contains @xmath195 and
call this space the * hereditary closure * of @xmath195 .
we define the * hereditary * -algebra of the game @xmath105 * to be the quotient @xmath197/{\mathcal{i}}^h({\mathcal{g}}).\ ] ] note that @xmath198 is a quotient of @xmath149 .
we say that a game has a perfect hereditary strategy if @xmath198 is nontrivial .
given graphs g and h , we write @xmath199 if @xmath200 is nontrivial .
we define the hereditary chromatic number of g by @xmath201 we define the positive cone in @xmath198 by setting @xmath202 so that @xmath198 is also a semi - pre-@xmath203-algebra .
the following is immediate : [ propercone ] let @xmath105 be a synchronous game , then @xmath204 thus , the `` positive '' cone on @xmath198 is now a * proper * cone and @xmath198 is an ordered vector space .
[ prop : k_n>k_c - hered ] if @xmath205 then @xmath206 .
consequently , @xmath207 suppose , that @xmath205 .
we let @xmath208 , where @xmath209 is defined as in [ sec : alg - sync - game ] .
then we have that @xmath210 for all @xmath211 set @xmath212 since @xmath213 we have that @xmath214 hence , @xmath215 is also a `` projection '' in the sense that @xmath216 now @xmath217 but also , @xmath218 hence , @xmath219 by definition , @xmath220 .
if @xmath221 , then @xmath222 and by proposition [ propercone ] , we have @xmath223 ; that is , @xmath224 which contradicts our hypothesis that @xmath205
. hence , @xmath225 .
this shows that @xmath226 . for the other inequality , note that if @xmath205 , then @xmath227 . by the results of @xcite ,
if @xmath0 is any graph with @xmath228 , then there is a unital * -homomorphism from @xmath148 into a c*-algebra with a trace .
the kernel of this homomorphism is a hereditary ideal and so must contain @xmath229 .
hence , this latter ideal is proper and @xmath230 .
thus , @xmath231 .
hence , @xmath232 .
the above proof also shows that : if @xmath233 and @xmath234 , then @xmath206 .
the next natural question is whether or not @xmath198 is a * pre-@xmath203-algebra * in the sense of @xcite .
the answer is that this cone will need to satisfy one more hypothesis .
let @xmath105 be a synchronous game and let @xmath235 denote the intersection of the kernels of all unital @xmath236-homomorphisms from @xmath131 $ ] into the bounded operators on a hilbert space ( possibly 0 dimensional ) that vanish on @xmath144 .
let @xmath237/ { \mathcal{i}}^c({\mathcal{g}})$ ] .
let @xmath105 be a synchronous game .
then @xmath238 and @xmath239 \ , : x^*x + { \mathcal{i}}({\mathcal{g } } ) \le \epsilon 1 + { \mathcal{i}}({\mathcal{g } } ) , \ , \forall \epsilon > 0 , \ , \epsilon \in { \mathbb{r } } \}.\ ] ] there exists a ( non - zero ) hilbert space @xmath240 and a unital * -homomorphism @xmath241 \to b({\mathcal{h}})$ ] that vanishes on @xmath144 if and only if @xmath242 .
the kernel of every * -homomorphism is a hereditary ideal and the intersection of hereditary ideals is a hereditary ideal , hence @xmath235 is a hereditary ideal containing @xmath144 .
so , @xmath238 .
we have that @xmath243 if and only if @xmath244 is in the kernel of every * -homomorphism of @xmath149 into the bounded operators on a hilbert space . in (
* theorem 1 ) it is shown that this is equivalent to @xmath183 being in the `` ideal of infinitesimal elements '' of @xmath149 , which is the ideal defined by the right - hand side of the above formula . the last result comes from the fact that the ideal of infinitesimal elements is exactly the intersection of the kernels of all such representations .
we say that a game @xmath105 has a * perfect c*-strategy * provided that @xmath245 is nontrivial .
following @xcite , we write @xmath246 provided that for given graphs @xmath0 and @xmath11 , the algebra @xmath247 is nontrivial .
we define the c*-chromatic number of g to be @xmath248 the following is immediate .
let @xmath0 be a graph
. then @xmath249 .
this motivates the following question .
let @xmath105 be a synchronous game .
is @xmath250 ? if @xmath251 $ ] , then does there exist a non - zero hilbert space @xmath240 and a unital * -homomorphism , @xmath241 \to b({\mathcal{h}})$ ] such that @xmath252 , that is , if @xmath251 $ ] , then is @xmath253 $ ] ? [ sec : computinghered ] here we mention some literature on determining if an ideal @xmath254 is hereditary and the issue of computing its `` hereditary closure . '' in the real algebraic geometry literature , a hereditary ideal is called a real ideal . for a finitely
generated left ideal @xmath255 in @xmath256 the papers @xcite present a theory and a numerical algorithm to test ( up to numerical error ) if @xmath255 is hereditary .
the algorithm also computes the `` hereditary radical '' of @xmath255 .
the computer algorithm relies on numerical optimization ( semidefinite programming ) and hence it is not exact but approximate . for two sided ideals
@xcite and @xcite contain some theory . also the first author and klep developed and crudely implemented a hereditary ideal algorithm under ncalgebra . however , it is too memory consuming to be effective , so we leave this topic for future work .
a moral one can draw from this literature is that computing hereditary closures is not broadly effective at this moment .
the clique number of a graph @xmath257 is defined as the size of the largest complete subgraph of g. it is not hard to see that g contains a complete subgraph of size c if and only if there is a graph homomorphism from @xmath10 to g. hence , there is a parallel theory of quantum clique numbers that we shall not pursue here , other than to remark that for each of the cases @xmath258 we define the * t - clique number * of g by @xmath259 so that @xmath260 lovasz @xcite introduced his theta function @xmath261 of a graph .
the famous lovasz sandwich theorem @xcite says that for every graph @xmath0 , if @xmath262 denotes its graph complement , then @xmath263 . in (
* proposition 4.2 ) [ thetainequality ] they showed the following improvement of the lovasz sandwich theorem : @xmath264 we shall show later , @xmath265 , while @xmath266 hence the sandwich inequality fails for the algebraic version .
this motivates the following problem : is @xmath267 for all graphs ?
it is a classic result that deciding if @xmath13 is an np - complete problem . in @xcite
it was shown that deciding if @xmath268 is np - hard , and , in particular , there is no known algorithm for deciding if this latter inequality is true . for these reasons
it is interesting to see what can be said about the new inequalities , @xmath269 and @xmath270 .
addressing the first two inequalities would require one to compute @xmath271 and @xmath272 , and unfortunately these ideals contain elements not just determined by simple algebraic relations .
however , studying @xmath273 and @xmath14 is rewarding , as we shall see now see . throughout the section , we use the notation @xmath274 .
let @xmath0 be a graph .
then @xmath275 if and only if @xmath0 is the empty graph .
hence , @xmath276 for each vertex we only have one idempotent @xmath277 and since these sum to the identity , necessarily @xmath278 .
but if there is an edge @xmath55 then @xmath279 .
[ prop : alg=2 ] let @xmath0 be a connected graph on more than one vertex .
then @xmath280 first assume that @xmath281 .
fix a vertex @xmath31 and set @xmath282 and @xmath283 note @xmath284 let @xmath61 , then @xmath285 and @xmath286 hence , @xmath287 and similarly , @xmath288 also , @xmath289 thus , whenever @xmath290 then @xmath291 i.e. , there are two projections and they flip . since @xmath0 is connected , by using a path from @xmath31 to an arbitrary @xmath32 we see that @xmath292 now we wish to 2-color @xmath0 .
define the color of any vertex @xmath32 to be @xmath293 if @xmath294 and 1 if @xmath295 this yields a 2-coloring , and since @xmath0 is connected on more than one vertex , there is no 1-coloring , showing @xmath296 .
conversely , if @xmath296 then @xmath0 is not the empty graph . since @xmath297 , by the previous result , @xmath298 .
[ prop : g - k-3-commute ] if @xmath61 then @xmath299 for all @xmath300 .
in particular , if @xmath0 is complete , then @xmath14 is abelian . for @xmath301 similarly , @xmath302 whenever @xmath303 now @xmath304 similarly , @xmath305 hence , for @xmath306 , @xmath307 , while when @xmath308 , @xmath309 @xmath310 for @xmath311 we have that @xmath312 by proposition [ prop : alg=2 ] . now
if @xmath313 then by proposition [ prop : alg=2 ] , we see @xmath314 which is a contradiction .
hence , @xmath315 finally , if @xmath316 , then by proposition [ prop : g - k-3-commute ] , we have that @xmath317 is a non - zero abelian complex * -algebra .
but every unital , abelian ring contains a proper maximal ideal @xmath318 , and forming the quotient we obtain a field @xmath319 .
the map @xmath320 embeds @xmath22 as a subfield .
now we use the fact that @xmath321 is generated by projections and that the image of each projection in @xmath319 is either 0 or 1 in order to see that the range of the quotient map is just @xmath22 .
thus , @xmath322 and we have a unital homomorphism @xmath323 and again using the fact that the image of each projection is 0 or 1 and that the projections commute , we see that @xmath324 is a * -homomorphism .
hence , by ( * ? ? ?
* theorem 4.12 ) , we have @xmath325 , a contradiction . thus , @xmath326 and the result follows . @xmath327 .
the original motivation for the construction of the algebra of a game comes from projective quantum measurement systems which are always given by orthogonal projections on a hilbert space , i.e. , operators satisfying @xmath328 .
this is why we have defined the algebra of a game to be a * -algebra .
but a natural question is whether or not one really needs a * -algebra or is there simply a free algebra with relations that suffices ? in this section we show that as long as one introduces the correct relations then the assumption that the algebra be a * -algebra is not necessary . to this end
let @xmath329 be the free unital complex algebra on @xmath330 generators and let @xmath331 where @xmath332 is the two - sided ideal generated by @xmath333 we let @xmath334 denote the coset of @xmath335 in the quotient so that @xmath336 there is an isomorphism @xmath337 $ ] with @xmath338 , where @xmath339 are defined as in the previous section .
let @xmath340 $ ] be the unital algebra homomorphism with @xmath341 then @xmath342 vanishes on @xmath332 and so induces a quotient homomorphism @xmath337 $ ] .
it remains to show that @xmath324 is one - to - one . to this end set @xmath343 and let @xmath344 .
it is readily checked that @xmath345 @xmath346 . since @xmath347 we have that @xmath348 generates @xmath349 .
now by the universal property of @xmath131 $ ] there is a homomorphism @xmath350 \to { \mathcal{b}}(n , m)$ ] with @xmath351 and hence , this is the inverse of @xmath324 . if a unital algebra contains 3 idempotents , @xmath352 with @xmath353 , then necessarily @xmath354 for @xmath306 . to see this note that @xmath355 is idempotent . squaring yields that @xmath356 .
thus , @xmath357 and @xmath358 , from which it follows that @xmath359 and so @xmath360 and the claim follows .
hence , when @xmath361 the condition that @xmath362 is a consequence of the other hypotheses and is not needed in the definition of the ideal @xmath332 .
however , heydar radjavi @xcite has constructed a set of 4 idempotents in a unital complex algebra which sum to the identity but for which @xmath363 for @xmath306 .
thus , for @xmath364 , it is necessary to include the relation @xmath365 in the ideal in order to guarantee that the quotient @xmath349 is isomorphic to @xmath131 $ ] , since these products are 0 in the latter algebra .
if a set of self - adjoint projections on a hilbert space , @xmath366 sum to the identity then it is easily checked that they project onto orthogonal subspaces and so @xmath367 .
thus , in any c*-algebra when self - adjoint idempotents sum to the identity , their pairwise products are 0 .
but the situation is not so clear for self - adjoint idempotents in a * -algebra and we have not been able to resolve this question .
so we ask : let @xmath368 be a unital * -algebra and let @xmath369 satisfy @xmath370 and @xmath371 .
then does it follow that @xmath372 ?
let @xmath121 be a symmetric synchronous game with @xmath124 and @xmath125 .
then @xmath149 is isomorphic to the quotient of @xmath373 by the 2-sided ideal generated by @xmath374 and @xmath375 note that @xmath365 for @xmath376 is in the ideal since @xmath377 .
we note that the hypothesis that the game be symmetric is needed , since in a * -algebra , the condition that @xmath378 implies that @xmath379 , while this relation would not necessarily be met in the quotient of the free algebra .
[ cor:1notin - i - fnm ] a symmetric synchronous game @xmath40 has a perfect algebraic strategy if and only if 1 is not in the 2-sided ideal of @xmath373 generated by @xmath380 and @xmath375 the following is important for grbner basis calculations .
let @xmath381 be any subfield .
we set @xmath382 equal to the free @xmath383-algebra on @xmath330 generators @xmath384 , so that @xmath385 .
given any symmetric synchronous game @xmath105 with @xmath129 inputs and @xmath130 outputs , we let @xmath386 be the 2-sided ideal generated by @xmath387 and @xmath375 we let @xmath388 . by corollary [ cor:1notin - i - fnm ] , @xmath105 has a perfect algebraic strategy if and only if @xmath389 , or equivalently @xmath390 .
we show that this computation is independent of the field @xmath383 .
[ prop : a_k - base - change ] if @xmath383 is a field containing @xmath21 , then @xmath391 furthermore , @xmath392 if and only if @xmath393 . by definition , we have a short exact sequence @xmath394 of @xmath21-vector spaces . since @xmath383 is flat over @xmath21 , we obtain a short exact sequence @xmath395 since the generators are independent of the field , one checks that @xmath396 and that the image of @xmath397 is equal to @xmath398 . since this latter map is injective , we see @xmath399 .
hence , the above short exact sequence shows @xmath400 .
lastly , @xmath401 is faithfully flat .
so , @xmath392 if and only if @xmath393 .
out old the fact that @xmath402 , means that @xmath403 . from this
it readily follows that @xmath404 for any @xmath405 , since we can express @xmath406 using any subset of 4 of the vertices .
thus , @xmath407 for any @xmath405 . in this section we give the details of a machine - assisted proof that shows that @xmath408 and hence @xmath265 .
since we have shown earlier that @xmath409 , this means that @xmath410 so that @xmath411 is not hereditarily closed . moreover , by proposition 3.1 , we have that @xmath412 . [ thm:4colors ] if @xmath0 is any graph , then @xmath413 . moreover , deciding whether @xmath414 is np - complete .
this section gives a machine - assisted proof which analyzes 4 algebraic colors .
we prove : [ thm:4colors ] for any graph @xmath0 , we have @xmath415 .
this theorem is equivalent to the statement that for any graph @xmath0 , the ideal satisfies @xmath416 .
we will prove this statement through the use of ( noncommutative ) grbner bases .
for a brief effective exposition to noncommutative grbner basis algorithms , see chapter 12.3 @xcite or @xcite . for those readers already familiar with ( commutative ) grbner bases , we explain the key differences with the noncommutative setting .
let @xmath417 be a two - sided ideal , and prescribe a monomial order .
a noncommutative grbner basis @xmath418 of @xmath255 is a set of generators such that the leading term of any element of @xmath255 is in the monomial ideal generated the leading terms of @xmath418 .
a noncommutative grbner basis is produced in the same way as in the commutative case .
let @xmath419 be the leading term of @xmath420 and notice that any two @xmath421 have as many as 4 possible least common multiples , each of which produces syzygyies .
one repeatedly produces syzygyies and reduces to obtain a grbner basis in the same way as the commutative setting .
however , unlike the commutative case , a grbner basis can be infinite .
very fortunately the grbner bases that arise in our coloring computations below are finite .
the key property we use is that @xmath185 is in @xmath255 if and only if the reduction of @xmath185 by a grbner basis for @xmath255 yields 0 .
recall that @xmath422 is generated by the following relations : @xmath423 to prove theorem [ thm:4colors ] we will make use of the following theorem : [ grob : bas ] for any @xmath424 a grbner basis for @xmath425 under the graded lexographic ordering with [ eq : order ] _ 0,0<_0,1<_0,2<_0,3<_1,0<_1,1 < < _ n-1,3 consists of relations of the following forms : [ it:1 ] @xmath426 with @xmath427 @xmath428 [ it:2 ] @xmath429 with @xmath430 [ it:3 ] @xmath431 [ it:4 ] @xmath432 with @xmath433 , @xmath434 [ it:5 ] @xmath435 @xmath436 with @xmath433 [ it:6 ] @xmath437 @xmath438 @xmath436 with @xmath433 [ it:7 ] @xmath439 @xmath440 @xmath441 @xmath442 with @xmath443 specifically @xmath444 has a grbner basis that does not contain 1 and thus @xmath445 . [ rem : gborder ] each of the relations correspond to a set of relations obtained by taking all choices of @xmath446 in @xmath447 .
however because of the monomial ordering chosen , the leading terms are always of the forms : @xmath448 @xmath449 @xmath450 @xmath451 @xmath452 @xmath453 @xmath454 additionally for every @xmath455 , all the vertices of @xmath456 which appear in the terms of relation ( i ) also appear in the leading term of ( i ) .
let the ideal generated by these relations be denoted by @xmath457 , we will first show that these relations form a grbner basis for @xmath457 , and then show that @xmath458 . before we begin our calculations pertaining to an algebra over @xmath22 we note that all of the coefficients that appear will be in @xmath459 .
section [ sec : field ] bears on this . to see that these relations form a grbner basis
we must show that the syzygy between any two polynomials in this list is zero when reduced by the list .
first by remark [ rem : gborder ] each of the relations has variables corresponding to at most three different vertices of @xmath456 and reducing by a relation will not introduce variables corresponding to different vertices .
thus when calculating and reducing the syzygy between any two relations , variables corresponding to at most 6 vertices of @xmath456 will be involved .
therefore we can verify that all syzygies reduce to zero by looking at the case @xmath460 which we verify using ncalgebra 5.0 and ncgb running under mathematica ( see notebook qcgb-9-20-16.nb , available at : https://github.com/ncalgebra/usernotebooks ) .
this proves that the relations
form a grbner basis .
we now show that @xmath458 .
we will first show that all of the generators of @xmath457 are contained in @xmath444 .
the elements of types ( 1 ) , ( 3 ) , and ( 4 ) are self - evidently in @xmath444 since they are elements of the generating set of @xmath444 . for type ( 2 ) we note that under the relations generating @xmath444 that @xmath461 and thus elements of type ( 2 ) are in @xmath444 . for type ( 5 ) we use the relations generating @xmath444 to get that @xmath462 @xmath463 @xmath464 finally type ( 6 ) is obtained by reducing @xmath465 @xmath466 using the relations of types ( 1)(5 ) , and type ( 7 ) is obtained by reducing @xmath467 @xmath468 using the relations of types ( 1)(5 ) .
these two reductions are verified with mathematica in qcgb-9-20-16.nb .
thus all of the generating relations of @xmath457 are in @xmath444 and we have that @xmath469
. next we will show that all the generators of @xmath444 are contained in @xmath457 .
the only generating relations of @xmath37 that are not immediately seen to be in @xmath457 are @xmath470 and @xmath471 to see that @xmath472 is in @xmath457 we consider @xmath473 .
this is an element of @xmath457 since @xmath474 is in @xmath457 , and when multiplied out all terms except @xmath472 are in @xmath475 , and thus @xmath472 is in @xmath457 , similarly @xmath476 is in @xmath457 . finally , we consider the equation @xmath477 @xmath478 @xmath479 the right - hand side is a sum of relations in @xmath457 and is thus in @xmath457 , and thus the left - hand side is also in @xmath457 , specifically @xmath480 is in @xmath457
. therefore all of the generating relations of @xmath444 are in @xmath457 , so that @xmath481 . since we have shown inclusion both ways , we have that @xmath482 and we are done .
[ lem : sublem ] if @xmath0 , @xmath11 are graphs such that @xmath483 and @xmath484 , then @xmath485 and thus @xmath486 .
the relations generating @xmath487 contains the relations generating @xmath488 and thus the result follows .
let @xmath0 be a graph on @xmath129 vertices . by theorem [ grob : bas ] , @xmath445 .
additionally @xmath489 , and thus by lemma [ lem : sublem ] , @xmath490 . therefore @xmath415 .
we do not know the complexity of deciding if @xmath491 , i.e. , of deciding if @xmath492 .
our analysis of `` algebraic colorability '' shows @xmath493 is too coarse a parameter to provide much information about graphs , after all every graph will be algebraically colorable by at most 4 colors , and so has @xmath493 has very little in common with any of the quantum chromatic numbers .
this problem does not occur with the hereditary chromatic number , which may be equal to the usual quantum chromatic number for all graphs , but unfortunately determining the elements of the hereditary closure of an ideal is difficult . in the forthcoming section , we add further physically motivated algebraic relations to @xmath494 , in order to obtain a chromatic number that is more amenable to algebraic analysis while still retaining a quantum flavor .
the new relations are additional commutation relations in @xmath12 and yield a new algebra @xmath495 that we call the _ locally commuting algebra_. this yields a new type of chromatic number , @xmath496 .
one goal of this section is to prove @xmath497 .
since our algebras were initially motivated by quantum chromatic numbers , it is natural to look to quantum mechanics for further relations to impose . in the case of a graph
, we can imagine each vertex as corresponding to a laboratory and think of two vertices as connected whenever those laboratories can conduct a joint experiment . in this case , all of the measurement operators for the two labs should commute , i.e. , whenever @xmath55 is an edge , then the commutator @xmath498 : = e_{v , i}e_{w , j } - e_{w , j}e_{v , i } = 0 $ ] .
note that this commutation rule is exactly the rule that we were able to derive in the case of three colors in proposition [ prop : g - k-3-commute ] .
this motivates the following definitions .
let @xmath499 be a synchronous game with @xmath124 and @xmath125 .
we say that @xmath118 are * adjacent * and write @xmath500 provided that @xmath501 and there exists @xmath119 such that @xmath502 .
we define the * locally commuting ideal * of the game to be the 2-sided ideal @xmath503 in @xmath131 $ ] generated by the set @xmath504 \mid \ v \sim w , \ , \forall a , b \in o \}.\ ] ] we set @xmath505/{\mathcal{i}}_{lc}({\mathcal{g}})$ ] and call this the * locally commuting algebra of @xmath105*. in the case that @xmath0 and @xmath11 are graphs and @xmath105 is the graph homomorphism game from @xmath0 to @xmath11 we set @xmath506 and @xmath507 we write @xmath508 provided that @xmath509 $ ] and set @xmath510 we similarly define @xmath511 note that in the case of the graph homomorphism game from @xmath0 to @xmath11 we have that @xmath512 and @xmath513 .
thus , the relationship @xmath514 extends the concept of adjacency to the inputs of a general synchronous game .
thus , @xmath515 is the universal * -algebra generated by self - adjoint projections @xmath516 satisfying * @xmath517 * @xmath518 * @xmath519=0 , \forall i , j$ ] and @xmath18 is the least @xmath1 for which such a non - trivial * -algebra exists .
we begin by showing that every graph homomorphism , in the usual sense , yields an @xmath520-morphism : [ l : to->lc ] if @xmath521 , then @xmath522 .
let @xmath523 be a graph homomorphism .
we must show @xmath524 .
consider the map @xmath525 sending @xmath526 to 1 and @xmath527 to 0 for @xmath528 .
it is easy to see this is a well - defined @xmath22-algebra map and hence surjective . as a result , @xmath524 .
[ cor : chi - lc>=chi ] we have @xmath529 and @xmath530 .
there is a graph homomorphism @xmath531 so by lemma [ l : to->lc ] , we have @xmath532 and hence @xmath529 . the inequality for @xmath533 is shown in an analogous fashion .
we are now ready to prove the main result of this section .
[ thm : chi_lc(k_n ) ] for every @xmath534 we have that @xmath535 .
the case @xmath536 is trivial to check .
assume that @xmath537 .
let us label the vertices by numbers @xmath406 to @xmath129 and set @xmath538 then @xmath539 and we want to show that @xmath540 is impossible .
suppose that @xmath541 , then since @xmath542 for @xmath543 we have that @xmath544 where the sum is over all @xmath129-tuples with @xmath545 . by the locally commuting hypothesis , all of the above projections commute
, so we may re - order the sum in any fashion . since @xmath540 , by the pigeon - hole principle
, each @xmath129-tuple must contain @xmath546 with @xmath547 .
but then @xmath548 .
hence , each product occurring in the above sum is 0 and so the sum is 0 .
thus , we have that @xmath549 , which shows that @xmath550 , that is , @xmath551 $ ] for @xmath221 .
we do not know if the lovsz sandwich result holds in this context , i.e. , if @xmath552 .
in this section we analyze the algebra @xmath495 more closely and obtain the value of @xmath18 for a few select graphs .
in particular , we are able to show , indirectly , that @xmath554 . throughout this section
we shall write @xmath555 to indicate that two algebras are isomorphic . we shall use @xmath556 to denote the abelian algebra of complex - valued functions on @xmath129 points .
if @xmath495 and @xmath568 are non - zero , then we must prove that @xmath569 is non - zero as well . to see this , consider the map @xmath570 given by @xmath571 and suppose that it vanishes on @xmath572 .
hence , there would be a well - defined map on the quotient , @xmath573 @xmath574 if @xmath575 in @xmath569 , then the same would be true in @xmath576 , since this map sends units to units .
[ thm : alc - functor ] the assignment @xmath583 @xmath584 is a functor , which is covariant in the first factor and contravariant in the second ; note that the category of graphs is with usual morphisms , not @xmath520-morphisms .
if @xmath585 is a morphism , then we have a map @xmath586 given by @xmath587 .
on the other hand , if @xmath588 is a morphism , then we have @xmath589 and so from the proof of lemma [ l : lc - comp ] , we have @xmath590 since @xmath591 is a morphism of graphs , we have a map @xmath592 as in the proof of lemma [ l : to->lc ] .
composing with the above , we have @xmath593 .
explicitly , this map is given by sending @xmath594 to @xmath595 .
we now show how the functor @xmath553 interacts with various natural graph operations . to begin , recall that if @xmath0 is a graph , its suspension @xmath596 is defined by adding a new vertex @xmath31 and an edge from @xmath31 to each of the vertices of @xmath0 .
given an algebra @xmath368 we shall let @xmath597 denote the algebra of @xmath1-tuples with entries from @xmath368 , i.e. , the tensor product @xmath598 where @xmath599 can be identfied with the algebra of @xmath22-valued functions on @xmath1 points . [ prop : sigmag - h - lc ] let @xmath0 and @xmath11 be any graphs , and let @xmath600 be the non - isolated vertices .
for @xmath601 we let @xmath602 denote the neighborhood of @xmath603 , i.e. , the induced subgraph of @xmath11 with vertices adjacent to @xmath603 ; notice @xmath604 unless @xmath603 has a self - edge .
then @xmath605 in particular , if @xmath11 is vertex transitive and @xmath603 is any vertex of @xmath11 with neighborhood @xmath606 , then @xmath607 let @xmath608 be the new vertex added to @xmath596 , i.e. , @xmath609 . since @xmath608 is adjacent to every vertex of @xmath0 , we see @xmath610 commutes with @xmath611 for all @xmath612 and @xmath613 . furthermore , the defining relations of @xmath553 tell us @xmath614 where @xmath615 denotes the kronecker delta function .
so , @xmath616/(\sum_x e_{ux}-1,e_{ux}e_{uy}=\delta_{x , y}e_{ux}).\ ] ] in other words , the @xmath610 for @xmath617 are commuting orthogonal idempotents , which shows @xmath618 where the last equality comes from the fact that @xmath619 for @xmath564 .
now note that @xmath620 remains non - zero in the quotient @xmath621 if and only if @xmath622 .
thus , @xmath623 which establishes the first assertion of the proposition .
the second assertion easily follows from the first since all neighborhoods are isomorphic .
one easily checks that @xmath636 for any graph @xmath0 .
in particular , our desired statement holds for @xmath536 .
the proof then follows from induction on @xmath129 by applying corollary [ cor : sigmag - lc ] and using that @xmath637 . in theorem [ thm : chi_lc(k_n ) ]
, we proved @xmath497 .
corollary [ cor : lc - kc ] gives another proof of this result which is more refined : the corollary tells us the specific structure of @xmath638 whereas the theorem merely tells us it is non - zero .
[ cor : lc - out - of - k_n ] if @xmath11 is a graph , then @xmath639 and for @xmath640 , we have @xmath641 where @xmath642 is an @xmath643-clique and @xmath644 . in particular , if @xmath11 is vertex transitive on @xmath1 vertices , then @xmath645 where @xmath642 is any @xmath643-clique in @xmath11 .
we show by induction that the @xmath649 form an @xmath650-clique and that @xmath651 . for @xmath652
this is just the definition . for @xmath653 ,
observe that by construction @xmath654 and since @xmath655 forms an @xmath656-clique , we see @xmath649 forms an @xmath650-clique .
next , @xmath657 is the set of @xmath658 that are adjacent to @xmath659 , which is the definition of @xmath660 .
this shows @xmath661 and by the @xmath536 case , each summand is isomorphic to @xmath662 copies of @xmath663 .
now notice that @xmath664 is independent of the order of the sequence , and hence this term arises @xmath665 times .
this yields the desired statement .
we next consider how @xmath553 interacts with the categorical product @xmath668 .
recall that if @xmath11 and @xmath669 are graphs then @xmath670 is a graph with vertex set @xmath671 and where @xmath672 if and only if @xmath673 and @xmath674 .
[ rmk : preserve - prod ] as shown in theorem [ thm : alc - functor ] , we can view @xmath553 as a functor .
theorem [ thm : preserve - prod ] can be interpreted as saying that the second factor of the functor @xmath676 preserves products . recall that since the second factor is contravariant , preserving products means that it takes products in graphs to coproducts in @xmath677-algebras , namely tensor products .
since we have a map @xmath678 , theorem [ thm : alc - functor ] tells us that we have a map @xmath679 .
similarly for @xmath669 , and hence a natural map @xmath680 . explicitly , it is given by @xmath681 we now construct an inverse map given by @xmath682 @xmath683 provided this is well - defined , it is indeed an inverse since @xmath684 where the last equality uses the fact that @xmath685 unless @xmath686 . also , composing the two maps in the opposite order yields the identity since @xmath687 the rest of the proof is devoted to showing that the map we constructed above is well - defined .
first note that @xmath688 next , @xmath689 if @xmath690 then without loss of generality @xmath433 .
so , @xmath691 next assume @xmath692 .
we need to check that the image of @xmath693 is equal to that @xmath694 . @xmath695 and since @xmath692 , this is equal to @xmath696 which is the image of @xmath694 .
finally , we must show that if @xmath692 and @xmath697 then @xmath693 maps to 0 . since @xmath697 , without loss of generality @xmath698 .
then the image of @xmath693 is @xmath699 since @xmath700 .
recall that the exponential of graphs @xmath702 is defined as follows : its vertex set consists of all functions @xmath703 and there is an edge @xmath704 if for all @xmath673 we have @xmath705 .
note that if @xmath706 is a graph homomorphism , then it has a self - edge .
[ prob : adjunction - exp ] if one allows @xmath11 to have self edges , then @xmath495 can be defined using the same relations at the beginning of this section .
one can then define a map @xmath708 given by @xmath709 is this map an isomorphism ?
in other words , are products and exponentials adjoints ?
in addition to the categorical product @xmath668 , there are several other kinds of products on graphs , which we now consider . both @xmath710 and @xmath711
have vertex set @xmath712 . in the former ,
@xmath672 if and only if @xmath713 and @xmath622 , or @xmath563 and @xmath714 . in the latter ,
@xmath672 if and only if @xmath563 and @xmath622 , or @xmath713 and @xmath622 , or @xmath563 and @xmath714 .
the products @xmath715 and @xmath716 are referred to as the cartesian and strong products , respectively .
we have at least @xmath719 maps @xmath720 , so lemma [ l : to->lc ] and corollary [ cor : chi_lc - under - maps ] show @xmath721 .
similarly for @xmath11 and so @xmath722 . to prove the result , it now suffices to show we have a map @xmath723 indeed , if @xmath724 and @xmath725 are non - zero , then so is @xmath726 since the above map would send 0 to 0 and 1 to 1 , and if @xmath727 in @xmath726 , then @xmath727 in @xmath728 , which is not the case .
taking @xmath729 , this would then show @xmath730 .
we now construct the above map .
we define it by : @xmath731 and show it is well - defined .
first suppose that @xmath732 and @xmath733 .
then @xmath734 and without loss of generality @xmath735 and @xmath736 . then @xmath737 since @xmath738 if @xmath428 , and if @xmath308 , then @xmath739 and so @xmath740 .
it suffices to construct a map @xmath758 we define it by @xmath759 one readily checks that this map is well - defined .
for example , in the case of the cartesian product @xmath760 we show that if @xmath761 and @xmath762 , then @xmath763 maps to 0 . without loss of generality
, we can assume that @xmath764 and @xmath765 .
then the image is @xmath766 , which is automatically 0 if @xmath747 .
so , we may assume @xmath734 , in which case @xmath767 since @xmath762 .
but then @xmath768 . in @xcite
, the authors showed that @xmath771 .
thus , separating @xmath772 from @xmath773 .
later , @xcite showed that @xmath774 , separating the potentially smaller @xmath9 from @xmath773 .
we show below that @xmath775 as well .
[ ex : c_5-boxtimes - k_2 ] let @xmath777 .
it is easy to see that @xmath778 and @xmath779 , so a priori @xmath496 could be 4 or 5 . we show @xmath780 we need to show that @xmath781 .
the graph @xmath0 is made up of 2 pentagons stacked on top of each other .
let one of the pentagons have vertices @xmath782 labeled clockwise and let the other pentagon have vertices @xmath783 with @xmath784 and @xmath785 having the same neighbors .
for ease of notation , we denote @xmath786 by @xmath787 .
note that @xmath788 multiplying on both the left and right by @xmath789 , we obtain @xmath790 similarly , we find @xmath791 for all @xmath300 . as a result , @xmath792 and so @xmath781 .
[ ex : c_5-boxtimes - k_3 ] let @xmath793 .
we see @xmath794 and @xmath795 , so a priori @xmath496 could be 6 , 7 , or 8 .
we show @xmath796 we must show @xmath797 .
we follow the same notational conventions as in example [ ex : c_5-boxtimes - k_2 ] .
let @xmath782 be the vertices of @xmath798 labeled clockwise and denote the next two copies of @xmath798 by @xmath799 resp .
@xmath800 where @xmath784 , @xmath785 , @xmath801 have the same neighbors in @xmath0 .
we also let @xmath802 . as in the previous example , @xmath803 where @xmath804 and analogously for @xmath805 . we show that every term occurring in the sum on the righthand side of the above equation is 0 .
the indices @xmath806 occurring in the sum @xmath642 must all lie in @xmath807 otherwise the term vanishes ( since it is multiplied by @xmath808 ) .
our goal is to show that all terms in the sum in the righthand side vanish , so we can fix a summand in @xmath642 and assume @xmath806 equal @xmath809 respectively .
then the indices in @xmath810 must be 3 of @xmath811 .
we also see that the indices in @xmath812 must be 3 of @xmath807 .
fix a summands @xmath813 and @xmath814 of @xmath810 and @xmath812 , respectively .
then @xmath815 has size at most 2 .
therefore , every summand @xmath816 of @xmath817 satisfies @xmath818 .
so , @xmath819 , and analogously we see @xmath820 for all @xmath806 .
so , @xmath821 showing that @xmath797 . as a result ,
@xmath822 .
since the definition of @xmath496 is not obviously related to representations on hilbert spaces , it is unclear how to relate it to @xmath823 for @xmath824 .
where does @xmath496 fit within this hierarchy ?
j. cimpric , j. w. helton , s.a .
mccullough , c. nelson ; a noncommutative real nullstellensatz corresponds to a noncommutative real ideal : algorithms .
( 3 ) 106 ( 2013 ) , no . 5 ,
10601086 t. cubitt , l. mancinska , d. roberson , s. severini , d. stahlke , and a. winter , _ bounds on entanglement assisted source - channel coding via the lovasz theta number and its variants _ , arxiv:1310.7120 [ quant - ph ] , oct .
2013 .
kadison and j.r .
ringrose , fundamentals of the theory of operator algebras .
i. elementary theory .
pure and applied mathematics , 100 . academic press ,
[ harcourt brace jovanovich , publishers ] , new york , 1983 .
xv+398 pp .
isbn : 0 - 12 - 393301 - 3 v. levandovskyy : non - commutative computer algebra for polynomial algebras : grbner bases , applications and implementation , _ ph.d .
thesis at universitt kaiserslautern , germany _ ( 2005 ) .
available at http://kluedo.ub.uni-kl.de/volltexte/2005/1883/ b. keller : algorithms and orders for finding noncommutative grbner bases , _ ph.d .
thesis at the virginia polytechnic institute and state university , usa _ ( 1997 ) .
available at + http://scholar.lib.vt.edu/theses/available/etd-405814212975790/ t. mora : an introduction to commutative and noncommutative grbner bases , second international colloquium on words , languages and combinatorics ( kyoto , 1992 ) .
_ theoret .
sci . _ * 134 * ( 1994 ) 131173 .
b. reinert : on grbner bases in monoid and group rings , _ ph.d .
thesis at universitt kaiserslautern , germany _ ( 1995 ) .
available at http://www-madlener.informatik.uni-kl.de/staff/reinert/privat/diss.long_en.html
|
we associate to each synchronous game an algebra whose representations determine if the game has a perfect deterministic strategy , perfect quantum strategy or one of several other strategies . when applied to the graph coloring game , this leads to characterizations in terms of properties of an algebra of various quantum chromatic numbers that have been studied in the literature .
this allows us to develop a correspondence between various chromatic numbers of a graph and ideals in this algebra which can then be approached via grbner basis methods .
|
a very efficient way to calculate the evolution of the parton densities in deeply inelastic scattering consist in solving the evolution equations in mellin@xmath2 space @xcite . with the advent of the 3loop anomalous dimensions
@xcite the scaling violations of the deep
inelastic structure functions can be analyzed at high precision .
the evolution equations are ordinary differential equations , which can be solved analytically .
the wilson coefficients @xcite and anomalous dimensions are given in terms of multiple nested harmonic sums @xcite , which obey algebraic relations @xcite .
each finite harmonic sum can be represented in terms of a mellin transform of a harmonic polylogarithm @xcite weighted by @xmath3 and a polynomial of harmonic sums of lower weight .
it turns out that for all basic functions at least up to weight five , as needed for the physical observables mentioned above , the harmonic polylogarithms are nielsen integrals @xcite . furthermore
, structural relations exist between various mellin transforms of weighted nielsen integrals @xcite .
they are due to relations at rational values of the mellin variable , differentiation w.r.t . the mellin variable , integration by parts relations and functional identities of nielsen integrals .
these relations , together with the quasi - shuffle algebra of harmonic sums allow to build the massless single
scale quantities in perturbative qed and qcd out of a low number of basic functions at a given order in perturbation theory .
the anomalous dimensions and wilson coefficients in mellin space , for instance , can be written in terms of these basic functions and a number of their derivatives , in addition to polynomials including euler s @xmath4functions as well as rational functions of the mellin variable for complex values of @xmath2 .
then , on this basis , results in @xmath5space , are obtained by means of a single inverse mellin transform , which has to be carried out numerically .
for the class of functions contributing to the 2loop wilson coefficients the analytic continuations were given in ref .
simple analytic continuations of nested sums are known for long , cf .
e.g. ref .
this direction has been followed in ref .
@xcite recently .
however , in many cases the numerical convergence is very slow . to obtain an appropriate representation for the 3loop anomalous dimensions and 2loop wilson coefficients
, we exploit the algebraic and structural relations between the nested harmonic sums . for the 2loop anomalous dimensions only one non
trivial function contributes aside from euler s @xmath6 functions @xcite .
all known massless 2loop wilson coefficients can be expressed by four more functions @xcite .
for the 3loop anomalous dimension the mellin transforms @xmath7(n ) = \int_0 ^ 1~dx~x^n~f(x)\ ] ] of the following eight functions form the set of additional basic functions : @xmath8 here the functions @xmath9 and @xmath10 denote polylogarithms @xcite and nielsen
integrals @xcite , respectively . in terms of harmonic polylogarithms
@xcite , they read @xmath11 where the algebra can be applied to products of harmonic polylogarithms to express them in the basis of single harmonic polylogarithms . in the present paper
we derive fast and precise numerical representations of these mellin transforms for complex values of @xmath2 .
previously numerical representations of the 3loop anomalous dimensions were given in @xcite .
following earlier investigations @xcite , we aim at numerical representations for the individual basic functions , the building blocks of the massless wilson coefficients and the anomalous dimensions , to high precision .
thus , in principle , the present approach is tunable to arbitrary accuracy .
as outlined in ref .
@xcite the mellin transforms containing the denominator @xmath12 are given by : @xmath13(n ) & = & \lnt \cdot f(1 ) - \int_0 ^ 1 dx ~x^{n-1}~\ln(1+x ) \left[n f(x ) + x f'(x)\right ] \nonumber\\ & = & \lnt \cdot f(1 ) - \sum_{k=1}^{20 } a^{(1)}_k \left ( n { \,\mbox{\bf m}}[f(x)](n+k-1 ) + { \,\mbox{\bf m}}\left[f'(x)\right](n+k)\right ) \ , .
\nonumber\\\end{aligned}\ ] ] accurate representations for the function @xmath14 can be given using the minimax method for numerical approximation : @xmath15 which we further improved to an accuracy of @xmath16 .
the coefficients read : @xmath17 for later use we also parameterize @xmath18 @xmath19 with @xmath20 the following mellin transforms contribute , cf . also refs .
@xcite , @xmath21(n-1 ) & = & { \z4 \over n } - { \z3 \over n^2 } + { \z2 \over n^3 } - { \s(1 ) \over n^4 } \\ { \,\mbox{\bf m}}\left[\li4'(x)\right](n-1 ) & = & { \,\mbox{\bf m}}\left[\li3(x)\right](n-2 ) \\ { \,\mbox{\bf m}}\left[s_{1,3}(x)\right](n-1 ) & = & { \z4 \over n } - { ( \s(1))^3 \over 6 \ * n^2 } - { \s(1 ) \ * \s(2 ) \over 2 \ * n^2 } - { \s(3 ) \over 3 \ * n^2 } \\ { \,\mbox{\bf m}}\left[s_{1,3}'(x)\right](n-1 ) & = & - { 1 \over 6 } \ * { \,\mbox{\bf m}}\left[\ln^3(1-x)\right](n-2 ) \\ { \,\mbox{\bf m}}\left[\li2 ^ 2(x)\right](n-1 ) & = & { 1 \over n } \ * \left ( \z2 ^ 2 - { 4 \ * \z3 \over n } - 2 \ * \z2 \ * { \s(1 ) \over n } + 2 \ * { \s(2,1 ) \over n } \right ) \nonumber\\ & & + { 2 \over n^3 } \ * \left ( ( \s(1))^2 + \s(2 ) \right ) \\ { \,\mbox{\bf m}}\left[(\li2 ^ 2(x))'\right](n-1 ) & = & - 2 \ * { \,\mbox{\bf m}}\left[\ln(1-x ) \li2(x)\right](n-2 ) \\ { \,\mbox{\bf m}}\left[s_{2,2}(x)\right](n-1 ) & = & { \z4 \over 4 \ * n } - { \z3 \over n^2 } + { ( \s(1))^2 + \s(2 ) \over 2 \ * n^3 } \\ { \,\mbox{\bf m}}\left[s_{2,2}'(x)\right](n-1 ) & = & { \,\mbox{\bf m}}\left[s_{1,2}(x)\right](n-2 ) \\ { \,\mbox{\bf m}}\left[s_{2,2}(-x)-\li2 ^ 2(-x)/2\right](n-1 ) & = & { 1 \over n } \ * \biggl [ - { 3 \over 4 } \ * \z2 ^ 2 + 2 \ * \li4(1/2 ) + { 7 \over 4 } \ * \z3 \ * \lnt - { 1 \over 2 } \ * \z2 \ * ( \lnt)^2 \nonumber\\ & & + { 1 \over 12 } \ * ( \lnt)^4 - { \,\mbox{\bf m}}\left[s_{1,2}(-x)\right](n-1 ) \biggr ] \nonumber\\ & & - { 1 \over n } \ * \left [ { \z2 ^ 2 \over 8 } + { \,\mbox{\bf m}}\left[\ln(1+x)\li2(-x)\right](n-1 ) \right ] \nonumber\\ \\ { \,\mbox{\bf m}}\left[\left(s_{2,2}(-x)-\li2 ^ 2(-x)/2\right)'\right](n-1 ) & = & { \,\mbox{\bf m}}\left[s_{1,2}(-x)\right](n-2 ) \nonumber\\ & & + { \,\mbox{\bf m}}\left[\ln(1+x ) \li2(-x)\right](n-2)\end{aligned}\ ] ] with @xmath22(n-1 ) & = & { \z3 \over n } - { \z2 \over n^2 } + { \s(1 ) \over n^3 } \\ { \,\mbox{\bf m}}\left[\ln^3(1-x)\right](n-1 ) & = & - { ( \s(1))^3 \over n } - 3 \ * { \s(1 ) \ * \s(2 ) \over n } - 2 \ * { \s(3 ) \over n } \\
{ \,\mbox{\bf m}}\left[s_{1,2}(x)\right](n-1 ) & = & { \z3 \over n } - { ( \s(1))^2 \over 2 \ * n^2 } - { \s(2 ) \over 2 \ * n^2 } \\ { \,\mbox{\bf m}}\left[s_{1,2}(-x)\right](n-1 ) & = & { \z3 \over 8 \ * n } - { ( \lnt)^2 \over 2 \ * n^2 } - { ( -1)^n \over n^2 } \ * \biggl [ \s(1,-1 ) + \lnt \ * \left ( \s(1 ) - \s(-1 ) \right ) \nonumber\\ & & - { 1 \over 2 } \ * ( \lnt)^2 \biggr ] \nonumber\\ & \simeq & { \z3 \over 8 \ * n } - { 1 \over 2 \ * n } \sum_{k=2}^{24 } { a^{(2)}_k \over n+k } \\ { \,\mbox{\bf m}}\left[\ln(1-x ) \li2(x)\right](n-1 ) & = & { 1 \over n } \ * \biggl [ - 2 \ * \z3 - \z2 \ * \s(1 ) + { 1 \over n } \ * \left((\s(1))^2 + \s(2 ) \right ) + \s(2,1 ) \biggr ] \nonumber\\ \]]@xmath23(n-1 ) & \simeq & \sum_{k=1}^{20 } a^{(1)}_k { \,\mbox{\bf m}}\left[\li2(-x)\right](n-1+k ) \\ { \,\mbox{\bf m}}\left[\li2(-x)\right](n-1 ) & = & - { \z2 \over 2 \ * n } + { \lnt \over n^2 } - { ( -1)^n \over n^2 } \ * ( \s(-1 ) + \lnt ) \ ] ] the functions @xmath24 and @xmath25 have been parameterized in ref .
@xcite @xmath26(n ) - \lnt \left(\s(1 ) - \s(-1 ) \right ) + \frac{1}{2 } ( \lnt)^2\\ \s(2,1 ) & = & { \,\mbox{\bf m}}\left[\left(\frac{\li2(x)}{1-x}\right)_+\right](n ) + \z2 \ * \s(1 ) \ , , \end{aligned}\ ] ] where @xmath27(n ) \simeq \frac{1}{2 } ( \lnt)^2 - \frac{n}{2}\sum_{k=2}^{24 } \frac{a^{(2)}_k}{n+k}~.\end{aligned}\ ] ] three functions of this type contribute . in case
the numerator function is analytic at @xmath28 one may represent the functions of this class using the minimax method directly .
both @xmath29 and @xmath30 have branch points at @xmath28 .
therefore we derive an analytic representation @xmath31 for @xmath32/(1-x)$ ] valid for the region @xmath33 and determine the remainder part using the minimax method : @xmath34 for @xmath35/(1-x)$ ] one obtains @xmath36 @xmath37 with @xmath38 the mellin transform of @xmath31 is given by a linear combination of @xmath39(n ) & = & b(n+1,m+1 ) \\ { \,\mbox{\bf m}}\left[(1-x)^m \ln(1-x ) \right](n ) & = & \left[\psi(m+1 ) - \psi(m+n+2)\right ] b(n+1,m+1 ) \\
{ \,\mbox{\bf m}}\left[(1-x)^m \ln^2(1-x ) \right](n ) & = & \bigl\{\left[\psi'(m+1 ) - \psi'(m+n+2)\right ] \nonumber\\ & + & \left[\psi(m+1 ) - \psi(m+n+2)\right]^2\bigr\ } b(n+1,m+1)~. \end{aligned}\ ] ] @xmath40/(1-x)$ ] obeys the representation ( [ eqr1 ] ) , where @xmath41 \nonumber\\ & & + \left ( - { 1 \over 7 } \ * x^{6 } + { 43 \over 42 } \ * x^{5 } - { 667 \over 210 } \ * x^{4 } + { 2341 \over 420 } \ * x^{3 } - { 853 \over 140 } \ * x^{2 } + { 617 \over 140 } \ * x - { 363 \over 140 } \right ) \ * \z3 \ ] ] @xmath42 and @xmath43 .
the nielsen integrals @xmath44 are analytic at @xmath45
. therefore @xmath46/$ ] @xmath47 can be represented by a polynomial directly , where @xmath48 we will do this for the functions @xmath49/(1-x)$ ] and @xmath50/(1-x)$ ] separately .
the following representation for @xmath51 $ ] is obtained : @xmath52(n ) = \sum_{k=0}^{23 } \frac{b^{(1)}_k}{n+k+1}~,\ ] ] and @xmath53 the corresponding polynomial in @xmath5 has an accuracy of @xmath54 in @xmath55 $ ] . similarly , for @xmath56(n)$ ] one obtains : @xmath57(n ) = \sum_{k=0}^{11 } \frac{b^{(2)}_k}{n+k+1}~,\ ] ] with @xmath58 the corresponding polynomial in @xmath5 has an accuracy of @xmath59 for @xmath60 $ ] . in table 1 the values of the maximum relative errors are given for the first 40 mellin moments .
depending on the function accuracies better than @xmath61 to @xmath62 are obtained .
we also compared the representation of the basic functions through numerical inversion of the mellin moments by a complex contour integral . in the range
@xmath63 $ ] maximal relative errors of @xmath64 were obtained , irrespective of the value of the basic functions . in some regions in @xmath5 even up to several order of magnitude
better results were obtained . [ cols="^,^,^ " , ] table 1 : relative accuracies of the representations .
single scale quantities in massless qed and qcd such as the wilson coefficients and anomalous dimensions can be represented by nested harmonic sums .
their representation may be reduced to a small set of basic functions in addition to euler s @xmath6 functions , each of the basic functions corresponding to one particular single harmonic sums .
this is possible due to algebraic and structural relations between the harmonic sums . up to the level of the 3loop anomalous dimensions
13 such basic functions occur . we have derived fast and precise numerical representations for the eight new functions contributing at the level of the 3loop anomalous dimensions in qcd . in the region
@xmath63 $ ] which is physically most interesting their relative accuracy is better than @xmath0 .
if needed , the procedure outlined in the present paper can readily be generalized to a higher level of accuracy at the expense of more coefficients in the representation . due to use of polynomial forms in the analytic continuations
very fast representations are obtained well suited for the use in qcd evolution codes .
the basic functions discussed in the present paper form a part of the building blocks for the 3loop coefficient functions and other higher order single scale quantities . * acknowledgment . * + this paper was supported in part by dfg sonderforschungsbereich transregio 9 , computergesttzte theoretische physik . s.m . acknowledges the hospitality of the fnal theory division during completion of the work .
10 m. glck , e. reya and a. vogt , z. phys . *
c48 * ( 1990 ) 471 . j. blmlein and a. vogt , phys . rev . *
d58 * ( 1998 ) 014020 , hep - ph/9712546 . s. moch , j.a.m .
vermaseren and a. vogt , nucl . phys .
* b688 * ( 2004 ) 101 , hep - ph/0403192 .
a. vogt , s. moch and j.a.m .
vermaseren , nucl . phys .
* b691 * ( 2004 ) 129 , hep - ph/0404111 .
d. i. kazakov and a. v. kotikov , nucl .
* b307 * ( 1988 ) 721 [ erratum - ibid .
* b345 * ( 1990 ) 299 ] .
j. sanchez guillen , j. miramontes , m. miramontes , g. parente and o. a. sampayo , nucl .
* b353 * ( 1991 ) 337 .
van neerven and e.b .
zijlstra , phys . lett .
* b272 * ( 1991 ) 127 ; * b273 * ( 1991 ) 476 ; * b297 * ( 1992 ) 377 ; nucl . phys .
* b383 * ( 1992 ) 525 . s. moch and j.a.m .
vermaseren , nucl . phys . *
b573 * ( 2000 ) 853 , hep - ph/9912355 .
vermaseren , int . j. mod
a14 * ( 1999 ) 2037 , hep - ph/9806280 .
j. blmlein and s. kurth , phys . rev .
* d60 * ( 1999 ) 014018 , hep - ph/9810241 .
j.m . borwein and r. girgensohn , electron .
j. combinatorics * 3 * ( 1996 ) r23 ( appendix by d.j .
broadhurst ) , cecm research report 95 - 053 .
j. blmlein , comput .
phys . commun .
* 159 * ( 2004 ) 19 , hep - ph/0311046 .
e. remiddi and j.a.m .
vermaseren , int . j. mod
a15 * ( 2000 ) 725 , hep - ph/9905237 .
n. nielsen , nova acta leopoldina ( halle ) * 90 * ( 1909 ) 123 .
klbig , siam j. math .
* 17 * ( 1986 ) 1232 .
j. blmlein , nucl .
* 135 * ( 2004 ) 225 , hep - ph/0407044 .
j. blmlein , comput .
* 133 * ( 2000 ) 76 , hep - ph/0003100 .
j. blmlein and s. kurth , ( 1997 ) , hep - ph/9708388 .
a.v . kotikov and v.n .
velizhanin , ( 2005 ) , hep - ph/0501274 .
a. gonzalez - arroyo , c. lopez and f.j .
yndurain , nucl . phys .
* b153 * ( 1979 ) 161 ; a. gonzalez - arroyo and c. lopez , nucl . phys . *
b166 * ( 1980 ) 429 .
j. blmlein and s. moch , to appear .
j. blmlein and v. ravindran , ( 2005 ) , hep - ph/0501178 .
l. lewin , ( north holland , new york , 1981 ) , isbn 0 - 444 - 00550 - 1 .
van neerven and a. vogt , nucl .
b568 * ( 2000 ) 263 , hep - ph/9907472 ; * b588 * ( 2000 ) 345 , hep - ph/0006154 ; * b603 * ( 2001 ) 42 , hep - ph/0103123 ; phys . lett . * b490 * ( 2000 ) 111 , hep - ph/0007362 .
|
we present for numerical use the analytic continuations to complex arguments of those basic mellin transforms , which build the harmonic sums contributing to the 3loop anomalous dimensions .
eight new basic functions contribute in addition to the analytic continuations for the 2loop massless wilson coefficients calculated previously .
the representations derived have a relative accuracy of better than @xmath0 in the range @xmath1 $ ] .
desy 05007 + sfb / cpp0505 + march 2005 + * analytic continuation of the harmonic sums * + * for the 3loop anomalous dimensions * + johannes blmlein and sven - olaf moch + _ platanenallee 6 , d-15735 zeuthen , germany _ +
|
a measurement of @xmath5 opens an important window on the non - perturbative strong physics involving heavy quarks .
the basic framework of the theory is well understood , however , there is still much speculation - predictions for the width range from @xmath34 to @xmath35 @xcite .
we know the @xmath0 width is dominated by strong decays since the measured electromagnetic transition rate is small , @xmath36% @xcite .
the level splitting in the @xmath37 sector is not large enough to allow real strong transitions .
therefore , a measurement of the width of the @xmath0 gives unique information about the strong coupling constant in heavy - light systems .
the total width of the @xmath0 is the sum of the partial widths of the strong decays @xmath3 and @xmath38 and the radiative decay @xmath39 .
we can write the width in terms of a single strong coupling , @xmath40 , and an electromagnetic coupling , @xmath41 : @xmath42 using the isospin relationship @xmath43 where @xmath44 is the fine structure constant , and the momenta are those for the indicated particle in the appropriate @xmath0 decay in its rest frame . the contribution of the electromagnetic decay can be neglected to note that the width of the @xmath0 only depends on @xmath40 @xcite . prior to this measurement ,
the @xmath0 width was limited to be less than @xmath45 at the @xmath46 confidence level by the accmor collaboration @xcite .
the limit was based on 110 signal events reconstructed in two @xmath47 decay channels with a background of 15% of the signal .
this contribution describes a measurement of the @xmath0 width with the cleo ii.v detector where the signal , in excess of 13,000 events , is reconstructed through a single , well - measured sequence , @xmath48 , @xmath49 .
consideration of charge conjugated modes are implied throughout this paper .
the level of background under the signal is less than 3% in our loosest selection .
figure [ fig : widths ] compares the cleo ii.v signal with signal from the previously published limit .
the challenge of measuring the width of the @xmath0 is understanding the tracking system response function since the experimental resolution exceeds the width we are trying to measure .
candidates with mismeasured hits , errors in pattern recognition , and large angle coloumb scattering are particularly dangerous because the signal shape they project is broad and the errors for these events can be underestimated , resulting in events that can easily influence the parameters of a breit - wigner fitting shape .
we generically term such effects `` tracking mishaps . ''
a difficulty is that there is no physical calibration for this measurement .
the ideal calibration mode would have a large cross - section , a width of zero , decay with a rather small energy release to three charged particles one of which has a much softer momentum distribution than the other two , and the other two would decay through a zero width resonance with a measurable flight distance .
such a mode would allow us to disentangle detector effects from the underlying width but no such mode exists . therefore , to measure the width of the @xmath0 we depend on exhaustive comparisons between a geant @xcite based detector simulation and our data .
we addressed the problem by selecting samples of candidate @xmath0 decays using three strategies .
first we produced the largest sample from data and simulation by imposing only basic tracking consistency requirements .
we call this the _ nominal _ sample .
second we refine the nominal sample selecting candidates with the best measured tracks by making very tight cuts on tracking parameters .
there is special emphasis on choosing those tracks that are well measured in our silicon vertex detector .
this reduces our nominal sample by a factor of thirty and , according to our simulation , has negligible contribution from tracking mishaps .
we call this the _ tracking selected _ sample .
a third alternative is to select our data on specific kinematic properties of the @xmath0 decay that minimize the dependence of the width of the @xmath0 on detector mismeasurements .
the nominal sample size is reduced by a factor of three and a half and , again according to our simulation , the effect of tracking problems is reduced to negligible levels .
we call this the _ kinematic selected _ sample . in all three samples the width
is extracted with an unbinned maximum likelihood fit to the energy release distribution and compared with the simulation s generated value to determine a bias which is then applied to the data .
these three different approaches yield consistent values for the width of the @xmath0 giving us confidence that our simulation accurately models our data .
the cleo detector has been described in detail elsewhere .
all of the data used in this analysis are taken with the detector in its ii.v configuration @xcite .
this work mainly depends on the tracking system of the detector which consists of a three layer , double sided silicon strip detector , an intermediate ten layer drift chamber , and a large 51 layer helium - propane drift chamber .
all three are in an axial magnetic field of 1.5 tesla provided by a superconducting solenoid that contains the tracking region .
the charged tracks are fit using a kalman filter technique that takes into account energy loss as the tracks pass through the material of the beam pipe and detector @xcite .
the data were taken in symmetric @xmath1 collisions at a center of mass energy around 10 gev with an integrated luminosity of 9.0/fb provided by the cornell electron - positron storage ring ( cesr ) .
the nominal sample follows the selection of @xmath50 candidates used in our @xmath51 mixing analysis@xcite .
our reconstruction method takes advantage of the small cesr beam spot and the kinematics and topology of the @xmath52 decay chain .
the @xmath53 and @xmath54 are required to form a common vertex .
the resultant @xmath47 candidate momentum vector is then projected back to the cesr luminous region to determine the @xmath47 production point .
the cesr luminous region has a gaussian width @xmath55 m vertically and @xmath56 m horizontally .
it is well determined by an independent method@xcite .
this procedure determines an accurate @xmath47 production point for @xmath47 s moving out of the horizontal plane ; @xmath47 s moving within 0.3 radians of the horizontal plane are not considered .
then the @xmath57 track is refit constraining its trajectory to intersect the @xmath47 production point .
this improves the resolution on the energy release , @xmath58 , by more than 30% over simply forming the appropriate invariant masses of the tracks .
the improvement to resolution is essential to our measurement of the width of the @xmath0 .
our resolution is shown in figure [ fig : errorcompare ] and is typically 150 kev .
the good agreement reflects that the kinematics and sources of errors on the tracks , such as number of hits and the effects of multiple scattering in detector material , agree between the data and the simulation . to further improve the quality of reconstruction in our sample
, we apply some kinematic cuts to remove a small amount of misreconstructed background .
figure [ fig : ppisoft ] shows the distribution of the momentum [ cols="^,^ " , ] [ tab : systematic ] uncertainties on the width of the @xmath0 and @xmath59 .
we have measured the width of the @xmath0 by studying the distribution of the energy release in @xmath3 followed by @xmath60 decay .
we have done this in three separate samples , one that is minimally selected , a second that reduces poorly measured tracks due to misassociated hits and non - gaussian scatters in the detector material , and a third that takes advantage of the kinematics of the decay chain to reduce dependence on mismeasurements of kinematic parameters .
the resolution on the energy release is well modeled by our simulation , with agreement between the sources of the resolution as predicted by the simulation and analytic calculations .
the largest sources of uncertainty are imperfect modeling of the dependence of the mean energy release on the kinematics of the decay chain , the simulation of the error on the energy release , and correlations among the parameters of the fit to the energy release distribution . with our estimate of the systematic uncertainties for each of the three samples being essentially
the same we chose to report the result for the sample with the smallest statistical uncertainty , the minimally selected sample , and obtain @xmath61 where the first uncertainty is statistical and the second is systematic .
we note that if we form an average value taking into account the statistical correlations among our three measures we get a result that is nearly identical with equation [ eq : result ] since the average is dominated by the result with the smallest statistical uncertainty .
this result is preliminary .
this preliminary measurement is the first of the width of the @xmath0 , and our measurement corresponds to a strong coupling@xcite @xmath62 this is consistent with theoretical predictions based on hqet and relativistic quark models , but higher than predictions based on qcd sum rules .
we gratefully acknowledge the effort of the cesr staff in providing us with excellent luminosity and running conditions .
m. selen thanks the pff program of the nsf and the research corporation , and a.h .
mahmood thanks the texas advanced research program .
this work was supported by the national science foundation , the u.s .
department of energy , and the natural sciences and engineering research council of canada .
r. ruckl , wue - itp-98 - 011 , cern - th/98 - 124 , hep - ph/9810338 is a recent survey summarizing and referencing previous theoretical work .
we follow his definition of the strong coupling @xmath40 .
p. singer hep - ph/9910558 , and j. l. goity and w. roberts jlab - thy-00 - 45 , hep - ph/0012314 appear since that survey .
j. bartelt _ et al . _
( cleo collaboration ) , phys .
lett . * 80 * , 3919 , ( 1998 ) .
s. barlag _ et al .
_ , phys . lett .
* b 38 * , 480 ( 1992 ) .
r. brun _ et al .
_ , geant3 users guide , cern
dd / ee/84 - 1 . y. kubota _ et al .
_ , ( cleo collaboration ) , nucl . instrum
. methods phys .
res . , a * 320 * , 66 ( 1992 ) ; t. hill , nucl .
methods phys .
res . , a * 418 * , 32 ( 1998 ) .
p. billior , nucl .
methods phys .
res . , a * 255 * , 352 ( 1984 ) .
r. godang _ et al . _
( cleo collaboration ) , phys .
lett . * 84 * , 5038 ( 2000 ) .
d. cinabro _ et al .
_ , clns 00/1706 , physics/0011075 .
|
we have made the first measurement of the @xmath0 width using 9/fb of @xmath1 data collected near the @xmath2 resonance by the cleo ii.v detector .
our method uses advanced tracking techniques and a reconstruction method that takes advantage of the small vertical size of the cesr beam spot to measure the energy release distribution from the @xmath3 decay .
our preliminary result is @xmath4 kev .
cleo conf 01 - 2 + * first measurement of @xmath5 * + cleo collaboration + ( february 5 , 2001 ) t. e. coan,@xmath6 v. fadeyev,@xmath6 y. s. gao,@xmath6 y. maravin,@xmath6 i. narsky,@xmath6 r. stroynowski,@xmath6 j. ye,@xmath6 t. wlodek,@xmath6 m. artuso,@xmath7 c. boulahouache,@xmath7 k. bukin,@xmath7 e. dambasuren,@xmath7 g. majumder,@xmath7 r. mountain,@xmath7 s. schuh,@xmath7 t. skwarnicki,@xmath7 s. stone,@xmath7 j.c .
wang,@xmath7 a. wolf,@xmath7 j. wu,@xmath7 s. kopp,@xmath8 m. kostin,@xmath8 a. h. mahmood,@xmath9 s. e. csorna,@xmath10 i. danko,@xmath10 v. jain,@xmath11 k. w. mclean,@xmath10 z. xu,@xmath10 r. godang,@xmath12 g. bonvicini,@xmath13 d. cinabro,@xmath13 m. dubrovin,@xmath13 s. mcgee,@xmath13 g. j. zhou,@xmath13 a. bornheim,@xmath14 e. lipeles,@xmath14 s. p. pappas,@xmath14 a. shapiro,@xmath14 w. m. sun,@xmath14 a. j. weinstein,@xmath14 d. e. jaffe,@xmath15 r. mahapatra,@xmath15 g. masek,@xmath15 h. p.
paar,@xmath15 d. m. asner,@xmath16 a. eppich,@xmath16 t. s. hill,@xmath16 r. j. morrison,@xmath16 r. a. briere,@xmath17 g. p. chen,@xmath17 t. ferguson,@xmath17 h. vogel,@xmath17 a. gritsan,@xmath18 j. p.
alexander,@xmath19 r. baker,@xmath19 c. bebek,@xmath19 b. e. berger,@xmath19 k. berkelman,@xmath19 f. blanc,@xmath19 v. boisvert,@xmath19 d. g. cassel,@xmath19 p. s. drell,@xmath19 j. e. duboscq,@xmath19 k. m. ecklund,@xmath19 r. ehrlich,@xmath19 p. gaidarev,@xmath19 l. gibbons,@xmath19 b. gittelman,@xmath19 s. w. gray,@xmath19 d. l. hartill,@xmath19 b. k. heltsley,@xmath19 p. i. hopman,@xmath19 l. hsu,@xmath19 c. d. jones,@xmath19 j. kandaswamy,@xmath19 d. l. kreinick,@xmath19 m. lohner,@xmath19 a. magerkurth,@xmath19 t. o. meyer,@xmath19 n. b. mistry,@xmath19 e.
nordberg,@xmath19 m. palmer,@xmath19 j. r. patterson,@xmath19 d. peterson,@xmath19 d. riley,@xmath19 a. romano,@xmath19 h. schwarthoff,@xmath19 j. g. thayer,@xmath19 d. urner,@xmath19 b. valant - spaight,@xmath19 g. viehhauser,@xmath19 a. warburton,@xmath19 p. avery,@xmath20 c. prescott,@xmath20 a. i. rubiera,@xmath20 h. stoeck,@xmath20 j. yelton,@xmath20 g. brandenburg,@xmath21 a. ershov,@xmath21 d. y .- j .
kim,@xmath21 r. wilson,@xmath21 t. bergfeld,@xmath22 b. i. eisenstein,@xmath22 j. ernst,@xmath22 g. e. gladding,@xmath22 g. d. gollin,@xmath22 r. m. hans,@xmath22 e. johnson,@xmath22 i. karliner,@xmath22 m. a. marsh,@xmath22 c. plager,@xmath22 c. sedlack,@xmath22 m. selen,@xmath22 j. j. thaler,@xmath22 j. williams,@xmath22 k. w. edwards,@xmath23 r. janicek,@xmath24 p.
m. patel,@xmath24 a. j. sadoff,@xmath25 r. ammar,@xmath26 a. bean,@xmath26 d. besson,@xmath26 x. zhao,@xmath26 s. anderson,@xmath27 v. v. frolov,@xmath27 y. kubota,@xmath27 s. j. lee,@xmath27 j. j. oneill,@xmath27 r. poling,@xmath27 a. smith,@xmath27 c. j. stepaniak,@xmath27 j. urheim,@xmath27 s. ahmed,@xmath28 m. s. alam,@xmath28 s. b. athar,@xmath28 l. jian,@xmath28 l. ling,@xmath28 m. saleem,@xmath28 s. timm,@xmath28 f. wappler,@xmath28 a. anastassov,@xmath29 e. eckhart,@xmath29 k. k. gan,@xmath29 c. gwon,@xmath29 t. hart,@xmath29 k. honscheid,@xmath29 d. hufnagel,@xmath29 h. kagan,@xmath29 r. kass,@xmath29 t. k. pedlar,@xmath29 j. b. thayer,@xmath29 e. von toerne,@xmath29 m. m. zoeller,@xmath29 s. j. richichi,@xmath30 h. severini,@xmath30 p. skubic,@xmath30 a. undrus,@xmath30 v. savinov,@xmath31 s. chen,@xmath32 j. fast,@xmath32 j. w. hinson,@xmath32 j. lee,@xmath32 d. h. miller,@xmath32 e. i. shibata,@xmath32 i. p. j. shipsey,@xmath32 v. pavlunin,@xmath32 d. cronin - hennessy,@xmath33 a.l .
lyon,@xmath33 and e. h. thorndike@xmath33 @xmath6southern methodist university , dallas , texas 75275 + @xmath7syracuse university , syracuse , new york 13244 + @xmath8university of texas , austin , texas 78712 + @xmath9university of texas - pan american , edinburg , texas 78539 + @xmath10vanderbilt university , nashville , tennessee 37235 + @xmath12virginia polytechnic institute and state university , blacksburg , virginia 24061 + @xmath13wayne state university , detroit , michigan 48202 + @xmath14california institute of technology , pasadena , california 91125 + @xmath15university of california , san diego , la jolla , california 92093 + @xmath16university of california , santa barbara , california 93106 + @xmath17carnegie mellon university , pittsburgh , pennsylvania 15213 + @xmath18university of colorado , boulder , colorado 80309 - 0390 + @xmath19cornell university , ithaca , new york 14853 + @xmath20university of florida , gainesville , florida 32611 + @xmath21harvard university , cambridge , massachusetts 02138 + @xmath22university of illinois , urbana - champaign , illinois 61801 + @xmath23carleton university , ottawa , ontario , canada k1s
5b6 + and the institute of particle physics , canada + @xmath24mcgill university , montral , qubec , canada h3a
2t8 + and the institute of particle physics , canada + @xmath25ithaca college , ithaca , new york 14850 + @xmath26university of kansas , lawrence , kansas 66045 + @xmath27university of minnesota , minneapolis , minnesota 55455 + @xmath28state university of new york at albany , albany , new york 12222 + @xmath29ohio state university , columbus , ohio 43210 + @xmath30university of oklahoma , norman , oklahoma 73019 + @xmath31university of pittsburgh , pittsburgh , pennsylvania 15260 + @xmath32purdue university , west lafayette , indiana 47907 + @xmath33university of rochester , rochester , new york 14627
|
one of the basic premises in the polymer physics is that the bonds connecting neighboring monomer units are uncorrelated beyond the persistent length along the chain @xcite .
this property allows us to consider only a flexible chain as long as we are interested in large scale properties of a polymer chain that is much longer than its persistent length . in this sense , it was a little embarrassing to realize that there is actually a long range correlation in the bond orientation of a polymer chain , and that the bond - bond correlation function decays not exponentially but as the power law @xcite .
the traditional picture for the bond - bond correlation is based on a simple calculation for the polymer with a fixed bond angle around freely rotating bonds . in this case
, one can calculate the bond - bond correlation function explicitly to show the exponential decay with a persistent length @xcite .
the existence of the long range correlation , however , has been pointed out @xcite , and it was demonstrated recently that the power law behavior is induced in the bond - bond correlation through the interaction between monomers separated by a long distance in the curvilinear coordinate along the chain @xcite .
the power law in the bond - bond correlation holds not only for an excluded volume chain , but also for a chain in a melt and in a theta solvent , where a polymer chain is supposed to behave as an ideal chain @xcite .
this has been confirmed both by numerical simulations and by theoretical analyses .
this deviation from the ideality of a chain in a melt has been seen also in the intramolecular structure factor .
for the ideal chain , the structure factor decays as @xmath3 in the intermediate range , @xmath4 , where @xmath2 and @xmath5 are the gyration radius and the bond length , respectively .
this @xmath3 decay comes from the fractal dimension of the ideal chain configuration .
the intramolecular structure factor of the polymer chain in a melt has been studied numerically and theoretically , and it has been found that there exists a substantial deviation from the ideal chain behavior @xcite . in this paper , we study the structure factor of a single polymer molecule in a theta solvent , that is another situation where a polymer chain is supposed to become ideal effectively . in a melt , the interaction between monomers is screened by the existence of other polymer chains , and the excluded volume effect is canceled exactly by the induced attraction due to the incompressibility of the system @xcite while the interaction in a theta solvent is being adjusted by some fine tuned external parameter , such as temperature , so that the excluded volume effect is compensated by the attractive part of the interaction . we study how this fine tuning of the parameter may affect on the virtual ideality of the structure factor .
this paper is organized as follow .
after quickly reviewing how the long range correlation comes into the bond - bond correlation in sec.2 , the model and the method of our simulations are described in sec.3 , and the simulations results are given in sec.4 .
the theoretical analysis is outlined in sec.5 and the results are discussed in connection with those for melt in sec.6 .
detailed expressions of the theoretical analysis are given in appendix .
let us quickly review how the long range correlation should arise in the bond - bond correlation along a polymer chain @xcite .
we consider a single polymer that consists of @xmath6 monomers .
let @xmath7 ( @xmath8 ) be the position of the @xmath9-th monomer in it , and the bond vector is denoted as @xmath10 we define a subchain as a part of the chain , and introduce the subchain vector as @xmath11 now , we assume that the bond - bond correlation depends only on the chemical distance between the bonds : @xmath12 where @xmath5 is the average bond length and @xmath13 denotes the ensemble average .
then , the size of the subchain does not depend on @xmath9 , and we have @xmath14 in the large @xmath15 approximation .
this gives @xmath16 with @xmath17 if the subchain size scales as @xmath18 with the exponent @xmath19 .
we have @xmath20 for the excluded volume chain , where @xmath21 . for the ideal chain with @xmath22 , eq.([bond - corr - scaling ] )
gives @xmath23 as it should , but for the case of apparent ideality of a polymer chain in a melt or a theta solvent , we have @xmath24 because of the correction term , @xmath25
we perform monte carlo simulations of the bond fluctuation model ( bfm ) on the three dimensional cubic lattice @xcite .
a polymer chain consists of @xmath6 monomers , and the @xmath26-th monomer is located at the center of a cubic cell @xmath27 , occupying the cell with 8 vertices at lattice sites .
the bond length between consecutive monomers along the chain should be in the range @xmath28 $ ] with the exception of @xmath29 .
the @xmath26-th and @xmath30-th monomers that are not consecutive along the chain , namely @xmath31 , interact each other through the `` quasi - lennard - jones '' potential energy @xcite @xmath32 where @xmath33 is the distance between the interacting monomers , and @xmath34 and @xmath35 are the boltzmann constant and the temperature , respectively .
the dimensionless parameter @xmath36 is proportional to the inverse temperature and characterizes the interaction @xcite .
we perform monte carlo simulations , using metropolis method along with the slithering snake algorithm to accelerate the relaxation towards equilibrium @xcite . in our system , there is only one polymer chain with @xmath6 monomer units .
one monte carlo step consists of @xmath6 trials of random displacement to one of the nearby sites for randomly chosen monomers , followed by @xmath6 slithering snake trials .
the quantities we study are the radius of gyration @xmath2 , @xmath37 the intrachain structure factor @xmath38 , @xmath39 and the bond - bond correlation @xmath40 averaged over the chain , @xmath41 with the average bond length @xmath5 , @xmath42 where the angular brackets represent the ensemble average .
note that the radius of gyration can be expressed using the bond - bond correlation by a similar equation as eq.([r(s)-p(r ) ] ) : @xmath43 .
\label{r_g - p}\ ] ]
the scaled radius of gyration @xmath44 are plotted against @xmath45 with @xmath46 , @xmath47 0.5 , and @xmath48 for various values of @xmath6 the inset shows the original plot of @xmath49 vs @xmath36 . , width=302 ]
first , we have to determined the theta point for our model .
it is often defined as the point where the second virial coefficient vanishes . in our simulations
, we define the theta point as the point where the radius of gyration @xmath2 behaves as an ideal chain , @xmath50 in the large @xmath6 limit .
the interaction parameter @xmath51 at the theta point is determined numerically from the data of @xmath2 for various values of @xmath36 and chain length @xmath6 by fitting them to the finite size scaling form @xmath52 with the two exponents , @xmath53 and @xmath54 .
the function @xmath55 is the scaling function that satisfies @xmath56 , and @xmath57 is the length scale proportional to the average bond length .
fig.[fig : r_g ] shows that the data fit well to the scaling form ( [ r_g - n ] ) with @xmath58 are plotted against @xmath15 for @xmath59 for various @xmath36 .
, width=302 ] where the bond - bond correlation shows the ideal behavior , @xmath60 , are plotted against @xmath61 . the linear extrapolation to
the infinite @xmath6 shows the theta point , @xmath62.,width=302 ] as we have discussed , the bond - bond correlation function decays as @xmath60 at the theta point .
fig.[fig : bond - corr ] shows the bond - bond correlation function @xmath40 for various temperatures near the theta point for @xmath59 .
it shows that the data fit to the form @xmath63 that is consistent with the theoretical results by shirvanyants , _ et al .
the theta point may be defined as @xmath64 where the bond - bond correlation decays as @xmath60 , but it turns out that @xmath65 depends on @xmath6 .
the @xmath6-dependence of @xmath65 plotted in fig.[fig : beta_c ] shows @xmath66 with the theta point in the infinite @xmath6 limit @xmath62 .
this value is close enough to the previous estimate of @xmath46 by the scaling plot of @xmath2 . in the rest of the paper , we use @xmath46 for the theta point .
note that the @xmath6-dependence of @xmath65 given by eq.([beta_c(n ) ] ) is consistent with the ideal chain behavior of @xmath2 of eq.([r_g - ideal ] ) at @xmath67 in respect to eq.([r_g - p ] ) . .
the simulation data at @xmath68 for the chain with the length @xmath691024 , 2048 , and 4096 are plotted along with the debye scattering function ( the dashed line ) . for the value of @xmath2
, the values obtained by the simulations are used for the simulation data and @xmath70 for the analytical expressions .
the red curve represent the theoretical estimate , eq.([delta_s(q ) ] ) , with @xmath71 and @xmath72 .
, width=302 ] the structure factor for the ideal chain , @xmath73 , in the large @xmath6 limit is given by @xmath74 with the radius of gyration for the ideal chain @xmath75 and the debye scattering function , @xmath76 in the intermediate length scale , i.e. , @xmath77 , this decays as @xmath78 the @xmath79-dependence comes from the scaling behavior of the ideal chain , therefore , the existence of plateau in the plot of @xmath80 , i.e. , kratky plot , has been considered to be a sign of the ideality in a polymer chain behavior .
fig.[fig : s_sim ] shows the kratky plot of our numerical simulations at the theta point for @xmath81 ; the wave number @xmath82 is scaled by the numerically obtained radius of gyration @xmath2 as @xmath83 . as the number of monomers @xmath6 increases , the curve in the smaller @xmath84 regime tends to converge to a common trend , but it is clearly different from that of the ideal chain ( the dashed line ) .
the kratky plot at the theta point shows a hump around @xmath85 and a dip around @xmath86 , and does not show the plateau as the one would expect for the ideal chain .
their general features are apparently similar to those found for a polymer chain in a melt @xcite .
now , we estimate the structure factor theoretically and compare it with those obtained by the simulations .
suppose @xmath87 be the interaction potential between the @xmath15-th and the @xmath88-th monomers with @xmath89 .
then the structure factor eq.([s(k ) ] ) is written as @xmath90 where @xmath91 and @xmath92 represents the statistical average for the ideal chain .
we now employ the approximation @xmath93 using the mayer function @xmath94 then up to the first order in @xmath95 , the correction in the structure factor @xmath96 is given by @xmath97 .
\label{delta_s - expand}\end{aligned}\ ] ] ) .
, width=302 ] there are four types of contribution in eq.([delta_s - expand ] ) depending upon the relative positions of @xmath9 , @xmath98 , @xmath88 , and @xmath15 ( fig.[diagrams ] ) . adopting the bead - spring model for the ideal chain average @xmath99 with the average bond length @xmath5 , and employing the further approximation valid for @xmath100
, we obtain the expression @xmath101 up to the second leading order in @xmath6 .
the functional forms for @xmath102 and @xmath103 are given in appendix .
the dimensionless parameters , @xmath104 characterize the interaction .
the parameter @xmath105 is twice of the second virial coefficient for the unlinked monomer gas @xcite , and is supposed to be close to zero for the theta solvent @xcite .
note that @xmath105 comes into @xmath106 as @xmath107 . )
are plotted with @xmath72 and @xmath71 , and 0 .
, width=302 ] behavior of the structure factor after subtracting @xmath108 with @xmath109 .
, width=302 ] fig.[fig : s_theor](a ) shows the kratky plots for the theoretical structure factor @xmath110 with @xmath111 by eq.([delta_s(q ) ] ) with @xmath72 and @xmath71 along with the curve for @xmath72 and @xmath112 , the debye function , and the simulation data for @xmath113 .
one can see that the curve for @xmath112 is almost proportional to the debye function and can not be fitted to the simulation data .
fig.[fig : s_theor](b ) shows @xmath114 with the simulation data for various values of @xmath6 .
the data converge to the theoretical curve quite well as @xmath6 increases . in fig.[fig :
s_asym ] , the large @xmath84 behavior of the structure factor are plotted in the logarithmic scale , after subtracting the leading order term of @xmath108 with @xmath109 .
it shows clearly that the second leading order is @xmath115 , which is consistent with the asymptotic expression we obtained in eq.([a.delta_s - large - q ] ) .
our findings are summarized as follows : ( i ) the theta point for the finite chain @xmath65 , where the bond - bond correlation decays as @xmath60 , depends on the chain length @xmath6 , and in the infinite chain length limit it converges to the theta point @xmath51 determined by the scaling behavior of the radius of gyration .
( ii ) the structure factors at the theta point @xmath51 obtained by the numerical simulations are distinctively different from that of the ideal chain , i.e. , ( ii - a ) in the intermediate range of @xmath116 , kratky plot of the structure factor shows a hump and a dip , and ( ii - b ) in the larger @xmath82 range , @xmath117 , the structure factor decays as @xmath79 with the next order term of @xmath118 .
( iii ) in the large @xmath6 limit , numerically obtained structure factors fit well to the simple perturbation expression up to the first order of the mayer function with the fitting parameters @xmath71 and @xmath72 . in comparison with the case of a melt
, the obtained structure factors at the theta point apparently resemble the ones of a melt in the existence of a hump and a dip @xcite . in the large @xmath82 region , however , the structure factors for the two cases differ ; in the case of melt , it has been shown by numerical simulations that structure factors decays as @xmath118 for the large @xmath82 region @xcite , in contrast with the present case , where we found the @xmath79 decay with the positive @xmath118 correction .
the @xmath118 decay in the melt was interpreted as a result of renormalization from the @xmath119 term obtained by the one - loop approximation @xcite while the @xmath79 term and the @xmath118 term at the theta point directly correspond to the @xmath120-term and the @xmath121-term , respectively , in our theoretical expression . in the real space
, the @xmath79 decay of the structure factor in large @xmath82 means the @xmath122 density correlation of the ideal chain in short @xmath88 .
the existence of the positive @xmath118 correction and the hump around @xmath0 implies that the density correlation does not decay as fast as that of the ideal chain around @xmath123 .
the simulated structure factors here at the theta point resemble those of a melt , but it is intriguing that we need to set @xmath124 for our theoretical expression of the structure factor to fit to the numerical results at the theta point @xmath51 .
the theta point is often considered as the point where the second virial coefficient @xmath125 vanishes .
actually , if we set @xmath112 , the structure factor shows kratky plateau and looks pretty much like the ideal one ( the dotted line in fig.[fig : s_theor](a ) ) , but this can not be fitted to the simulation results by adjusting @xmath120 alone even with an arbitrary factor .
theoretically , if we consider higher order corrections , the theta point should correspond to the vanishing point of the interaction parameter @xmath126 @xmath127 where @xmath128 is the excluded volume parameter with the additive renormalization by the higher order cluster contributions , and may be given in the form @xmath129 with a constant @xmath130 .
however , such renormalization effect could be partially taken into account by replacing @xmath105 with @xmath128 , and one would expect the structure factor at the theta point should be given by the one with @xmath131 in our expression . in the bond - bond correlation @xmath40 , we observed the analogous @xmath6-dependence , which can be interpreted in the same way .
shirvanyants et al .
@xcite have obtained the theoretical expression corresponding to eq.([p(s ) ] ) in a similar approximation to the one we employed for @xmath38 , and showed that the coefficients for the @xmath132 and the @xmath60 terms in eq.([p(s ) ] ) , i.e. @xmath133 and @xmath134 , are proportional to the parameters @xmath105 and @xmath120 , respectively .
the observed @xmath6-dependence in @xmath65 as eq.([beta_c(n ) ] ) in our simulations should be a result of higher order effects , that could be obtained by replacing the coefficient of @xmath132-term with the renormalized one as eq.([ren - b ] ) . regarding the origin for the deviation of the structure factor from the debye function
, an obvious possibility could be that inaccurate estimate for the theta point @xmath51 , i.e. our estimate @xmath46 is not close enough to the theta point for the structure factor to be debye - like even though the two independent estimates from the radius of gyration @xmath2 and the bond correlation coincide with each other within our numerical precision .
this might happen due to the slow convergence caused by the logarithmic correction at the theta point .
another possibility , perhaps more interesting one , would be that the tricritical fluctuations at the theta point produce a non - trivial contribution to the structure factor .
the authors thank dr .
takahiro sakaue for critical reading of the manuscript .
in the appendix , we describe the calculation of eq.([delta_s - expand ] ) to obtain the explicit expressions for @xmath102 and @xmath103 in eq.([delta_s(q ) ] ) .
the contribution from the diagram ( i ) in fig.[diagrams ] is given by @xmath135 \nonumber \\
& \equiv & { 2\over n}\sum_{1\le s < m < n < r\le n } h(q ; r - s , n - m),\end{aligned}\ ] ] where @xmath136 with the free propagator @xmath137 .\ ] ] @xmath138 can be estimated as @xmath139 -1 \right ) f(\mib r ) \nonumber\\ & \approx & \left({3\over 2\pi}\right)^{3/2 } \exp\left[-{1\over 6}q^2 a^2 l_2\right ] { 1\over l_1^{3/2 } } \biggl [ -\left ( \exp\left[{q^2a^2\over 6}{l_2 ^ 2\over l_1}\right]-1 \right ) b \nonumber\\ & & \hspace{1 cm } -\biggl\ { { 3/2\over l_1 } \left ( \exp\left[{q^2a^2\over 6}{l_2 ^ 2\over l_1}\right]-1 \right ) % \nonumber\\ & & \hspace{2 cm } + { 1\over 6}q^2 a^2 \exp\left[{q^2a^2\over 6}{l_2 ^ 2\over l_1}\right ] \left({l_2\over l_1}\right)^2 \biggr\ } a \biggr].\end{aligned}\ ] ] in the last expression , we have expanded the cosine up to @xmath140 , which is valid for @xmath100 . for large @xmath6
, we can replace the summation by integral and obtain , to the leading orders in @xmath6 , @xmath143 with @xmath144@xmath145 , \\
g_{ii}(q ) & = & -5\left({3\over 2\pi}\right)^{3/2 } \zeta({3/2 } ) \left [ \left({2\over q^4}+{1\over q^2}\right ) e^{-q^2 } -\left({2\over q^4}-{1\over q^2}\right ) \right ] , \\
g_{iii}(q ) & = & -2\left({3\over 2\pi}\right)^{3/2 } { 1\over q^2 } \int_0 ^ 1 dp \int_0 ^ 1 dl { 1\over \sqrt p } \left [ 1-p + { e^{-q^2(1-p)}-1\over q^2 } \right],\end{aligned}\ ] ] where @xmath146 is riemann s zeta function and @xmath147 . with these functions ,
the correction of the structure function is given by @xmath148 with @xmath149 in the leading orders in @xmath6 .
for large @xmath84 , i.e. , @xmath150 , we have @xmath151 \label{a.delta_s - large - q}\ ] ] with @xmath152 , and for small @xmath84 , i.e. , @xmath153 , @xmath154 with @xmath155 . from eq.([a.delta_s - small - q ] ) , the correction for the radius of gyration @xmath2 is obtained as @xmath156 r_{g0}^2 .\ ] ]
|
we study the configurational properties of single polymers in a theta solvent by monte carlo simulation of the bond fluctuation model .
the intramolecular structure factor at the theta point is found to be distinctively different from that of the ideal chain .
the structure factor shows a hump around @xmath0 and a dip around @xmath1 in the kratky plot with @xmath2 being the radius of gyration .
this feature is apparently similar to that in a melt .
the theoretical expression by the simple perturbation expansion to the first order in terms of the mayer function can be fitted to the obtained structure factor quite well , but the second virial coefficient can not be set to zero .
|
actual problems of science , engineering and society can be so complex , that their mathematical portrait requires more than three dimensions .
quantum world gives us a perfect example of essentially high dimensional systems , described by a joint wavefunction ( or density matrix ) of all particles .
a simple system of @xmath0 spin-@xmath1 particles is an entanglement of @xmath2 possible states , and should be described by the same amount of numbers , which creates out - of - memory errors on a typical workstation for @xmath3 even with a brute force of modern supercomputers , standard numerical methods can not honestly simulate protein - size molecules ( @xmath4 @xmath5 ) , since the complexity and storage explode exponentially with @xmath6 to overcome this problem , known as the _ curse of dimensionality _ , we use data - sparse representations for high - dimensional vectors and matrices , and develop special algorithms to work with them . proposed in 1992 , the _ density matrix renormalization group _ ( dmrg ) algorithm @xcite and the _ matrix product states _ ( mps ) formalism @xcite suggest to represent a wavefunction @xmath7 in the following tensor - product form @xmath8 in numerical linear algebra this format was re - discovered as the _ tensor train _ ( tt ) decomposition @xcite . a single tt core ( or _ site _ ) @xmath9 $ ] is described by @xmath10 numbers , where @xmath11 denotes the number of possible states for the @xmath12th particle ( the _ mode size _ ) , and @xmath13 is the tt rank ( or _ bond dimension _ ) .
the total number of representation parameters scales as @xmath14 @xmath15 @xmath16 and is feasible for computations with @xmath17 to the dmrg algorithm was originally proposed to find the _ ground state _ , i.e. the minimal eigenpair of a hermitian matrix @xmath18 this problem is equivalent to the minimization of the rayleigh quotient @xmath19 substituting @xmath20 with @xmath21 and applying the same algorithm , we can solve linear systems @xmath22 with hermitian positive definite matrix @xcite .
this framework can be extended to a broad class of problems .
since @xmath7 is a huge high - dimensional vector , the solution is sought in the structured format with some tt ranks @xmath23 defined _ a priori _ or chosen adaptively .
the simultaneous optimization over all sites is a highly nonlinear and difficult problem .
as it is usual in high - dimensional optimization , we substitute it by a sequence of partial optimizations , each over a particular ( small ) group of variables . for our problem , it is natural to group the variables according to the tensor format , e.g. optimize over the components of a single site @xmath24 at a time .
the tt format is linear in each site , i.e. @xmath25 where @xmath26 is the @xmath27 _ frame matrix _ , which linearly maps the elements of @xmath24 to the full vector @xmath28 this turns every partial optimization into a local problem of the same type , as the original one , @xmath29 where @xmath30 is the @xmath31 _ reduced matrix _ , which inherits the properties of @xmath32 i.e. is hermitian . since the frame matrix @xmath26 has a structured tt representation ( which is the same as with @xmath24 substituted by the identity matrix ) , the reduced matrix @xmath33 can be assembled avoiding the exponential costs .
finally , introducing simple orthogonality conditions for all sites but @xmath34 we can make the whole matrix @xmath26 orthogonal @xcite . as a consequence , @xmath33 becomes better conditioned than @xmath32 and the reduced functional writes @xmath35 for the ground state problem , and @xmath36 for the linear system .
each optimization is now a classical problem of a tractable size , that can be solved by classical algorithms of numerical linear algebra .
each local step finds @xmath37 where the subspace @xmath38 is of dimension @xmath10 , see fig [ fig:3d ] ( left ) . here and later by @xmath39
we denote the subspace of columns of a matrix @xmath40 if tt ranks are fixed , the local convergence of such scheme can be analysed using standard methods of multivariate analysis @xcite .
however , in numerical practice the tensor ranks of the solution are not known in advance , and fixed - rank optimization with wrong ranks would not be efficient .
the dmrg scheme with variable tt ranks is more advantageous , but the theoretical analysis is even more difficult .
when we allow tt ranks to grow , the dimensions of subspaces @xmath41 grow as well , and we can use different strategies to expand the subspaces . originally , the one - site dmrg scheme ( dmrg1 ) increased the rank @xmath13 by adding ( random ) orthogonal vectors to @xmath41 , but this algorithm often got stuck far from the ground state .
the problem was solved using _ two sites _ instead of one in the optimization step @xcite .
the _ two - site _ dmrg algorithm ( dmrg2 ) merges blocks @xmath24 and @xmath42 and solves the local optimization problem in @xmath43 see fig .
[ fig:3d ] ( right ) . here
@xmath44 is the @xmath45 matrix , which has the same tt representation as with blocks @xmath24 and @xmath46 replaced by the identities .
the dmrg2 converges remarkably well ( and is in fact a method of choice ) for 1d systems with short range interactions , but the cost is approximately @xmath47 times larger than in the dmrg1 . for systems with long - range interactions two neighboring sites do not provide sufficient information , and dmrg2 can stagnate as well .
to simulate such systems faster and more accurately , better methods to choose search subspaces are required .
the _ gradient direction _ is central in the theory of optimization methods , and many algorithms use the gradient or its approximate surrogates . in 2005 , s. white proposed the _ corrected one - site _ dmrg algorithm ( dmrg1c ) , which adds auxiliary direction to improve the convergence and reduce the computational cost , see @xcite and ( * ? ? ?
* sec 6.3 ) for more details . in this paper
we compare the dmrg1c with the _ alternating minimum energy _ ( amen )
the amen algorithm was recently proposed in @xcite for the solution of linear equations , and the version for the ground state problem appears immediately when we choose @xmath20 as a target function . in the next section
we compare the ideas and implementation aspects of both methods and explain the motivation behind amen from numerical linear algebra perspective . in sec .
[ sec : num ] we reproduce a numerical experiment of s. white from @xcite , and demonstrate that amen can solve it better than dmrg1c .
both dmrg1c and amen combine the local optimization with the step that injects the auxiliary information .
both algorithms are _ local _ , i.e. modify only one block @xmath24 at a time ( cf . the non - local `` als@xmath48 '' algorithm in @xcite ) .
both methods sequentially cycle over tt blocks ( @xmath49 ) . in the following we assume that was just solved for @xmath34 and
consider the step that corrects @xmath24 before the optimization passes to the next block @xmath50 this step does not change the vector @xmath51 ( for amen ) , or perturbs it slightly ( for dmrg1c ) , and therefore has a minor direct effect on @xmath52 however , it inserts additional directions to @xmath53 that improves the convergence of @xmath20 to its global minimum .
it is crucial how exactly the block @xmath24 is modified , and which vectors end up in @xmath54 after that . in the following we discuss these details , which constitute the main difference between the dmrg1c and the amen .
following the _ power iteration _ method , the dmrg1c algorithm of s. white targets in addition to the solution @xmath7 the first krylov vector @xmath57 the amen algorithm uses the gradient direction @xmath58 in exact arithmetics this makes no difference , since @xmath59 in practical computations both @xmath60 and @xmath61 are perturbed by inevitable machine rounding errors , perturbations associated with the use of tensor format , and additional errors which appear when a surrogate formula ( like ( * ? ? ?
* eq . @xmath62 ) ) is used to speed up the computations .
the dmrg1c algorithm is derived from perturbation arguments , valid in the vicinity of the minimum of @xmath52 when @xmath7 approaches the ground state , the angle between @xmath7 and @xmath63 vanishes , and any perturbation in @xmath64 yields a random new direction .
this creates a certain gap between the theory supporting the dmrg1c , and the practice . following the _
steepest descent _
algorithm , the amen uses orthogonal vectors @xmath61 and @xmath65 and @xmath66 is much more stable to perturbations of @xmath67 ( in general , the krylov vectors @xmath68 form an extremely unstable basis , and orthogonalization is crucial . ) the steepest descent algorithm with @xmath61 substituted by @xmath69 converges as long as @xmath70 for the linear systems this fact is elegantly proven in @xcite , and the convergence rate of perturbed method is estimated .
an eigenvalue counterpart follows similarly , and the rate of convergence in @xmath71 can be estimated from the spectral range of @xmath18 this makes the approach implemented in the amen algorithm preferable both theoretically and in practice .
the computation of full vectors @xmath63 and @xmath73 is not possible due to their exponentially large size . since @xmath7 and @xmath74 are both in tt format , we can avoid the curse of dimension and represent @xmath63 and @xmath73 by the tt format
. however , the tt ranks of @xmath64 can be as large as product of tt ranks of @xmath74 and @xmath65 which makes the calculations difficult . to reduce these costs ,
s. white suggests in the * dmrg1c * the following scheme .
the tt format is divided in two parts : left blocks ( number @xmath75 ) are referred to as _ system _ , and right blocks ( @xmath76 ) as _ environment_. the tt format for the matrix @xmath74 is written accordingly , @xmath77 or shortly @xmath78 similarly , eq . reduces to @xmath79 the targeting of @xmath63 is substituted by targeting of all @xmath80 although in general @xmath81 it can be argued that the set @xmath82 contains a sufficient subspace information . to show this , we write @xmath83 and consider vectors @xmath60 and @xmath84 as system - by - environment matrices @xmath85 and @xmath86 of size @xmath87 now @xmath88 where @xmath89 contains the coefficients of the required linear combination in the exact arithmetics the _
system_-related components of @xmath60 belong to @xmath90 each @xmath84 is easier to compute than @xmath91 because it does not depend on the environment part @xmath92 the total dimension of @xmath93 grows in each step , and to keep tt ranks and storage moderate , we have to truncate it .
the approximation step in the dmrg1c replaces @xmath93 with a subspace @xmath94 of a smaller dimension , using a classical _ singular value decomposition _ ( svd ) , or schmidt decomposition technique .
the _ dominant subspace _
@xmath72 is spanned by the first singular vectors of the matrix @xmath95 where all target vectors are concatenated with empirically chosen weighting coefficients @xmath96 .
the method assumes that the vector @xmath63 is likely to belong to @xmath97 this assumption makes perfect sense if @xmath60 is a random sample from @xmath93 for a random @xmath98 , @xmath99 is more likely to end up in the dominant subspace of @xmath40 however , the target vector @xmath63 does not belong to @xmath72 in general , for any choice of weights @xmath100 the reason is that @xmath60 depends crucially on @xmath101 whereas this information is dropped for the sake of faster computations in @xmath84 and hence @xmath93 and @xmath72 . selecting @xmath89 in , we may come across any vector in @xmath93 , even the smallest singular vector .
that is , for each choice of @xmath102 and @xmath7 there is a _ counterexample _ of a hamiltonian , for which the slightest truncation of @xmath93 loses the system - related part of the target vector @xmath57 the * amen * approximates @xmath56 into its own tt format using any compression tool .
either the svd based technique , which computes the approximation @xmath103 up to any prescribed tolerance @xmath104 , or a faster ( but heuristic ) _ alternating least squares _ ( als ) method may be used . in any case
, we may generate an approximation @xmath69 with a desired accuracy , which guarantees the convergence of the steepest descent method with the imperfect direction @xmath105 this fact provides the theoretical bounds for the global convergence rate of the whole amen scheme , similarly to @xcite .
the last but not the least detail is how exactly the information about the auxiliary direction is injected in the algorithm . to show this in isolation from the other dissimilarities outlined above , we assume that in both methods we target in addition to @xmath7 _ only one _ vector @xmath106 to simplify the presentation we also consider the @xmath107 case , and write @xmath108 and @xmath109 where `` @xmath110 '' and `` @xmath111 '' denote the first and the second blocks , respectively .
the dmrg1c _ averages _ the subspaces @xmath112 and @xmath113 by computing the dominant subspace @xmath114 of the gram matrix as follows , @xmath115 where @xmath116 as shown in previous subsection , this procedure does not guarantee that @xmath7 or @xmath117 ends up in @xmath118 unless @xmath119 the tt core @xmath120 is replaced by the vectors of @xmath121 that introduces a @xmath122 perturbation to @xmath7 and probably worsen @xmath52 it is clear though that @xmath123 should vanish when we approach the exact solution , but the general recipe is not known .
the amen avoids the outlined difficulties by _ merging _ @xmath124 and zero - padding the second block .
values of @xmath7 and @xmath20 are preserved , no rescaling is required , and both @xmath125 the downside is that we choose @xmath126 each time we expand the subspaces . however , when we use the approximate gradient direction @xmath127 the low - rank @xmath69 usually suffice , e.g. with @xmath128 we can also truncate the tt - ranks at the end of each iteration and control the perturbation to @xmath52
following s. white @xcite , we consider the spin-1 periodic heisenberg chain , @xmath129 where @xmath130 , are the @xmath131 pauli matrices for spin-@xmath132 particles .
the number of spins @xmath0 is set to @xmath133 , i.e. the wavefunction belongs to the @xmath134-dimensional hilbert space .
this example is particularly illustrating , since the mismatch between the linear tt model and the cycle structure of complicates the problem the solution has large tt ranks , and both the one and two site dmrg converge slowly .
the way how the tt ranks are chosen during the algorithm is also very important .
we first adopt the rank selection strategy from @xcite , and compare the dmrg2 , the dmrg1c and the amen algorithms .
the results are shown in fig .
[ fig : heisen ] ( top left ) , which overlays ( * ? ? ?
3 ) with the amen behavior . in fig .
[ fig : heisen ] ( top right ) the convergence of @xmath135 to the reference value @xmath136 ( computed in @xcite by the dmrg1c with tt ranks @xmath137 ) is given w.r.t . the cumulative cpu time .
we see that both dmrg methods correctly reproduce the experiment from @xcite : the two - site dmrg stagnates at a high error level , while the corrected dmrg converges significantly faster .
the amen method manifests practically the same efficiency .
since it searches in a larger subspace , it is even more accurate w.r.t .
iterations , but becomes slightly slower during the optimization of inner tt blocks .
however , letting it to increase the ranks ( each fourth iteration ) yields sharper error decays . to free the algorithm from tuning parameter , we prefer to choose the ranks adaptively to the desired accuracy . with this
we also avoid artificial rank limitation , which pollutes the convergence .
therefore , in the second experiment we use the same algorithms but perform the truncation of tt blocks via the svd using the relative frobenius - norm accuracies @xmath138 and @xmath139 .
the results are shown in fig .
[ fig : heisen ] ( bottom ) .
we see that when ranks are chosen adaptively , the amen rapidly becomes faster than the other algorithms .
even the dmrg1c stagnates relatively early , since the correction @xmath140 contaminates the dominant basis of the ground state .
moreover , since the @xmath141-truncation eliminates the correction if @xmath142 , it is worthless to decrease the scale @xmath123 ( cf .
fig [ fig : heisen ] , top left ) . both the adaptivity and speed speak in favour of such truncation :
the same accuracy levels are achieved several times faster than in the fixed - rank experiment ( e.g. 10 vs. 100 sec . for @xmath143 and @xmath138 ) .
larger time spent by amen in the latter iterations is compensated by a significantly better accuracy , which is close to the optimal level @xmath144 .
height 2pt depth -1.6pt width 23pt , _ alternating minimal energy methods for linear systems in higher dimensions .
part ii : faster algorithm and application to nonsymmetric systems _
, arxiv preprint 1304.1222 , 2013 .
|
given in the title are two algorithms to compute the extreme eigenstate of a high - dimensional hermitian matrix using the tensor train ( tt ) / matrix product states ( mps ) representation .
both methods empower the traditional alternating direction scheme with the auxiliary ( e.g. gradient ) information , which substantially improves the convergence in many difficult cases .
being conceptually close , these methods have different derivation , implementation , theoretical and practical properties .
we emphasize the differences , and reproduce the numerical example to compare the performance of two algorithms . _
keywords : _ high dimensional problems , dmrg , mps , tensor train format , extreme eigenstate .
_ msc : _
15a18 , 15a69 , 65f10 , 65f15 , 82b28 , 82b20 _ pacs : _
02.10.xm , 02.60.dc , 75.10.pq , 05.10.cc
|
every animal must eat in order to survive . for certain predator species this necessarily implies to chase and bring down a sufficient amount of prey .
with predators always on the lookout for food , prey must constantly be on the alert . while scattering and zigzagging to confuse the predator is a popular method of herd animals to escape @xcite , if the escape paths are not well co - ordinated individual prey
may also block each other .
the self volume effect is also relevant in the hunt of killer cells ( macrophages , for instance ) in biological organisms attacking bacteria colonies or biofilms . in this paper
we study the influence of self volume effects on a herd of non - communicating prey with the autonomy of taking decisions on the run , as quantified by the typical time to catch a prey . in the study of the dynamics of predator - prey systems
one is generically interested in the likelihood for the survival of the prey as a function of the parameters of the dynamics of both prey and predator .
prototype mathematical models of predator - prey systems are reaction - diffusion models @xcite , in which both species are assumed to move randomly . in one dimension
the survival probability of a diffusing prey exposed to a number of diffusing predators decays as a power law in time @xcite . in two dimensions
the predators catch the prey with probability one , but the mean life time of the prey is infinite .
the survival probability of a lamb in the presence of @xmath0 lions in two dimensions decays logarithmically slowly as @xmath1 @xcite .
in contrast , in dimensions @xmath2 the capture is unsuccessful as a consequence of the transience of random walks @xcite .
other features considered in predator - prey models include finite life times of the species @xcite or the presence of a third party in the form of a repellent obstructing the predator to reach the prey @xcite .
moreover , three groups of species hunting each other were modelled @xcite , owing to the fact that most animal predators are prey of other animals themselves .
finally , effects of safe havens for prey animals may be considered @xcite .
while such continuum random walk models revealed various interesting results it is clear that the escape and pursuit dynamics is at least partially deterministic , that is , both predator and prey hunt or escape in some sense intelligently . a way to improve the mathematical modelling is to assume that both species can see each other within a certain radius of vision and try to use this as an advantage in the escape and pursuit process @xcite .
in such a model the motion consists of random walks which turn into directed ballistic transport once predator and prey spot each other .
as shown in @xcite the probability to escape can be greatly enhanced if the prey can see the predator and has the possibility to run away . during the pursuit
the prey s movement is superdiffusive . in this scenario
a total of three predators may be necessary to catch a single prey @xcite .
predators may also optimise their search by sharing information @xcite . while the assumption of some level of intelligence certainly makes the model more realistic ,
there is still one aspect that has up to now been ignored .
namely , in reality prey are impenetrable bodies .
thus , in an abundant population of prey ( a lion chasing a herd of antelopes , a wolf charging at a flock of sheep , or a killer cell attacking a bacteria colony or biofilm ) the prey species may obstruct each other while trying to escape .
the self volume ( non - phantom ) constraint greatly influences the single species and collective dynamics of random walkers @xcite leading to qualitative differences in the walkers motion .
therefore , the dynamics and survival probability in predator prey systems at intermediate and higher prey densities is expected to be equally affected .
recently a herd of prey chased by a pack of predators including self volume effects was studied @xcite .
as a result the prey s survival time was found to increase if the prey aim for a specific type of clustering . in this paper
we study the success of a single predator hunting a flock of prey on a two - dimensional square lattice with periodic boundary conditions taking into account the prey s self volume .
in addition , both species move intelligently in that they can influence their movement by visual perception within their sighting range .
the paper is structured as follows : first we introduce our model .
next we present the numerical and analytical results for the mean first capture time , which is the time the predator needs to catch the first prey , as a function of prey density and the respective sighting ranges .
we find that the mean first capture time as a function of prey density follows a power law .
the ( non - universal ) exponent depends on the sighting ranges of both predator and prey . for the analytical calculations we split the predator s motion into a diffusing part and a ballistic part , representing the search for the prey and the direct chase , respectively .
we then present a study of the mean distance between predator and nearest prey , which is found to decrease exponentially in time . using the mean distance
we show that we can capture its dynamics in terms of a simple ornstein - uhlenbeck process : the relative motion of predator and nearest prey can thus effectively be viewed to be a random process confined by an harmonic potential . neglecting all other prey , the model parameters such as sighting ranges and prey density can be absorbed into the associated spring constant .
to study the success of a single predator hunting a herd of prey we create an agent - based simulation in which predator and prey move on a two dimensional square lattice with periodic boundary conditions .
each species has its specific sighting range @xmath3 in which it can see the other species as depicted in fig [ model ] .
distances as well as sighting ranges are measured as chemical distances @xmath4 of the added bond lengths , with lattice spacing @xmath5 equal to unity .
the predator starts from the centre of the lattice and the prey are initially randomly distributed excluding the centre of the lattice
such that the occupancy of a single site is less or equal to a single prey .
predators and prey move with the autonomy of decision in the following sense . if no prey is in the sighting range of the predator and , for a given prey , the predator is not in its sighting range , both participants perform a nearest neighbour random walk .
if a prey comes into the sighting range of the predator , the predator chooses a site randomly , subject to the condition that the distance @xmath6 to the prey necessarily decreases . every lattice site that minimises the distance to the prey
is chosen with the same probability , lattice sites that increase the distance can not be chosen .
analogously , if the predator is spotted the prey chooses a site randomly , subject to the condition that the distance to the predator necessarily increases .
if two or more prey are within the same distance to the predator the latter chooses randomly which prey to pursue . due to the self volume of the prey ,
the prey s motion is restricted . in principle , there exist two possible ways to implement the self volume .
either the prey chooses only from empty sites and always executes a jump as long as there is at least one empty nearest neighbour site . or the prey blindly chooses a nearest neighbour site but only jumps if the chosen site is unoccupied ; otherwise , if it is occupied , the prey retains its location .
we chose the latter scenario , as this appears closer to the situation encountered for confused prey or for moving bacteria . using this update strategy
, we simultaneously choose the individual moves for the prey and the predator . in each round of motion
updates for the prey we randomly choose a sequence of individuals , thus avoiding any bias among individuals @xcite . according to this random sequence we then check whether the individual prey are allowed to jump given the actual positions of all other prey .
the motion of the predator takes into account the positions of all prey at the end of the previous update .
once all inidividual jumps of prey and predator are determined , all positions of the entire predator - prey system are updated simultaneously .
the time unit is chosen arbitrarily and relates to the diffusion constant @xmath7 @xmath8 where @xmath9 is the lattice spacing .
corresponds to the diffusion coefficient for a single prey or predator moving on the lattice . ]
after the individual steps of all participants are accomplished , we check if the predator caught a prey . if the first prey is caught the simulation terminates .
the mean first capture time and the mean distance are obtained from @xmath10 realisations and the first passage density is obtained from @xmath11 runs .
we start by quantifying the success of the predator by computing the mean first capture time @xmath12 , that is , the typical time the predator needs to catch the first prey . in mathematical terms
this corresponds to the prey s survival time .
as one can easily imagine the mean first capture time depends crucially on both sighting ranges @xmath13 and @xmath14 as well as on the prey density @xmath15 , where @xmath0 is the number of prey and @xmath16 is the number of lattice sites .
a higher prey density reduces the prey s survival expectation .
one reason is that the probability that initially one prey sits close to the predator is higher and therefore the prey gets spotted earlier .
the second reason is that a chased prey gets trapped more easily if there are more prey that occupy nearest neighbour sites and therefore lead to a frustration of the prey s mobility . in this setup
we distinguish two limiting cases : a single prey ( @xmath17 ) with sighting range greater than two can never get caught , its life time is infinite .
conversely , if every lattice spacing is occupied by a prey ( @xmath18 ) then the predator needs exactly one time step to catch the first prey . for arbitrary densities , as a result of extensive simulations we find from fig .
[ fig : capturetime ] that the mean first capture time as a function of the prey density follows a power law behaviour @xmath19 in which different combinations of sighting ranges lead to different slopes .
furthermore , there appears a crossover between two regimes for larger sighting ranges of the prey , in which we find different slopes for the low and intermediate density range and the high density range ; see , for instance , the square symbols in figs .
[ fig : capturetime ] b ) and c ) . in more detail , while the predator s sighting range only slightly influences the prey s survival , as shown in fig . [
fig : capturetime]a ) , the prey can increase their life expectancy significantly by a finite sighting range of at least two , compare fig .
[ fig : capturetime]b ) , even in the case of a long sighting range of the predator , see fig . [
fig : capturetime]c ) . in both figures
[ fig : capturetime]b ) and [ fig : capturetime]c ) a significant variation at intermediate @xmath3 values is distinct .
we note that a short sighting range of the prey ( @xmath20 ) has no advantage over a vanishing one .
an explanation can be found `` microscopically '' .
there are two possibilities for a prey to get caught .
first , a prey gets stuck and , despite his eyesight , can not evade the encounter with the predator ; or , second , predator and prey simultaneously jump on the same lattice site and collide randomly , see fig . [
fig : randomcollicion ] . with sighting range zero or one
a prey can not foresee a random collision , because the distance decreases instantly from two to zero .
thus , the prey needs at least a sighting range of two to prevent such a situation .
these random collisions further lead to the fact that the predator is more successful with an even sighting range @xmath21 , @xmath22 being an integer number , than with a higher odd one @xmath23 .
since the random collision is a natural and frequent way to get caught , we decided to eliminate these effects by treating only odd sighting ranges .
= 1 ) , depicted by the red dot , can get caught by the predator ( blue cross ) despite his field of vision by random collision , due to simultaneous jumps to the same lattice site.,width=302 ] we note that we did not include error bars in our figures .
a stochastic variable with exponential ( poissonian ) probability density function @xmath24 has the mean @xmath25 and variance @xmath26 .
the mean first capture time presented in this section is the first moment of the exponentially distributed first passage density obtained in section [ section_density ] .
the standard deviation of this poissonian process @xmath27 is equal to the mean @xmath25 , which is indeed confirmed from our numerical results with a sample size of @xmath10 per data point .
repeated simulations produced practically indistinguishable results .
in comparison to an ensemble of non - interacting random walkers self volume effects and the autonomy of decision - making of the participants limit the possibilities of analytical calculations .
we succeeded in calculating the distribution of the time for catching the first prey only in the case of blind prey .
as the results are nevertheless instructive we discuss this case here in some detail . the autonomy to switch the mode of motion of the predator can be included by dividing the process into two subprocesses .
the first one describes the diffusing predator while looking out for a prey .
the second subprocess portrays the direct chase of the prey , which can in fact be considered as a ballistic motion in chemical space such that the predator still has the option of choosing sites in different directions .
the first subprocess describes the random motion of the predator while looking out for a prey .
according to the model during that time the predator performs a nearest neighbour random walk on the lattice .
we are interested in the first passage density function of the predator to find the first prey , that is , until the first prey enters the predator s sighting range .
for simplification we use a continuous radial coordinate and ignore the fact that the participants move on a lattice .
we assume that there exists an effective radius @xmath28 around the predator in which he will not encounter a prey .
this radius has a natural lower bound which is the initial distance between predator and nearest prey ( at time @xmath29 ) , calculated in section [ meandistanceanalytics ] .
if the predator hits this effective radius , he spots the prey and will from there on switch his motion to the direct chase calculated in the next subsection .
we consider the predator as a diffusing particle in two dimensions and calculate his first passage time to escape a sphere with radius @xmath30 . for simplification
we let the particle diffuse between concentric spheres with an inner reflecting boundary at radius @xmath31 , which will later tend to zero , and an outer absorbing boundary at radius @xmath32 , representing the point where the predator spots a prey .
@xmath32 is thus the distance between predator and prey minus the sighting range of the predator .
the predator starts inside the interval @xmath33
. we will later let @xmath34 tend to @xmath31 to capture the predator s starting position correctly .
the diffusing particle can be described by the radial diffusion equation @xmath35 for the probability density function @xmath36 to find the predator at radius @xmath37 at time @xmath38 .
the initial condition we choose as @xmath39 , that is , the particle starts at @xmath40 .
we impose the absorbing boundary condition @xmath41 at @xmath32 and the reflecting boundary condition @xmath42_{r_-}=0 $ ] at @xmath43 .
after laplace transform @xmath44 and with @xmath45 the diffusion equation is reduced to the ordinary differential equation @xmath46 for @xmath47 and @xmath48 this is the modified bessel equation of zero order with known solution @xmath49 for @xmath50 @xcite , where @xmath51 and @xmath52 are the modified bessel functions of first and second kind .
we solve this equation by imposing the continuity condition @xmath53 and the jump - discontinuity @xmath54 where @xmath55 is the solution in the range @xmath47 and @xmath56 is the solution in the range @xmath48 . with the shorthand notations
@xmath57 and @xmath58 @xcite the solution yields in the form @xmath59 if the particle starts at the inner boundary @xmath60 , corresponding to @xmath61 in the reduced coordinates , @xmath62 we calculate the flux through the outer boundary as @xmath63 when @xmath61 approaches zero , we therefore find that @xmath64 where on the right hand side we restored the original variables .
this is but the first passage time density function in laplace space of the predator to spot a prey . from that time
the predator will chase the prey directly , this part being calculated in the next subsection .
the second subprocess , which describes the predator s movement from the moment of spotting the prey until the prey is caught , can be reduced to a one - dimensional problem .
remember that the decision for every step of the predator is constrained by the following rule : the distance to the prey has to necessarily decrease .
for the prey , analogously , the goal is to increase the distance .
consequently after a combined predator and prey step the distance between predator and prey can either stay the same or decrease by one lattice spacing if the chosen site of the prey is already occupied and the prey remains at its site .
the first capture time can thus be calculated exactly from the number of times a prey remains at its location .
a large sighting range of the prey renders the analysis of the chasing process more difficult as all prey try to escape from the predator and will eventually build a cluster that moves away from the predator . due to the random order of the updates , one can not say which of the prey remains sitting .
therefore , we confine ourselves to the case of blind prey . in this case
the prey undergoes normal diffusion and the predator moves constantly towards the prey .
we therefore consider the predator to be a moving cliff towards a diffusing particle , the blind prey .
the survival probability of a diffusing particle in presence of a ballistically moving cliff decays exponentially @xcite , @xmath65 the associated first passage density @xmath66 then becomes @xmath67 with the laplace transform @xmath68 in the long time limit corresponding to a small @xmath69 expansion , we finally get @xmath70 . using the first passage time densities of the subprocesses of search and chase we calculate the total first capture time density function in the next subsection . the distribution of the first capture time is now given by the convolution of results @xmath71 and @xmath70 , @xmath72 which designates the probability that the predator spots the first prey at time @xmath73 and catches the prey during the time span @xmath74 . in laplace space this convolution simplifies to the product @xmath75 .
the inverse laplace transform can be obtained in the long time limit , corresponding to taking @xmath76 .
we thus need to invert @xmath77 where @xmath78 and @xmath79 .
taking the leading terms for small @xmath69 , @xmath80 with @xmath81 , the inverse laplace transform yields the final result , @xmath82 this density of first capture is thus an exponential distribution , where the rate @xmath83 is a function of the prey density and the predator s sighting range .
[ fp ] shows the numerical data of our simulation for the case of blind prey and a short sighting range of the predator ( @xmath84 ) .
the exponential form ( [ eq_scaling_density ] ) agrees quite well with the data over the whole density range .
we now turn to study the dynamics of the mean distance between the predator and the nearest prey in more detail . in fig .
[ meandistance ] our simulation results for this mean distance are plotted for a low prey density @xmath85 in the upper row ( panels a ) , c ) and e ) ) and for an intermediate density @xmath86 in the lower row ( panels b ) , d ) and f ) ) .
the distance decreases exponentially in time except for the case when a blind predator is combined with a low prey density ( fig .
[ meandistance ] a ) ) or with a very good eye - sight of the prey ( fig .
[ meandistance ] b ) ) . in these cases
the distance is approximately constant in the shown time window . in case of identical
sighting ranges the distance between short - sighted species decreases faster than the distance between blind species .
this phenomenon is due to the random collisions explained in section [ section_mfct ] .
as intuitively expected , the distance between the predator and the nearest prey decreases faster in the case of a large sighting range of the predator .
however when the prey s sighting range is large it softens the decay of the distance .
a high prey density also leads to a faster decay of the distance between the predator and the nearest prey , because it implies more prey - prey obstruction events for the chased prey , and with every one such event the distance is reduced by one lattice spacing . a naive model that captures the effective interaction between two diffusive particles such as the predator and the nearest prey turns out to be the ornstein - uhlenbeck process @xcite .
it is defined in terms of the stochastic differential equation @xmath87 with non - negative parameters @xmath88 , @xmath89 , and @xmath90 .
@xmath91 denotes the wiener process @xcite .
the ornstein - uhlenbeck process describes the relaxation of the variable @xmath92 with initial value @xmath93 to the mean value @xmath94 in the presence of gaussian white noise .
the first moment is given by the exponential decay @xmath95+x_0\exp(-ct).\ ] ] comparing the first moment to the observed simulated decay of the mean distance between the predator and the nearest prey ( fig .
[ meandistance ] ) , @xmath96 we see that the mean distance decreases as a special case of the ornstein - uhlenbeck process with vanishing excentricity parameter , @xmath97 .
a popular application of the ornstein - uhlenbeck process in physics is a hookean spring with spring constant @xmath98 , whose dynamics is highly overdamped with friction coefficient @xmath99 in the presence of thermal fluctuations .
therefore we can imagine the predator and the nearest prey to be connected by a hookean spring and being driven by an external wiener noise .
the corresponding mean relaxes to zero .
the equilibrium length of the spring is therefore zero .
the bottom of the corresponding harmonic potential thus represents the capture of the prey by the predator . due to the analogy
, the respective sighting ranges and the prey density affect the stiffness of the spring .
the spring constant is easily related to the decay rate @xmath100 of the mean distance , @xmath101 . as shown in fig .
[ springconstants ] the fitted values for the spring constant display the power law behaviour @xmath102 the spring constant corresponds to the slopes of the functions in fig .
[ meandistance ] , extracted from the exponential fit and plotted as a function of the prey density .
it relates to the mean first capture time discussed in section [ section_mfct ] in the following way .
the exponential decay of the mean distance between predator and nearest prey has a mean life time related to the decay rate @xmath103 since the nearest prey is the one that will get caught , its mean first capture time is related to the mean life time of the mean distance and consequently to the inverse of the decay rate , compare figs .
[ fig : capturetime ] and [ springconstants ] .
we finally mention an analytical approximation for the distance between the predator and the nearest prey . since we want to capture the whole dynamics we first need to determine the initial distance between predator and nearest prey at time @xmath29 . in the simulation
we place the predator in the centre and place the prey randomly around him including the self volume interaction
. then we measure the distance between the predator and the nearest prey . in section [ searching ]
we used an effective radius @xmath104 within which the predator does not encounter a prey .
although we can not calculate this effective radius a natural lower bound is the initial mean distance @xmath105 between the predator and the nearest prey . within this distance
there is no prey present and therefore it is impossible for the predator to encounter a prey .
we determine the initial distance between the predator and the nearest prey on a square lattice with edge length @xmath106 .
the predator sits in the centre of the lattice and the prey are randomly distributed on the remaining @xmath107 sites . as the prey have a self volume , a lattice site can only be occupied by a single prey .
the probability for the distance between predator and nearest prey @xmath108 to be equal to @xmath6 is @xmath109 we then calculate the probability @xmath110 using combinatorics .
the detailed calculation can be found in appendix a. for the probability function of the distance between predator and nearest prey we obtain @xmath111\big/\sum\limits _ { i=1}^{d_{\mathrm{max}}}{n_r(d_i ) \choose n_p } , \label{eq_pdf_initial}\ ] ] where we define @xmath112 as the maximal possible distance between the predator and the nearest prey .
the expectation value of the initial distance from the predator to the nearest prey , @xmath113 then yields in the form @xmath114 the probability distribution of the initial distance to the nearest prey is shown in fig .
[ fig_dmin ] and the related initial mean distance as a function of prey density can be seen in fig .
[ fig_meandistance ] .
we simulated both the initial distance distribution and the initial mean distance between the predator and the nearest prey by placing all participants on the lattice under the model conditions with @xmath10 iterations .
both analytical and numerical results show excellent agreement in figs .
[ fig_dmin ] and [ fig_meandistance ] .
we studied the predator - prey dynamics of a single predator hunting a herd of prey on a square lattice with decision - making species .
while many predator - prey models deal with collective predation @xcite or the search for the optimal number of predators given the number of prey @xcite , we chose a model consisting of one predator and many prey , which is often found in nature .
solitary hunters such as tigers , bears , or sea turtles often have herd animals as their target .
a tiger , for example , hunts a herd of antelopes or a flock of sheep , a bear fishing a salmon out of a swarm , or sea turtles eating jellyfish , shrimp , and fish living in schools .
similarly individual killer cells in biological organisms may attack a colony of bacteria or a biofilm .
a major ingredient of our model is the self volume of the prey , such that no two prey are allowed on a given lattice site .
we showed that in the case of impenetrable prey the predator hunts more successfully if the prey have worse eyesight .
moreover , we found that the predator benefits more from a deterioration of the prey s eyesight than from an improvement of his own eyesight .
while trapping reaction models obtain a minor influence of the prey s long time survival probability by their diffusion constant @xcite we found the prey s sighting range and thereby motion predominating their survival probability . due to self volume interactions the prey are forced to improve their eye - sight , and with a good field of vision can drastically increase their chances of survival even in the range of high densities .
the prey only profit from a sighting range of at least two .
a very short eyesight does not at all improve its survival probability with respect to being blind .
this is attributed to random collisions between predator and prey . using a simplified analytic approach we showed that in the long time limit the first passage density of the predator to catch a blind prey decays exponentially in time with a non - linear dependence of the decay rate on the prey density .
the effective motion during the chase ( described in terms of the distance between the predator and the chased prey ) can be effectively described as a linear relaxation process in an harmonic potential with a stochastic driving where the density and sighting ranges determine the stiffness of the corresponding hookean spring .
all non - linear effects entering the motion due to self volume interactions can thus effectively be described with a single parameter .
there exist a range of further open questions . to imitate natural environment one could extend the dynamics by introducing ( time or sighting range dependent ) waiting times
. one could choose different rates of motion for predator and prey as well or even distribute the rates within the prey to simulate old , sick or infant animals .
additionally , many prey live in herds , so one could let the prey be clustered as the initial condition .
last but not least , communication between the prey is a reasonable thing to assume .
once one of the prey spots the predator , immediately all of them are informed ( similar to stamping of rabbits or the cheeping calls of groundhogs ) , that is , a collective response of prey .
we finally note that random search processes with non - brownian search dynamics are also widely discussed in literature .
while brownian motion is an advantageous process to find nearby targets @xcite , it is known that pure stochastic motion leads to oversampling of the area on longer time scales .
hence , the optimal number of encounters with prey can be found by switching between search modes @xcite .
representative for such a process is for example the intermittent search strategy which combines phases of slow motion , allowing the searcher to detect the target , and phases of fast motion during which targets can not be detected @xcite .
another widely applicable process concerning optimal search strategies are lvy flights , which are based on random walk processes with long- tailed jump length distributions and are known to be an efficient strategy for finding a target of unknown place @xcite . a species which is known to move in lvy patterns are wandering albatrosses @xcite or marine predators as sharks , bony fishes , sea turtles and penguins @xcite .
it would thus be interesting to study effects of self volume in these models as well .
we thank andrey cherstvy for discussions .
we acknowledge funding through an alexander von humboldt fellowship and arrs program no .
p1 - 0002 ( a.g . ) and an academy of finland fidipro grant ( r.m . ) .
we determine the initial distance between the predator and the nearest prey on a two dimensional square lattice with edge length @xmath106 .
the predator sits in the centre of the lattice and the prey are randomly distributed on the remaining @xmath107 sites .
as the prey have a self volume , a lattice site can only be occupied by a single prey .
the probability for the minimal distance between predator and nearest prey @xmath108 to be equal to @xmath6 is @xmath115 we calculate the probability function @xmath116 using combinatorics . if @xmath117 all sites within distance @xmath6 ( up to distance @xmath118 ) must be unoccupied . to obtain the number of these sites we count all sites at exactly distance @xmath6 and add them from distance @xmath119 up to @xmath118 .
the number of sites at distance @xmath6 can be shown to be @xmath120 counting all empty sites within the distance @xmath6 from the predator leads to @xmath121 that is explicitly , @xmath122 due to the predator sitting in the centre there are in general @xmath107 possible sites for the prey to be placed on . under the assumption that the minimal distance is @xmath6 , i.e. , @xmath123 sites are vacant , there are @xmath124 remaining sites for the prey .
the probability for the minimal distance to be greater or equal @xmath6 is the number of possibilities to place the prey at the remaining sites @xmath125 over the possibilities to place the prey at sites greater equal every possible distance ( @xmath119 to @xmath112 ) @xmath126 for the probability function of @xmath108 using eq .
( [ eq_appendix1 ] ) we obtain @xmath127 where we define @xmath112 as the maximal possible distance between the predator and the nearest prey .
it is determined by the number of prey ( due to the self volume of the prey ) and can be calculated by allocating all prey as greatest distance as possible starting at @xmath128 .
then the first fully unoccupied diamond at distance @xmath6 is the maximal possible distance @xmath112 .
there exist the following condition to place all prey @xmath129 , @xmath130 we then get the maximal possible distance as a function of prey , @xmath131 \left\lfloor\frac{\sqrt{1 + 2(n_s - n_p)}-1}{2}\right\rfloor , & n_p\geq n_s/2 \end{array}\right.\ ] ] where @xmath132 is the floor function .
we now obtain the expectation value of the initial distance from the predator to the nearest prey , @xmath133 such that @xmath134
we here present plots depicting the dependence of the parameter @xmath135 from fig .
[ fp ] versus the prey density ( fig . [ fig_lambda ] ) as well as of the scaling exponents @xmath136 and @xmath137 from figs .
[ fig : capturetime ] and [ springconstants ] .
99 r. martnez - garca , j .
m. calabrese , t. mueller , k. olson , c. lpez .
2013 optimizing the search for resources by sharing information : mongolian gazelles as a case study .
lett . _ * 110 * 248106 .
a. ramanantoanina , c. hui , a. ouhinou .
2011 effects of density - dependent dispersal behaviours on the speed and spatial patterns of range expansion in predator - prey metapopulations . _
ecol . model .
_ * 222 * 3524 .
|
we study the degree of success of a single predator hunting a herd of prey on a two dimensional square lattice landscape .
we explicitly consider the self volume of the prey restraining their dynamics on the lattice .
the movement of both predator and prey is chosen to include an intelligent , decision making step based on their respective sighting ranges , the radius in which they can detect the other species ( prey can not recognise each other besides the self volume interaction ) : after spotting each other the motion of prey and predator turns from a nearest neighbour random walk into direct escape or chase , respectively .
we consider a large range of prey densities and sighting ranges and compute the mean first passage time for a predator to catch a prey as well as characterise the effective dynamics of the hunted prey .
we find that the prey s sighting range dominates their life expectancy and the predator profits more from a bad eyesight of the prey than from his own good eye sight .
we characterise the dynamics in terms of the mean distance between the predator and the nearest prey .
it turns out that effectively the dynamics of this distance coordinate can be captured in terms of a simple ornstein - uhlenbeck picture . reducing the many - body problem to a simple two - body problem by imagining predator and nearest prey to be connected by a hookean bond , all features of the model such as prey density and sighting ranges merge into the effective binding constant .
|
the emergence of synchronization in ensembles of dynamical units is a universal phenomenon that is vital to the functionality of many natural and man - made systems @xcite .
in addition to the ability of the individuals that make up such systems to operate in unison , in many instances the particular frequency or velocity with which they evolve is crucial .
for example , the sources and loads that make up power grids must reach consensus to avoid power failures , but reaching a common frequency alone is not enough ; the system is most efficient near a certain reference frequency of approximately @xmath0 - @xmath1 hz , and may fail if the collective dynamics are too far from this range @xcite . in a wide variety of disciplines , from biology and neuroscience to mechanical and electrical engineering
, there are vital systems whose functionality is jeopardized if the collective frequency or velocity differs too much from a given reference frequency ; examples include brain dynamics , cardiac excitation , consensus networks , and coordination of muscle movements in the digestive track @xcite . in the case of cardiac excitation , for instance
, rapid oscilations can give rise to dynamical instabilities that often precede ventricular fibrillation and eventually heart failure .
in the majority of works studying the dynamics of network synchronization , it is often assumed that the collective frequency of the synchronized state is precisely the mean natural frequency of the individual units @xcite . in other words
, the synchronized state reaches an oscillation frequency that is equal to the unweighted average of the oscillation frequencies of the individual elements when acting in isolation , i.e. , uncoupled . in this article
we study the collective frequency of self - organizing systems of oscillators and show that it is _ not _ in general equal to the mean of the individuals natural frequencies .
we find that collective frequency variation is a consequence of the directedness of network and heterogeneity of the dynamics . for networks lacking either ,
e.g. , undirected networks or identical oscillators , we find that that the collective frequency does recover the mean oscillator frequency @xcite .
importantly , systems with directed connections and non - identical agents are ubiquitous @xcite , and therefore collective frequency variation is a fundamental yet unexplored property of real - world self - organizing systems @xcite . to investigate this phenomenon
, we consider the general linearized dynamics of @xmath2 coupled units , @xmath3 , for @xmath4 , given by @xmath5 where @xmath6 is the natural frequency of oscillator @xmath7 , @xmath8 is the global coupling strength , and @xmath9 is the network laplacian matrix .
the entries of @xmath9 are defined @xmath10 , where @xmath11 is the network adjacency matrix and @xmath12 is the in - degree of node @xmath7 .
we also define the out - degree of node @xmath7 , @xmath13 .
we assume the network encoded by @xmath14 to be strongly - connected @xcite . in principle
, our analysis allows the network to be directed and weighted , although unless otherwise noted we will focus on the case of unweighted edges : @xmath15 if a directed link @xmath16 exists , and otherwise @xmath17 .
we note that there are several ways to define a laplacian matrix for directed networks @xcite ; we study a version that is appropriate for the dynamics of interest .
these linearized dynamics represent a versatile description of a wide range of dynamical processes on networks @xcite .
for instance , eq . ( [ eq:01 ] ) can be obtained from linearizing self - organizing systems around the synchronized manifold , for instance the kuramoto model which serves as a model for a wide range of synchronization phenomena including power grid dynamics @xcite , as well as other systems with more general coupling which are utilized in modeling excitable- and reaction - diffusion - type systems @xcite .
this linear relaxation has been found to accurately capture the dynamics of the system , provided that initial conditions are within the basin of attraction of the synchronized state @xcite . in the case of a network of coupled oscillators
, this tends to be particularly robust , capturing the dynamics provided that the overall coupling is not too small in comparison to the spread in the natural frequencies ( which we illustrate in sec .
[ sec5 ] ) .
we study the frequency - synchronized state , given by @xmath18 , and quantify the collective frequency variation by examining @xmath19 , where @xmath20 denotes the collective frequency of the synchronized population and @xmath21 is the mean natural frequency .
we call this difference the _ collective frequency variation_. we show that under generic conditions which are present in most practical application , when the frequencies @xmath6 are non - identical and the in- and out - degrees @xmath22 and @xmath23 are not perfectly balanced , then @xmath24 . however , when the in- and out - degrees match for each node in the network , @xmath25 , then the collective frequency variation vanishes , i.e. , @xmath26 , for any choice of frequencies .
we calculate the collective frequency variation directly from eq .
( [ eq:01 ] ) and show that @xmath19 is given by a weighted average of the natural frequency vector , where the weights correspond to entries of the first left singular vector @xmath27 of @xmath9 that is associated with the trivial singular value @xmath28 .
we find that @xmath27 represents an out - flow centrality measure , and in fact the entries of @xmath27 are often well - approximated by the out - to - in - degree ratio , @xmath29 .
interestingly , the first - left - singular - vector centrality is a reverse analogue of google s pagerank centrality @xcite , which provides a cornerstone to google s ranking of webpages and favors nodes with strong in - flow @xcite .
these findings reveal an interesting and surprising link between synchronization dynamics and pagerank , paving a path for new theoretical exploration and the possibility of applying well - established pagerank methods to synchronization .
we will also demonstrate the presence of collective frequency variation in real - world uk and scandinavian power grid networks .
however we emphasize that our findings fit in a much broader and more interdisciplinary framework .
the remainder of this article is organized as follows . in sec .
[ sec2 ] we derive the collective frequency variation of a network . in sec .
[ sec3 ] we study the range of possible collective frequency variation for a given network structure . in sec .
[ sec4 ] we show that the weights that contribute to the collective frequency admit a centrality that is the reverse analogue of google s pagerank centrality . in sec .
[ sec5 ] we study collective frequency variation in the power grid as a real example . in sec .
[ sec6 ] we conclude with a discussion of our results .
we begin by writing eq .
( [ eq:01 ] ) in vector form , @xmath30 our aim is to calculate the collective frequency of the synchronized population , and therefore we propose the ansatz @xmath31 where @xmath32 is a vector encoding the steady - state value of each @xmath3 in an appropriate rotating frame , @xmath33^t$ ] , and @xmath20 is the collective frequency . to proceed
, we will utilize the pseudoinverse @xmath34 of the laplacian matrix , which satisfies @xmath35 and @xmath36 @xcite . in the undirected case
, @xmath34 can be found using the eigenvalue decomposition of @xmath9 , whereas in the more general case of a directed network , @xmath34 is formulated in terms of the singular value decomposition ( svd ) of @xmath9 . in particular , if @xmath37 , where @xmath38 are the singular values which are ordered @xmath39 and make up the diagonal entries of @xmath40 , and @xmath41 and @xmath42 are the corresponding left and right singular vectors that make up the columns of @xmath43 and @xmath44 , respectively , then the pseudoinverse is given by @xmath45 .
an important distinction between @xmath9 and @xmath34 is that , while @xmath9 maps all constant vectors to zero since its rows sum to zero , this is not generally true of @xmath34 , whose nullspace is nontrivial .
furthermore , the sets of singular vectors @xmath46 and @xmath47 ( appropriately normalized ) each form an orthonormal basis for @xmath48 . proceeding with the analysis
, we insert eq .
( [ eq:03 ] ) into eq .
( [ eq:02 ] ) and rearrange to obtain @xmath49 left - multiplying by @xmath50 , and using that @xmath35 , we find @xmath51 equations ( [ eq:04 ] ) and ( [ eq:05 ] ) thus imply that @xmath52 next , since @xmath28 , the matrix @xmath53 can be simplified to @xmath54 .
finally , we left - multiply eq .
( [ eq:06 ] ) by @xmath55 , rearrange , and subtract @xmath56 from the right- and left - hand sides to obtain @xmath57 where @xmath58 denotes the inner product .
this result is in good agreement with previous research on consensus systems . in particular , by differentiating eq .
( [ eq:02 ] ) with respect to time , using the initial condition @xmath59 , and noting that the first left singular vector and first left eigenvector are equal , we find that our derivation of eq .
( [ eq:07 ] ) provides a complementary derivation of eq .
( 23 ) in ref .
@xcite .
equation ( [ eq:07 ] ) gives the collective frequency variation @xmath19 of a synchronized population as the projection of the natural frequency vector @xmath60 ( shifted to have zero mean ) onto the first left singular vector @xmath27 .
the physical interpretation of eq .
( [ eq:07 ] ) is that the collective frequency variation is a weighted average of the natural frequencies , wherein the weights are proportional to the entries of @xmath27 .
thus , nodes with large entries in @xmath27 contribute more to the collective frequency variation than those with small entries , allowing for non - zero values of @xmath19 provided that the entries of @xmath27 are not identical .
furthermore , we can formulate the full range of collective frequencies for a given network as the maximum of @xmath61 over all choices of @xmath62 with some fixed variance . as we will show below
, the first left singular vector @xmath27 induces a centrality measure for the network that is related to the out - flow of each node .
interestingly , we will show that this centrality is analogous to a `` reverse '' pagerank .
in fact , it is equivalent to google s pagerank centrality for the network obtained by reversing the direction of each link in the original network .
we demonstrate our main result , eq . , with a simple example using two small networks of size @xmath63 , which are illustrated in fig .
[ fig1](a ) and ( b ) .
both networks contain @xmath64 links , yielding a mean in- and out - degree of @xmath65 ; however , in network ( a ) the links are made randomly so the in- and out - degrees at each node are not necessarily equal , while network ( b ) is balanced so that the links are made to satisfy @xmath66 for all @xmath7 , but is still directed . for visual distinction ,
each node s area is proportional to the out - to - in - degree ratio @xmath67 .
next , we draw a set of normally distributed natural frequencies with mean @xmath68 and variance @xmath69 and calculate for each network the collective frequency variation @xmath19 using eq .
( [ eq:07 ] ) for @xmath70 different permutations of these frequencies . in fig .
[ fig1](c ) , we plot the observed density @xmath71 for networks ( a ) and ( b ) ( solid blue and dashed red , respectively ) . in the generic case ,
network ( a ) , where in- and out - degrees are not necessarily equal at each node , we observe a wide range of collective frequencies , while for network ( b ) , where the balance @xmath72 is maintained , the collective frequency is zero in each case , resulting in a delta function @xmath73 .
this example highlights two important properties .
first , the collective frequency variation is intimately linked with the directedness of a network : once the balance @xmath72 is broken , a non - zero value of @xmath19 should be expected .
second , the precise value of @xmath19 depends not only on the network and set of natural frequencies , but the arrangement of natural frequencies ( dynamical heterogeneity ) on the network . therefore , for a fixed network and set of oscillator frequencies , depending on how the oscillators are assigned on the network , the system s collective frequency may either be faster or slower than the mean frequency .
a natural question to ask of a given network is : what is the range of possible collective frequency variations ?
we formalize this by considering for a given network , the magnitude of the maximum collective frequency variation across all frequency vectors with fixed variance @xmath74 , i.e. , @xmath75 . inspecting eq .
( [ eq:07 ] ) , it is straight - forward to see that the collective frequency variation is maximized when the shifted natural frequency vector @xmath60 is aligned with the first left singular vector @xmath27 .
thus , the choices of @xmath62 that maximize @xmath61 with mean @xmath56 and variance @xmath74 are precisely @xmath76 where @xmath77 and the + and - symbols correspond to maximizing and minimizing @xmath19 , respectively ( that is , assuming @xmath78 for each @xmath7 ) .
this yields a collective frequency variation range of @xmath79 and various mean degrees , the collective frequency variation range @xmath75 for @xmath69.,title="fig : " ] and various mean degrees , the collective frequency variation range @xmath75 for @xmath69.,title="fig : " ] to investigate how the range of collective frequency variation depends on network structure , we consider a variety of erds - rnyi @xcite ( er ) and scale - free ( sf ) networks .
er networks are constructed using a link probability @xmath80 that describes the probability of directed link @xmath16 existing .
sf networks are built using the configuration model @xcite for target in- and out- degrees drawn from the distribution @xmath81 for @xmath82 , where @xmath83 is an enforced minimum degree .
the mean degree for er and sf networks can be tuned according to @xmath84 and @xmath85 , respectively . in our experiment , we fix @xmath86 and construct networks of size @xmath87 with various mean degrees and compute the collective frequency range according to eq . with @xmath69 . in fig .
[ fig2](a ) and ( b ) , we plot the results for over @xmath88 er and sf network realizations , respectively ; we denote the mean and standard deviations using the symbols and dashed curves , respectively . for both network families
, the collective frequency variation range tends to increase as the networks become more sparse .
the central difference we observe is that both the mean collective frequency variation range and its standard deviation tend to be larger for sf networks than for er networks .
this suggests that structural heterogeneity has an amplifying effect on the range of collective frequency variation for a network
however this effect can be mitigated on average by saturating the network structure : as the average connectivity increases , the range of collective frequency variation diminishes . to better understand the role of network structure in determining collective frequency variation ,
we ask the following : for which network structures is the collective frequency variation exactly zero ?
that is , which network structures yield @xmath89 regardless of the choice of @xmath62 ? from eq .
( [ eq:07 ] ) , it follows that @xmath89 for any @xmath62 whenever the entries of @xmath27 are all identical , i.e. , @xmath90 .
we note that since @xmath91 , where @xmath92 , and @xmath28 , then @xmath27 must satisfy @xmath93 , or equivalently @xmath27 is the leading right eigenvector of @xmath94 . at each entry , we must have @xmath95 , and therefore by inserting @xmath96 ( for any @xmath97 ) it is easy to see then that @xmath90 implies that the network must be degree - balanced , i.e. , @xmath72 for all @xmath7 .
the converse follows from a simple application of the perron - frobenius theorem @xcite .
specifically , @xmath90 is a solution of the leading right eigenvalue equation for @xmath94 , and the perron - frobenius theorem implies that it is in fact the unique solution , provided that the network is strongly connected .
therefore , any given network generically has zero collective frequency variation if and only if @xmath72 for all @xmath7 .
given the non - uniformity of each oscillator s contribution to a network s collective frequency variation , we now turn our attention to the properties of the first left singular vector @xmath27 , which dictates the contribution of each oscillator to the collective frequency variation .
first , we note that the entries @xmath98 are positive , and thus @xmath27 induces a centrality measure for the network .
the positiveness of the entries follows from applying the perron - frobenious theorem @xcite to the irreducible and non - negative matrix @xmath99 and noting that @xmath27 is the leading right eigenvector of the matrix .
the role of @xmath27 as the leading right eigenvector of @xmath94 also elucidates its structural properties . in particular , google s pagerank centrality which tends to favor nodes with strong in - flow is given by the leading right eigenvector @xmath100 of the matrix @xmath101 , where @xmath102 and @xmath103 is a damping factor @xcite .
formally the pagerank of a network represents the steady - state of a markovian random - walk on the network . when the damping factor is set to zero and each directed link
is reversed , the matrix @xmath104 from which pagerank is calculated is equal to @xmath105 ( for which @xmath27 is the leading right eigenvector ) .
thus , the centrality induced by the first left singular vector represents a reverse pagerank , i.e. , the steady - state of a markovian random walk on the network with each link reversed .
of the first left singular vector vs the out - to - in - degree ratio @xmath67 for an er network of size @xmath87 and @xmath106 .
( b ) the relationship between pagerank entries @xmath107 ( damped and undamped cases are plotted with red triangles and blue dots , respectively ) and first - left - singular - vector entries for the same network .
the expected inverse relationship @xmath108 is plotted as a black curve.,title="fig : " ] of the first left singular vector vs the out - to - in - degree ratio @xmath67 for an er network of size @xmath87 and @xmath106 .
( b ) the relationship between pagerank entries @xmath107 ( damped and undamped cases are plotted with red triangles and blue dots , respectively ) and first - left - singular - vector entries for the same network .
the expected inverse relationship @xmath108 is plotted as a black curve.,title="fig : " ] to provide further insight into the structure of @xmath27 , we consider instead @xmath109 , where @xmath110 is the mean - field counterpart to @xmath14 .
in particular , the corresponding mean - field approximation of @xmath27 , which satisfies @xmath111 , is precisely @xmath112 where @xmath113^{-1/2}$ ] is a normalizing factor .
thus , the centrality induced by @xmath114 can be approximated by the out - to - in - degree ratio @xmath67 a local indicator of the out - flow at a given node . in fig .
[ fig3](a ) , we plot the entries @xmath98 vs @xmath67 for an er network of size @xmath87 with @xmath115 , and we denote the mean field approximation given by eq .
( [ eq:10 ] ) with a dashed black line . in fig .
[ fig3](b ) , we compare the centrality induced by @xmath27 to pagerank centrality induced by @xmath116 ; we plot the entries @xmath107 vs @xmath98 for both a damped case ( @xmath117 ) and the undamped case ( @xmath118 ) in red triangles and blue dots , respectively .
the black curve indicates an approximate inverse relationship between the entries of @xmath100 and @xmath119 .
specifically , we use an approximation similar to the derivation of @xmath120 to find @xmath121 , which implies that the mean field approximations satisfy @xmath122 where the right - hand side is a constant . the strong agreement between eq .
( [ eq:11 ] ) and the actual entries of @xmath27 and @xmath116 illustrates the strong and opposite relationship between the centrality induced by the first left singular vector @xmath119 and pagerank @xmath100 .
this relationship between synchronization and pagerank that is revealed by the collective frequency variation of a network represents a new direction for network analysis and , immediately , the potential for applying pagerank - based techniques to self - organizing networks .
pagerank and random walker dynamics remains one of the most popular topics of research connecting various disciplines , and has a rich literature @xcite .
specifically , various algorithms and techniques exist for analysis and optimization which might be applied to manipulate a network s collective frequency variation . for instance , given the inverse relationship between the first left singular vector centrality and pagerank , we expect that increasing ( decreasing ) a node s pagerank corresponds to decreasing ( increasing ) its contribution to the collective frequency variation .
the power grid represents a prime example of a network of self - organizing dynamical systems whose functionality we rely on everyday without robust synchronization near a specified range ( @xmath123 50 - 60 hz ) , our power supply is jeopardized .
power grids @xcite are widely modeled using the following system of second - order differential equations : @xmath124 where @xmath125 represents the mechanical phase of oscillator @xmath7 , @xmath126 and @xmath127 represent the inertial and damping constants , respectively , @xmath128 represents the generated or consumed power of oscillator @xmath7 , @xmath8 is the global coupling strength , and @xmath129 is a phase - lag parameter for the interaction between oscillators @xmath7 and @xmath130 .
although the adjacency matrix @xmath14 is taken to be undirected , the presence of heterogeneity in the damping coefficients yields an effective directedness in the network coupling .
specifically , dividing eq .
( [ eq:12 ] ) through by @xmath127 and linearizing around the synchronized state @xmath131 , where we expect @xmath132 , yields the system @xmath133 where the new laplacian @xmath134 is defined @xmath135 , where @xmath136 , @xmath137 , @xmath138 , and @xmath139 .
note in particular that the effective coupling matrices are directed , i.e. , @xmath140 and @xmath141 .
different power grid models treat the inertial term in eq .
( [ eq:12 ] ) differently @xcite . in certain models
the inertial term @xmath126 depends on the role of oscillator @xmath7 : if oscillator @xmath7 is a source , or power generator , @xmath126 is nonzero , but if it is a load , or power consumer , @xmath126 is zero and thus the equation for oscillator @xmath7 is a first - order differential equation .
some models treat all @xmath126 s as non - zero , resulting in a full system of second - order differential equations , and others treat all @xmath126 s as zero , resulting in a full system of first - order differential equations .
we note that regardless of the treatment of the inertial terms , in the synchronized state @xmath142 holds for all @xmath7 , and therefore the collective frequency of the synchronized state is preserved , and thus eq .
( [ eq:07 ] ) holds . to demonstrate the presence of collective frequency variation in a real - world setting , we consider the power grid model in eq .
( [ eq:12 ] ) on empirical power grids .
specifically , we consider course - grain versions of the uk and scandinavian power grids @xcite , which we illustrate in fig .
[ fig4](a ) and ( b ) , respectively , and which consist of @xmath143 and @xmath144 nodes and @xmath145 and @xmath146 links , respectively .
it is well - known that in real - world power grids the power @xmath128 of sources and loads are positive and negative with respect to their mean , and damping coefficients are all positive , and with an appropriate rescaling of time can be set to have mean one @xcite .
therefore , we draw each @xmath128 from the bimodal normal distribution @xmath147 and each @xmath127 from the gamma distribution @xmath148 . for simplicity ,
inertial coefficients @xmath126 and phase lags @xmath129 are all set to zero .
we simulate eq .
( [ eq:12 ] ) using @xmath149 , @xmath150 , @xmath151 , and @xmath152 on both the uk and scandinavian power grid networks , calculating the collective frequency variation @xmath19 from direct observation , and compare to the theoretical prediction of eq .
( [ eq:07 ] ) given eq .
( [ eq:13 ] ) in fig . [
fig4](c ) for @xmath153 realizations of the parameters @xmath128 and @xmath127 . the dashed black curve ( which is almost completely covered ) underscores perfect agreement . in fig .
[ fig4](d ) we plot the distribution of collective frequencies found in the @xmath153 trials on each network , demonstrating that collective frequency variation can be a significant effect in important , real - world networks such as power grids .
furthermore , our numerical exploration indicate that by appropriately adding and/or deleting links , the collective frequency variation can be either amplified or mitigated , suggesting that the collective frequency variation could be tuned with a collection of judiciously chosen perturbations to the network structure .
in this article , we have studied the collective frequency of self - organizing systems in general directed networks . in particular , we have shown that in generic directed networks the collective frequency variation is nonzero and is given by a weighted average of the natural frequencies . in other words , the collective frequency of the synchronized state is not equal to the mean of the oscillators individual natural frequencies .
the weights that determine the collective frequency variation are associated with the left singular vector @xmath27 of the laplacian matrix @xmath9 corresponding to the singular value @xmath154 .
this formalism allows us to define and calculate the full range of collective frequency variations possible for any given network .
we have shown that the only networks with generically zero collective frequency variation are degree - balanced networks in which the in- and out - degrees match for every node ( i.e. , @xmath72 ) .
we have found that the first left singular vector in fact induces a centrality measure on the network .
this centrality is intimately linked with the directedness of the network and measures an effective out - flow at each node .
interestingly , we have found that this centrality is a reverse analogue of pagerank centrality @xcite ; pagerank is a cornerstone to google s ranking of webpages and is well - known to quantify the in - flow at each node @xcite .
moreover , we have shown that the mean field approximations to the first - left - singular - vector centrality and the pagerank centrality are precisely the inverse of one another .
we believe that these results will have significant impact on the study of self - organizing processes on networks , since in many application the collective dynamics of the synchronized state , i.e. , the collective frequency , plays an important role in the functionality of the system . as a prime example we have considered the dynamics of two real - world power grids a particularly important complex network of oscillators ( i.e. , sources and loads ) that governs the flow of energy @xcite .
in particular , power grids must synchronize to avoid power failures , but must also evolve close enough to a reference frequency of approximately @xmath0 - @xmath1 hz @xcite .
we have demonstrated that , despite the fact that power grid networks are structurally undirected , dynamical heterogeneity yields an effectively directed network structure , and therefore allows significant collective frequency variation .
however , we emphasize that our results have broader applications than just power grid dynamics .
in fact , the collective frequency of an ensemble plays a crucial roll in the functionality of a wide range of systems from disciplines including biology , neuroscience , and engineering .
examples of systems whose functionality can be compromised if the collective frequency differs too much from a given reference include oscillations of brain waves , propagation of activity through cardiac tissue , consensus in sensor networks , and the coordination of muscle contractions in the digestive track @xcite .
moreover , these results demonstrate a novel relationship between a widely used topological quantity used to rank the importance of nodes and the dynamical process of synchronization .
the implications point towards a new method of ranking nodes using synchronization a notion consistent with other findings where synchronization can be utilized to uncover topological properties of networks @xcite .
a particularly interesting finding is the link between the synchronization dynamics of a network ensemble and the role of pagerank in determining each oscillator s contribution to the collective frequency .
this link opens the possibility for analysis and optimization of the synchronization properties of networks using pagerank a topic with a large body of literature and
well - established algorithms for optimization @xcite . in particular
, we expect that pre - existing methods for optimizing the pagerank in networks can be applied to manipulate the collective frequency of generic various oscillator networks .
we thank m. timme for the uk and scandinavian power grid networks .
dt acknowledges support from nih award number r01hd075712 .
js acknowledges funding from the simons foundation grant no .
318812 and the army research office grant no .
w911nf-12 - 0276 .
aa acknowledges support by the european commission fet - proactive project multiplex ( grant no .
317532 ) , the icrea academia , the james s. mcdonnell foundation , and by fis2015 - 71582-c2 - 1 .
99 s. h. strogatz , _ sync : the emerging science of spontaneous order _ ( hypernion , new york , 2003 ) .
a. pikovsky , m. rosenblum , and j. kurths , _ synchronization : a universal concept in nonlinear sciences _ ( cambridge university press , 2003 ) .
a. arenas , a. daz - guilera , j. kurths , y. moreno , and c. zhou , synchronization in complex networks , phys . rep . * 469 * , 93153 ( 2008 ) .
m. rohden , a. sorge , m. timme , and d. witthaut , self - organized synchronization in decentralized power grids , phys .
* 109 * , 064101 ( 2012 ) .
a. e. motter , s. a. myers , m. anghel , and t. nishikawa , spontaneous synchrony in power - grid networks , nat . phys .
* 9 * , 191197 ( 2013 ) .
e. g. antzoulatos and e. k. miller , increases in functional connectivity between the prefrontal cortex and striatum during category learning , neuron * 83 * 216225 ( 2014 ) .
a. karma and r. f. gilmour , nonlinear dynamics of heart rhythm disorders , phys . today * 60 * , 5157 ( 2007 ) .
r. olafti - saber , j. a. fax , and r. m. murray , consensus and cooperation in networked multi - agent systems , proc .
ieee * 95 * , 215233 ( 2007 ) .
r. r. aliev , w. richards , and j. p. wikswo , a simple nonlinear model of electrical activity in the intestines , j. theor .
* 7 * , 2128 ( 2000 ) .
y. kuramoto , _ chemical oscillations , waves , and turbulence _
( springer , new york , 1984 ) .
e. ott and t. m. antonsen , low dimensional behavior of large systems of globally coupled oscillators , chaos * 18 * , 037113 ( 2008 ) .
j. g. restrepo , e. ott , and b. r. hunt , onset of synchronization in large networks of coupled oscillators , phys .
e * 71 * , 036151 ( 2005 ) .
j. g. restrepo , e. ott , and b. r. hunt , synchronization in large directed networks of coupled phase oscillators , chaos * 16 * , 015107 ( 2006 ) .
s. assenza , r. gutirrez , j. gmez - gardees , v. latora , and s. boccaletti , emergence of structural patterns out of synchronization in networks with competitive interactions , sci . rep . * 1 * , 00099 ( 2011 )
. m. e. j. newman , the structure and function of complex networks , siam rev . *
45 * , 167256 ( 2003 ) .
f. chung , laplacians and the cheeger inequality for directed graphs , ann
. comb . * 9*(1 ) , 119 ( 2005 ) . c. grabow , s. grosskinsky , and m. timme , small - world spectra in mean - field theory , phys .
lett . * 108 * , 218701 ( 2012 ) . c. grabow , s. grosskinsky , j. kurths , and m. timme , collective relaxation dynamics of small - world networks , phys .
e * 91 * , 052815 ( 2015 ) .
f. drfler , m. chertkov , and f. bullo , synchronization in complex oscillator networks and smart grids , proc .
sci . * 110 * , 20052010 ( 2013 ) .
p. s. skardal and a. arenas , control of coupled oscillator networks with application to microgrid technologies , sci .
adv . * 1 * , e1500339 ( 2015 ) .
n. kopell and g. b. ermentrout , in _ handbook of dynamical systems ii _ , edited by b. fiedler ( elsevier , new york , 2002 ) , vol .
2 , pp . 354 .
p. s. skardal , d. taylor , and j. sun , optimal synchronization of complex networks , phys . rev
. lett . * 113 * , 144101 ( 2014 ) .
p. s. skardal , d. taylor , j. sun , and a. arenas , erosion of synchronization in networks of coupled oscillators , phys .
e * 91 * , 010802(r ) ( 2015 ) .
p. j. menck , j. heitzig , n. marwan , and j. kurths , how basin stability complements the linear - stability paradigm , nat . phys .
* 9 * , 8992 ( 2013 ) .
s. brin and l. page , the anatomy of a large - scale hypertextual web search engine , computer networks and isdn systems * 30 * , 107117 ( 1998 ) .
d. f. gleich , pagerank beyond the web , siam rev .
* 57 * , 321 ( 2015 ) .
a. ben - israel and t. n. e. grenville , _ generalized inverses _ ( springer , new york , 1974 ) .
r. olafti - saber and r. m. murray , consensus problems in networks of agents with switching topology and time - delays , ieee trans .
control * 49 * , 15201533 ( 2004 ) .
p. erds and a. rnyi , on the evolution of random graphs , publ .
inst . hung .
sci . * 5 * , 1761 ( 1960 ) .
m. molloy and b. reed , a critical point for random graphs with a given degree sequence , random struct .
* 6 * , 161180 ( 1995 ) . c. r. maccluer , the many proofs and application of perron s theorem , siam rev . *
42 * , 487498 ( 2000 ) .
a. n. langville and c. d. meyer , _ google s pagerank and beyond : the science of search engine rankings _ ( princeton university press , 2006 ) .
t. nishikawa and a. e. motter , comparative analysis of existing models for power - grid synchronization , new j. phys .
* 17 * , 015012 ( 2015 ) .
i. simonsen , l. buzna , k. peters , s. bornholdt , and d. helbing , transient dynamics increasing network vulnerability to cascading failures , phys . rev .
lett . * 100 * , 218701 ( 2008 ) .
d. witthaut , m. rohden , x. zhang , s. hallerberg , and m. timme , critical links and nonlocal rerouting in complex supply networks , phys . rev . lett .
* 116 * , 138701 ( 2016 ) .
s. lozano , l. buzna , and a. daz - guilera , role of network topology in the synchronization of power systems , euro .
j. b * 85 * , 231238 ( 2012 ) .
f. drfler and f. bullo , synchronization and transient stability in power networks and non - uniform kuramoto oscillators , siam j. control optim . *
50 * , 16161642 ( 2012 ) .
a. schnitzler and j. gross , normal and pathological oscillatory communication in the brain , nat .
* 6 * , 285296 ( 2005 ) .
a. karma , physics of cardiac arrhythmogenesis , annu .
rev . condens .
matter phys .
* 4 * , 313337 ( 2013 ) .
a. arenas , a. daz - guilera , and c. j. prez - vicente , synchronization reveals topological scales in complex networks , phys .
lett . * 96 * , 114102 ( 2006 ) .
|
a wide range of natural and engineered phenomena rely on large networks of interacting units to reach a dynamical consensus state where the system collectively operates . here
we study the dynamics of self - organizing systems and show that for generic directed networks the collective frequency of the ensemble is _ not _ the same as the mean of the individuals natural frequencies .
specifically , we show that the collective frequency equals a weighted average of the natural frequencies , where the weights are given by an out - flow centrality measure that is equivalent to a reverse pagerank centrality .
our findings uncover an intricate dependence of the collective frequency on both the structural directedness and dynamical heterogeneity of the network , and also reveal an unexplored connection between synchronization and pagerank , which opens the possibility of applying pagerank optimization to synchronization .
finally , we demonstrate the presence of collective frequency variation in real - world networks by considering the uk and scandinavian power grids .
|
observations with the rossi x - ray timing explorer ( rxte ) of neutron - star low - mass x - ray binaries ( lmxbs ) have led to the discovery of kilohertz quasi - periodic oscillations ( khz qpos ) in about 20 sources ( van der klis 2000 for a review ) .
the khz qpo frequency is always correlated with position in the tracks traced out by these sources in x - ray color - color diagrams .
this usually implies khz qpo frequency is correlated with x - ray count rate , but anticorrelation also occurs ( wijnands et al .
1997 , homan et al .
2001 , this paper ) . on time scales
longer than hours to days the count - rate / frequency correlations usually break down ( e.g. , mndez 1999 ; see also van der klis 2001 ) .
usually a pair of khz qpo peaks is observed ( called lower and upper khz qpo in order of increasing frequency ) , with a frequency separation @xmath2 that varies in anticorrelation with , but by much less than , the khz qpo frequencies themselves . in sco x-1 ( van der klis et al .
1996 , 1997 ) for variations in frequency by @xmath0500hz , @xmath2 varies by @xmath070hz .
numerous models have been proposed to explain the khz qpos ( see van der klis 2000 ) .
most associate one of the two khz qpo frequencies with the keplerian frequency at the inner edge of the accretion disk .
miller , lamb & psaltis ( 1998 ) describe how the khz qpo frequency can be affected by radiative stresses ( both drag and radial force ) .
the @xmath06hz normal - branch oscillation ( nbo ) seen in the most luminous lmxbs , the z sources , is probably related to near - eddington accretion ( e.g. hasinger et al .
1990 ; but see wijnands , van der klis and rijkhorst 1999 ) .
several models have been proposed for the nbo ( see [ disc ] ) . in the z source sco x-1 ,
the nbo occurs simultaneously with khz qpos ( van der klis 1996 , 1997 ) . at a fractional rms amplitude of 3 - 5% it dominates the 4 - 8 hz x - ray variability .
if the nbo - modulated flux originates from within or near the inner disk radius then the associated 6-hz modulation of the drag on the inner disk could modulate the orbital frequency there . if khz qpo frequencies are indeed related to orbital frequency , a modulation with nbo phase would be expected . if , on the other hand , the 6-hz modulation originates further out , or the khz qpo frequencies have another origin , then such modulation of the khz qpo frequencies would not be predicted .
so , clues to the origin of both khz qpos and slower fluctuations can be obtained by monitoring the khz qpos on subsecond time scales . in this letter , we perform the first study of this kind .
we investigate the response of the khz qpos to the flux variations due to the nbo in sco x-1 .
we find , that while the response of the khz qpo frequencies to the 6-hz flux variations is similar to ( but probably less strong than ) that to slower variations in flux , the power ratio of the two khz peaks varies in the _ opposite _ sense .
we discuss how these findings can help our understanding of both nbo and khz qpos .
we analyzed 260 kev high time - resolution data taken of sco x-1 between feb 1996 and july 1998 with the rxte proportional counter array ( pca ) . based on 1/16 hz resolution fourier power spectra of individual @xmath024 ks continuous data segments , we selected segments where the upper khz qpo was significant at @xmath3 , and the nbo was strong ( peak leahy power @xmath45.0 ) , narrow ( fwhm @xmath024 hz ) , and between 6 and 8 hz ( table 1 ) .
we used data from all active pcus ( 3 - 5 ) .
the 260 kev count rate was @xmath5 c / s per pcu .
lower khz qpos were detected in about 1/3 of the segments .
we made ffts with length @xmath6 s and time resolution 1/4096 s. the frequency resolution of the resulting power spectra is 32 hz and the nyquist frequency 2048 hz .
we compared the counts in non - overlapping pairs of successive power spectra .
let the number of counts in the @xmath7th pair of successive power spectra be @xmath8 for the higher count - rate power spectrum of the pair , and @xmath9 for the lower count - rate one .
we then selected those pairs for which the statistic @xmath10 exceeds a given pre - set value .
@xmath11 is a measure for the significance of the count - rate difference between the two power spectra in each pair .
we used those selected pairs to construct a high - rate sample of power spectra , with corresponding selected @xmath12 values , and a low - rate sample , with corresponding @xmath13 values .
two criteria were used in the pair selection : @xmath14 and @xmath15 .
we also combined the two to get the sample corresponding to @xmath16 .
we averaged the high - rate and low - rate power spectra separately and fitted the average power spectra in the 3001800 hz range with two lorentzians ( for the two khz qpos ) plus a line ( for the dead - time modified poisson noise ) .
all the uncertainties were determined using @xmath17 .
the difference @xmath18 is related to the fourier power around frequency @xmath19 .
hence , with @xmath6s we sample the fluctuations near 16 hz . to cover lower frequencies we constructed longer segments of duration @xmath20 where @xmath21 , which sample frequencies of @xmath22 hz .
we calculated the total counts and the average of the @xmath23 power spectra in each @xmath20 wide segment , and constructed high - rate and low - rate average power spectra for these longer segments in the same way as above for @xmath24 .
the frequency and amplitude of the khz qpos were then determined by fitting each high - rate and low - rate average power spectrum for each n as described above .
counting statistics of course contribute to the count differences between the segments , but at near - nbo frequencies the nbo by far dominates the variability .
monte - carlo simulations show that the contribution of counting statistics to the count differences is @xmath02% for @xmath25 , and @xmath08% for @xmath15 ; around the nbo frequency these numbers drop to @xmath00.5% and @xmath06% , respectively . with the exception of figure 3 , where we report absolute count rates ,
the average count differences and derived quantities reported in this paper have all been corrected for this counting - statistics bias associated with our selection procedure .
the khz qpos in the average power spectra were significant at @xmath26 , except the lower khz qpo in the high - rate sample for n=1 , 2 , 3 and 4 ( @xmath27 ) .
below we refer to the frequencies of the lower and upper khz qpo as @xmath28 and @xmath1and mark parameters for the high - rate and low - rate samples with `` h '' , and `` l '' .
fig . 1 , shows results for @xmath29 .
fig.1a is the average power spectrum of the observations , showing the nbo peak .
fig.1b shows the difference @xmath30 between @xmath31 in the low- and high - rate samples . near the nbo frequency
, @xmath32 is 1020 hz larger than @xmath33 . as the counting - statistics - corrected average count - rate difference between the high and the low - rate samples @xmath34 is different at each frequency ( i.e. , each @xmath23 ) , we show in fig.1c @xmath30 normalized to count - rate difference ( in units of counts per 1/32 s ) .
also when normalized this way , there is a clear anticorrelation between @xmath1 and count rate .
the lower khz frequencies @xmath35 and @xmath36 are consistent with being the same near the nbo frequency , and the frequency separation @xmath2 between the two peaks is larger in the low - rate sample : @xmath37 . however , as shown in fig.1d , this is only a 2-@xmath38 effect and could be affected by the strength variations in the peaks reported below .
fig.2 shows the average high- and low - rate power spectra for n=2 and @xmath39 .
the raw measured integrated leahy powers are [email protected] and [email protected] at low count rate and [email protected] and [email protected] at high count rate for lower and upper khz peak , respectively .
the dashed line marks @xmath31 of the low - rate sample , @xmath41 hz . at high count rate
only the upper khz qpo , at @xmath42 hz , is significant .
taken at face value , fig .
2 suggests that on the nbo time scale , going to higher count rate , the lower khz qpo weakens by a large factor , while the upper one remains comparatively constant in strength ( it weakens by a factor of only 1.5 ) .
however , in view of the large and uncertain deadtime effects at these count rates , we can not exclude that instead the upper peak becomes much stronger but the entire high - rate spectrum is suppressed by differential deadtime ( van der klis et al .
the ratio between the two khz qpo powers is less sensitive to dead time ( see van der klis et al .
for @xmath43 the ratio @xmath44 of lower to upper khz qpo power is lower at high rate sample than at low rate , but not significantly so ( 1@xmath38 ) . at @xmath15 ,
the ratio @xmath44 is [email protected] at low rate , @xmath45 larger than the value of [email protected] at high rate .
the average count rates and @xmath1 values for n=2 using selections @xmath14 and @xmath15 are shown in fig.3 .
no correction for counting - statistics bias was performed ; the outer and inner points should be 6% and 0.5% closer together , respectively .
the anticorrelation between @xmath1 and the average count rate is obvious .
it is of interest to compare these short - term correlations to those on longer time scales .
the khz qpo frequencies vary too little in our observations ( @xmath4610701093hz ) to determine the flux - frequency relation on the 10100 s time scales on which the z track is traced out .
we can measure this relation in an observation on april 24 , 1997 , when sco x-1 was on its normal branch , covering a wider range of frequencies ( @xmath4610001100hz ) .
the nbo was weaker and broader then . using 128-sec data segments , we find that also on these longer time scales , @xmath31 and count rate are anticorrelated .
this is as expected , as on the normal branch ( nb ) of sco x-1 count rate and position on the z track are anticorrelated ( priedhorsky et al .
1986 ; dieters & van der klis 2000 , see also van der klis 2001 ) .
the fractional frequency shift @xmath47 is [email protected] times the fractional count - rate variation @xmath48 on these longer time scales . on the nbo time scale
this is [email protected] , about 1/3 of that on longer time scales .
( deadtime corrections on @xmath48 were neglected here , as both @xmath49 and @xmath50 were nearly equal in the two cases so that the corrections do not affect the final conclusion . ) in principle this difference could be caused by our method to select high and low count rates caused by the nbo .
however , as we corrected our results for counting - statistics bias , to the extent that the dependence of khz qpo frequency on the nbo flux variations can be considered linear ( which is likely as the variations in nbo amplitude and khz qpo frequency are both @xmath515% ) , these numbers can be directly compared .
of course , they were derived from different data sets .
we find that the power ratio @xmath44 , in contrast to what is seen on the nbo time scale , on long time scales increases with count rate , from about 0.5 at @xmath52 1100 hz to about 1.0 at @xmath521000 hz ( this confirms what is indicated by the results shown in figure 3d in van der klis et al . 1997 ; however , in our current analysis , contrary to that work , we completely cover the 10001100hz range without gaps ) .
the quantity @xmath53 is 0.6 times @xmath54 on long time scales ; on the nbo time scale this is @xmath552 .
so , a similar anticorrelation between khz qpo frequency and flux occurs in sco x-1 on the nb both on long time scales and on the nbo time scale , but the weakening of the lower khz qpo relative to the upper khz qpo that we observe towards higher count rate on the nbo time scale is _ opposite _ to that observed on longer time scales .
in response to the @xmath06 hz flux variations due to the nbo in sco x-1 the upper khz qpo frequency @xmath1 decreases with count rate . this is qualitatively , yet not quantitatively consistent with the behavior of sco x-1 on the normal branch on longer time scales , which , however , due to statistics we could only measure using a different method and in another data set , and at khz qpo frequencies including , but mostly below those where we measured the short - term correlation .
the lower khz qpo weakens relative to the upper one when flux increases on the nbo time scale .
this is _ opposite _ to what ( on the nb for @xmath56hz ) is observed on longer time scales .
the weakness of the lower khz qpo in the high count - rate phase of the nbo cycle limits the accuracy of the determination of its frequency shifts .
although the frequency separation between the two peaks appears different between the high count - rate phase and the low count - rate phase , we can not rule out the possibility that it remains constant . in any case ,
our results show that both frequency and amplitude of khz qpos are affected by flux variations on subsecond time scales , and that in some respects khz qpos react differently to nbo - related flux variations than to slower ones .
the frequency shift of 22 hz we observe in the upper khz qpo peak on the nbo time scale is of the same order as its width ( fwhm@xmath070 hz ) .
so , the width of the upper khz qpo peak usually reported is due in part to short - time - scale variations in khz qpo frequency that went unnoticed up to now . as our measurements show that the khz qpo signals also change in relative strength on this time scale
, the short - term fluctuations must also affect the @xmath2 measured on longer time scales : due to what is happening on the nbo time scale , in power spectra averaged over longer time intervals the khz qpo peaks appear closer together than they really are .
the fact that the lower khz qpo becomes _ weaker _ with respect to the upper khz qpo with increasing count rate on the nbo time scale , but _
stronger _ on longer time scales , strongly suggests that two different mechanisms produce the count rate variations on these two time scales , which interact with the khz qpo production mechanism in different ways .
the coupling between the khz qpos and the nbo provides clues to the origin of both phenomena .
one possibility for
the nbo is that it is a disk oscillation mode in a subsonic region of a thick disk , as suggested in the sound wave model of alpar et al .
( 1992 ) . in this model ,
the nbo frequency is of the same order as the orbital frequency in a thick disk .
if the upper khz qpo is also related to an orbital frequency , the nbo is clearly generated at a much larger radius .
the mass accretion rate in the inner disk region could then be modulated by this oscillation , and in that way affect @xmath1 . the same relation between @xmath47 and @xmath48
would then be expected to hold not only qualitatively but also quantitatively on both long time scales and the nbo time scale , which seems inconsistent with our results .
however , our measured long - term and short - term values do not refer to the same data set and were measured using different methods .
another possibility is that the nbo is associated with a radial accretion mode , as proposed in the radiation feed - back model ( fortner , lamb and miller 1989 , miller and park 1995 ) .
the x - ray flux modulation in this model arises either by modulation of the mass inflow rate of the radial inflow , or by modulation of its compton scattering opacity . as shown in fig.3 , when @xmath31 is around 1080 hz , the frequency decreases by @xmath02% for an increase in average count rate by @xmath05% . from equation 39 in miller et al .
( 1998 ) , at an accretion rate of @xmath57 eddington and a constant radius near the the marginally stable orbit ( this may approximately represent the accretion state of sco x-1 when the nbo is observed ) , a flux variation of 6.5% would yield an orbital frequency change of @xmath0 5% due to a change in radial radiation force , which is of the same order of magnitude as observed .
both radiation drag and radial force effects would lead to increase radius when flux increases .
this suggests that the frequency shifts may be caused by the modulation in the radiative stresses caused by the nbo .
for this the inner boundary of the radial inflow , where it produces its radiation , should be well inside the inner disk boundary , some 16 km for a 1.4 m@xmath58 neutron star at a @xmath1 of 1080hz .
this is consistent with the idea that this flow produces its radiation at or close to the neutron star surface .
the quantitative difference between the @xmath31 variation on the nbo time scale and that on longer time scales may come from the fact that the nbo is only related to the radial inflow , which takes place outside the equatorial plane , while the long - term variation is related to the accretion flow through the disk , accreting in the equatorial regions .
the radiation from near the equator may interact more strongly with the inner disk than that emitted from higher latitudes .
a third possibility is that the nbo originates from the neutron star surface by a mechanism such as g - modes ( bildsten and ushomirsky 1996 ) .
this can lead to radiative stresses which modulate the upper khz qpo frequency .
finally , the lapidus , nobili & turolla ( 1994 ) mechanism for photospheric oscillations during x - ray bursts , which relies on a near - eddington flux , could perhaps also work in sco x-1 to generate a flux variation at the nbo frequency emerging from the neutron star surface .
we have found that the flux - frequency correlation on the nbo time scale and on longer time scales has the same sign in sco x-1 on the normal branch : frequency drops when count rate increases .
the situation on sco x-1 s normal branch is actually unusual ; what is more commonly observed ( on time scales of hours to days ) is that frequency increases with count rate .
it is interesting to ask what would be observed if in such a case the dependence of khz qpo frequency on nbo phase could be measured .
if the general idea is correct that the radiative stresses on the inner disk edge , modulated at the nbo frequency by some other mechanism than a modulation of the disk accretion rate , lead to the observed frequency shifts of the upper khz qpo , then this anticorrelation would be expected to hold for all nbo signals produced in the same way as in sco x-1 .
the prediction is that in the more usual situation of a positive long - term correlation between khz qpo frequency and count rate , on the nbo time scale there would still be an anticorrelation , so that then the long and short - term effects would be of opposite sign .
however , there is evidence that nbos arise variously as a consequence of luminosity and radiative - transfer fluctuations ( fortner et al . 1989 ) , so that the complete picture may be more complex than suggested by this line of reasoning .
as the frequency correlations between qpos and noise among neutron star and black hole x - ray binaries occur over a wide range of source types and luminosities , further study of how these features are correlated on short time scales will help reveal the origin of qpos and noise components , and understand the tracks in the plot of khz qpo frequency vs. x - ray flux .
the study of the khz qpo or high frequency qpo in response to qpo and noise components at lower frequencies in black hole and neutron star x - ray binaries are particularly promising approaches .
we are grateful to jeroen homan , mariano mndez , marc klein - wolt for carefully reading the manuscript and useful comments on this work .
wy thanks eric ford and robert fender for assistances on the use of the computation facilities in the x - ray group .
pgj was supported by nwo spinoza grant 08 - 0 to e.p.j.van den heuvel .
wy was supported by nwo grant 614.051.002 and the national natural science foundation of china and the chinese academy of sciences .
this work has made use of data obtained through the high energy astrophysics science archive research center online service , provided by the nasa / goddard space flight center .
hasinger , g. , van der klis , m. , ebisawa , k. , dotani , t. & matsuda , k. 1990 , a&a , 235 , 131 fortner , b. , lamb , f. k. & miller , g. 1989 , nature , 342 , 775 homan , j. , van der klis , m. , jonker , p.g . ,
wijnands , r. , kuulkers , e. , mndez , m. , lewin , w.h.g .
, 2001 , apj submitted ; astro - ph/0104323 lapidus , i. , nobili , l. & turolla , r. 1994 , , 431 , l103 ccc observation i d & segments & data modes + 10056 - 01 - 04 - 02 & 1,2,3,4 & b+2lld + 20053 - 01 - 01 - 03 & 1,2 & b+sb+2lld + 20053 - 01 - 01 - 05 & 3,4 & b+sb+2lld + 20053 - 01 - 02 - 03 & 1,2 & b+sb+2lld + 20053 - 01 - 02 - 030 & 2,3,4,5 & b+sb+2lld + 30035 - 01 - 02 - 00 & 1 & b+sb+2lld + 30035 - 01 - 02 - 000 & 4,5 & b+sb+2lld + 30035 - 01 - 03 - 00 & 1,2,3,4,5,6 & b+sb+2lld + 30035 - 01 - 07 - 00 & 1,2,3 & b+sb+2lld +
|
we analysed rxte data of sco x-1 which show khz quasi - periodic oscillations ( qpos ) and the @xmath068 hz normal - branch oscillation ( nbo ) simultaneously . using power spectra of 0.030.5 s data segments
, we find that both the upper khz qpo frequency @xmath1 and the ratio of lower to upper khz qpo amplitude are anticorrelated to variations in the x - ray count - rate taking place on the nbo time scale .
the frequency dependence is similar to ( but probably weaker than ) that found on longer time scales , but the power ratio dependence is _ opposite _ to it .
a model where radiative stresses on the disk material , modulated at the nbo frequency , lead to changes in @xmath1 can explain the data ; this implies some of the nbo flux changes originate from inside the inner disk radius .
we discuss how these findings affect our understanding of khz qpos and of the low - frequency variability of low - mass x - ray binaries .
|
alice ( a large ion collider experiment ) is a general - purpose heavy - ion experiment designed to study the physics of strongly interacting matter and the quark - gluon plasma in nucleus - nucleus collisions at the lhc @xcite .
it will study the global properties with hadron production and correlations , and probe the properties of the medium with the collision products such as heavy quarkonia , open charm and beauty , light vector mesons , thermal leptons , direct-@xmath0 , jets and high-@xmath1 hadrons .
one of the powerful probes of the created qcd medium is heavy quarkonia , whose suppression or enhancement @xcite is sensitive to the screening of color charge due to deconfinement and to statistical recombination .
thus measurements of leptons from their decay are crucial .
an important task of the transition radiation detector ( trd ) for the alice experiment is to supplement the time projection chamber ( tpc ) electron / pion identification by a pion rejection factor of the order of 100 at momenta in excess of 1 gev / c .
in addition , by measurement of energy loss , the trd improves the identification of other charged particles @xcite .
the trd also provides space points for the global central barrel tracking together with the inner tracking system ( its ) and the tpc . due to its large lever arm
it improves the overall momentum resolution , especially at high momentum , where the resolution is expected to be @xmath23.5% at 100 gev / c @xcite .
additionally , the trd provides a fast trigger for single / pairs of electrons and cluster of high-@xmath1 tracks , thus it allows us to study rare probes such as high-@xmath1 @xmath3/@xmath4 , @xmath5 and high-@xmath6 jets .
the required pion rejection of a factor 100 is obtained by production of transition radiation ( tr ) which is only generated by electrons within the relevant momentum range .
the relativistic electron ( @xmath0@xmath71000 ) radiates photons in the x - ray range when it traverses a boundary between media of different refraction indices .
these tr photons ( @xmath8 30 kev ) are absorbed by the high-@xmath9 gas mixture ( xe - based ) of high photo - absorption cross section .
[ design_principle ] illustrates the working principle and design of the trd chamber . considering tr photon detection and tracking ,
the chamber consists of polypropylene fibres / form sandwich radiator ( 48 mm ) , drift ( 30 mm ) and multi - wire proportional ( 7 mm ) section with cathode readout pad . on top of the signal due to the ionization of the gas by traversing charged particle , the characteristic peak
is produced by the tr photons at large drift time for electron as it is shown in fig .
[ pulse_height_prototype ] .
this allows us to separate electrons from pions .
the 540 chambers of the trd are arranged in 18 supermodules , which surrounds the tpc in the central barrel of alice , containing 5 stacks along the longitudinal and 6 layers along the radial direction .
it has 694 m@xmath10 active area and 28 m@xmath10 gas volume of xe / co@xmath11 ( 85/15 ) .
the total radiation length is @xmath224% of @xmath12 and the total weight is @xmath230 tons .
the trd front - end electronics ( fee ) is directly mounted on the back panel of the chamber .
groups of 18 pads are connected by short cables to a multi - chip module ( mcm ) which comprises the pre - amplifier and shaper amplifier ( pasa ) and the tracklet processor ( trap ) @xcite .
the pasa has a 120 ns shaping time , a gain of 12.4 mv / fc , and an equivalent noise charge of 850 electrons at 25 pf input capacitance .
the trap chip comprises 21 channels each with a 10 mhz 10 bits adc , four stages of digital filters , event buffers , readout interface and local tracking unit ( preprocessor and four 120 mhz cpus ) which allows to calculate the inclination of a track in the bending direction as well as to measure the total charge deposited along the track .
one read - out board ( rob ) consists of 17 or 18 mcms and 6 or 8 robs are mounted on each chamber .
each chamber has one linux based detector control system ( dcs ) board which controls fee and two optical readout interface modules ( ori ) for data shipping . the local track segments ( tracklets ) and the raw data acquired by the trap are sent to the global tracking unit ( gtu ) via 2.5 gbps ori .
based on the tracklets collected , the gtu performs transverse momentum reconstruction and electron identification . after finding high-@xmath1 tracks and identifying electrons ,
various trigger schemes are applied for di - electron decays and jets .
then this trigger contribution is sent to the alice central trigger processor ( ctp ) within 6.1 @xmath13 to drive the level-1 trigger decision .
such fast processing is possible due to the massive parallel hardware architecture of the gtu whose core is fpga - based .
it is capable of processing up to 20k tracklets within 2 @xmath13 ( maximum 16k tracklets for @xmath14 = 8000 events with a @xmath1 threshold of 2.3 gev / c ) .
the gtu also forwards raw data to the daq when level-2 is accepted .
supermodule integration is done layer - wise with 30 chambers and each layer undergoes tests of electronics , gas tightness , high - voltage distribution and cooling .
it is concluded by several days of cosmics data taking to produce a calibration data set .
the first supermodule was installed in october 2006 .
noise measurements after installation show that we achieved an average noise level of 1.1 adc counts , which is close to the design goal .
the fraction of dead channels is less than 0.1% . in total , four supermodules were installed in alice in 2008 and participated in the cosmic ray data taking together with other detectors . besides the proof of combined operation , these runs are used to gather reference data for alignment and calibration from traversing cosmic particles .
the cosmic trigger decision was based on coincident hits in the time of flight ( tof ) detector and gtu level-1 trigger .
for the first time we used the full chain of trigger sequence for the trd within the alice setup and the gtu level-1 trigger was the first running level-1 trigger in alice .
level-1 rejection from level-0 ( originated by tof coincident hits ) was a factor 20 and the level-1 rate was @xmath20.05 hz with a purity better than 85% .
in total 55k tracks were recorded in the trd under tight constraints on the cosmic ray topology , requiring tracks close to horizontal at 60 m below the earth surface . before the lhc injection test in 2008 , we were ready for data taking .
the tracking and pid algorithms of the trd rely on the knowledge of several calibration constants which depend on environment temperature and gas pressure , gas composition and chamber geometry .
the calibration constants are the drift velocity of the electrons , the time - offset of the signal and the gas gain . during data taking ,
a first calibration is performed as an online procedure and later also done offline .
the drift velocity and time offset are determined with the average signal as a function of time ( fig .
[ pulse_height ] ) . with the assumption that the cosmic rays are uniformly distributed over the detector , a gas gain variation of about 16%
was found over the chambers .
[ trd_gain_driftvelocity ] shows the gain factors for the four supermodules .
the results obtained online were consistent with those obtained offline ( fig .
[ trd_gain_driftvelocity ] ) after the tracking in a second pass calibration .
the extracted drift velocity values has a variation of 3.3% over the chambers .
the obtained spatial resolution within the trd chambers is @xmath15350 @xmath16 at 0@xmath17 , which is close to the design goal . as a part of commissioning , in 2004 and 2007
there were test beam measurement at cern ps with electron and pion beams .
the likelihood distributions for six layers , based on the total energy deposit in one layer are shown in fig .
[ liklihood ] for the momentum of 2 gev / c .
cuts of given electron efficiency are imposed on the likelihood value and the pion efficiency @xmath18 is calculated .
the momentum dependence of the pion efficiency calculated with different likelihood methods and for the neural networks is shown in fig .
[ pion_efficiency ] .
it demonstrates that we exceed the design goal of factor 100 pion rejection for isolated tracks for momenta less than 10 gev / c .
seven out of the total 18 supermodules of the trd will be ready in the alice setup when cosmic ray data taking resumes in july 2009 and will contribute to physics results with beams expected to start in fall 2009 .
the completion of the trd setup with all 18 supermodules installed is planned during the next long shutdown of the lhc .
we acknowledge the alice collaboration , the installation and support team at cern . this work has been supported by german bnbf .
|
the transition radiation detector ( trd ) for the alice experiment at the large hadron collider ( lhc ) identifies electrons in p+p and in the challenging high multiplicity environment of heavy - ion collisions and provides fast online tracking for the alice level1 trigger .
the trd is designed to have excellent position resolution and pion rejection capability .
presently , six of the 18 trd supermodules are installed in the alice central barrel . in 2008 ,
four supermodules were installed and commissioning of the detector using cosmic ray tracks was successfully performed .
we briefly describe the design of the detector and report on the performance and current understanding of the detector based on these data .
|
the long - distance color - dependent forces among quarks and gluons are the key quantities for the understanding of the internal structure of hadron as well as color confinement dynamics .
the quarks and gluons are described in the quantum chromodynamics ( qcd ) , and have a color charge based on the su(3 ) group , with which the quark combination yields many color - dependent forces including both attractive and repulsive forces .
they make the hadron structure ( quark bound state ) more complicated .
therefore , it is essential to know the basic behavior of the color - dependent force at short and large distances
. the color - dependent force will be important when one investigates the multiquark hadron ( made of more than four quarks ) and the exotic meson , etc .
although many candidates of those particles have been recently reported by experimental groups @xcite , it is still a hard task to theoretically understand those new particles .
they may consist of many quarks with various color - dependent forces that make their internal structure more complicated than the existing baryons .
the quark interaction at short distances can be characterized by a coefficient @xmath0 of a one - gluon exchange potential .
this is a basic assumption if we construct a quark model to deal with the baryon system .
however , owing to the quark confinement , it is not obvious how quarks behave for large quark separation .
we thus need a lattice simulation to obtain a nonperturbative feature for the long - range color - dependent force . however , the lattice study along this line is sparse , although there are many lattice studies about the @xmath1 potential @xcite obtained by the gauge - invariant wilson - loop operator .
this operator mixes color - singlet with color - octet contributions ; therefore , we can not separately extract the color - octet potential from the wilson loop . in spite of this
, the color - octet process has been discussed in the detailed analysis of the @xmath2 photoproduction to explain the deviation between the experimental data and the theoretical predictions @xcite . on the other hand , there is no lattice calculation for @xmath3 sector as well ; however , the diquark ( color-@xmath4 @xmath3 ) is also an important ingredient on hadron physics .
the existence of the diquark has been believed by many physicists and applied to work out unsolved problems @xcite .
note that the @xmath3 potential in the lattice theory is not formulated gauge - invariantly . in this study , we employ a color - dependent polyakov line correlator ( plc ) with the coulomb gauge fixing in order to clarify the color - dependent forces @xcite . in the quark - gluon plasma phase ,
there have been already some numerical studies of the color - screened color - dependent force with the coulomb and landau gauge fixings @xcite .
investigating the confining color - dependent forces we use coulomb gauge qcd @xcite , theoretical background of which has been recently studied well @xcite .
in this theory , the plc potential can be separated into the vacuum polarization part and the color - coulomb instantaneous part ; the latter relatively produces us a clear numerical signal even in the confinement phase @xcite .
the color - coulomb instantaneous potential is not the same as the wilson - loop potential .
however , it yields an upper bound of a linearly - rising quark potential @xcite and is the most important quantity on the coulomb gauge confinement scenario @xcite .
the infrared singularity of this potential is caused from accumulation of the faddeev - popov ( fp ) ghost eigenvalue at the vanishing momentum @xcite .
coulomb gauge qcd may provide us the most favorable framework when one considers the hadron phenomenological studies such as a constituent quark model .
the coulomb gauge as a physical gauge gives a positive - definite fock space that is suitable for constructing effective hamiltonians .
moreover , the color - coulomb instantaneous potential without the vacuum polarization ( or a retarded effect ) is required to make a quark bound state in analogy with the quantum electrodynamics ( qed ) .
this formulation is applicable in the heavy - quark system with no retarded effect as well as in the light - quark system with the constituent quarks due to chiral - symmetry breaking @xcite . in this paper
, we study the long - distance divergence behavior of the color - dependent forces . because the color - nonsinglet state can not exist in nature
, singularities in such states will emerge on the lattice as a finite - volume effect .
this has not been verified in the previous calculation [ 13 ] , and thus we should investigate their divergence behavior on a variety of lattice sizes .
we also discuss the color - factor dependence for the degree of the divergence and will confirm this point by the numerical lattice simulation . in section
ii , we briefly summarize coulomb gauge qcd and the definition of the color - dependent potentials with the plcs . in section iii , we give lattice numerical results and fitting analyses .
section iv is devoted to our summary .
coulomb gauge qcd has been quantized through the faddeev - popov technique @xcite and renormalizability of this theory has been also proved in terms of the hamiltonian and lagrangian formalism @xcite . the use of the coulomb gauge as a physical gauge leads us to classify transverse gluon modes and an instantaneous interaction , which is required to make quark bound states in analogy with qed .
the hamiltonian of qcd in the coulomb gauge can be given by @xmath5 where @xmath6 , @xmath7 and @xmath8 are the transverse electric field , the transverse magnetic field and the color - charge density , respectively .
the function @xmath9 in the second term is made by the faddeev - popov ( fp ) operator in the spatial direction , @xmath10 , @xmath11 .
\label{instinteraction}\ ] ] from the partition function with the hamiltonian eq .
( [ part ] ) , one can evaluate the time - time gluon propagator composed of the following two parts : @xmath12 where @xmath13 equation ( [ vi ] ) is the instantaneous color - coulomb potential at equal time and causes antiscreening , so that this potential should be a confining potential to attract quarks in hadrons and is the most important quantity on the coulomb gauge confinement scenario .
note that eq .
( [ vi ] ) in the case of quantum electrodynamics ( qed ) as a nonconfining theory is identified as a coulomb propagator @xmath14 or a coulomb potential @xmath15 .
simultaneously , eq . ( [ d00 ] ) stands for a vacuum polarization , @xmath16 which brings about color - screening effect owing to the negative sign , producing the reduction of a color - confining force .
note that this term is associated with a quark - pair creation from vacuum when dynamical quarks exist .
consequently , we can consider that the instantaneous interaction is somewhat more classical than the time - dependent vacuum part .
one can define the color - dependent potentials on a lattice with the polyakov line ( pl ) correlators@xcite . in this study
we separate the original potential obtained from the pl correlators into the color - coulomb instantaneous and color vacuum polarization ( retarded ) parts with the coulomb gauge fixing .
moreover , the color - coulomb potential defined by the link - link correlator as will be described below gives clear signals into practical numerical calculation even in the quenched lattice simulations .
we employ a partial - length polyakov line ( ppl ) which can be defined as @xcite @xmath17 here @xmath18 is an @xmath19 link variable in the temporal direction and @xmath20 , @xmath21 , @xmath22 and @xmath23 represent the lattice cutoff , the gauge coupling , the time component of a gauge potential and the temporal - lattice size .
a ppl correlator in the color - singlet channel is given by @xmath24 \right>,\label{pots}\ ] ] where @xmath25 stands for @xmath26 . from eq .
( [ pots ] ) one evaluates a color - singlet potential on a lattice , @xmath27\label{pot1}.\ ] ] for the smallest temporal - lattice extension , i.e. , @xmath28 , we define @xmath29\label{pot2}.\ ] ] here @xmath30 in the coulomb gauge is assumed to be the color - coulomb instantaneous potential @xmath31 .
the @xmath32 in the limit @xmath33 becomes the usual polyakov line correlator .
these two potentials are expected to satisfy zwanziger s inequality , @xmath34 @xcite , where @xmath35 is the physical potential extracted from the wilson loop .
we apply the above discussion to the other @xmath19 color - dependent potentials between two quarks @xcite .
a color - octet correlator on @xmath1 is given by @xmath36 and @xmath3 correlators in the symmetric - sextet and antisymmetric - triplet channels ( @xmath37 ) are also given as @xmath38 @xmath39 in the same way as described in eqs .
( [ pot1 ] ) and ( [ pot2 ] ) we obtain the color - dependent potentials in each color channel .
the above four potentials are classified in terms of the color ( the quadratic casimir ) factor on color @xmath19 group in the fundamental representation : @xmath40 for color - singlet , color - octet , color - triplet and color - sextet channels , respectively .
these coefficients appear as the proportional constant of the one - gluon exchange potential . and
@xmath3 sectors in relation to a so - called @xmath0 quark model . therefore , in this study , we do not calculate the gauge - invariant wilson - loop potentials in higher representations for @xmath1 sector as introduced in refs .
@xcite ] since the color - dependent potentials defined by the ppl correlators are not gauge invariant , we have to fix a gauge .
we use the coulomb gauge realized on a lattice as @xmath41 by repeating the following gauge rotations : @xmath42 where @xmath43 @xmath44 is a gauge rotation matrix , where the parameter @xmath45 is chosen suitably depending on the lattice size , etc . ] and @xmath46 are spatial lattice link variables .
the thermalized lattice configuration can be gauge fixed iteratively @xcite . because the coulomb gauge fixing does not fully fix a gauge
one can still perform a time - dependent gauge rotation on the coulomb - gauge fixed links , @xmath47 thereby , @xmath48 and @xmath49 constructed by ppl are not invariant under this transformation eq .
( [ gt ] ) .
accordingly , when performing numerical simulations for the octet and two @xmath3 correlators with @xmath48 and @xmath49 , we should additionally implement a global temporal - gauge fixing on the coulomb - gauge fixed links as @xmath50 where @xmath51 is a spatial lattice volume .
note that this gauge fixing does not affect an intrinsic coulomb gauge feature .
in coulomb gauge qcd , we find that infrared divergences cancel only for color - singlet interactions .
let us write the color - charge density for two - quark system as @xmath52 , where @xmath53 are the generators of the color-@xmath19 group .
the instantaneous potential with @xmath54 ( eq . ( [ instinteraction ] ) ) is given by @xmath55 where @xmath56 is the expectation value of the faddeev - popov operator and @xmath57 means the deviations of the factorization , e.g. , @xmath58 if there is no quantum correction ; this argument has been done in refs .
@xcite . here
@xmath59 is reduced to each color factor that appeared in eq .
( [ colorfac ] ) . on the other hand ,
the self - energy interaction not depending on the distance @xmath25 is written as @xmath60 where the casimir invariant @xmath61 in the fundamental representation of the @xmath19 group .
the infrared ( not ultraviolet ) divergence emerges in both eps .
( [ self1 ] ) and ( [ self2 ] ) .
if the term @xmath62 behaves as @xmath63 that would be responsible for the linear confinement @xmath64 , as has been analyzed in the refs .
@xcite , then the infrared singularities unrelated to the linear potential arise from two terms : @xmath65 consequently , they are completely cancelled in the case of the color - singlet representation since @xmath66 . meanwhile ,
the other cases are proportional to the following factors : @xmath67 , @xmath68 , @xmath69 for @xmath70 , @xmath4 and @xmath71 , respectively , implying that the color - sextet channel may diverge most strongly .
we carry out @xmath19 lattice gauge simulations in the quenched approximation to calculate the color - decomposed ppl correlators .
the lattice update is done by the heat - bath monte carlo algorithm with a plaquette wilson gauge action .
the lattice configuration numbers for the @xmath72 , @xmath73 and @xmath74 lattices are 600 , 700 and 320 ; additionally , in order to investigate the volume dependence , we added the @xmath75 and @xmath76 lattices with @xmath77 configurations .
the lattice coupling constant @xmath78 for all the lattices is fixed to @xmath79 corresponding to the lattice cutoff @xmath80 @xcite .
figure [ 4pots ] shows numerical results for the color - coulomb instantaneous potential @xmath30 , in the color - singlet , color - octet , color - sextet , color - triplet ( antisymmetric ) channels .
we find that both the color - singlet @xmath81 and color - antitriplet @xmath82 yield attractions at all distances , and in particular , are linearly rising potential at large distances . on the other hand ,
the color - octet @xmath83 and color - sextet @xmath84 potentials are repulsive forces although the variation of those potentials on distances is small . in order to investigate the magnitude of the string tensions , we calculate an effective string tension for the color - singlet and color - antitriplet channels .
the confining potential can be described by the function @xmath85 where @xmath86 is a self - energy constant term , k corresponds to the string tension , and the last term is the coulomb term . here
the effective string tension is defined as @xmath87 in lattice units , which should be a constant for large quark separations if it is a confining potential with a finite string tension . in fig . [ effstring ]
we find that the ks become stable over approximately @xmath88 as the lattice size increases .
when we use the data for @xmath89 ( @xmath90 for the @xmath72 lattice ) , we plotted in fig .
[ ratios ] the ratio @xmath91 , which is found to be close to @xmath92 ; for example , we obtain the @xmath93 and @xmath94 from the @xmath73 lattice .
note that our definition of the color - coulomb instantaneous part on a lattice in terms of the ppl correlator ( eq .
( [ pot2 ] ) ) does not completely exclude a vacuum polarization effect . for the color - octet and sextet channels it is hard to discuss a precise value of both string tensions under the present statistical accuracy .
in practice , from @xmath73 lattice data and under the fitting range @xmath95 , we obtain the octet string tension @xmath96 and the sextet string tension @xmath97 although the volume dependence does not appear large as shown in fig .
this indicates that the long - range behavior of the octet and sextet channels is quantitatively a minor contribution to two - quark system .
however , in order to work out more quantitative measurement for the repulsive channels ( and also the attractive channels ) , one may need enough statistics on a larger lattice .
here we consider the infrared divergence property of the color - dependent potentials while the string tensions obtained from them have little volume dependence as shown in the previous section .
the short - distance coulomb term @xmath98 , which is not related to the color confinement , may not matter on this argument .
figure [ divfor4pot ] shows our nonperturbative numerical results , by which we find that the color - singlet potential has little volume dependence and the absolute value of the color - nonsinglet potentials increases with the lattice volume ; this tendency remains at the distances @xmath99 and @xmath71 as displayed in fig .
[ divfor4pot2 ] .
thereby we conclude that the color - nonsinglet potentials diverge regardless of the distance .
moreover , in order to estimate the degree of divergence for the color - nonsinglet channels , we assume the fitting function @xmath100 for the potential data in the distance @xmath95 ( long range ) .
the corresponding self - energy term @xmath86s fitted well are plotted in fig .
[ divc ] ; this graph shows that the divergence contribution increases monotonically with the volume .
therefore we adopt the linear function @xmath101 where @xmath102 may correspond to the constant meaning the degree of divergence .
the fitted results are summarized in table i , which gives the conclusion that the magnitude of the divergence on the finite lattice depends on the ratio being @xmath103 for @xmath4 , @xmath70 and @xmath71 as observed in the section ii .
e. however , note that they will diverge equally to infinity in the large volume limit , as shown in figs . [ divfor4pot ] and [ divc ] .
this indicates that the color - nonsinglet quarks can not exist independently ; finally , they will become a color - singlet state with finite energy to compensate insufficient color degrees of freedom .
this numerical result does not contradict the dual - ginzburg - landau picture , in which one understands that the color - singlet flux between two quarks is shrunk like a string while the color - nonsinglet flux will radiate rather than produce a closed string .
a singularity of the color flux not making the closed string can be expected to emerge as the volume dependence on the lattice gauge theory .
.table of fitted results to estimate the degree of divergence in color - nonsinglet channels .
@xmath104 is assumed as a fitting function to fit the data of fig .
[ divc ] . [ cols="^,^,^,^",options="header " , ]
we have studied the long - distance color - dependent forces between two quarks in the quenched @xmath19 lattice simulation with the polyakov line correlator .
we here focus the color - coulomb instantaneous term in coulomb gauge qcd , which has been discussed in the gribov - zwanziger confinement scenario and is required to make the hadron bound state consisting of quarks .
our numerical simulation shows that the color - singlet @xmath1 channel as well as the color - antitriplet @xmath3 ( diquark ) channel causes a linearly confining potential at large distances .
the other color - octet and color - sextet channels at large distances yield also repulsive forces .
in addition , we find that the string tensions in the color - singlet and color - antitriplet channels have not significant volume dependence we also investigated the infrared divergence of color - nonsinglet potentials and find that the divergence on the finite lattice seems to be proportional to the color ( casimir ) factor ; however , the divergence of the color - singlet channel is not left over , in the infinite volume limit the color - dependent potentials except the color - singlet channel will diverge ; in particular , the color - sextet channel diverges most strongly as expected .
this conclusion is consistent with the dual - ginzburg - landau picture of the color confinement .
this approach , in which we take notice of the color - coulomb instantaneous part , may be suitable for further work to investigate three - quark color - dependent forces for the understanding of baryons as well as new multiquark particles .
it is also necessary to study how the vacuum polarization term affects the instantaneous forces because it is reported in refs .
@xcite that the octet ( adjoint ) channel calculated by the full - length polyakov line correlator with the vacuum polarization gives the complicated distance dependence at large distances .
the simulation was performed on sx-5 and sx-8 ( nec ) vector - parallel computers at the rcnp of osaka university .
we appreciate the warm hospitality and support of the rcnp administrators .
in eq . ( [ self1 ] ) , if the coulomb kernel @xmath105 produces a confining force , it is reported in ref .
@xcite that the ghost factor @xmath56 and the laplacian factor @xmath57 behave as approximately @xmath106 and @xmath107 , respectively .
thus we assume that @xmath105 is @xmath108 . omitting @xmath109 term
one also calculates eq .
( [ self1 ] ) : @xmath110 we here do the taylor expansion in the infrared limit @xmath111 for @xmath112 . in the above equation
, the first term disappears because it is a odd function on @xmath113 ; the second term proportional to @xmath108 causes the infrared divergence ; the third term produces a ( linear ) potential ; the other terms are irrelevant .
finally , the second term relating to the divergence is @xmath114 in the same way , the relevant contribution from the self - energy of eq .
( [ self2 ] ) is also given by @xmath115 e. braaten and s. fleming , phys .
lett . * 74 * , 3327 ( 1995 ) , e. braaten and y.q .
chen , phys .
* 76 * , 730 ( 1996 ) , m. cacciari and m. krmer , phys .
. lett . * 76 * , 4128 ( 1996 ) .
r.a briere , et al .
, cleo collaboration , phys .
d70 , 072001 ( 2004 ) .
bodwin , j. lee , and d.k .
sinclair , phys .
d72 , 014009 ( 2005 ) .
k. nagata , a. hosaka , l. j. abu - raddad , phys .
c72 , 035208 ( 2005 ) .
k. nagata and a. hosaka , j. phys .
g32 ( 2006 ) 777 - 786 . l. j. abu - raddad , a. hosaka , d. ebert , h. toki , phys .
c66 , 025206 ( 2002 ) .
k. nagata and a. hosaka , prog .
( 2004 ) 857 - 870 .
a. nakamura and t. saito , prog.theor.phys . 112
( 2004 ) 183 - 188 .
a. nakamura and t. saito , prog.theor.phys . 111
( 2004 ) 733 - 743 .
y. maezawa , n. ukita , s. aoki , s. ejiri , t. hatsuda , n. ishii , and k. kanaya , phys .
d 75 ( 2007 ) 074501 .
m. dring , k. hbner , o. kaczmarek , f. karsch phys.rev . d75 ( 2007 ) 054504 .
d. zwanziger , nucl .
b 518 ( 1998 ) 237 - 272 .
l. baulieu , d. zwanziger , nucl .
b548 ( 1999 ) 527 - 562 .
a. nigawa , phys .
d74 , 045021 ( 2006 ) ; a. nigawa , m. inui , and h. kohyama , phys . rev .
d 74 , 105016 ( 2006 ) .
j. greensite and .
olejnk , phys .
d*67 * , 094503 ( 2003 ) .
j. greensite , .
olejnk and d. zwanziger , phys .
d*69 * , 074506(2004 ) .
a. nakamura , t. saito , prog.theor.phys .
115 ( 2006 ) 189 - 200 .
y. nakagawa , a. nakamura , t. saito , h. toki , d. zwanziger , phys.rev .
d73 ( 2006 ) 094504 .
s. deldar , phys .
d62 , 034509 ( 2000 ) .
bali , phys .
d62 , 114503 ( 2000 ) .
m. faber , j. greensite , .
olejnk , phys .
d57 , 2603 ( 1998 ) .
y. koma , e .-
ilgenfritz , h. toki , and t. suzuki , phys .
d64 , 011501(r ) ( 2001 ) .
|
we study the color - dependent confining forces between two quarks by the quenched lattice simulations of coulomb gauge qcd .
the color - singlet and color - antitriplet instantaneous potentials yield attractive forces .
the ratio of the string tensions obtained from them is approximately 2 and have little volume dependence .
meanwhile , the color - octet and color - sextet channels give a minor contribution for two - quark system .
we finally find that the infrared self - energy of the color - nonsinglet channels diverges in the infinite volume limit ; however , the degree of the divergence on the finite lattice can be understood in terms of color factors .
|
owing to the progress of experimental techniques , information concerning the excited states of light unstable nuclei has rapidly increased .
recently , exotic clustering in the light unstable nuclei has become one of the attractive subjects in 3experimental and theoretical research . since
, in light stable nuclei , it has already been known that clustering is one of the essential features of nuclear dynamics , not only in excited states , but also in ground states , it is natural to expect cluster features in light unstable nuclei .
pioneering theoretical studies have suggested the development of cluster structures with a 2@xmath1 core in be and b isotopes @xcite . especially , highly excited states with developed cluster structures in @xmath4be and @xmath5be have been studied by microscopic calculations @xcite . in the case of @xmath0be , the existence of cluster states was suggested in experimental measurements of the excited states @xcite . in recent experiments of @xmath3he+@xmath3he and @xmath6he+@xmath7he breakup reactions @xcite ,
many new excited states were discovered above the threshold energies .
some of the states are candidates of exotic cluster states , since the measured spin - parities of those excited states indicate a rotational band with a large moment of inertia .
it is an interesting subject to investigate clustering aspects in @xmath0be .
although the molecular states in @xmath0be were theoretically suggested with a potential model with he clusters @xcite , they have not yet been studied by microscopic calculations . in the recent investigations with the generator
coordinate method of a coupled - channel two - body cluster model , using microscopic @xmath3he+@xmath3he and @xmath6he+@xmath7he wave functions , a strong mixing of both configurations was predicted @xcite .
the ground and the low - lying states of @xmath0be present us with other attractive subjects concerning the vanishing of a neutron magic number , 8 .
the vanishing of the neutron magic number is already known in a neighboring nucleus , @xmath5be .
the vanishing in @xmath0be was predicted in the early theoretical works @xcite .
the abnormal configurations(@xmath8 ) in the low - lying states were experimentally studied by @xmath4be@xmath9be reaction @xcite .
the abnormal structure of the ground state was also suggested in the analyses of the beta - decay strength from the ground state of @xmath0be into @xmath0b @xcite .
moreover , the vanishing in @xmath0be was supported by a recent measurement of the spin - parity for a low - lying @xmath10 state @xcite .
a new low - lying @xmath11 state @xcite is also important to solve the inversion mechanism . in the recent theoretical studies ,
the structure of low - lying states were microscopically described by a molecular - orbital model by n. itagaki et al .
@xcite , and by a method of antisymmetrized molecular dynamics by one of the authors ( y.k . ) @xcite .
they were also discussed by f. m. nunes et al . within a three - body ( n+n+@xmath4be ) model with core excitation @xcite .
our aim is to conduct systematic research of the structures of the ground and excited states of @xmath0be based on microscopic theoretical calculations while focusing on clustering aspects .
first of all , we have studied the systematics of the level structures , including the experimentally observed excited states .
we have searched for cluster and non - cluster states to solve the following problems concerning clustering aspects in @xmath0be .
do their cluster structures appear in @xmath0be ?
if they do appear , what are the characteristics of their structures in unstable nuclei compared with those of stable nuclei ?
the roles of the valence nucleons in the cluster states are interesting problems .
we have also investigated the mechanism of the development and breaking of clustering .
the important point is that the theoretical approach should be free from model assumptions , such as the stability of the mean - field and the existence of inert cores or clusters , because we have to describe various structures covering developed cluster structures as well as shell - model - like structures in the ground and excited states .
it is difficult to study developed clustering in excited states with such mean - field approaches as the traditional shell model and the hartree - fock model .
the cluster structures of the excited states of @xmath12be and @xmath4be were successfully explained by cluster models @xcite while assuming a 2@xmath1 core and surrounding neutrons .
however , we think that a description based on cluster models simplifying the system as a 2@xmath1 core with valence neutrons or 2-he clusters is not sufficient for a systematic investigation of @xmath0be because of many valence neutrons in the system . we have applied a theoretical approach of antisymmetrized molecular dynamics ( amd ) . the amd method has already proved to be a useful theoretical approach for the nuclear structure @xcite . within the amd framework
, we do not need such model assumptions as inert cores , clusters , or axial symmetries , because the wave function of the nuclear system is written by slater determinants , where the spatial part of each single - particle wave function for a nucleon is expressed by a localized gaussian wave packet . due to the flexibility of the amd wave function
, we successfully described the structure changes between shell - model - like states and cluster states as functions of the neutron number in light unstable nuclei . owing to progress in computational power ,
it has become possible to study excited states by extended amd calculations .
this method is based on variational calculation after spin - parity projection ( vap calculation ) within the framework of amd , which has already been confirmed to be powerful for studying the excited states of light nuclei , as shown in studies of the stable nucleus @xmath0c @xcite , and also unstable nuclei(@xmath4be and @xmath5be ) @xcite .
the authors and their collaborators succeeded to describe various structures of the excited states and to reproduce many kinds of experimental data for nuclear structures of these nuclei with the amd method . in the present work ,
the structures of the ground and excited states of @xmath0be were analyzed by performing a variational calculation after spin - parity projection based on the amd method . in the next section ( sec .
[ sec : formulation ] ) , we explain the formulation of amd for a study of the nuclear structure of excited states .
the adopted effective interactions are explained in sec .
[ sec : interaction ] . in sec .
[ sec : results ] , we present the calculated results concerning such observables as the energy levels , radii and @xmath13 decays as well as the @xmath14 and @xmath15 transitions compared with the experimental data . in the discussion(sec . [
sec : discuss ] ) , the intrinsic structures and the rotational band structures are described .
the single - particle behavior of the valence neutrons is analyzed .
we discuss the systematics of the development of cluster states in be isotopes , which can be classified according to the neutron orbits surrounding the 2@xmath1 core .
we investigate the inter - cluster motions between he clusters extracted from the obtained @xmath0be wave functions , and calculate the decay widths for the @xmath16he and @xmath17he channels by using the method of reduced width amplitudes .
the formulation of amd for a nuclear structure study of ground and excited states is explained in @xcite . in particular , the formulation of the present calculations is same as that described in ref .
@xcite . in this section ,
i briefly review the formulation .
the wave function of a system is written by a superposition of the amd wave functions(@xmath18 ) .
the amd wave function of a nucleus with a mass number @xmath19 is a slater determinant of gaussian wave packets .
the @xmath20th single - particle wave function is a product of the spatial wave function , the intrinsic spin function and the iso - spin function .
the spatial part is presented by variational complex parameters , @xmath21 , @xmath22 , @xmath23 , which indicate the center of the gauusian wave packets .
the orientation of the intrinsic spin part is expressed by a variational complex parameter @xmath24 , and the iso - spin function is fixed to be up(proton ) or down(neutron ) in the present calculations .
thus , an amd wave function is expressed by a set of variational parameters , @xmath25 , which indicate the centers of gaussians of the spatial part and the spin orientations of the intrinsic spin part of the single - particle wave functions .
when we consider a parity - eigen state projected from an amd wave function , the total wave function is written by two slater determinants . in the case of a total - angular - momentum eigen state ,
the wave function of the state is represented by the integral of the rotated amd wave functions .
the expectation values of a given tensor operator for the total - angular - momentum projected states are calculated by evaluating the integral with a sum over mesh points of the euler angles . in principle
, the total wave function can be a superposition of independent amd wave functions .
we can make the superposition of the spin - parity projected amd wave functions ( @xmath26 ) as follows : @xmath27 we performed a variational calculation for a trial wave function to find the state which minimizes the energy of the system , @xmath28 by the method of frictional cooling , which is one of the imaginary time methods . regarding the frictional cooling method in amd , the reader is referred to two papers @xcite . in order to obtain the wave function for the lowest @xmath29 state
, we varied the parameters @xmath30 and @xmath24(@xmath31 ) to minimize the energy expectation value of a spin - parity projected amd wave function , @xmath32 , by using the frictional cooling method .
that is to say , we performed energy variation after spin - parity projection(vap ) for an amd wave function . with the vap calculation for the @xmath29 eigen state , @xmath33 , with an appropriate @xmath34
, we obtained a set of parameters , @xmath35 , which presents the wave function of the first @xmath29 state . in order to search for the parameters @xmath36 of the @xmath37th @xmath29 state
, the wave functions were superposed so as to be orthogonal to the lower states .
the parameters @xmath38 for the @xmath37th @xmath29 state were provided by varying @xmath39 so as to minimize the energy of the wave function orthogonalized to the lower states .
after the vap calculation of the @xmath40 states for various @xmath41 , @xmath37 and @xmath42 , we obtained the optimum intrinsic states , @xmath43 , which approximately describe the corresponding @xmath40 states . in order to obtain more precise wave functions , we superposed the spin - parity eigen wave functions projected from all of the obtained intrinsic states .
namely , we determined the final wave functions for the @xmath44 states by simultaneously diagonalizing the hamiltonian matrix , @xmath45 ) for all of the obtained intrinsic states , and to ( @xmath46 ) . compared with the experimental data , such as the energy levels and @xmath15 transitions , the expectation values were calculated with the final states after diagonalization
the adopted interaction is the sum of the central force , the spin - orbit force and the coulomb force .
the central force is chosen to be the mv1 force of case 3 @xcite , which contains a zero - range three - body force , @xmath47 , as a density - dependent term in addition to the two - body interaction , @xmath48 , of the modified volkov no.1 force .
the spin - orbit force of the g3rs force @xcite is adopted .
the structures of the excited states of @xmath0be were studied based on the vap calculation within the framework of amd . in this section , we present theoretical results , such as the energy levels , radii , and transitions ( @xmath13 , @xmath15 , @xmath14 and @xmath49 ) , while comparing them with the experimental data . a detailed analysis of the structures of the states is given in the next section .
the adopted interaction parameters in the present work are those used in refs.@xcite , which reproduced the abnormal spin - parity 1/2@xmath50 of the ground state of @xmath5be .
namely , the majorana , bartlett and heisenberg parameters in the central force are @xmath51 , @xmath52 , and the strength of the spin - orbit force is chosen to be @xmath53 mev .
the width parameter ( @xmath54 ) was chosen to be 0.17 @xmath55 , which gave a minimum energy of @xmath0be in a simple amd calculation without the spin projection .
the wave functions for the lowest @xmath29 states were obtained by the vap calculation of @xmath26 by choosing @xmath56= @xmath57 , @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 , @xmath66 @xmath67 , @xmath68 , @xmath69 , @xmath70 , @xmath71 , @xmath72 , @xmath73 . after obtaining the lowest states(@xmath74 ) , we calculated the second and third @xmath29 states ( @xmath11 , @xmath75 , @xmath76 , @xmath77 , @xmath78 , @xmath79 , @xmath80 ) with the vap calculation of the superposed wave functions orthogonal to the obtained lower @xmath29 states .
the obtained amd wave functions are considered to approximately describe the intrinsic states of the corresponding @xmath44 states .
the final wave functions of the @xmath44 states were determined by superposing the spin - parity eigen states projected from these obtained amd wave functions so as to simultaneously diagonalize the hamiltonian matrix and the norm matrix . in principle
, the number of the superposed amd wave functions is 24 , which is the number of calculated levels . in the present calculations
, we omitted several of them which are not important , just to save computational time .
namely , we diagonalize the positive - parity(negative - parity ) states projected from 22(18 ) amd wave functions .
the theoretical binding energy of @xmath0be was found to be 61.9 mev(60.4 mev ) after(before ) diagonalization , which underestimated the experimental value , 68.65 mev .
we found that the binding could be improved by changing the majorana parameter ( @xmath81 ) of the interaction to a smaller value .
for example , a theoretical value of 66.1 mev was obtained by using @xmath82 with a simple vap calculation before diagonalization .
nevertheless , we adopted the parameter @xmath51 in the present work , because this value reproduces the parity inversion of @xmath5be @xcite , which should be important for describing the ground - state properties of the neighboring nucleus , @xmath0be .
we checked that the change in the @xmath81 parameter has no significant effect on the excitation energies , @xmath83 and @xmath84 , of @xmath0be , at least .
in fact , the calculated excitation energies of @xmath83 before diagonalization are 3.0 mev and 2.8 mev , in the cases of @xmath82 and @xmath51 , respectively .
the theoretical level scheme is shown in fig.[fig : be12sped ] .
the calculations suggest that many excited states appear in the low - energy region .
this is the first theoretical work which systematically reproduces the experimental energy levels of all the spin - assigned states , except for the @xmath10 state . by analyzing the intrinsic amd wave functions of the states
, we can consider that there exist rotational bands ( @xmath85=@xmath86 , @xmath11 , @xmath75 and @xmath2 ) , which consist of the following states : \{@xmath86 , @xmath87 , @xmath88 , @xmath89 and @xmath90 } , \{@xmath11 , @xmath76 } , \{@xmath75 , @xmath91 , @xmath78 and @xmath79 } and \{@xmath2 , @xmath92 , @xmath93 , @xmath94 , @xmath95 } , respectively the energy - spin systematics for positive - parity states with natural spins are shown in fig .
[ fig : be12rot ] .
it is surprising that the newly observed levels , @xmath96 at 13.2 mev and @xmath97 at 16.1 mev @xcite , correspond well to the @xmath78 and @xmath79 states obtained in the present results .
an interesting point is that these excited states belong not to the yrast band , @xmath98 , but to the new excited @xmath99 band with developed cluster structure .
the theoretical results predict the existence of many positive - parity states belonging to the @xmath98 and @xmath100 bands in the lower energy region .
even though @xmath0be has a neutron magic number of 8 , the calculated ground @xmath101 band has a large moment of inertia , because the band head @xmath86 state is not the ordinary state with a closed neutron @xmath102-shell , but an intruder state with a developed cluster structure .
it has a prolate deformed structure , which is dominated by @xmath103-@xmath104 configurations , and reaches the band terminal at the @xmath90 state , because the @xmath105 state is the highest - spin state in the 2@xmath106 configurations . the third @xmath107 band ( @xmath99 )
was found to be represented by other @xmath103-@xmath104 configurations than those in the @xmath98 band .
the @xmath99 band has an extremely large moment of inertia because of a remarkably developed @xmath3he+@xmath3he clustering , and reaches the band terminal @xmath79 state accompanying the spin - alignment of nucleons in the high - spin region . on the other hand , the main components of the @xmath11 and @xmath76 states are of the @xmath108 configurations , with the closed neutron @xmath102-shell , and constitute the @xmath100 band .
the band head @xmath11 state of this band , which was theoretically predicted to appear just above the intruder ground band by itagaki et al.@xcite and kanada - enyo et al .
@xcite , was recently discovered in an observation of coincidence gamma rays by shimoura et al .
@xcite .
the theoretical excitation energy of the @xmath2 state is larger than the measured @xmath10 state at 2.68 mev @xcite . in spite of the overestimation of the excitation energy ,
we regard this @xmath2 state in the @xmath109 as the @xmath10 state at 2.68 mev because of the large @xmath14 transition strength , as shown later . in order to improve the excitation energy of this @xmath10 state , it is important to take the mixing of the other configuration into account .
for example , by mixing a @xmath10 state with @xmath110 , which is obtained at slightly higher energy than the @xmath10 state with @xmath109 in vap calculations , the @xmath2 state gains about 1 mev .
another reason for the overestimation may be connected to the fact that the present interactions give a too small value for the effective energy difference between the @xmath111 and @xmath112 orbits , which was found in the calculations of @xmath5be in ref .
@xcite .
= 0.49 = 0.49 the theoretical values of the root - mean - square radii of the density distributions of point - like nucleons , protons , and neutrons are listed in table [ tab : radii ] together with the experimental matter radius @xcite deduced from the reaction cross sections .
because of deformations , the proton and neutron radii in the ground state are larger compared with those in the @xmath11 state . since the theoretical values concerning the difference @xmath113 between the proton and neutron radii are about 0.3 fm , @xmath0be is a candidate of the neutron skin nucleus .
compared with the experimental data , the present result for @xmath86 is larger because the present parameter @xmath51 is considered to be too large to quantitatively reproduce the radii of @xmath102-shell nuclei .
the proton and neutron densities as functions of the radius are shown in fig .
[ fig : rdens ] . in both the @xmath86 and @xmath11 states ,
the excess neutrons enhance the neutron density in the surface region . in the @xmath86 state , due to cluster development , the proton density increases in the surface region while it decreases at the center of the nucleus .
it should be noted that , even if a nucleus has a neutron halo ( the long tail of the neutron density in the outer region ) , the details of the halo structure are not expressed in the present model space because of a limitation of the gaussian forms in amd wave functions . .
[ tab : radii ] the root - mean - square radii of the density distribution of point - like nucleons , protons , and neutrons .
the experimental matter radius deduced from the reaction cross sections is taken from ref.@xcite . [
cols="^,^,^,^ " , ]
we studied the structures of the ground and excited states of @xmath0be based on the framework of the amd method .
this was the first microscopic calculation which systematically reproduced the energy levels of all of the spin - assigned states in @xmath0be , except for the @xmath10 state .
one of the present discoveries is the @xmath99 band with the @xmath3he+@xmath3he cluster structure , which corresponds well to the states recently observed in the he - he break - up reactions .
the ground - state properties , such as the @xmath13-decay strength and the @xmath14 transition strength , were reproduced by the present calculations , which suggest that the ground state is dominated by a deformed intruder state with the developed cluster structure .
the theoretical results predicted many low - lying excited states .
it was found that most of them have a 2@xmath1 core , while the @xmath114 state has a no @xmath1-cluster structure .
we analyzed the structures and the single - particle wave functions of the intrinsic states , and found rotational bands with @xmath115 , which are dominated by the @xmath116 , @xmath108 , @xmath117 , @xmath116 and @xmath116 configurations , respectively .
the @xmath116 and @xmath117 states have deformed structures with a 2@xmath1 core . in the positive - parity orbits occupied by the valence neutrons
, we found the molecular orbits , @xmath118-orbits and the new @xmath119-orbits .
the positive - parity molecular orbits play an important role in the development of cluster structure .
especially , the reason for the intruder ground state can be understood by the energy gain of the @xmath118-orbit in the developed cluster state .
although the molecular orbits are helpful to understand the structures of the band head states , the model of the 2@xmath1 and 4 neutrons in the molecular orbits is too simple to describe all of the excited states , because such phenomena as the intrinsic spin alignment , the dissociation or breaking of 2@xmath1 core and the mixing of the molecular orbits appear in @xmath0be .
we discussed the inter - cluster motions in the @xmath3he+@xmath3he and @xmath6he+@xmath7he channels , and estimated the partial width of these decay channels with the method of reduced width amplitudes .
the analyzed results of the partial widths strongly suggest that the neutrons of the @xmath99 band move around either of two @xmath1 clusters , while the neutrons of the @xmath98 band move over the whole system .
we also discussed the systematics of the cluster development in the neutron - rich be isotopes according to the classification of the states with the number of neutrons in the @xmath118-orbits . concerning the vanishing of the neutron magic number @xmath120 in
the be isotopes , we suggested that the reason for the intruder ground states may be explained by the restoration of the natural distance between the 2 @xmath1 clusters .
the authors would like to thank dr .
n. itagaki for many discussions .
they are also thankful to prof .
w. von oertzen for helpful comments .
valuable comments of prof .
s. shimoura and dr .
a. saito are also acknowledged .
the computational calculations in this work were supported by the supercomputer project nos .
58 and 70 of high energy accelerator research organization(kek ) , and also supported by research center for nuclear physics in osaka university and yukawa institute for theoretical physics in kyoto university .
this work was partly supported by japan society for the promotion of science and a grant - in - aid for scientific research of the japan ministry of education , science and culture .
this work was partially performed in the `` research project for study of unstable nuclei from nuclear cluster aspects '' sponsored by institute of physical and chemical research ( riken ) .
9 m. seya , m. kohno , and s. nagata , prog .
theor . phys . * 65 * , 204 ( 1981 ) .
y. kanada - enyo , h. horiuchi and a. ono , phys .
c * 52 * , 628 ( 1995 ) ; y. kanada - enyo and h. horiuchi , phys .
c * 52 * , 647 ( 1995 ) . w. von oertzen , z. phys .
a * 354 * , 37 ( 1996 ) ; * 357 * , 355(1997 ) .
k. arai , y. ogawa , y. suzuki and k. varga , phys .
c * 54 * , 132 ( 1996 ) .
a. dot , h. horiuchi , and y. kanada - enyo , phys .
c * 56 * , 1844 ( 1997 ) .
y. kanada - enyo , h. horiuchi and a. dot , j. phys .
g , nucl . part .
* 24 * 1499 ( 1998 ) .
y. kanada - enyo , h. horiuchi and a. dot , phys .
* c 60 * , 064304(1999 ) .
n. itagaki and s. okabe , phys .
c * 61 * , 044306 ( 2000 ) ; n. itagaki , s. okabe and k. ikeda , phys . rev .
c * 62 * , 034301 ( 2000 ) .
y. ogawa , k. arai , y. suzuki , and k. varga , nucl . phys . * a673 * 122 ( 2000 ) .
y. kanada - enyo and h. horiuchi , phys .
c * 66 * , 024305(2002 ) .
y. kanada - enyo , phys . rev .
* c 66 * , 011303(2002 ) .
korsheninnikov , et al . ,
* b * 343 , 53(1995 ) . m. freer , et al .
lett . * 82 * , 1383 ( 1999 ) ; m. freer , et al . ,
c * 63 * , 034301 ( 2001 ) .
a. saito , et al .
sympo . on clustering aspects of quantum many - body systems _ , eds a. ohnishi , n. itagaki , y. kanada - enyo and k. kato , ( world scientific publishing co. ) . m. ito , y. sakuragi , y. hirabayashi , phys.rev.*c 63 * , 064303(2001 ) .
p. descouvement and d. baye , phys .
b * 505 * , 71(2001 ) .
f. c. barker , j. phys .
g , nucl . part . phys . * 2 * , l45 ( 1976 ) .
h. t. fortune and g .- b .
liu , phys . rev . * c * 50 , 1355 ( 1994 ) .
t. suzuki and t. otsuka , phys .
rev . * c * 56 , 847(1997 ) .
h. iwasaki , et al .
phys.lett.*b*491 , 8(2000 ) .
s. shimoura , et al . ,
preprint , cns - rep-47(2002 ) .
f. m. nunes , i. j. thompson , j. a. tostevin , nucl .
a703 * , 593 ( 2002 ) .
s. okabe , y. abe , and h. tanaka , prog .
phys.*57 * , 866(1977 ) ; s. okabe , y. abe , prog .
phys.*61 * , 1049(1979 ) .
y. kanada - enyo and h. horiuchi , prog .
. phys . * 93 * , 115 ( 1995 ) .
y. kanada - enyo , phys .
lett . * 81 * , 5291 ( 1998 ) .
t. ando , k.ikeda , and a. tohsaki , prog .
. phys . * 64 * , 1608 ( 1980 ) .
n. yamaguchi , t. kasahara , s. nagata , and y. akaishi , prog .
. phys . * 62 * , 1018 ( 1979 ) ; r. tamagaki , prog .
. phys . * 39 * , 91 ( 1968 ) .
i. tanihata et al . , phys.lett.b 206 , 592(1988 ) .
chou , e. k. warburton and b. a. brown , phys . rev .
c*47 * , 163(1993 ) .
w. von oertzen , nuovo cimento 110 , * 895 * , ( 1997 ) .
y. kanada - enyo and h. horiuchi , prog .
suppl.*142 * , 205(2001 ) . h. horiuchi , chapter iii of suppl . of prog .
no.62(1977 ) .
|
the structures of the ground and excited states of @xmath0be were studied with antisymmetrized molecular dynamics .
the ground state was found to be a state with a developed 2@xmath1 core with two neutrons occupying the intruder orbits .
the energy levels of the newly measured spin - assigned states were described well , except for the @xmath2 state .
the calculations indicated that many exotic cluster structures appear in the low - energy region .
the widths concerning @xmath1 and @xmath3he decays were discussed by using reduced width amplitudes .
|
in this paper , we are concerned about the construction problem of homologically interesting connected quasi - fuchsian subsurfaces in closed hyperbolic @xmath0-manifolds .
we show that in a closed hyperbolic @xmath0-manifold , it is always possible to construct an oriented compact connected @xmath1-injectively immersed quasi - fuchsian subsurface which is virtually bounded by prescribed multicurves and which virtually represents a prescribed rational relative second homology class ( theorem [ main - qfsurface ] ) .
the following two results are motivational special cases of theorem [ main - qfsurface ] . for simplicity we state them first .
[ qfsurfaceclosed ] every rational second homology class of a closed hyperbolic @xmath0-manifold has a positve integral multiple represented by an oriented connected closed @xmath1-injectively immersed quasi - fuchsian subsurface .
[ qfsurfaceboundary ] every rationally null - homologous , @xmath1-injectively immersed oriented closed @xmath2-submanifold in a closed hyperbolic @xmath0-manifold has an equidegree finite cover which bounds an oriented connected compact @xmath1-injective immersed quasi - fuchsian subsurface . here
the closed @xmath2-submanifold being @xmath1-injectively immersed means that all components are homotopically nontrivial , and a finite cover being equidegree means that the covering degree does not vary over different components of the @xmath2-submanifold
. however , we do not require the finite cover to be connected restricted to any component of the closed @xmath2-submanifold .
corollary [ qfsurfaceclosed ] was a question that was recently ( and informally ) raised by william thurston .
note that if not requiring the subsurface to be connected , one may easily obtain a componentwise quasi - fuchsian embedded incompressible subsurface representing a second homology class that is nontrivial and non - fibered , or obtain a componentwise @xmath1-injectively immersed quasi - fuchsian representative subsurface , using the cooper long reid construction @xcite in the fibered case or the kahn
markovic construction @xcite in the trivial case . in the paper @xcite
, danny calegari and alden walker show that in a random group at any positive density , many second homology classes can be rationally represented by quasiconvex ( closed ) surface subgroups ( cf .
remark 6.4.2 of @xcite ) .
corollary [ qfsurfaceboundary ] answers a question of calegari in the case of closed hyperbolic @xmath0-manifold groups .
calegari proved the surface group case @xcite but his question remains widely open for hyperbolic groups in general .
next , we state our main result theorem [ main - qfsurface ] .
a compact immersed subsurface @xmath3 of a closed hyperbolic @xmath0-manifold @xmath4 is _ quasi - fuchsian _ if it is an essential subsurface of a closed @xmath1-injectively immersed quasi - fuchsian subsurface of @xmath4 .
perhaps it would be better to call @xmath3 ` quasi - schottky ' if it is quasi - fuchsian with nonempty boundary .
[ main - qfsurface ] let @xmath4 be a closed hyperbolic @xmath0-manifold , and @xmath5 be the ( possibly empty ) union of finitely many mutually disjoint , @xmath1-injectively embedded loops .
then for any relative homology class @xmath6 , there exists an oriented connected compact surface @xmath3 , and an immersion of the pair @xmath7 such that @xmath8 is @xmath1-injective and quasi - fuchsian , and that @xmath3 represents a positive integral multiple of @xmath9 .
the reader is referred to subsection [ subsec - descriptionoftheproblem ] for more explanation about the formulation .
in fact , the proof also implies that the claimed immersed subsurface is nearly geodesic and nearly regularly panted ( cf .
section [ sec - methodology ] ) .
in the course of proving theorem [ main - qfsurface ] , we revisit the techniques developed in the work of kahn markovic in @xcite , with an attempt to distill the topological ingredients from those papers .
specifically , we quote theorems 2.1 , 3.4 , and 4.2 from @xcite as black boxes , so details in quasi - conformal geometry or dynamics of frame flow are not involved in our proofs . in order to prove results in this paper
we are not required to generalize the quantitative aspect of the good correction theory , which is thereom 3.3 ( 3 ) of @xcite , so randomization techniques are not required for our discussion . on the other hand , we recall the gluing construction of @xcite , and reform the topological part of the good correction theory of @xcite .
the treatment of this paper is completely self - contained except for the quoted results from @xcite .
the connectedness of the surface @xmath3 in the conclusion of theorem [ main - qfsurface ] comes from improving the gluing construction of @xcite .
the idea of the construction of @xcite is to build a closed @xmath1-injectively immersed quasi - fuchsian subsurface in a closed hyperbolic @xmath0-manifold by gluing a sufficiently large finite collection of nearly regular pairs of pants with nearly evenly distributed feet .
a crucial criterion was proved in ( * ? ? ?
* theorem 2.1 ) , asserting that a nearly unit shearing gluing yields the @xmath1-injectiveness and the quasi - fuchsian property . in section [ sec - methodology ] , we will review the program in more details with emphasis on the boundary operator on measures of nearly regular pairs of pants .
however , the criterion of @xcite does not necessarily produce a connected surface , so we provide a slightly stronger criterion ( theorem [ gluing ] ) which ensures connectedness of the output .
the new criterion will be proved in section [ sec - quasifuchsianconnectedgluing ] by applying a trick called _
hybriding_. on the other hand , the assumptions of the new criterion are not hard to be satisfied , for instance , cf .
theorem [ homologyviapants ] .
the control of the homology class of the surface @xmath3 in the conclusion of theorem [ main - qfsurface ] comes from extending and strengthening the non - random good correction theory of @xcite in the @xmath0-dimensional case . for an oriented closed hyperbolic @xmath0-manifold @xmath4
, we will reformulate the good pants homology introduced in @xcite as the _ nearly regularly panted cobordism group _ @xmath10 ( definition [ pantedcobordismgroup ] ) . in section [ sec - pantedcobordismgroup ] , we will find a canonical isomorphism @xmath11 between @xmath10 and the first integral homology of the special orthonormal frame bundle @xmath12 over @xmath4 ( theorem [ theorempantedcobordism ] ) .
this isomorphism fully characterizes the structure of @xmath10 , and improves the treatment of non - random good correction theory of @xcite in that it accounts for the torsion part which was previously ignored . in section [ sec - pantifyingsecondhomologyclasses ]
, we will further show that any second integral homology class of @xmath4 can be represented by an oriented closed nearly regularly panted subsurface ( theorem [ secondhomologyclass ] ) . with an extra property called _
nearly regularly panted connectedness _ introduced in section [ sec - pantedconnectedness ] , our study of nearly regularly panted cobordisms can be summarized by the following theorem [ main - pantedsurface ] , stated in a form analogous to theorem [ main - qfsurface ] ( cf .
section [ sec - methodology ] for the notations ) . note that corollaries [ qfsurfaceboundary ] and [ qfsurfaceclosed ] are also parallel to theorems [ theorempantedcobordism ] and [ secondhomologyclass ] in their statements respectively .
these results are all based on geometric constructions using @xmath13-framed segments as we will study in section [ sec - basicconstructions ] .
[ main - pantedsurface ] let @xmath4 be closed hyperbolic @xmath0-manifold .
for any universally small positive constant @xmath14 and any sufficiently large positive constant @xmath15 depending only on @xmath4 and @xmath14 , the following holds .
there exists a nontrivial invariant @xmath16 valued in @xmath17 , defined for all null - homologous oriented @xmath18-multicurve @xmath19 in @xmath4 , satisfying the following . 1 . for any null - homologous oriented @xmath18-multicurve @xmath20 , @xmath21 2 .
the invariant @xmath16 vanishes if and only if @xmath19 bounds a connected compact oriented @xmath18-panted subsurface @xmath3 immersed in @xmath4 .
when @xmath16 vanishes , every relative homology class @xmath22 with @xmath23 equal to the fundamental class @xmath24\in h_1(l;{{\mathbf z}})$ ] is represented by a connected compact oriented @xmath18-panted immersed subsurface @xmath3 bounded by @xmath19 .
in fact , the universal bound for @xmath14 can be explicitly taken to be @xmath25 .
note that the @xmath18-panted immersed subsurface @xmath3 is not required to be @xmath1-injective or quasi - fuchsian .
the invariant @xmath16 is defined as @xmath26_{r,\epsilon})$ ] , where @xmath24_{r,\epsilon}$ ] is the @xmath18-panted cobordism class of @xmath19 , so @xmath16 lies in a canonical submodule of @xmath27 isomorphic to @xmath17 .
the proofs of theorems [ main - qfsurface ] and [ main - pantedsurface ] will be completed in section [ sec - boundedquasifuchsiansubsurfaces ] . a few further questions will be proposed in section [ sec - conclusions ] . * acknowledgement*. the authors thank hongbin sun for pointing out errors in a previous draft of this paper , and danny calegari for valuable comments .
the authors also thank the anonymous referees for suggestions and corrections .
for a typical construction problem of quasi - fuchsian subsurfaces in a closed hyperbolic @xmath0-manifold , such as addressed in theorem [ main - qfsurface ] , one may generally follow two steps : first , decide a suitable finite collection of ( oriented ) nearly geodesic pairs of pants whose cuff lengths are nearly equal ; secondly , glue these pairs of pants up along boundary in a well controlled fashion to output a connected quasi - fuchsian subsurface .
the second step is supposed to be automatic once we have fed in the collection of pairs of pants as initial data , so the real task is to provide such a collection . regarding the collection as a finite measure over the set of pairs of pants , we will translate the compatibility condition for the gluing into a linear system of equations of that measure , and we will introduce properties on solutions to ensure a suitable gluing . in other words , we will be interested in certain solutions of the linear system of equations associated to a _ boundary operator _ on measures of pants .
the purpose of this section is to set up the framework , and to divide the discussion into several aspects that can be treated separatedly in the rest of this paper . throughout this subsection ,
@xmath4 will be a closed hyperbolic @xmath0-manifold . identifying the universal cover @xmath28 of @xmath4 as the @xmath0-dimensional hyperbolic space @xmath29
, we will regard the deck transformation group @xmath30 as a torsion - free cocompact discrete subgroup of the group of isometries @xmath31 .
let @xmath32 be the topological circle with a fixed orientation .
an _ oriented curve _ in @xmath4 , or simply a _ curve _ , is the free homotopy class of a @xmath1-injective immersion @xmath33 .
we often abuse the notations for curves and their representatives , and write a curve as @xmath34 every curve can be homotoped to a unique oriented closed geodesic in @xmath4 with the length parametrization up to a rotation , so we define the _ visual torus _
@xmath35 of @xmath36 to be the unit normal vector bundle of the geodesic representative .
we think of the visual torus to be a holomorphic torus , and the name comes from the fact that we may alternatively define @xmath35 as follows .
let @xmath37 be any elevation of a curve @xmath36 in @xmath29 .
as @xmath37 is a quasi - geodesic with endpoints @xmath38 on the sphere at infinity @xmath39 , we may define @xmath35 to be the holomorphic cylinder @xmath40 quotiented by the stabilizer @xmath41 .
the two definitions of @xmath35 are certainly equivalent , but the latter might be more natural from a perspective of geometric group theory . the collection of curves in @xmath4 will be denoted as @xmath42 , or simply @xmath43 .
the quotient of @xmath43 under the free involution induced by orientation reversion of curves is the collection of _ unoriented curves _ in @xmath4 , and we will denote it as @xmath44 .
let @xmath45 be a topological pair of pants , namely , a compact three - holed sphere . for convenience
, we will fix an orientation of @xmath45 .
an _ unmarked oriented pair of pants _ in @xmath4 , or simply a _
pair of pants _ , is the homotopy class of a @xmath1-injective immersion @xmath46 , up to orientation - preserving self - homeomorphisms of @xmath45 .
we often abuse the notations for homotopy classes and their representatives , and write a pair of pants as @xmath47 the _ cuffs _ of @xmath45 are the three boundary curves of @xmath45 , and the _ seams _ of @xmath45 are three mutually disjoint , properly embedded arcs connecting the three pairs of cuffs , which are unique up to orientation - preserving self - homeomorphisms of @xmath45 .
every pair of pants can be homotoped so that the cuffs are the unique geodesic closed curves , and that the seams are the unique geodesic arcs orthogonal to the adjacent cuffs , or possibly points in the degenerate case .
we say a pair of pants in @xmath4 is _ nonsingular _ if no seam degenerates to a point under the straightening as above .
the collection of nonsingular pants in @xmath4 will be denoted as @xmath48 , or simply @xmath49 .
the quotient of @xmath49 under the free involution induced by orientation reversion of pants is the collection of _ unoriented nonsingular pants _ in @xmath4 , and we will denote it as @xmath50 .
suppose @xmath51 is a nonsingular pair of pants , straightened so that the cuffs are geodesic and the seams are geodesic and orthogonal to the cuffs .
for every pair of cuffs @xmath36 and @xmath52 , the seam @xmath53 from @xmath36 to @xmath52 defines a unit normal vector @xmath54 at @xmath36 , pointing along @xmath53 towards @xmath52 .
we call @xmath55 the _ foot _ of @xmath56 at @xmath36 toward @xmath52 , and it is the ` visual direction of the nearest point ' as we observe @xmath52 from @xmath36 .
there are exactly six feet of @xmath56 , two at each cuff toward the other two cuffs respectively . throughout this paper , a measure is always considered to be nonnegative .
let @xmath57 denote all finitely - supported finite measures on the set of nonsingular pants @xmath49 in @xmath4 .
we usually write a nontrivial element of @xmath57 as a finite formal sum of elements of @xmath49 with positive coefficients .
similarly , let @xmath58 denote all finitely - supported finite measures on the set of curves @xmath43 in @xmath4 .
there is a natural _ boundary operator _
@xmath59 defined by assigning @xmath60 to be the sum of the cuffs of @xmath56 .
we will consider two related notions : the footed boundary @xmath61 , which is a geometric refinement of @xmath13 remembering the feet ; and the net boundary @xmath62 , which is an algebraic reduction of @xmath13 forgetting the orientation .
let @xmath63 denote all borel measures on the visual torus of any curve @xmath36 in @xmath4 , and let @xmath64 denote the direct sum of @xmath63 as @xmath36 runs over all curves @xmath43 .
the _ footed boundary operator _ is the homomorphism : @xmath65 defined by assigning @xmath66 to be one half of the sum of the atomic measures supported at the six feet of @xmath56 , where @xmath67 is any nonsingular pants .
the normalization coefficient @xmath68 has been chosen so that the total measure on @xmath35 of each cuff @xmath36 is equal to @xmath2 .
let @xmath69 denote all finite measures on the set of unoriented curves @xmath44 in @xmath4 .
we identify @xmath69 as the subspace of @xmath58 fixed under the free involution induced by the orientation reversion @xmath70 , in other words , regard the atomic measure supported on the unoriented class @xmath71 as the measure @xmath72 .
the _ net boundary operator _ is the homomorphism : @xmath73 defined by @xmath74 if we regard @xmath75 as the subspace of @xmath57 fixed under the orientation reverion , then @xmath75 lies in the kernel of @xmath62 .
we have the following commutative diagram relating various operators : @xmath76^{\mathrm{tot}}\\ { { \mathcal{m}}}({{\mathbf{\pi } } } ) \ar[ru]^{{{\partial^\sharp } } } \ar[r]^{\partial } \ar[rd]_{{{\partial^\flat } } } & { { \mathcal{m}}}({{\mathbf{\gamma } } } ) \ar[d]^{\mathrm{net } } \\ & { { \mathcal{m}}}(|{{\mathbf{\gamma}}}| ) } \ ] ] here @xmath77 is the componentwise total @xmath78 , and @xmath79 is the unorientation reduction defined by linearly extending @xmath80 for all @xmath81 . for most of our treatment
we will focus on pairs of pants in @xmath4 that are nearly geodesic with cuffs of nearly equal length , or _ nearly regular pants _ as we will introduce below .
first recall that for a boundary - framed segment in @xmath29 ( with the canonical orientation ) , the ( oriented geometric ) _ complex length _ of it can be defined as a complex value in @xmath82\,{{\mathrm{i}}}.\ ] ] more precisely , an oriented _ @xmath13-framed segment _ is an oriented geodesic arc with a unit normal vector at each endpoint , so the real part of the complex length is the usual length of the geodesic arc , and the imaginary part is the signed angle from the initial normal vector to the parallel transportation of the terminal normal vector to the initial point of the geodesic arc , with respect to the initial tangent vector . the complex length does not change if we reverse the orientation of the @xmath13-framed segment .
it is clear that the complex length of @xmath13-framed segments also makes sense in any oriented hyperbolic @xmath0-manifold @xmath4 . for a geodesic loop in @xmath4
, we may pick a normal vector at a point , and define the _ complex length _ of the geodesic loop as the complex length of the boundary - framed segment obtained from cutting the geodesic loop along the chosen point , and endowing both the end - points with the same chosen normal vector .
the definition is clearly independent of the choices of the point or the normal vector .
let @xmath4 be a closed hyperbolic @xmath0-manifold .
suppose @xmath51 be a nonsingular pair of pants , straightened by homotopy so that the cuffs and seams are geodesic and orthogonal as before .
observe that each cuff @xmath36 of @xmath56 is bisected into two boundary - framed segments with the boundary framing given by the two feet .
in fact , these two boundary - framed segments , called _ half cuffs _ , have the same complex length . we define the _ complex half length _ of the cuff @xmath36 of @xmath56 to be the complex length of either of the half cuffs , denoted as @xmath83 .
we denote the complex length of @xmath36 as @xmath84 .
when the imaginary part of @xmath83 is at most the right angle , @xmath84 is equal to twice @xmath83 .
[ nearlyregular ] let @xmath4 be a closed hyperbolic @xmath0-manifold .
suppose @xmath18 is any pair of positive constants . 1 .
we say that a curve @xmath85 is _ @xmath18-nearly hyperbolic _ , if @xmath86 the subcollection of @xmath18-nearly hyperbolic curves in @xmath4 will be denoted as @xmath87 2 .
we say that a nondegenerate pair of pants @xmath51 is _ @xmath18-nearly regular _ , if for each cuff @xmath36 of @xmath56 , @xmath88 the subcollection of @xmath18-nearly regular pants in @xmath4 will be denoted as @xmath89 .
we often simply say _ nearly hyperbolic _ or _ nearly regular _ with the usage explained in the following convention [ nearly ] .
[ nearly ] when ambiguously saying _ nearly _ instead of _
@xmath18-nearly _ , we suppose that @xmath18 are understood from the context . presumably , @xmath14 will be universally small , and @xmath15 will be sufficiently large , depending on @xmath4 and @xmath14 .
this precisely means that for some universal constant @xmath90 to be determined , @xmath14 is assumed to satisfy @xmath91 , and that for any given closed hyperbolic @xmath0-manifold @xmath4 , and for some constant @xmath92 to be determined , @xmath15 is assumed to satisfy @xmath93 . from definition [ nearlyregular ]
, it follows that the restriction of the boundary operator yields : @xmath94 and similarly for @xmath95 and @xmath96 . for any finite collection of pairs of pants in a closed hyperbolic @xmath0-manifold @xmath4 , we can try to glue them along common cuffs with opposite induced orientations , and this will give rise to a panted subsurface in @xmath4 , precisely as follows .
[ pantedsubsurfacedefinition ] an _ @xmath18-nearly regularly panted subsurface _ of @xmath4 , or simply an _ @xmath18-panted subsurface _ , is a ( possibly disconnected ) compact oriented surface @xmath3 with a pants decomposition , and with an immersion @xmath97 into @xmath4 such that the restriction of @xmath8 to each component pair of pants is @xmath18-nearly regular .
let @xmath98 be the integral measure such that for each @xmath99 , there are exactly @xmath100 copies of @xmath56 in all component pairs of pants of @xmath3 immersed via @xmath8 .
then we say that the @xmath18-panted subsurface is _ subordinate to _ @xmath101 .
in general , the panted subsurface would be neither @xmath1-injective quasi - fuchsian nor connected .
however , we wish to introduce conditions on @xmath101 to ensure that some quasi - fuchsian connected panted subsurface therefore exists and is subordinate to @xmath101 .
recall that for a metric space @xmath102 , and for a positive number @xmath103 , two borel measures @xmath104 are said to be _ @xmath103-equivalent _ , if @xmath105 equals @xmath106 , and if for every borel subset @xmath107 of @xmath108 , @xmath109 , where @xmath110 is the @xmath103-neighborhood of @xmath107 .
note that @xmath103-equivalence is a symmetric relation . for any nonvanishing finite borel measure @xmath101 on @xmath108 , we may speak of @xmath103-equivalence after normalization , namely , after dividing @xmath101 by @xmath105 .
[ gluingconditions ] let @xmath4 be a closed hyperbolic @xmath0-manifold . let @xmath98 be a measure of nearly regular pants . 1 .
we say that @xmath101 is _ ubiquitous _ , if @xmath101 is positive at every @xmath99 , and if @xmath111 is positive at every @xmath112 .
we say that @xmath101 is _ irreducible _ , if for any nontrivial decomposition @xmath113 , @xmath114 and @xmath115 have adjacent supports , namely , that there is a curve @xmath112 which lies in the support of @xmath116 and the orientation - reversal of which lies in the support of @xmath117 .
we say that @xmath101 is _ @xmath18-nearly evenly footed _ ,
if for every curve @xmath112 on which @xmath111 is nonvanishing , the normalization of @xmath118 is @xmath119-equivalent to the normalization of the lebesgue measure , with respect to the euclidean metric on the visual torus @xmath35 induced from the unit normal vector bundle of the geodesic representative of @xmath36 .
we often simply say _ nearly evenly footed _ following convention [ nearly ] .
we say that @xmath101 is _ rich _ , if the net boundary of @xmath101 at any unoriented curve is a relatively small portion compared to the cancelled part , or specifically for our application , that @xmath120 . here
@xmath121 means the unoriented class @xmath71 for any curve @xmath122 .
the following criterion about connected quasi - fuchsian gluing will be proved in section [ sec - quasifuchsianconnectedgluing ] .
[ gluing ] let @xmath4 be a closed hyperbolic @xmath0-manifold . for any universally small positive @xmath14 and any sufficiently large positive @xmath15 depending on @xmath4 and @xmath14 , the following statement holds .
for any nontrivial rational measure @xmath98 , if @xmath101 is irreducible , @xmath18-nearly evenly footed , and rich , then there exists an oriented , connected , compact , @xmath1-injectively immersed quasi - fuchsian subsurface : @xmath123 which is @xmath18-nearly regularly panted subordinate to a positive integral multiple of @xmath101 . for a closed hyperbolic @xmath0-manifold @xmath4
, we wish to understand the structure of the boundary operator @xmath124 .
more specifically , the following theorem [ homologyviapants ] should be viewed from this perspective .
let @xmath125 be a collection of distinct curves , invariant under orientation reversion .
we write @xmath126 for the corresponding unoriented curves , namely , the quotient of @xmath127 by orientation reversion .
let @xmath128 denote the subset of @xmath129 consisting of measures @xmath101 with the net boundary @xmath130 supported on ( possibly a proper subset of ) the unoriented curves @xmath131 .
choosing a collection of mutually disjoint , embedded unoriented loops @xmath132 representing elements of @xmath131 , we write @xmath133 for @xmath134 .
note that @xmath133 is well defined up to natural isomorphisms for different choices of the loops @xmath135 .
thus there is a natural homomorphism between semimodules over the semiring of nonnegative real numbers : @xmath136\,:\,{{\mathcal{zm}}}({{\mathbf{\pi}}}_{r,\epsilon},|\mathcal{l}|)\to h_2(m,|\mathcal{l}|;\,{{\mathbf r}}).\ ] ] in particular , when @xmath127 is empty , we denote the kernel of the homomorphism @xmath137 $ ] above as @xmath138 nevertheless , @xmath139 is naturally contained in @xmath140 for any @xmath127 as well .
[ homologyviapants ] let @xmath4 be a closed hyperbolic @xmath0-manifold . for any sufficiently small positive @xmath14 depending on @xmath4 , and any sufficiently large positive @xmath15 depending on @xmath4 and @xmath14 , the following statements hold .
suppose @xmath125 is a collection of distinct curves invariant under orientation reversion .
there is a short exact sequence of semimodules over the semiring of nonnegative real numbers : @xmath141 2 .
there exists a nontrivial measure @xmath142 which is ubiquitous , irreducible , @xmath18-nearly evenly footed , and rich .
moreover , every measure in @xmath140 can be adjusted to satisfy the same properties , by adding some measure in @xmath139 . furthermore ,
the same statements hold for rational coefficients instead of real coefficients as well .
in fact , the constant @xmath14 is required to be bounded by the injectivity radius of @xmath4 so as to guarantee the existence of @xmath143 .
theorem [ homologyviapants ] will be proved in section [ sec - boundedquasifuchsiansubsurfaces ] . as theorem [ homologyviapants ] feeds theorem [ gluing ] with workable input , homologically
interesting connected quasi - fuchsian subsurfaces can be produced in closed hyperbolic @xmath0-manifolds under fairly general conditions .
in this section , we prove theorem [ gluing ] , restated as proposition [ gluingprop ] in terms of gluing .
let @xmath4 be a closed hyperbolic @xmath0-manifold , and let @xmath18 be a pair of positive constants
. given a panted surface @xmath3 of which the pants structure is given by a union of disjoint simple closed curves @xmath144 , we may cut @xmath3 along @xmath145 to obtain a disconnected surface @xmath146 whose components are all pairs of pants .
denote the union of all the new boundary components of @xmath146 coming from the cutting as @xmath147 .
then the panted surface @xmath3 can be recovered as the quotient of @xmath146 by an orientation - reversing involution @xmath148 , which sends any preimage component of @xmath145 to its opposite boundary component . with this in mind , we introduce the notion of gluing as follows . for any integral measure @xmath98 ,
let @xmath146 be the finite disjoint union of copies of nearly regular pants prescribed by @xmath101 , namely , such that for any @xmath99 , there are exactly @xmath100 copies of @xmath56 in @xmath146 . by a _ gluing _ of @xmath146 , we mean a pair @xmath149 , such that @xmath150 is a subunion of cuffs , and that @xmath151 is a free involution which sends each cuff @xmath152 to its orientation - reversal , regarded as in @xmath153 .
we say that @xmath149 is _ maximal _ if @xmath154 can not be extended to any subunion of cuffs @xmath155 larger than @xmath156 .
since the quotient of @xmath146 by @xmath154 yields a compact oriented @xmath18-panted subsurface @xmath157 , the quotient image of any cuff @xmath152 in @xmath3 will be called a _ glued cuff_. [ gluingprop ] let @xmath4 be a closed hyperbolic @xmath0-manifold . for any universally small positive @xmath14 and any sufficiently large positive @xmath15 depending on @xmath4 and @xmath14 ,
the following holds .
if a rational nontrivial pants measure @xmath98 is irreducible , @xmath18-nearly evenly footed , and rich , then possibly after passing to a positive integral multiple of @xmath101 , the prescribed oriented compact surface @xmath146 admits a maximal gluing @xmath149 , which yields a @xmath1-injectively immersed , quasi - fuchsian , and connected subsurface @xmath97 . the key technique to ensure the connectedness of the resulting surface is a trick called _ hybriding_. to illustrate the idea , the reader may assume for simplicity that @xmath130 is zero , so that any panted surface @xmath3 resulted from a maximal gluing is closed .
we say that a gluing is _ nearly unit shearing _ , if for any glued cuff @xmath158 on the resulting @xmath18-panted surface @xmath3 , the feet of the pair of pants on one side of @xmath158 is almost exactly opposite to the feet of the pair of pants on the other side of @xmath158 after a parallel transportation along @xmath158 of distance @xmath2 ( definition [ nearlyunitshearing ] ) . by the construction of @xcite , the assumption that @xmath101 is @xmath18-nearly evenly footed implies that such a maximal gluing always exists , resulting in a @xmath1-injectively immersed surface which is quasi - fuchsian .
since @xmath3 might be disconnected , we wish to slightly modify the gluing without affecting the nearly unit shearing property , nevertheless the number of components of @xmath3 can be decreased in that case .
denote the components of @xmath3 as @xmath159 , where @xmath160 is at least two .
if two components of @xmath3 , say @xmath161 and @xmath162 , has glued cuffs @xmath163 and @xmath164 that are homotopic to each other , supposing that @xmath165 is nonseparating on @xmath166 , we may modify the gluing by cutting @xmath166 along @xmath165 , and regluing in a cross fashion .
then the new resulting surface @xmath167 has a connected component @xmath168 instead of the previous two components @xmath161 and @xmath162 .
we say that @xmath168 is obtained by _ hybriding _ @xmath161 and @xmath162 along @xmath169 and @xmath170 .
to preserve the nearly unit shearing property , we need to require that the feet of pants on one side of @xmath169 is almost the same as the feet of pants on the same side of @xmath170 .
such @xmath166 and @xmath165 can be found by the following argument .
first , the irreducibility of @xmath101 implies that there is some curve class @xmath112 , such that there are at least two distinct components of @xmath3 that have glued cuffs homotopic to @xmath36 . because the footed boundary @xmath171 restricted to the unit normal vector bundle @xmath35 over @xmath36
is nearly evenly distributed , the connectedness of @xmath35 implies that at least distinct two components @xmath161 and @xmath162 of @xmath3 ( not necessarily the components that we started with ) have glued cuffs @xmath169 and @xmath170 homotopic to @xmath36 with their feet on the same side very close to those of each other .
therefore , performing the hybriding on these @xmath166 along @xmath165 will decrease the number of components of the resulting surface , preserving the nearly unit shearing property . iterating the process until the resulting surface become connected ,
then we are done .
a minor point here is that the hybriding trick also require that @xmath165 be nonseparating on @xmath166 .
in fact , with the somewhat technical assumption that @xmath101 is rich , the nonseparating property of glued cuffs can be satisfied if we pass to a cyclic finite cover of @xmath3 , and hence pass to positive multiple of @xmath101 .
note also that the arguments above certainly works as well when @xmath3 has boundary . in practice , one need to be slightly careful to control the error so that the new resulting surface remains @xmath18-panted , but the general idea of hybriding follows the outline above . roughly speaking , the assumption that @xmath101 is nearly evenly footed allows us to control the shape of @xmath3 along each glued cuff , which ensures the @xmath1-injectivity and the quasi - fuchsian property ; the assumption that @xmath101 is rich allows us to construct @xmath3 so that glued cuff is nonseparating in @xmath3 , so combined with the assumption that @xmath101 is irreducible , we may perform a hybriding trick to obtain a connected @xmath3 , possibly after passing to a further positive integral multiple of @xmath101 . in the rest of this section , we prove proposition [ gluingprop ] . in subsection [ subsec - nonseparatinggluedcuffs ]
we explain how to control the gluing so that the glued cuffs are nonseparating ; in subsection [ subsec - gluingwithnearlyunitshearing ] , we review the nearly unit shearing condition that is used in @xcite ; subsection [ subsec - hybridingdisconnectedcomponents ] is the hybriding argument ; subsection [ subsec - proofofpropositiongluingprop ] summarizes the proof of proposition [ gluingprop ]
. it will be convenient to introduce a measure @xmath172 naturally associated to any subunion of cuffs @xmath173 .
this measure records the contribution to the footed boundary @xmath174 from those pairs of pants which contain components of @xmath156 . more concretely ,
each component @xmath175 lies in a unique pair of pants @xmath176 . if @xmath152 is a copy of @xmath177 , and if @xmath178 is a copy of @xmath179 , we define the _ marked footed boundary _
@xmath180 to be sum of the two feet ( as atomic measures ) of @xmath56 at the cuff corresponding to @xmath181 .
note that potentially @xmath178 could have other cuffs which are copies of @xmath36 but which might not come from @xmath156 , so we need to specify @xmath158 rather than just mentioning @xmath36 .
we define @xmath182 in particular , we also write @xmath183 the lemma below essentially follows from the condition that @xmath101 is rich .
[ gluedcuffs ] possibly after passing to a positive integral multiple of @xmath101 , we may assume that the prescribed disjoint union of pants @xmath146 admits a subunion of cuffs @xmath173 , satisfying the following : * any gluing @xmath149 of @xmath146 along @xmath156 is maximal ; * restricted to any @xmath35 , the measure @xmath184 is a positive rational multiple of @xmath174 ; * any pair of pants @xmath185 contains at least two cuffs from @xmath156 . for simplicity , we write @xmath186 for @xmath187 , and @xmath188 for @xmath100 .
let @xmath189 be the number of times that a curve @xmath36 occurs as the cuff of a pair of pants @xmath56 .
for any curve @xmath112 , let @xmath190 let @xmath191 if @xmath192 ; otherwise , let @xmath193 . since @xmath101 is rich , it is clear by definition [ gluingconditions ] that @xmath194 whenever @xmath192 .
it follows that @xmath195 because every pair of pants has only three cuffs .
furthermore , possibly after passing to a positive integral multiple of @xmath101 , we may assume that @xmath101 and all @xmath196 are integral .
therefore , in the disjoint union of pairs of pants @xmath146 prescribed by @xmath101 , we may find mutually disjoint subunions @xmath197 prescribed by @xmath196 , and for each component @xmath198 , we may mark one cuff @xmath199 which is a copy of @xmath36 . when @xmath200 has more than one cuffs homotopic to @xmath36 , we require the marking to be evenly weighted on these cuffs : in other words , if @xmath200 has only two cuffs @xmath169 and @xmath170 homotopic to @xmath36 , among all the pairs of pants in @xmath197 that are copies of @xmath200 , there will be half of them having the copy cuff @xmath169 marked , and the other half having the copy cuff @xmath170 marked ; similarly will we mark those pairs of pants @xmath200 with all the cuffs homotopic to @xmath36 .
let @xmath150 be the union of all the unmarked cuffs .
it is straightforward to check that the three listed properties about @xmath156 are satisfied by our construction .
in fact , for any @xmath201 , suppose without loss of generality that @xmath202 . because @xmath156 has exactly @xmath203 components homotopic to @xmath36 and exactly @xmath203 components homotopic to @xmath204 , any gluing @xmath149 is maximal .
the measure @xmath205 restricted to @xmath35 equals @xmath206 times @xmath207 and restricted to @xmath208 equals @xmath207 , both proportional to @xmath207 . for any @xmath209
, we marked at most one cuff @xmath199 in the construction above , so it contains at least two cuffs from @xmath156 .
suppose @xmath149 is a gluing of @xmath146 prescribed by @xmath101 .
for a disjoint union of pants @xmath210 prescribed by a positive integral multiple of @xmath101 , we say that a gluing @xmath211 _ covers _
@xmath149 , if @xmath212 is the preimage of @xmath156 under the natural covering @xmath213 , and if the following diagram commutes : @xmath214^{\tilde{\phi } } \ar[d]^{\kappa } & \tilde{\mathcal{c}}\ar[d]^{\kappa}\\ \mathcal{c } \ar[r]^{\phi } & \mathcal{c } } \ ] ] in this case , the associated surface @xmath215 naturally covers @xmath3 as well .
[ nonseparatinggluedcuffs ] let @xmath216 be a subunion of cuffs satisfying the conclusion of lemma [ gluedcuffs ] .
suppose that @xmath149 is a gluing of @xmath146 prescribed by @xmath101 .
then the disjoint union of pants @xmath210 prescribed by @xmath217 admits a gluing @xmath211 covering @xmath149 , such that every glued cuff in the resulted surface @xmath215 is nonseparating . the glued cuffs induces a decomposition of @xmath3 into pairs of pants , and let @xmath218 be the ( possibly disconnected ) dual graph . by lemma [ gluedcuffs ] ,
the valence of any vertex of @xmath218 is at least two .
it follows from an easy construction that @xmath218 admits a double cover @xmath219 in which every edge is nonseparating . in other words
, there is a double cover @xmath215 of @xmath3 induced from a cover @xmath211 of the gluing @xmath149 , in which every glued cuff is nonseparating .
note also that @xmath210 can be identified with @xmath215 cut along the glued cuffs , so it is prescribed by @xmath217 . to control the shape of @xmath3 along a glued cuff
, we will require the gluing @xmath154 to be _ nearly unit shearing _ , which can be described on the visual tori as follows .
observe that for any nearly purely hyperbolic curve @xmath112 ( or indeed for any curve ) , there is a natural action of the additive group of complex numbers on @xmath35 .
more precisely , for any @xmath220 , there is an isomorphism between holomorphic tori : @xmath221 satisfying that for any @xmath222 , @xmath223 parallel transports any normal vector along @xmath36 by signed distance @xmath160 , and that for any @xmath224 , @xmath225 rotates the direction of any normal vector by a signed angle @xmath226 .
it is clear that the kernel of the action is the lattice in @xmath227 generated by @xmath228 and the complex length @xmath84 of @xmath36 .
there is also a canonical anti - isomorphism @xmath229 taking any unit normal vector @xmath230 to the opposite vector @xmath231 at the same point @xmath232 , where @xmath121 is regarded as the unoriented geodesic representative .
note that an anti - isomorphism is orientation - reversing .
we will think of the composition of the bar anti - isomorphism with @xmath233 as the model of a @xmath234-shearing gluing along @xmath36 , denoted as : @xmath235 in other words , a nearly unit shearing gluing @xmath149 should behave very much like @xmath236 along each glued cuff .
[ nearlyunitshearing ] a gluing @xmath149 of @xmath146 is said to be _
@xmath18-nearly unit shearing _ ,
if for every pair of cuffs @xmath237 with @xmath238 equal to @xmath239 , the feet measure @xmath240 is @xmath119-equivalent to @xmath241 on @xmath242 , where @xmath112 is the curve class of @xmath158 .
a reader familiar with the kahn
markovic construction should recognize the definition above as equivalent to the condition @xmath243 in ( * ? ? ?
* theorem 2.1 ) .
the lemma below is a consequence of the condition that @xmath98 is @xmath18-nearly evenly footed .
[ nearlyunitshearinggluing ] with the notations of proposition [ gluingprop ] , there is a gluing @xmath149 of @xmath146 , which is @xmath18-nearly unit shearing .
let @xmath173 be a subunion of cuffs as ensured by the conclusion of lemma [ gluedcuffs ] .
the lemma follows from the hall marriage argument , cf .
* theorem 3.2 and subsection 3.5 ) .
[ resultingqf ] there exists @xmath90 depending on @xmath4 , and for any @xmath91 , there exists @xmath244 depending on @xmath4 and @xmath14 , such that for any @xmath93 , the following holds . if a gluing @xmath149 of @xmath146 is @xmath18-nearly unit shearing , then the resulting surface @xmath97 is @xmath1-injectively immersed and quasi - fuchsian .
this is exactly ( * ? ? ?
* theorem 2.1 ) if @xmath3 is closed . in the general case ,
recall that a compact immersed subsurface @xmath3 of @xmath4 is _ quasi - fuchsian _ in our sense if it is an essential subsurface of a closed immersed quasi - fuchsian subsurface @xmath167 of @xmath4 ( section [ sec - introduction ] ) .
we may take a ubiquitous @xmath18-nearly evenly footed measure @xmath245 , for instance , as guaranteed by theorem [ homologyviapants ] ( 2 ) .
then for a sufficiently large integer @xmath246 , we may assume that @xmath247 is still ubiquitous and @xmath18-nearly evenly footed .
let @xmath248 be the disjoint union of @xmath18-nearly regular pants prescribed by @xmath249 .
we may identify @xmath146 as a subunion of components of @xmath248 .
it is not hard to see that the gluing @xmath149 can be extended to be a gluing @xmath250 which is still @xmath18-nearly unit shearing , provided @xmath246 sufficiently large .
then the gluing @xmath250 yields a possibly disconnected , componentwise @xmath1-injectively immersed quasi - fuchsian closed subsurface @xmath251 by ( * ? ? ?
* theorem 2.1 ) .
the subsurface @xmath3 obtained via the gluing @xmath149 of @xmath146 is an essential subsurface of @xmath167 , so it is @xmath1-injectively immersed and quasi - fuchsian .
the following lemma uses the condition that @xmath101 is irreducible .
[ hybriding ] suppose that @xmath149 is a gluing of @xmath146 prescribed by a positive integral multiple of @xmath101 which is @xmath18-nearly unit shearing with all glued cuffs nonseparating on the resulting surface @xmath3 .
then possibly after passing to a further positive multiple of @xmath101 , there is a gluing @xmath252 , which is @xmath253-nearly unit shearing with all glued cuffs nonseparating on the resulting surface @xmath167 , and moreover , @xmath167 is connected . for simplicity ,
we rewrite the positive integral multiple of @xmath101 prescribing @xmath146 as @xmath101 . note that we may still assume @xmath101 to be irreducible , @xmath18-nearly evenly footed , and rich .
we also observe that if @xmath3 has @xmath160 components @xmath159 , then for any positive integral multiple @xmath254 , there is a gluing @xmath211 covering @xmath149 such that the resulting surface @xmath215 is an @xmath255-fold cover of @xmath3 with @xmath160 components as well . indeed ,
each component @xmath256 of @xmath215 can be chosen as the @xmath255-fold cyclic cover of @xmath166 dual to a glued cuff @xmath257 , and @xmath215 has an induced pants decomposition that describes @xmath211 .
moreover , @xmath211 is clearly @xmath18-nearly unit shearing with all glued cuffs nonseparating as well .
therefore , possibly after passing to a positive integral multiple of @xmath101 , and considering the gluing @xmath211 instead of @xmath149 , we may further assume that once any component @xmath166 has a glued cuff homotopic to a curve @xmath112 , there are at least @xmath160 glued cuffs of @xmath166 homotopic to @xmath36 .
let @xmath258 be the decomposition of @xmath3 into connected components .
then there is an induced decomposition of @xmath146 into subunion of pairs of pants @xmath259 , such that components of each @xmath260 is projected to be @xmath166 under the gluing .
it follows that @xmath101 equals @xmath261 , where the measure @xmath262 prescribes @xmath260 .
similarly , there is an induced decomposition of @xmath263 such that each @xmath264 is the subunion of cuffs of @xmath260 , which is invariant under @xmath154 .
it follows that @xmath265 consider a simplicial graph @xmath108 as follows of @xmath160 vertices @xmath266 . for any @xmath267 ,
the vertices @xmath268 and @xmath269 are connected by an edge if and only if there is a pair of cuffs @xmath270 and @xmath271 representing the same curve class @xmath112 , such that @xmath272 and @xmath273 are @xmath274-equivalent on @xmath208 .
we hence fix a choice of @xmath275 as above , rewriting as @xmath276 . with the assumption that any glued cuff of @xmath166 or @xmath277 has at least @xmath160 copies on the same component
, we may assume that @xmath278 are mutually distinct components of @xmath156 , and similarly for @xmath279 .
observe that @xmath108 is connected .
in fact , let @xmath280 be the components of @xmath108 , and let @xmath281 be the subset of indices so that @xmath282 if and only if @xmath283 .
suppose on the contrary that @xmath284 .
we write the @xmath274-neighborhood of the support of @xmath285 as @xmath286 , and write the union of @xmath287 as @xmath288 .
because @xmath101 is @xmath18-nearly evenly distributed and rich ( definition [ gluingconditions ] ) , @xmath289 is either the emptyset or @xmath35 , for any curve @xmath112 . if @xmath289 equals @xmath35 , then for all the @xmath286 that meets @xmath35 in a nonempty set , the connectedness of @xmath35 implies that the corresponding vertices @xmath268 must lie on the same component of @xmath108 .
therefore , writing @xmath290 for the sum of @xmath262 for @xmath282 , it follows that the supports of @xmath291 are mutually disjoint subsets of @xmath153 .
however , @xmath101 equals @xmath292 .
this is contrary to the assumption that @xmath101 is irreducible ( definition [ gluingconditions ] ) . guided by the simplicial graph @xmath108 together with the decorating data @xmath276
, we choose a maximal tree @xmath293 of @xmath108 and perform the hybriding construction accordingly along the edges of @xmath293 .
in other words , we obtain a new gluing @xmath294 as follows . for any cuff @xmath152 ,
if @xmath158 is some @xmath278 corresponding to an edge of @xmath293 , we define @xmath295 to be @xmath296 , and @xmath297 to be @xmath298 ; otherwise , we define @xmath299 to be @xmath239 as before . because @xmath278 and @xmath279 are projected to nonseparating glued cuffs of @xmath166 and @xmath277 respectively , and because @xmath293 is a maximal tree of the connected graph @xmath108 , the new gluing @xmath252 of @xmath146 results in a connected surface @xmath167 , by induction on the number of components @xmath160 .
because @xmath300 and @xmath301 are @xmath274-equivalent on @xmath302 , where @xmath303 denotes the homotopy class represented by both @xmath278 and @xmath279 , and because @xmath304 and @xmath300 as @xmath149 is @xmath18-nearly unit shearing , the construction of @xmath305 implies that @xmath252 is @xmath253-nearly unit shearing .
this completes the proof .
we summarize the proof of proposition [ gluingprop ] as follows .
as @xmath101 is irreducible , @xmath18-nearly evenly footed , and rich , possibly after passing to a positive integral multiple of @xmath101 , we may construct an @xmath253-nearly unit - shearing gluing @xmath149 of @xmath146 prescribed by @xmath101 , so that the resulting surface @xmath3 is connected ( lemmas [ gluedcuffs ] , [ nonseparatinggluedcuffs ] , [ hybriding ] ) .
if @xmath14 is sufficiently small so that @xmath306 as in lemma [ resultingqf ] , and if @xmath15 is sufficiently large depending only on @xmath14 , then the induced immersion @xmath157 is @xmath1-injective and quasi - fuchsian .
this completes the proof of proposition [ gluingprop ] .
in this section , we study the techniques to construct @xmath18-panted surfaces in oriented closed hyperbolic @xmath0-manifolds via @xmath13-framed bigons and tripods , which generalizes the constructions of @xcite in the @xmath307-dimension case .
in fact , our attempt is to develop a theory about the geometry of @xmath13-framed segments in a closed oriented hyperbolic @xmath0-manifold , which seems to be generalizable to any closed oriented hyperbolic manifold . following the spirit of euclid ,
objects to be studied in the geometry are shapes that can be constructed via @xmath13-framed segments , so our discussion falls naturally into two parts , about _ shapes _ and about _ constructions_. in the first part , for our purpose of application , we will provide an approximate formula that calculates the length and phase of sufficiently tame reduced concatenations of approximately consecutive chains and cycles , which should be compared to the cosine law in elementary euclidean geometry . in the second part , we will define a list of basic constructions , and derive several more efficient constructions by composing the basic ones .
these basic constructions should be regarded as axioms that can be implemented in an oriented closed hyperbolic @xmath0-manifold .
the axiomatic approach to constructions brings at least two benefits .
first , it highlights the connection principle ( lemma [ connectionprinciple ] ) as a featuring axiom ( definition [ axiomsofconstructions ] ( 5 ) ) in the theory .
secondly , it allows us to understand the limitation of constructions .
for instance , as the spine principle ( lemma [ spineprinciple ] ) implies , any construction provides no additional information about the second homology of the @xmath0-manifold @xmath4 .
the second point will be of particular importance to us because it suggests that certain _ a priori _ knowledge about the fundamental group of @xmath4 is demanded so as to construct any homologically interesting @xmath18-panted surface . in the treatment of sections [ sec - pantedcobordismgroup ] and [ sec - pantifyingsecondhomologyclasses ] ,
this piece of information will be supplemented by a finite presentation of @xmath30 .
the central problem of interest to us is the following : given a geodesic immersed graph @xmath308 and a simplical @xmath2-cycle @xmath158 of @xmath309 realized by a union of immersed @xmath18-curves up to homotopy in @xmath4 , if @xmath158 is null - homologous in @xmath309 , does @xmath158 bound an @xmath18-panted subsurface @xmath3 of @xmath4 ?
we remind the reader that @xmath3 is not required to be connected or @xmath1-injective ( definition [ pantedsubsurfacedefinition ] ) , so controlling feet for the gluing is not part of the question . for the simplest example , take @xmath309 to be a ( pointed ) @xmath18-curve @xmath36 and @xmath158 to be the sum of @xmath36 and its orientation reversal @xmath204 . if @xmath56 is an @xmath18-pair - of - pants with a cuff @xmath36 , which exists by a derived construction called _ splitting _ ( construction [ splitting ] ) , then @xmath3 can be taken as the union of @xmath56 and @xmath310 glued along the two pairs of cuffs other than @xmath36 and @xmath204 . in subsection [ subsec - derivedconstructions
] , we will develop several more derived constructions , namely , _ swapping _
, _ rotation _ , and _ antirotation _ , where @xmath309 is the carrier graph of a swap pair or a rotation pair under certain assumptions of nice geometry , ( definitions [ swappair ] and [ rotationpair ] ) .
these derived constructions are fundamental preparation for sections [ sec - pantedcobordismgroup ] and [ sec - pantifyingsecondhomologyclasses ] . in subsection [ subsec - terminology ]
, we introduce some basic concepts in the geometry of @xmath13-framed segments . in subsection
[ subsec - lengthandphaseformula ] , we state and prove the length and phase formula ( lemma [ lengthandphaseformula ] ) . in subsection [ subsec - principlesofconstruction ]
, we introduce the basic constructions in terms of the constructible classes , and show the connection principle ( lemma [ connectionprinciple ] ) and the spine principle ( lemma [ spineprinciple ] ) . in subsection [ subsec - derivedconstructions ] , we discuss the derived constructions .
suppose @xmath4 is an oriented hyperbolic @xmath0-manifold .
we introduce several basic concepts in the geometry of @xmath13-framed segments .
[ partialframedsegment ] an _ oriented @xmath13-framed segment _ in @xmath4 is a triple @xmath311 such that @xmath312 is an immersed oriented compact geodesic segment , and that @xmath313 and @xmath314 are two unit normal vectors at the initial endpoint and the terminal endpoint , respectively . *
the _ carrier segment _ is the oriented segment @xmath312 ; * the _ initial endpoint _ @xmath315 and the _ terminal endpoint _ @xmath316 are the initial endpoint and the terminal endpoint of @xmath312 , respectively ; * the _ initial framing _ @xmath317 and the _ terminal framing _
@xmath318 are the unit normal vectors @xmath319 and @xmath320 , * the _ initial direction _ @xmath321 and the _ terminal direction
_ @xmath322 are the unit tangent vectors in the direction of @xmath312 at the initial point and the terminal point , respectively . the _ orientation reversal _ of @xmath323 is defined to be @xmath324 where @xmath325 is the orientation reversal of @xmath312 . the _ framing rotation _ of @xmath323 by an angle @xmath326
is defined to be @xmath327 where @xmath328 means the cross product in the tangent space .
in particular , the _ framing flipping _ of @xmath323 is defined to be framing rotation by @xmath329 , denoted as @xmath330 it follows from the definition that @xmath331 and in particular , framing flipping commutes with orientation reversion .
[ lengthandphase ] for an oriented @xmath13-framed segment @xmath323 in @xmath4 , the _ length _ of @xmath323 , denoted as @xmath332 is the length of the unframed segment @xmath312 carrying @xmath323 , and the _ phase _ of @xmath323 , denoted as @xmath333 is the angle from the initial framing @xmath319 to the transportation of @xmath320 to the initial point of @xmath312 via @xmath312 , signed with respect to the normal orientation induced from @xmath334 and the orientation of @xmath4 .
we may combine the length and phase into a complex value known as the _ phasor _ of @xmath323 , defined as @xmath335 the value of a phasor always lies outside the unit circle of @xmath227 . for an oriented closed geodesic curve @xmath158 in @xmath4 , we will also speak of its _ length _ , _ phase _ , or _ phasor _ , by taking an arbitrary unit normal vector @xmath336 at a point @xmath337 , and regarding @xmath158 as a @xmath13-framed segment obtained from cutting at @xmath338 and endowed with framing @xmath336 at both endpoints .
it follows from the definition that length and phase are invariant under orientation reversal and under framing rotation .
let @xmath339 , and @xmath340 , and @xmath341 be constants .
let @xmath4 be an oriented hyperbolic @xmath0-manifold
two oriented @xmath13-framed segments @xmath323 and @xmath342 are said to be _
@xmath103-consecutive _ if the terminal endpoint of @xmath323 is the initial endpoint of @xmath342 , and if the terminal framing of @xmath323 is @xmath103-close to the initial framing of @xmath342 .
we simply say _ consecutive _ if @xmath103 equals zero . the _ bending angle _ between @xmath323 and @xmath342 is the angle between the the terminal direction of @xmath323 and the initial direction of @xmath342 , which is valued in @xmath343 $ ] .
a _ @xmath103-consecutive chain _ of oriented @xmath13-framed segments is a finite sequence @xmath344 such that each @xmath345 is @xmath103-consecutive to @xmath346 .
@xmath103-consecutive cycle _ if furthermore @xmath347 is @xmath103-consecutive to @xmath348 .
a @xmath103-consecutive chain or cycle is said to be _
@xmath349-tame _ , if each @xmath345 has length greater than @xmath350 , and the bending angle at each joint point is less than @xmath226 .
3 . for an @xmath349-tame @xmath103-consecutive chain @xmath351 , the _ reduced concatenation _ , denoted as @xmath352 is the oriented @xmath13-framed segment as follows .
the unframed oriented segment of @xmath353 is the geodesic path which is homotopic to the piecewise geodesic path obtained from concatenating the unframed oriented segments carrying @xmath345 , relative to the initial point of @xmath348 and the terminal point of @xmath354 ; the initial framing of @xmath353 is the closest unit normal vector to the initial framing of @xmath348 ; the terminal framing of @xmath353 is the closest unit normal vector to the terminal framing of @xmath354 .
4 . in the case of @xmath349-tame @xmath103-consecutive cycles , the _ reduced cyclic concatenation _ , denoted as @xmath355,\ ] ] is the unframed oriented closed geodesic curve free - homotopic to the concatenation of the unframed oriented segments defined similarly as above , assuming the result not contractible to a point .
we introduce @xmath356-tame bigons and tripods .
these objects should be thought of as nearly hyperbolic curves and geodesic @xmath307-simplices in the context of @xmath13-framed segment geometry .
we also introduce swap pairs and rotation pairs which will be used in subsection [ subsec - derivedconstructions ] .
[ bigon ] an _ @xmath356-tame bigon _ is an @xmath356-tame @xmath103-consecutive cycle of two oriented @xmath13-framed segments @xmath357 of phase @xmath103-close to @xmath358 , with respect to the path - metric distance on @xmath359 valued in @xmath343 $ ] .
we usually say that the reduced cyclic concatenation @xmath360 $ ] is a _
@xmath356-tame bigon _ with the cycle understood .
furthermore , it is said to be _
@xmath361-nearly regular _ if the edges @xmath362 and @xmath363 have length @xmath103-close to @xmath364 .
note that framing flipping does not change the reduced cyclic concatenation , namely , @xmath365 $ ] is the same as @xmath360 $ ] .
however , the orientation of @xmath366 $ ] is exactly opposite to that of @xmath360 $ ] .
[ swappair ] a _ @xmath356-tame swap pair _ of bigons is a pair of @xmath356-tame bigons @xmath360 $ ] and @xmath367 $ ] , such that @xmath368 $ ] and @xmath369 $ ] also form @xmath356-tame bigons , and that @xmath362 and @xmath370 have length and phase @xmath103-close to each other respectively , and that the same holds for @xmath363 and @xmath371 . in this case , we say that the new pair of @xmath356-tame bigons @xmath368 $ ] and @xmath369 $ ] is the @xmath103-swap pair resulted from _ swapping _
@xmath360 $ ] and @xmath367 $ ] , and vice versa . [ tripod ] an _ @xmath356-tame tripod
_ , denoted as @xmath372 is a triple @xmath373 of oriented @xmath13-framed segments of length at least @xmath350 and of phase @xmath103-close to @xmath358 , such that @xmath374 is @xmath103-consecutive to @xmath375 with bending angle @xmath103-close to @xmath376 , for @xmath377 .
furthermore , it is said to be _
@xmath361-nearly regular _ if the legs @xmath378 have length @xmath103-close to @xmath364 , for @xmath377 . for each @xmath377
, @xmath378 will be referred to as a _ leg _ of @xmath379 , and @xmath380 will be referred to as a _ side _ of @xmath379 .
note that the initial framings of @xmath378 are @xmath103-close to each other , so approximately the ordered initial directions rotates either couterclockwise or clockwise around any of them .
we say that @xmath381 is _ right - handed _ if it is the former case , or _
left - handed _ if the latter .
note that framing flipping switches the chirality and the side order of a nearly regular tripod , namely , the chirality of @xmath382 is exactly opposite to that of @xmath379 .
however , the chirality of @xmath383 is the same as that of @xmath379 , where the indices are considered to be in @xmath384 .
[ rotationpair ] a _ @xmath356-tame rotation pair _ of tripods is a pair of @xmath356-tame tripods @xmath379 and @xmath385 , where @xmath379 is @xmath386-nearly regular , and @xmath385 is @xmath387-nearly regular . moreover , the chains @xmath388 are @xmath103-consecutive and @xmath356-tame for all @xmath389 .
in particular , the terminal endpoints of all the legs are the same .
the reader should not confuse an @xmath356-tame rotation pair with a pair of well - connected tripods in the sense of @xcite .
rotation pairs are introduced for the derived constructions rotation and antirotation ( subsection [ subsec - derivedconstructions ] ) . in a quite different form
, these constructions will play a role similar to the rotation lemmas in ( * ? ? ? * subsection 8.1 ) , ( cf .
remark [ remarktriangularrelation ] ) . for sufficiently tame concatenations of approximately consecutive chains and cycles , the change of length and phase under reduction of the concatenation can be approximately calculated .
recall that for any bending angle @xmath390 , the _ limit inefficiency _ associated to @xmath391 is defined as @xmath392 the function @xmath393 is unbounded , strictly convex , and increasing on @xmath394 , and the geometric meaning is explained by lemma [ tametriangle ] ( 2 ) . in particular , @xmath395 , @xmath396 , and @xmath397 .
[ lengthandphaseformula ] given any positive constants @xmath103 , @xmath226 , @xmath19 where @xmath341 and @xmath398 , the following statements hold in any oriented hyperbolic @xmath0-manifold . 1 .
if @xmath351 is an @xmath349-tame @xmath103-consecutive chain of oriented @xmath13-framed segments , denoting the bending angle between @xmath345 and @xmath346 as @xmath399 , then @xmath400 and @xmath401 where @xmath402 on @xmath359 is understood as the distance from zero valued in @xmath343 $ ] .
2 . if @xmath351 is an @xmath349-tame @xmath103-consecutive cycle of oriented @xmath13-framed segments , denoting the bending angle between @xmath345 and @xmath346 as @xmath403 $ ] with @xmath404 equal to @xmath405 by convention , then @xmath406 ) -\sum_{i=1}^m\ell(\mathfrak{s}_i)+\sum_{i=1}^{m } i(\theta_i)\right|\,<\ , \frac{me^{(-l+10\log2)/2}\sin(\theta/2)}{l-\log2},\ ] ] and @xmath407 ) -\sum_{i=0}^m\varphi(\mathfrak{s}_i)\right|\,<\ , m(\delta+e^{(-l+10\log2)/2}\sin(\theta/2)),\ ] ] where @xmath402 on @xmath359 is understood as the distance from zero valued in @xmath343 $ ] .
[ remarklengthandphaseformula ] in this paper , we will only apply the formula for @xmath226 equal to @xmath408 , @xmath409 , or constant multiples of @xmath103 .
moreover , for any positive constant @xmath103 , @xmath410 where @xmath411 and @xmath412 , we will always ensure that the participating chains or cycles are @xmath413-consecutive , @xmath414-tame .
then the error bounds can be replaced by @xmath415 and @xmath416 for ( 1 ) and ( 2 ) , respectively .
the proof relies on a lemma in elementary hyperbolic geometry , which provides some key estimation for tame concatenation of segments .
[ tametriangle ] given any constants @xmath341 , and @xmath417 , suppose that @xmath418 is a geodesic triangle in hyperbolic space , where @xmath419 , @xmath420 , and @xmath421 , then 1 .
2 . @xmath423 .
to prove the inequality ( 1 ) , it suffices to assume that @xmath424 since @xmath425 , which equals @xmath426 , achieves its unique maximum in this case .
let @xmath4 be the midpoint of @xmath427 . in the right triangle @xmath428
, it follows from the dual law of cosines that @xmath429 therefore , @xmath430 since the assumption @xmath431 implies @xmath432 we have @xmath433 the same estimation holds for @xmath434 , so @xmath435 which is the inequality ( 1 ) . to prove the inequality ( 2 ) , consider the inscribed circle @xmath436 of @xmath418 , denoting the tangent point of @xmath436 with @xmath427 , @xmath437 , and @xmath438 as @xmath439 , @xmath440 , and @xmath441 , respectively .
then @xmath442 which approaches the supremum as @xmath443 and @xmath444 tend to @xmath445 , and achieves the unique minimum when @xmath424 .
a direct computation shows that the supremum is exactly @xmath446 , so the upper bound of inequality ( 2 ) holds . for the lower bound , it hence suffices to assume that @xmath424 .
we write @xmath447 for the triangle with ideal points @xmath448 and @xmath449 , and let @xmath450 be the inscribed circle which is tangent to @xmath451 , @xmath452 and @xmath453 at @xmath454 , @xmath455 and @xmath456 , respectively .
note that now @xmath439 is the midpoint of @xmath427 , and similarly for @xmath454 .
it is also clear that @xmath457 in the right triangle @xmath458 , it follows from the dual law of cosines that @xmath459 therefore , @xmath460 since @xmath461 , we obtain @xmath462 on the other hand , the difference between the area of @xmath447 and @xmath418 is the area of the quadrilateral @xmath463 , which clearly equals the value of @xmath464 in @xmath418 . because @xmath465 and @xmath466 are perpendicular to @xmath467 , there is the comparison of area : @xmath468 therefore , the estimations of @xmath464 and @xmath469 above imply @xmath470 we obtain @xmath471 which verifies the lower bound in the inequality ( 2 ) . since it follows from certain standard estimation argument provided lemma [ tametriangle ] , we only sketch the proof . to see the statement ( 1 ) , we write @xmath472 and @xmath473 for the error terms that we will estimate . for each @xmath474 , let @xmath475 be any unit vector perpendicular to both @xmath345 and @xmath346 at their joint point , and let @xmath476 , @xmath477 by convention .
let @xmath478 be the @xmath13-framed segment with the same carrier segment as that of @xmath345 , and with @xmath479 and @xmath480 be @xmath481 and @xmath475 respectively .
then @xmath482 is a consecutive @xmath349-chain .
note that @xmath483 equals @xmath484 .
since @xmath344 is @xmath103-consecutive , @xmath485 is @xmath103-close to @xmath486 , so the new error of phase differ from the old by @xmath487 it is clear that @xmath488 if @xmath255 equals @xmath2 , we have already done , as @xmath489 and @xmath490 are both @xmath358 by definition .
if @xmath255 is greater than @xmath2 , we may consider a chain @xmath491 as follows . for each @xmath492 , write @xmath478 as the concatenation of two consecutive oriented @xmath13-framed segments @xmath493 and @xmath494 of equal length and phase .
for @xmath495 , let @xmath496 be @xmath497 .
let @xmath498 be @xmath499 , and @xmath500 be @xmath501 , or in the case that @xmath255 equals @xmath307 , let @xmath498 be @xmath502 .
it follows immediately from lemma [ tametriangle ] that @xmath491 is @xmath503-consecutive and @xmath504-tame , where @xmath505 moreover , @xmath506 is @xmath507-close to @xmath508 .
it is also clear that @xmath509 is @xmath510-close to @xmath511 .
thus , @xmath512 and @xmath513 if @xmath255 equals @xmath307 , we have done since the chain @xmath491 has only one term . if @xmath255 is greater than @xmath307 , we may further obtain a chain @xmath514 from @xmath491 , in a similar way as we obtain the latter from @xmath344 .
proceed iteratively to obtain new chains @xmath515 from @xmath516 until @xmath517 equals @xmath2 .
the chain @xmath515 is @xmath518-consective and @xmath519-tame where @xmath520 for @xmath521 .
thus for @xmath521 , we have @xmath522 and @xmath523 where @xmath524 by convention . summing up the error of length in each step yields that @xmath525 and that @xmath526 therefore , we have the estimations of the statement ( 1 ) . the statment ( 2 ) can be proved similarly .
we first obtain a consecutive cycle @xmath482 from @xmath344 , then construct iteratively the cycles @xmath527 from @xmath528 by joining consequential midpoints , starting with @xmath529 which is @xmath482 .
similar estimations as before hold in this case , and as @xmath160 tends to infinity , the ( non - reduced ) cyclic concatenation @xmath528 converges to @xmath530 $ ] geometrically . note that the summation of errors in this case will be a series which converges absolutely , but the upper bound will stay unchanged except for the replacement of @xmath517 by @xmath255 . combining the estimations as before yields the estimations in the statement ( 2 ) .
we close this subsection with a lemma of elementary hyperbolic geometry , which will be used later . recall that the _ fermat point _ of a geodesic triangle @xmath418 in hyperbolic space is the point @xmath3 minimizing @xmath531 , which lies on the @xmath307-simplex bounded by @xmath418 .
if the inner angles of @xmath418 are all smaller than @xmath532 , then @xmath3 lies inside @xmath418 and @xmath533 , @xmath534 , @xmath535 form an angle of @xmath532 pairwise .
[ fermatpoint ] for any positive constant @xmath536 , if the radius @xmath537 ball centered at each vertex of @xmath418 is separated by a hyperplane from the other vertices , then @xmath3 lies inside @xmath418 , and @xmath538 , @xmath539 , and @xmath540 are all greater than @xmath536 . under the assumption , for any point @xmath200 with @xmath541 bounded by @xmath536 ,
the angle @xmath542 is at most @xmath532 , so the fermat point lies outside the @xmath536 neighborhood of @xmath107 .
similar statements hold for @xmath543 and @xmath145 .
thus @xmath3 lies inside @xmath418 of distance at least @xmath536 from the vertices .
before we discuss how to construct various @xmath18-panted surfaces with @xmath13-framed segments ( subsection [ subsec - derivedconstructions ] ) , in this subsection , we wish to formally discuss what we mean by a _ construction _ ( definition [ constructibleextensions ] ) .
we will enumerate our basic constructions as axioms ( definition [ axiomsofconstructions ] ) .
these constructions are realizable in an oriented closed hyperbolic @xmath0-manifold @xmath4 essentially because of the connection principle ( lemma [ connectionprinciple ] )
. then we will state the spine principle ( lemma [ spineprinciple ] ) , which morally says that since we are only drawing auxiliary @xmath13-framed segments in any such construction , we will not gain any new knowledge about the second homology of @xmath4 .
this observation will be of fundamental importance when we pantify a second homology class ( section [ sec - pantifyingsecondhomologyclasses ] ) . in practice
, it will be convenient to describe constructions more naturally in terms of @xmath13-framed segments , and at the end of this subsection , we will explain how to translate between the natural description and the formal description in terms of constructible extensions .
however , the reader may safely skip the discussion of this subsection until section [ sec - pantifyingsecondhomologyclasses ] .
we provide a formal definition of constructible objects in terms of partially-@xmath544 spaces over an oriented closed hyperbolic @xmath0-manifold .
[ partiallydeltaspace ] let @xmath4 be an oriented hyperbolic @xmath0-manifold .
a _ partially-@xmath544 space over @xmath4 _ is a triple @xmath545 as follows .
the space @xmath546 is a cw subspace of a cw space @xmath108 , enriched with a @xmath544-complex structure ; the map @xmath547 is geodesic restricted to simplices of @xmath546 , possibly degenerate .
we often simply mention a partially-@xmath544 space @xmath108 with @xmath546 and @xmath548 implicitly assumed .
a _ partially combinatorial _ map between two partially-@xmath544 space @xmath108 and @xmath549 is a cw map @xmath550 , combinatorial with respect to the @xmath544-complex structure of @xmath546 and @xmath551 , such that @xmath552 . for a partially-@xmath544 space @xmath309 over @xmath4 , an _ extension _ of @xmath309
is a partially-@xmath544 space @xmath108 together with a partially combinatorial embedding @xmath553 .
we list our basic constructions by the following axioms , and verify that they are all possible in the situation that we will be concerned with .
denote by @xmath554 $ ] the standard @xmath555-simplex in @xmath556 spanned by the standard basis vectors @xmath557 .
[ axiomsofconstructions ] for an oriented hyperbolic @xmath0-manifold @xmath4 and a pair of positive constants @xmath356 , the _ axioms of constructions _ are the following statements : 1 .
_ vertex creation_. suppose that @xmath558 is a point .
if @xmath108 is a partially-@xmath544 space over @xmath4 , then there exists a partially-@xmath544 space @xmath559 as follows .
the space pair @xmath560 is @xmath561 where @xmath54 is a new vertex ; the map @xmath562 is an extension of @xmath548 such that @xmath563 .
_ simplex subdivision_. suppose that @xmath564 is a unit vector at a point @xmath558 , and that @xmath565 is a positive real constant .
if @xmath108 is a partially-@xmath544 space over @xmath4 with an @xmath555-simplex @xmath566\to x_\delta$ ] , such that @xmath567 , then there exists a partially-@xmath544 space @xmath559 as follows .
the space pair @xmath560 is @xmath568,\ , x_\delta\cup_{\sigma}[\varepsilon_0,\cdots,\varepsilon_{n+1}])$ ] , where @xmath569 $ ] denotes the mapping cone of @xmath570 , and similarly for @xmath569 $ ] ; the map @xmath562 is the extension of @xmath548 so that @xmath571}$ ] is a geodesic segment of length @xmath565 with the initial point @xmath338 and the initial direction @xmath572 , and that @xmath573}$ ] is geodesic .
_ simplex filling_. if @xmath108 is a partially-@xmath544 space over @xmath4 , and if there is a combinatorial map @xmath574 \to x_\delta$ ] from the last @xmath555 faces of an @xmath555-simplex @xmath554 $ ] to @xmath546 , then there exists a partially-@xmath544 space @xmath559 over @xmath4 as follows .
the space pair @xmath560 is @xmath575,\ , x_\delta\cup_{\phi}[\varepsilon_0,\cdots,\varepsilon_{n}])$ ] ; the map @xmath562 is the extension of @xmath548 so that @xmath576}$ ] is geodesic .
free loop reduction_. if @xmath108 is a partially-@xmath544 space over @xmath4 , and if there is an edge @xmath577 of @xmath546 with its boundary attached the same vertex @xmath54 and with @xmath578 nondegenerate , then there exists a partially-@xmath544 space @xmath559 over @xmath4 as follows .
the space pair @xmath560 is @xmath579 , where @xmath580 is a new vertex , @xmath581 is a new edge with its boundary attached to @xmath54 , and @xmath107 is a new oriented annulus with the boundary the disjoint union of @xmath582 and the orientation reversal of @xmath583 ; the map @xmath562 is an extension of @xmath584 such that @xmath585 carries the closed geodesic in @xmath4 freely homotopic to @xmath586 .
edge connection_. suppose that @xmath587 and @xmath588 are pairs of orthogonal unit vectors at points @xmath589 respectively , and that @xmath565 is a complex constant of modulus at least @xmath19 .
if @xmath108 is a partially-@xmath544 space over @xmath4 with ( not necessarily distinct ) vertices @xmath590 of @xmath546 such that @xmath591 and @xmath592 , then there exists a partially-@xmath544 space @xmath559 over @xmath4 as follows .
the space pair @xmath560 is @xmath593 , where @xmath577 is a new edge from @xmath54 to @xmath594 ; the map @xmath562 is an extension of @xmath584 such that @xmath595 carries an oriented @xmath13-framed segment @xmath323 from @xmath338 to @xmath596 satisfying the following .
* the oriented @xmath13-framed segment @xmath323 has length and phase @xmath103-close to @xmath597 and @xmath598 , respectively .
the initial direction and framing of @xmath323 are @xmath103-close to @xmath572 and @xmath599 , respectively .
the terminal direction and framing of @xmath323 are @xmath103-close to @xmath600 and @xmath601 , respectively .
[ constructibleextensions ] suppose that @xmath309 is a partially-@xmath544 space over @xmath4 .
we define the _ class of constructible extensions of @xmath309 _ with respect to @xmath356 to be the smallest class @xmath602 of extensions of @xmath309 , satisfying that @xmath603 , and that @xmath604 implies @xmath605 for any @xmath559 obtained from @xmath108 by one of the axioms of constructions above .
a partially-@xmath544 space @xmath288 over @xmath4 is said to be _
constructible from @xmath309 _ if there exists a constructible extension @xmath606 of @xmath309 and a partially combinatorial map @xmath607 , such that @xmath608 .
in practice , @xmath309 will serve as the object that a construction starts with , where @xmath609 records the piece of information that are available for the construction ; @xmath108 will serve as the recipe of the construction ; and @xmath288 will serve as the resulting object of the construction , where @xmath610 is meant to record certain shape properties that result should satisfy . for any oriented closed hyperbolic @xmath0-manifold @xmath4 , the axioms of constructions listed in definition [ axiomsofconstructions ] are all realizable for any positive constant @xmath103 universally small positive , and
any positive constant @xmath19 is sufficiently large depending only on @xmath103 and @xmath4 .
in fact , the first four axioms are true for any hyperbolic manifold , and the last axiom follows from the connection principle ( lemma [ connectionprinciple ] ) , which holds for oriented closed hyperbolic @xmath0-manifolds .
we emphasize the following connection principle as it is the fundamental reason for all our constructions of nearly regular pants to work .
for example , it implies that @xmath153 and @xmath611 are nonempty for an oriented closed hyperbolic @xmath0-manifold @xmath4 , provided that @xmath14 is universally small positive and that @xmath15 is sufficiently large depending only on @xmath4 and @xmath14 .
[ connectionprinciple ] for any universally small positive @xmath103 , and for any sufficiently large positive @xmath19 depending only on @xmath103 and @xmath4 , the following statement holds . if @xmath587 and @xmath588 are pairs of orthogonal unit vectors at points @xmath589 respectively , and if @xmath565 is a complex number of modulus at least @xmath19 , then there exists an oriented @xmath13-framed segment @xmath323 from @xmath338 to @xmath596 satisfying the following . *
the oriented @xmath13-framed segment @xmath323 has length and phase @xmath103-close to @xmath597 and @xmath598 , respectively .
the initial direction and framing of @xmath323 are @xmath103-close to @xmath572 and @xmath599 , respectively . the terminal direction and framing of @xmath323 are @xmath103-close to @xmath600 and @xmath601 , respectively .
this follows from the fact that the frame flow is mixing ( * ? ? ?
* theorem 4.2 ) .
the argument is the same as that of ( * ? ? ?
* lemma 4.4 ) .
the spine principle says that any constructible extension of a partially-@xmath544 space over an oriented hyperbolic @xmath0-manifold is _
relatively @xmath2-spined _ , or precisely as follows .
[ spineprinciple ] if @xmath612 is a constructible extension of a partially-@xmath544 space @xmath613 over an oriented hyperbolic @xmath0-manifold @xmath4 , then the defining inclusion @xmath614 can be extended to be a homotopy equivalence @xmath615 where @xmath616 is obtained from @xmath309 by attaching cells of dimension at most @xmath2 .
this follows immediately from inspecting the construction axioms listed in definition [ constructibleextensions ] . in the rest of this paper
, we will often describe a construction without explicitly writing down the associated partially-@xmath544 spaces .
instead , the hypothesis of a construction will be stated in terms of @xmath13-framed segments . using simplex subdivision of dimension @xmath358 , @xmath2 , or @xmath307
, we can construct * a segment emanating from a constructed point in a given direction ; * division of a constructed geodesic segment in a given ratio ; or * a tripod carried by a constructed @xmath307-simplex with a given center .
we can enrich these with @xmath13-framings to turn them into @xmath13-framed objects .
for example , by saying that we construct a right - handed @xmath13-framed right - handed tripod @xmath379 carried by an immersed oriented @xmath307-simplex in a hyperbolic @xmath0-manifold @xmath4 , we precisely mean a procedure as follows . start with a given constructible extension @xmath108 of a given partially-@xmath544 space @xmath309 over @xmath4 , together with a given @xmath307-simplex @xmath617\to x_\delta$ ] , such that the immersed @xmath307-simplex is given by @xmath618 ; let @xmath559 be an extension of @xmath108 attaching a new @xmath0-simplex @xmath619\to x'$ ] ; we consider the leg @xmath378 to be carried by the geodesic segment @xmath620}$ ] . by suitably choosing a tangent direction @xmath572 of @xmath618 and the length @xmath565
, we can control the place of the center of the tripod , and hence the shape of @xmath379 .
in fact , there is a unique way to assign @xmath13-framings of @xmath378 so that @xmath621 are all consecutive for @xmath377 .
the other two constructions in dimension @xmath358 and @xmath2 can be understood in a similar fashion . using simplex filling of dimension @xmath2 and free loop reduction
, we can construct * the reduced concatenation of a chain of @xmath13-framed segments ; and * the cyclically reduced concatenation of a cycle of @xmath13-framed segments .
we consider these to be carried by @xmath2-chains and @xmath2-cycles of a given constructible extension of @xmath108 . by iterately applying simplex filling for a finite number of times ,
we can obtain the reduced concatenation of any chain of @xmath13-framed segments , or obtain a cycle of a single @xmath13-framed segment from any cycle of @xmath13-framed segments . in the latter case , we can then obtain the cyclically reduced concatenation by further applying free loop reduction once . in subsection [ subsec - derivedconstructions ] , we will describe more constructions of @xmath18-panted surfaces @xmath622 with the boundary prescribed in terms of @xmath13-framed segments under various hypotheses . to translate any description there into one in terms of partially-@xmath544 spaces ,
the instruction is as follows .
the hypothesis describes a partially-@xmath544 space @xmath613 .
namely , @xmath609 is considered to be a @xmath2-complex such that each @xmath13-framed segment in the hypothesis corresponds to an oriented edge , and any two edges have identified endpoints if and only if the corresponding @xmath13-framed segments are declared to be nearly consecutive ; @xmath309 is the same as @xmath609 ; and @xmath623 is considered to be the obvious map that send any @xmath2-simplex to the carrier segment of its defining @xmath13-framed segment . in the same fashion , the recipe of the construction applies the axioms of the construction ( definition [ axiomsofconstructions ] ) step by step , by indicating at each step that an auxiliary point , segment , or @xmath13-framed segment should be drawn , so the procedure gives rise to an extension @xmath108 of @xmath309 .
the result of the construction can then be translated in terms of a partially-@xmath544 space @xmath624 constructible from @xmath309 , where @xmath625 has a preferred @xmath2-complex structure since it is prescribed by the @xmath13-framed segments from the hypothesis .
in fact , one can always write down the partially combinatorial map @xmath626 explicitly , where @xmath108 is the extension of @xmath309 from the recipe of the construction . in this subsection , we exhibit several constructions of @xmath18-panted surfaces using @xmath13-framed segments that will be applied in the rest of this paper .
these constructions are all _ derived _ from the basic constructions listed in definition [ axiomsofconstructions ] , and are all _ definite _ in the sense that the number of @xmath13-framed segments involved are universally bounded . throughout this subsection ,
we assume @xmath4 to be an oriented closed hyperbolic @xmath0-manifold , and * @xmath356 is any pair of positve constants where @xmath627 and @xmath628 , and @xmath19 satisfies the conclusion of the connection principle ( lemma [ connectionprinciple ] ) with respect to @xmath103 and @xmath4 .
* @xmath18 is any pair of positive constants where @xmath629 .
the constants have been chosen according to the length and phase formula ( definition [ lengthandphaseformula ] and remark [ remarklengthandphaseformula ] ) .
for each of the constructions of this subsection , we will need additional assumptions comparing @xmath14 and @xmath103 .
however , for later applications , it will be convenient to adopt a uniform requirement : * @xmath630 .
the splitting construction below gives rise to a nearly regular pair of pants by adding a bisecting segment to a nearly purely hyperbolic curve .
the reader should compare it with ( * ? ? ?
* lemma 3.2 and remark ) .
[ splitting ] suppose @xmath631 .
suppose that @xmath632 $ ] is an @xmath356-tame bigon with the length and phase of @xmath323 and @xmath342 @xmath103-close to each other , and that @xmath632\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] .
then a pair of pants @xmath99 can be constructed , with one cuff @xmath632\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] , and with the other two cuffs in @xmath633 . by the connection principle ( lemma [ connectionprinciple ] ) ,
draw an oriented @xmath13-framed segment @xmath634 from @xmath316 to @xmath315 as follows : * the initial and terminal directions are @xmath103-close to @xmath635 and @xmath636 respectively , and the initial and terminal framings are @xmath103-close to @xmath318 and @xmath317 respectively , and the length and phase are @xmath103-close to @xmath637 and @xmath638 respectively .
then a pair of pants @xmath99 can be constructed , such that one cuff of @xmath56 is @xmath632\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] , and that the other two cuffs are @xmath639},\ , \overline{[\bar{\mathfrak{m}}\mathfrak{s}']}\in{{\mathbf{\gamma}}}_{r,10\delta}$ ] .
the verification follows from the length and phase formula ( lemma [ lengthandphaseformula ] ) .
the swapping construction below allows us to exchange the arcs of two nearly purely hyperbolic curves if they fellow travel near a pair of common points , as long as the result are still purely hyperbolic curves
. the reader should compare it with the geometric square lemma ( * ? ? ?
* lemma 5.4 ) .
our construction simplifies some of the arguments there . [
swapping ] suppose @xmath640 .
suppose that @xmath360 $ ] and @xmath367 $ ] is a @xmath641-tame swap pair of bigons , and that @xmath360,\,[\mathfrak{a}'\mathfrak{b}'],\ , \overline{[\mathfrak{a}\mathfrak{b}']},\ , \overline{[\mathfrak{a}'\mathfrak{b}]}\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] .
then an oriented connected compact @xmath18-panted surface @xmath3 can be constructed , with exactly four boundary components @xmath360,\,[\mathfrak{a}'\mathfrak{b}'],\ , \overline{[\mathfrak{a}\mathfrak{b}']},\ , \overline{[\mathfrak{a}'\mathfrak{b}]}$ ] .
we first prove a preliminary case when the joint points are @xmath103-closely antipodal , and then derive the general case from the preliminary case .
* step 1*. suppose @xmath631 .
suppose @xmath360 $ ] and @xmath367 $ ] are a @xmath356-tame swap pair of bigons , and that @xmath360,\,[\mathfrak{a}'\mathfrak{b}'],\ , \overline{[\mathfrak{a}\mathfrak{b}']},\ , \overline{[\mathfrak{a}'\mathfrak{b}]}\in{{\mathbf{\gamma}}}_{r,10\delta}$ ] .
moreover , suppose that the length and phase of @xmath362 , @xmath363 , @xmath370 , @xmath371 are all @xmath103-close to each other .
we construct an oriented connected compact @xmath18-panted surface @xmath3 with exactly four boundary components @xmath360,\,[\mathfrak{a}'\mathfrak{b}'],\ , \overline{[\mathfrak{a}\mathfrak{b}']},\ , \overline{[\mathfrak{a}'\mathfrak{b}]}$ ] as follows . by the connection principle ( lemma [ connectionprinciple ] ) , draw an oriented @xmath13-framed segment @xmath634 from @xmath642 to @xmath643 as follows : * the initial and terminal directions are @xmath103-close to @xmath644 and @xmath645 respectively , and the initial and terminal framings are @xmath103-close to @xmath646 and @xmath647 respectively , and the length and phase are @xmath103-close to @xmath648 and @xmath649 respectively . with @xmath650 standing for @xmath362 or @xmath370 , and @xmath651 standing for @xmath363 or @xmath371
, there is a unique pair of pants @xmath652 determined by its cuffs @xmath653,\ , \overline{[\mathfrak{a}^\circ\mathfrak{m } ] } , \overline{[\mathfrak{b}^\circ\bar{\mathfrak{m } } ] } \in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] .
note that the curve @xmath654}$ ] appears in exactly two pairs of pants as a cuff , and that the same holds for @xmath655}$ ] .
thus the four pairs of pants @xmath656 can be glued along these cuffs , yielding the desired @xmath18-panted surface @xmath3 , which is a torus with four holes .
* step 2*. for @xmath640 , suppose that @xmath360 $ ] and @xmath367 $ ] is a @xmath641-tame swap pair of bigons .
suppose @xmath360,\,[\mathfrak{a}'\mathfrak{b}'],\ , \overline{[\mathfrak{a}\mathfrak{b}']},\ , \overline{[\mathfrak{a}'\mathfrak{b}]}\in{{\mathbf{\gamma}}}_{r,10\delta}$ ] .
we construct an oriented connected compact @xmath18-panted surface @xmath3 with exactly four boundary components @xmath360 $ ] , @xmath367 $ ] , @xmath657}$ ] , @xmath658}$ ] . by edge division
, we may decompose @xmath362 as the concatenation @xmath659 of two consecutive @xmath13-framed subsegments of equal length and phase , and similarly decompose @xmath363 , @xmath370 , @xmath371 as @xmath660 , @xmath661 , @xmath662 .
moreover , we may require the length and phase of @xmath663 to be @xmath103-close to that of @xmath664 , and the length and phase of @xmath665 to be @xmath103-close to @xmath666 and @xmath667 respectively , and the length and phase of @xmath668 @xmath103-close to @xmath669 and @xmath670 respectively .
choose a pair of auxiliary points @xmath589 .
draw five auxiliary @xmath13-framed segments @xmath671 , @xmath672 , @xmath673 , @xmath674 , @xmath675 as follows .
* the initial and terminal endpoints of @xmath671 are @xmath338 and @xmath596 respectively , and the length and phase of @xmath671 are @xmath103-close to @xmath676 and @xmath358 respectively . * the initial and terminal endpoints of @xmath672 are @xmath677 and @xmath338 respectively , and the length and phase of @xmath672 is @xmath103-close to @xmath678 and @xmath679 respectively .
the same holds for @xmath673 with @xmath680 replaced by @xmath681 . * the initial and terminal endpoints of @xmath674 are @xmath596 and @xmath682 respectively , and the length and phase of @xmath674 is @xmath103-close to @xmath678 and @xmath683 respectively .
the same holds for @xmath675 with @xmath684 replaced by @xmath685 . with @xmath686 standing for @xmath687 or @xmath688 , and similarly for the notations @xmath689 , @xmath690 , @xmath691 , @xmath650 , and @xmath651 , there are four pair of pants @xmath692 , each determined by its cuffs @xmath653,\ , \overline{[(\bar{\mathfrak{s}}_a\mathfrak{a}_+)^\circ(\mathfrak{b}_-\mathfrak{s}_b)^\circ\bar{\mathfrak{r } } ] } , \overline{[(\mathfrak{a}_-\mathfrak{s}_a)^\circ\mathfrak{r}(\bar{\mathfrak{s}}_b\mathfrak{b}_+)^\circ ] } \in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] .
decompose @xmath671 as the concatenation @xmath693 of two consecutive @xmath13-framed segments of phase @xmath103-close to @xmath358 , such that @xmath694 is @xmath103-close to @xmath695 and @xmath696 is @xmath103-close to @xmath697 .
note that the chains and cycles involved are all @xmath698-tame and @xmath699-consecutive since @xmath360 $ ] and @xmath367 $ ] are @xmath641-tame bigons .
since @xmath640 , the preliminary case of step 1 applies to the @xmath700-swap pair @xmath701 $ ] and @xmath702 $ ] .
thus there is an @xmath18-panted four - hole torus @xmath703 with boundary components @xmath701 $ ] , @xmath702 $ ] , @xmath704}$ ] , @xmath705}$ ] .
similarly , there is an @xmath18-panted four - hole torus @xmath706 with boundary components @xmath707 $ ] , @xmath708 $ ] , @xmath709}$ ] , @xmath710}$ ] .
glue up @xmath711 , @xmath712 , @xmath713 , @xmath714 , @xmath703 , and @xmath706 along their oppositely oriented common boundary components .
the result is an @xmath18-panted surface @xmath3 with exactly four boundary components @xmath360 $ ] , @xmath367 $ ] , @xmath657}$ ] , @xmath658}$ ] , as desired .
when two nearly regular tripods of opposite chiralities have legs almost opposite to the ones of each other , one can naturally build a nearly regular pair of pants spined on the union of the two tripods .
this will be the first statement of the rotation construction below .
however , there is another case when the tripods have identical chirality .
thus we have two rotation constructions .
note that in the identical chirality case , we take two copies of the curves arising from the construction . this turns out to be necessary due to theorem [ theorempantedcobordism ] .
[ rotationconstruction ] suppose @xmath379 and @xmath385 are a @xmath715-tame rotation pair of tripods .
suppose @xmath716\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] for @xmath389 .
then an oriented connected compact @xmath18-panted surface @xmath3 can be constructed satisfying the following . 1 .
if @xmath379 and @xmath385 are of opposite charalities , then @xmath3 is a pair of pants @xmath99 with cuffs @xmath717 $ ] , for @xmath377 .
2 . suppose @xmath718 .
if @xmath379 and @xmath385 are of identical charality , then @xmath3 has exactly six boundary components , namely , two copies of each @xmath717 $ ] , for @xmath377 . to prove the statement ( 1 ) , suppose that @xmath379 and @xmath385 are of opposite chiralities .
this is the simple case because @xmath3 can be naturally chosen as the pair of pants @xmath99 with cuffs @xmath717 $ ] for @xmath377 .
the spine of @xmath56 is the figure-@xmath226 graph which is approximately the union of the carrier segments of @xmath719 for @xmath377 . to prove the statement ( 2 ) ,
suppose that @xmath379 and @xmath385 are of identical chirality . considering the framing flipping @xmath382 and @xmath720 instead if necessary , we may assume that the tripods are both right - handed without loss of generality .
* step 1*. suppose @xmath721 .
suppose @xmath379 and @xmath385 are a @xmath700-tame rotation pair of right - handed tripods , and that @xmath717\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] for @xmath377 .
moreover , suppose that @xmath385 can be written as @xmath722 where @xmath723 is a @xmath103-consecutive @xmath641-tame chain for each @xmath377 and where @xmath724 is @xmath725-nearly regular .
we construct an @xmath18-panted surface @xmath3 with exactly six boundary components , which are two copies of each @xmath717 $ ] for @xmath377 .
draw an auxiliary oriented @xmath13-oriented segment @xmath323 satisfying the following . * the length and phase of @xmath323 is @xmath103-close to @xmath726 and @xmath358 respectively .
the initial and terminal endpoints of @xmath323 coincide with @xmath727 and @xmath728 respectively .
the initial and terminal directions of @xmath323 are @xmath103-close to @xmath729 and @xmath730 respectively .
the initial framing of @xmath323 is @xmath103-close to @xmath731 , and the terminal framing of @xmath323 is @xmath103-close to @xmath732 .
it follows that @xmath733 is an @xmath734-nearly regular left - handed tripod , where @xmath735 is a @xmath103-consecutive @xmath641-tame chain for each @xmath377 .
observe the following @xmath18-panted surfaces .
since @xmath718 , by swapping ( construction [ swapping ] ) of @xmath700-tame swap pairs , for each @xmath377 , there is an @xmath18-panted surface @xmath736 with boundary components the curves @xmath737 $ ] , @xmath738 $ ] , @xmath739}$ ] , @xmath740}$ ] in @xmath153 ; also by swapping ( construction [ swapping ] ) , for each @xmath377 , there is an @xmath18-panted surface @xmath741 with boundary components the curves @xmath737 $ ] , @xmath742 $ ] , @xmath743}$ ] , @xmath744}$ ] in @xmath153 ; applying rotation in the opposite chirality case ( statement ( 1 ) ) to the @xmath700-tame rotation pair @xmath733 and @xmath379 , there is a pair of pants @xmath99 with boundary components the curves @xmath745}$ ] in @xmath153 , where @xmath746 runs over @xmath384 ; for another copy @xmath747 of @xmath56 , we may rewrite the boundary components of @xmath747 as @xmath748}$ ] in @xmath153 , where @xmath746 runs over @xmath384 . furthermore , observe the following common boundary components of opposite orientations .
the third curve of @xmath749 is the orientation reversal of the third curve of @xmath750 ; the fourth curve of @xmath751 can be rewritten as @xmath752}$ ] , which is clearly the orientation reversal of the fourth curve of @xmath753 ; the orientation reversal of each curve of @xmath60 appears exactly once as the second curve in @xmath749 , and similarly for @xmath754 and @xmath750 .
gluing the @xmath18-panted surfaces @xmath736 , @xmath741 , @xmath56 , and @xmath747 along these oppositely oriented common boundary components , the result is an oriented connected compact surface @xmath3 with exactly six boundary components , namely , two copies for each @xmath755 $ ] , where @xmath746 runs over @xmath384 . since @xmath737 $ ] equals @xmath756 $ ] under the additional assumption of the current step , the @xmath18-panted surface @xmath3 is as desired . *
step 2*. suppose @xmath718 .
suppose @xmath379 and @xmath385 are a @xmath641-tame rotation pair of right - handed tripods , and that @xmath717\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] for @xmath377 .
we construct an @xmath18-panted surface @xmath3 with exactly six boundary components , which are two copies of each @xmath717 $ ] for @xmath377 .
since @xmath385 is @xmath387-nearly regular where @xmath757 , an auxiliary @xmath758-nearly regular right - handed tripod @xmath759 can be drawn so that @xmath379 and @xmath759 form a @xmath760-tame rotation pair , and that @xmath759 satisfies the additional assumption of step 1 .
observe the following @xmath18-panted surfaces . applying step 1 to @xmath379 and @xmath759
, there is an @xmath18-panted surface @xmath167 with six boundary components , namely , two copies of each curve @xmath761 $ ] in @xmath153 for @xmath377 ; by rotation in the opposite chirality case ( statement ( 1 ) ) , applied to the @xmath700-tame rotation pair @xmath385 and @xmath762 , there is a pair of pants @xmath99 with boundary components @xmath763 $ ] in @xmath153 for @xmath377 ; also by rotation in the opposite chirality case ( statement ( 1 ) ) , applied to the @xmath760-tame rotation pair @xmath759 and @xmath762 , there is a pair of pants @xmath764 with boundary components @xmath765}$ ] in @xmath153 for @xmath377 ; since @xmath718 , by swapping ( construction [ swapping ] ) of @xmath760-tame swap pairs , for each @xmath377 , there is an @xmath18-panted surface @xmath736 with boundary components the curves @xmath717 $ ] @xmath766 $ ] @xmath767}$ ] @xmath768}$ ] in @xmath153 .
gluing the @xmath18-panted surfaces @xmath167 , and two copies of the @xmath18-panted surfaces @xmath56 , @xmath747 , @xmath769 , @xmath770 , and @xmath771 , along their oppositely oriented common boundary components , the result is an oriented connected compact surface @xmath3 with exactly six boundary components , namely , two copies for each @xmath717 $ ] for @xmath377 as desired .
this completes the proof of the statement ( 2 ) .
the antirotation construction is a variation of rotation when we join the legs in an ` unnatural ' way .
as before , we have two cases depending on the chiralities of the pair of tripods in consideration . [ antirotationconstruction ]
suppose that @xmath379 and @xmath385 are a @xmath715-tame rotation pair of tripods , and that @xmath716\in{{\mathbf{\gamma}}}_{r,\epsilon}$ ] for @xmath389 .
then an oriented connected compact @xmath18-panted surface @xmath3 can be constructed satisfying the following . 1 .
suppose @xmath718 . if @xmath379 and @xmath385 are of opposite charalities , then @xmath3 has exactly six boundary components , namely , two copies of each @xmath772 $ ] in @xmath153 , for @xmath377 .
2 . suppose @xmath640
. if @xmath379 and @xmath385 are of identical charality , then @xmath3 has exactly three boundary components namely , @xmath772 $ ] in @xmath153 , for @xmath377 . because @xmath773 $ ] and @xmath774 $ ] form a @xmath715-tame swap pair , by swapping ( construction [ swapping ] ) , there is an oriented connected compact @xmath18-panted surface @xmath775 with exactly four boundary components @xmath773 $ ] , @xmath774 $ ] , @xmath776}$ ] , @xmath777}$ ] .
thus , it suffices an oriented connected compact @xmath18-panted surface @xmath167 satisfying the following . 1 . if @xmath379 and @xmath385 are of opposite charalities , then @xmath167 has exactly six boundary components , namely , two copies of @xmath778 $ ] , @xmath779 $ ] , and @xmath780 $ ] .
2 . if @xmath379 and @xmath385 are of identical charality , then @xmath167 has exactly three boundary components namely , @xmath778 $ ] , @xmath779 $ ] , and @xmath780 $ ] .
in fact , this follows immediately from rotation ( construction [ rotationconstruction ] ) with respect to the @xmath715-tame rotation pair @xmath379 and @xmath781 .
note that the chirality of @xmath781 is exactly opposite to that of @xmath385 .
this completes the proof .
in this section , we introduce the @xmath18-panted cobordism group of oriented @xmath18-multicurves in an oriented closed hyperbolic @xmath0-manifold .
this will serve as a correction theory which will reduce the relative case of theorem [ homologyviapants ] ( 1 ) to the absolute case .
let @xmath4 be an oriented closed hyperbolic @xmath0-manifold . by an _
@xmath18-nearly hyperbolic multicurve _ , or simply an _
@xmath18-multicurve _ , we mean a ( possibly disconnected ) nonempty immersed oriented closed @xmath2-submanifold of @xmath4 of which all the components are @xmath18-nearly hyperbolic curves .
[ pantedcobordismgroup ] an _ @xmath18-nearly regularly panted cobordism _ , or simply an _ @xmath18-panted cobordism _ , between two @xmath18-multicurves @xmath782 in @xmath4 is an @xmath18-panted subsurface @xmath3 such that @xmath625 is the disjoint union @xmath783 , where @xmath784 denotes the orientation - reversal of @xmath785 .
we say that @xmath782 are _ @xmath18-panted cobordant _ if there exists an @xmath18-panted cobordism between them . assuming that being @xmath18-panted cobordant is an equivalence relation on the set of @xmath18-multicurves , we denote by @xmath10 , or simply @xmath786 , the set of all @xmath18-panted cobordism classes of @xmath18-multicurves , and by @xmath24_{r,\epsilon}$ ] the cobordism class of any @xmath18-multicurve @xmath19 .
an @xmath18-panted cobordism is not necessarily connected or @xmath1-injective , ( cf .
definition [ pantedsubsurfacedefinition ] ) . for any universally small positive @xmath14 and any sufficiently large positive @xmath15 depending only on @xmath4 and @xmath14
, it will be verified that being @xmath18-panted cobordant is an equivalence relation for @xmath18-multicurves .
furthermore , @xmath786 is an abelian group with the addition induced by the disjoint union operation between @xmath18-multicurves and the inverse induced by the orientation reversion of @xmath18-multicurves , ( subsection [ subsec - phi ] ) .
[ theorempantedcobordism ] let @xmath4 be an oriented closed hyperbolic @xmath0-manifold . for any universally small positive @xmath14 , and any sufficiently large positive @xmath15 depending only on @xmath4 and @xmath14
, there is a canonical isomorphism @xmath787 where @xmath12 denotes the bundle over @xmath4 of special orthonormal frames with respect to the orientation of @xmath4 .
moreover , for all @xmath24_{r,\epsilon}\in{{\mathbf{\omega}}}_{r,\epsilon}(m)$ ] , the image of @xmath26_{r,\epsilon})$ ] under the bundle projection is the homology class @xmath24\in h_1(m;{{\mathbf z}})$ ] .
the last part of the statement implies that the integral module @xmath10 is equivalent to @xmath788 as a split extension of @xmath789 by @xmath17 ( subsection [ subsec - phi ] ) .
therefore , the isomorphisms between @xmath10 and @xmath788 that are extension equivalences are in natural bijection to @xmath790 , where the canonical isomorphism @xmath11 corresponds to @xmath358 . in fact , one may require @xmath14 to be at most @xmath25 and @xmath15 to be at least @xmath791 as provided by lemma [ setupdata ] , then the conclusion of theorem [ theorempantedcobordism ] holds .
the idea of theorem [ theorempantedcobordism ] is developed from the ( non - random ) correction theory part of @xcite . in that paper , the notion _ good pants homology _ of an oriented closed hyperbolic surface @xmath792 is informally introduced , and with our notations above , the good pants homology of @xmath792 there means precisely the rational @xmath18-panted cobordism group @xmath793 , cf .
* definition 3.2 ) .
the proof of the good correction theorem ( * ? ? ?
* theorem 3.2 ) essentially implies that there is an isomorphism @xmath794 , such that @xmath795_{r,\epsilon})$ ] equals @xmath796 $ ] .
in fact , most part of the proof of ( * ? ? ? * theorem 3.2 ) can be extended directly to the @xmath0-dimensional case , yielding an isomorphism @xmath797 as above .
motivated by pushing the result to the integral coefficient case , the main innovation of theorem [ theorempantedcobordism ] lies in the observation that in many senses , it should be more natural to replace @xmath798 with @xmath27 .
one reason , for instance , is that @xmath18-multicurves and @xmath18-pants admit certain canonical lifts into @xmath12 because of their geometry ; another reason is that technically , passing to @xmath12 resolves certain ambiguity in the definition of the inverse of @xmath11 ; the reader may also observe a vague analogy between the statement of theorem [ theorempantedcobordism ] and the thom
pontrjagin correspondence in cobordism theory , thinking of @xmath27 as @xmath799 modulo the action of conjugations .
the proof of theorem [ theorempantedcobordism ] is organized slightly differently from the treatment of @xcite .
fix a basepoint @xmath800 of @xmath4 and a special orthonormal frame @xmath801 as a basepoint of @xmath12 .
we will construct the homorphism @xmath11 and a homomorphism @xmath802 , which descends to be a homomorphism @xmath803 by abelianization .
we will verify that @xmath804 and that @xmath805 surjects @xmath786 .
this will imply that @xmath11 is an isomorphism with the inverse @xmath806 . to briefly describe @xmath11 ,
note that any curve @xmath112 has a framing over its geodesic representative given by ( nearly ) parallel transporting a frame around @xmath36 .
the _ canonical lift _
@xmath807 of @xmath36 , well defined up to homotopy , is then a framing that differs from the the parallel - transportation framing by a loop of matrices @xmath808 that represents the nontrivial element of @xmath809 ( definition [ canonicallift ] ) .
the homomorphism @xmath810 will hence be uniquely defined so that @xmath811_{r,\epsilon})$ ] equals @xmath812 $ ] .
the definition of @xmath805 will depend on a choice of a finite triangular generating set @xmath813 of @xmath814 , together with some other setup data . here _ triangular _ means that all the relators of length at most @xmath0 in the generating set gives rise to a finite presentation of @xmath814
. we will define @xmath815 ( or decorated with some setup data , @xmath816 ) to be represented by some @xmath18-multicurve constructed from @xmath817 , so that @xmath818 $ ] is obviously equal to @xmath819 $ ] in @xmath27 . with the derived constructions of subsection [ subsec - derivedconstructions ]
, we will verify that @xmath820 in @xmath786 whenever there is a triangular relation @xmath821 .
it will follow that @xmath822 is a well defined homomorphism .
we point out that the triangular generating set of @xmath814 , or more essentially , a finite triangular presentation of @xmath823 , is the source of topological information which makes the construction of @xmath805 possible .
this idea will be further investigated in section [ sec - pantifyingsecondhomologyclasses ] when we pantify second homology classes .
the rest of this section is devoted to the proof of theorem [ theorempantedcobordism ] . in subsection [ subsec - phi ]
, we define the homomorphism @xmath11 . in subsection
[ subsec - theinverseofphi ] , we define the homomorphism @xmath805 and verify that @xmath806 is the inverse of @xmath11 . in subsection
[ subsec - proofoftheorempantedcobordism ] , we summarize the proof of theorem [ theorempantedcobordism ] . throughout this section , after fixing a basepoint @xmath800 of @xmath4 , we will no longer distinguish a nontrivial element of @xmath823 from its pointed geodesic loop representative , so it makes sense to speak of the length , or the initial or terminal direction of a nontrivial element in @xmath823 . in this subsection
, we define the homomorphism @xmath11 .
we also need to mention some basic facts about the @xmath18-panted cobordism gorup @xmath786 , and about the special orthonomal framing bundle @xmath12 .
the following assumptions of constants will be adopted throughout this subsection : * @xmath356 is any pair of positve constants where @xmath824 and @xmath628 , and @xmath19 satisfies the conclusion of the connection principle ( lemma [ connectionprinciple ] ) with respect to @xmath103 and @xmath4 .
* @xmath18 is any pair of positive constants where @xmath14 is at most @xmath25 . *
@xmath825 and @xmath826 .
we will need some stronger assumptions in the next subsection , which will be specified by lemma [ setupdata ] .
[ pantedcobordismequivalence ] being @xmath18-panted cobordant is an equivalence relation on the set of @xmath18-multicurves .
the relation of being @xmath18-cobordant is clearly symmetric and transitive , so it remains to check the reflexivity .
any @xmath18-curve can be written as the cyclic concatenation of two consecutive @xmath13-framed segments of the same length and phase , which is @xmath356-tame by the assumptions of constants for this subsection . then by splitting ( construction [ splitting ] ) , any @xmath18-curve @xmath36 is the cuff of an @xmath18-pair - of - pants @xmath56 . taking two
oppositely oriented copies of @xmath56 and gluing along the two pair of cuffs other than the pair of cuff @xmath36 and its orientation reversal yields an @xmath18-panted cobordism between @xmath36 and itself .
this implies that any @xmath18-multicurve is @xmath18-panted cobordant to itself .
[ pantedcobordismgroupstructure ] the set of @xmath18-panted cobordism classes of @xmath18-multicurves form a finitely generated abelian group @xmath786 .
the addition is induced by the disjoint union operation , and the inverse is induced by orientation reversion .
the connection principle ( lemma [ connectionprinciple ] ) implies that @xmath153 and @xmath611 are nonempty finite sets .
hence @xmath786 is nonempty .
define the addition on @xmath786 by @xmath24_{r,\epsilon}+[l']_{r,\epsilon}=[l\sqcup l']_{r,\epsilon}$ ] .
it is straightforward to check that @xmath786 forms an abelian group under the addition finitely generated over @xmath153 .
in fact , we have a natural presentation of @xmath786 given by the exact sequence @xmath827 the zero of @xmath786 is represented by the sum of the three cuffs for any @xmath99 . the same argument as the proof of lemma [ pantedcobordismequivalence ] implies that @xmath24_{r,\epsilon}=-[\bar{l}]_{r,\epsilon}$ ] .
it suffices to assume that @xmath19 has only one component .
then @xmath828_{r,\epsilon}=-[\bar{c}]_{r,\epsilon}$ ] if there is a pair of pants @xmath829 with a cuff @xmath158 , for each component @xmath158 of @xmath19 .
the condition certainly holds for universally small @xmath14 if @xmath15 is sufficiently large ( construction [ splitting ] ) .
in fact , one may take two oppositely oriented copies of @xmath830 of @xmath831 , and glue them along their two cuffs other than @xmath158 or @xmath832 .
the resulting @xmath18-panted surface has exactly two boundary component @xmath158 and @xmath832 .
[ finercurves ] @xmath786 is generated by the @xmath18-panted cobordism classes of @xmath833-multicurves .
this follows immediately from splitting ( construction [ splitting ] ) .
fix an orthonormal frame @xmath834 at a fixed basepoint @xmath800 of @xmath4 , and regard @xmath835 as a basepoint of the total space @xmath12 of the bundle over @xmath4 of special orthonormal frames with respect to the orientation of @xmath4 .
one may naturally identify @xmath12 as @xmath836 . since the tangent bundle of a closed orientable @xmath0-manifold is always trivializable
, @xmath12 is a trivial @xmath837-principal bundle over @xmath4 .
there are canonical short exact sequences @xmath838 and @xmath839 both of which are splitting but not naturally . note that @xmath840 is the center of @xmath814 .
we will usually write @xmath841 for the nontrivial central element .
certain noncentral elements of @xmath814 , namely , the @xmath103-sharp elements in the sense of definition [ sharpelement ] below , can be naturally represented by their associated oriented @xmath13-framed segments .
this provides a convenient way to understand such elements , which will be especially useful when we construct the inverse of @xmath11 .
[ sharpelement ] for any positive constant @xmath103 which is at most @xmath25 , a noncentral element @xmath842 is said to be _
if its image @xmath843 in @xmath823 has the initial and terminal directions @xmath103-close to @xmath844 and @xmath845 , respectively . for a @xmath103-sharp @xmath846
, we will say that an oriented @xmath13-framed segment @xmath847 is _ associated to _
@xmath846 , and vice versa , if @xmath847 satisfies the following . *
the carrier segment of @xmath847 is @xmath843 .
the phase of @xmath847 is @xmath103-close to @xmath358 .
the initial and terminal framings of @xmath847 are @xmath103-close to each other . *
the element @xmath846 is the homotopy class represented by a closed path of frames based at @xmath848 as follows .
the path first flows @xmath835 along @xmath843 by parallel transportation , and then rotates @xmath849 counterclockwise about @xmath850 , and then returns to @xmath835 along a @xmath103-short path within @xmath851 .
we will justify the existence of the associated @xmath13-segment in lemma [ sharpelementlemma ] . to understand the issue which the definition addresses , consider for the moment a general nontrivial element @xmath852 represented by a closed geodesic path based at @xmath853 .
the parallel transportation of @xmath835 along @xmath843 gives rise to a path of frames @xmath854\to{{\mathrm{so}}}(m)$ ] which departs from @xmath835 and arrives at another frame @xmath855 at @xmath853 .
concatenate this path of frames with any path of frames @xmath856\to{{\mathrm{so}}}(m)|_{{*}}$ ] , then we obtain a closed path based at @xmath835 which represents a lift @xmath842 of the element @xmath852 .
note that there are two possible lifts of @xmath843 , namely , @xmath846 and @xmath857 .
in general , there is no natural way to distinguish these two possibilities by choosing the second path without ambiguity .
however , when the angle of @xmath843 at @xmath853 is sufficiently sharp , there are essentially ( up to @xmath103-small difference ) two ways to enrich @xmath843 with a @xmath13-framing subject to the first condition of definition [ sharpelement ] , and each of them specifies a lift of @xmath843 in @xmath814 by the description in second condition of definition [ sharpelement ] .
we point out that our biased choice of the @xmath849 _ counterclockwise _ rotation in definition [ sharpelement ] determines our choices of chirality for tripods in the rest of this section .
the essential difference between distinct chiralities is revealed by lemma [ sharptripodlemma ] and remark [ sharptripodlemmaremark ] .
[ sharpelementlemma ] for any @xmath103-sharp element @xmath842 , there is an oriented @xmath13-framed segment @xmath847 associated to @xmath846 , unique up to @xmath103-small change of framings at endpoints .
moreover , @xmath858 is associated to @xmath857 , and @xmath859 is associated to @xmath860 .
let @xmath861 be a unit vector orthogonal to @xmath844 such that the parallel transportation of @xmath861 along @xmath843 to the other end is @xmath103-close to @xmath861 .
up to @xmath103-small change , there are only two possible such vectors , namely , @xmath862 . enriching @xmath843 with initial and terminal framings both @xmath103-close to @xmath862 yield oriented @xmath13-framed segments @xmath863 satisfying the first part of the listed properties in definition [ sharpelement ] .
it is clear that exactly one of @xmath863 fulfills the second part of the listed properties , so we pick it as @xmath847 . the ` moreover ' part is straightforward from the construction above as well .
[ sharptripodlemma ] for any positive constant @xmath103 which is at most @xmath25 , suppose that @xmath864 is a @xmath865-tame left - handed tripod with coincident terminal points of legs @xmath853 and with the terminal directions of legs @xmath103-close to @xmath845 . then there exist angles @xmath866 satisfying @xmath867 such that @xmath868 is @xmath103-close to the directed angle from @xmath869 to @xmath870 with respect to the common orthogonal vector at @xmath800 which is @xmath103-close to @xmath845 , and that each @xmath871 ( definition [ partialframedsegment ] ) is associated to a @xmath103-sharp element @xmath872 for @xmath377 . for any such @xmath873 as above
, the triangular relation @xmath874 is satisfied .
moreover , adding two of the three @xmath873 by @xmath329 yields another triple of angles satisfying the conditions above , with two corresponding @xmath817 changed into @xmath875 .
[ sharptripodlemmaremark ] if @xmath864 is right - handed , we must either replace the triangular relation in the conclusion with the twisted triangular relation @xmath876 , or alternatively , replace the equation for the anlges with @xmath877 . for each @xmath377
, pick a unit vector @xmath475 at @xmath800 orthogonal to @xmath844 , such that @xmath475 is @xmath103-close to @xmath878 .
let @xmath879 be the angle from @xmath475 to @xmath880 with respect to @xmath844 , and let @xmath881 be half of @xmath882 , valued in @xmath883 . choose a lift @xmath884 for each @xmath881 ,
so that @xmath885 .
note that any other lift can be obtained from changing two @xmath873 by adding @xmath329 .
it is clear that @xmath871 is associated to a @xmath103-sharp @xmath872 , and @xmath886 equals either @xmath887 or @xmath888 .
we claim that it is the former case .
it suffices to verify that @xmath886 is trivial in @xmath889 .
the argument is routine and easy so we only include an outline below .
let @xmath890 be a path of frames from @xmath891 to @xmath892 that first flows along @xmath893 by parallel transportation , and then rotates @xmath849 counterclockwise about @xmath880 , and then rotates to @xmath892 via a @xmath103-short path in @xmath851 . here
@xmath894 is the image of @xmath895 , also regarded as a pointed geodesic loop . since @xmath475 are all @xmath103-close to the normal vector of the @xmath307-simplex @xmath570 spanned by the concatenation of @xmath896 , the loop of frames @xmath897 obtained by concatenation is homotopic to the constant loop @xmath898 in @xmath899 as @xmath864 is left - handed .
( compare with the right - handed case where the resulting loop would be the @xmath900 counterclockwise rotation of the frame @xmath898 about @xmath901 . ) in particular , @xmath902=0\ ] ] in @xmath889 .
let @xmath903 be a unit vector at @xmath800 orthogonal to @xmath844 , such that @xmath903 is @xmath103-close to both the initial and terminal framings of @xmath871 .
let @xmath904 be the path of frames in @xmath851 from @xmath905 to @xmath906 by a rotation of angle @xmath103-close to @xmath907 , and let @xmath908 be the path of frames in @xmath851 from @xmath909 to @xmath905 by a rotation of angle @xmath103-close to @xmath907 .
then the loop of frames @xmath910 based at @xmath905 can be conjugated to the loop based at @xmath835 representing @xmath895 described in definition [ sharpelement ] .
thus @xmath911=[\hat{\xi}_{i+2}\hat{\beta}_{i+2}\hat{\eta}_{i+2}]\ ] ] in @xmath889 .
note that @xmath912 is a loop of frames in @xmath851 that rotates counterclockwise about @xmath844 of angle @xmath913 .
as @xmath885 modulo @xmath914 , the winding number of this loop around @xmath844 is even
. hence @xmath915=0\ ] ] in @xmath27 . therefore , in @xmath27 , @xmath916&= & [ \hat{\xi}_{0}\hat{\beta}_{0}\hat{\eta}_{0}]+[\hat{\xi}_{1}\hat{\beta}_{1}\hat{\eta}_{1}]+[\hat{\xi}_{2}\hat{\beta}_{2}\hat{\eta}_{2}]\\ & = & [ \hat{\xi}_{0}\hat{\beta}_{0}\hat{\eta}_{0}\hat{\xi}_{1}\hat{\beta}_{1}\hat{\eta}_{1}\hat{\xi}_{2}\hat{\beta}_{2}\hat{\eta}_{2}]\\ & = & [ \hat{\beta}_{0}\hat{\beta}_{1}\hat{\beta}_{2}]+[\hat{\xi}_0\hat{\eta}_0\hat{\xi}_1\hat{\eta}_1\hat{\xi}_2\hat{\eta}_2]\\ & = & 0 .
\end{aligned}\ ] ] this implies that @xmath917 in @xmath814 .
there is a canonical way to lift @xmath18-curves and @xmath18-pants into @xmath12 , up to homotopy .
[ canonicallift ] for any curve @xmath112 , a _
canonical lift _ of @xmath36 is a loop of frames @xmath918 as follows .
choose a point @xmath338 on the geodesic representative of @xmath36 , and a normal vector @xmath599 of @xmath36 at @xmath338 .
let @xmath572 be the direction vector of @xmath36 at @xmath338 .
the frame @xmath919 is an element of @xmath920 . with these notations ,
a base - point free loop of frames @xmath921 starts from @xmath922 , and then flows once around @xmath36 by parallel transportation , and then rotates @xmath900 counterclockwise about @xmath599 , and then rotates back to @xmath922 along an @xmath14-short path within @xmath920 . for any pair of pants @xmath99 , a _
canonical lift _ of @xmath56 is a lift @xmath923 of @xmath56 , such that the three cuffs are canonically lifted .
the definition is justified by the following lemma [ canonicalliftlemma ] .
similar to the explanation after definition [ sharpelement ] , the nearly geodesic assumption helps to distinguish the two possible lifts of a @xmath18-curve @xmath36 in @xmath12 up to homotopy , namely , the canonical lift above and the lift by almost parallel transportation along @xmath36 .
the fundamental difference between the two possible lifts is that if we lift all the cuffs of @xmath56 in the latter way , it would be impossible to extend the lift to @xmath56 up to homotopy .
[ canonicalliftlemma ] for any positive constant @xmath14 which is at most @xmath25 , suppose @xmath112 and @xmath924 .
the canonical lifts @xmath921 and @xmath925 as described in definition [ canonicallift ] exist and are unique up to homotopy .
[ canonicalliftremark ] however , if the canonical lifts of @xmath925 on the cuffs have been chosen , @xmath925 is unique only up to homotopy relative to cuffs together with @xmath17-dehn twists in the fiber @xmath837 near the boundary . in other words ,
the relative homotopy class of @xmath925 is determined by any class of @xmath926 that projects to the fundamental class @xmath927 $ ] in @xmath928 .
the existence of @xmath921 is by definition .
the uniqueness follows from the fact that the set of homotopy classes of framings of @xmath929 is bijective to @xmath930\cong { { \mathbf z}}_2 $ ] . to see the existence of @xmath925 , note that the pull - back tangent bundle @xmath931 , namely , @xmath932 , is isomorphic to @xmath933 .
consider a trivialization of @xmath934 , for example , by embedding @xmath45 into the plane and endowing with the standard framing of @xmath935 . by direct summing with the trivialization induced by the orientation of @xmath56 ,
the trivialization of @xmath934 naturally induces a framing of @xmath931 up to homotopy .
the restriction of this framing on any cuff @xmath36 of @xmath56 is the canonical lift of @xmath36 .
thus this framing of @xmath931 is a canonical lift @xmath936 of @xmath56 by definition . to see the uniqueness of @xmath925 , note that the set of homotopy classes of framings of @xmath937 is bijective to @xmath938\cong { { \mathbf z}}_2\oplus{{\mathbf z}}_2 $ ] .
thus the homotopy class of a framing of @xmath937 is uniquely determined by its restriction to the cuffs .
this completes the proof . under the assumptions of constants introduced at the beginning of this subsection , we construct @xmath939 as follows .
suppose @xmath36 is a geodesic representative of a curve in @xmath153 .
we define @xmath811_{r,\epsilon})$ ] in @xmath27 to be represented by the canonical lift of @xmath36 ( definition [ canonicallift ] ) . for an @xmath18-multicurve @xmath19
, we define @xmath940_{r,\epsilon})\,\in\,h_1({{\mathrm{so}}}(m);\,{{\mathbf z}})\ ] ] to be the sum of @xmath11 defined for each of its components .
we verify that @xmath11 is well defined .
[ welldefinedphi ] the homology class @xmath26_{r,\epsilon})$ ] in @xmath788 constructed above depends only on the @xmath18-panted cobordism class @xmath24_{r,\epsilon}\in{{\mathbf{\omega}}}_{r,\epsilon}$ ] of @xmath19 . moreover , the induced map @xmath11 from @xmath786 to @xmath788 is a homomorphism .
if @xmath24_{r,\epsilon}$ ] vanishes in @xmath786 , there exists an @xmath18-panted surface @xmath941 bounded by @xmath19 .
the canonical lifts of pairs of pants of @xmath3 ( definition [ canonicallift ] ) yield a canonical lift @xmath942 , whose restriction to the boundary are the canonical lifts of components of @xmath19 .
this implies that @xmath18-panted cobordant multicurves yield homologous canonical lifts , or in other words , that @xmath26_{r,\epsilon})$ ] in @xmath788 depends only @xmath24_{r,\epsilon}$ ] .
the ` moreover ' part is straightforward from the definition .
[ canonicallydefinedphi ] for any @xmath18-panted cobordism class @xmath24_{r,\epsilon}$ ] , the image of @xmath26_{r,\epsilon})$ ] under the natural projection from @xmath788 to @xmath798 is the homology class @xmath24 $ ] .
this follows immediately from the construction of @xmath11 .
fix an orthonormal frame @xmath943 at a fixed basepoint @xmath800 of @xmath4 as before . in this subsection
, we construct a homomorphism @xmath944 which , descending to the abelianization , @xmath945 yields the inverse of @xmath11 .
we need to choose some setup data including a triangular finite generating set of @xmath823 , and define @xmath805 on a subset of @xmath823 that contains the triangular generating set .
we verify that @xmath805 extends as a homomorphism by showing that it vanishes on words corresponding to the triangular relations .
we also verify that @xmath806 is the inverse of @xmath11 by showing that it is surjective and is the pre - inverse of @xmath11 . given any small positive number @xmath14 at most @xmath25 , we need to fix some setup data @xmath946 in terms of @xmath823 , restrained by some environmental data @xmath947 , which is a pair of constant fulfilling the conditions of the connection principle ( lemma [ connectionprinciple ] ) . such a collection of data is provided by lemma [ setupdata ] below , together with a positive constant @xmath948 such that the homomorphism @xmath805 from @xmath814 to @xmath786 can be constructed for any constant @xmath15 greater than @xmath791 .
we naturally identify the universal cover of @xmath4 as the hyperbolic @xmath0-space @xmath29 with a given basepoint @xmath949 together with an orthonormal frame @xmath950 , so @xmath823 naturally acts on @xmath29 by isometries .
for any positive constant @xmath160 , we denote by @xmath951 the open ball of radius @xmath160 centered at @xmath949 , and by @xmath952 the open half - space containing @xmath949 bounded by the hyperplane tangent to @xmath953 at the point in the direction @xmath954 .
define @xmath955 and @xmath956 to gain some intuition about these subsets , one may verify that elements of @xmath957 are represented by geodesic closed paths in @xmath4 based at @xmath800 which have length at most @xmath160 , and elements of @xmath958 are represented by geodesic closed paths which form an angle at most @xmath959 at @xmath800 . in the following , a _
triangular generating set _ of @xmath823 means a generating set such that there is a finite set of defining relations in these generators ( or their inverses ) consisting of words of length at most @xmath0 , ( cf .
subsection [ subsec - triangularpresentationcomplexes ] for discussion in more details ) ; the conjugation @xmath960 induced by an element @xmath961 acts on @xmath823 by @xmath962 .
[ setupdata ] let @xmath4 be an oriented closed hyperbolic @xmath0-manifold .
given any positive constant @xmath14 at most @xmath25 , there is a collection of data depending only on @xmath4 and @xmath14 as follows . 1 .
there exist positive constants @xmath103 and @xmath19 .
the constant @xmath103 is at most @xmath963 ; the constant @xmath19 is at least @xmath964 , and satisfies the conclusion of the connection principle ( lemma [ connectionprinciple ] ) with respect to @xmath103 and @xmath4 .
2 . there exist positive constants @xmath965 and @xmath410 .
the constant @xmath965 is at least @xmath966 , and @xmath967 contains a generating set of @xmath823 ; the constant @xmath410 is at least @xmath965 , and @xmath968 contains a triangular generating set of @xmath823 which further contains @xmath967 .
3 . there exists a conjugation @xmath960 of @xmath823 such that @xmath969 is contained in @xmath970 .
moreover , @xmath971 can be chosen in @xmath970 .
4 . there exists a positive constant @xmath948 such that @xmath972 contains @xmath969 .
in particular , the derived constructions of subsection [ subsec - derivedconstructions ] can be applied with respect to the constants @xmath356 of lemma [ setupdata ] ( 1 ) . with lemma [ setupdata ] ( 2 ) and ( 3 ) , we have a triangular generating set @xmath969 of @xmath973 , which are represented by ( unframed ) @xmath103-sharp elements , ( cf .
lemma [ sharpck ] ) . this will enable us to define the homomorphism @xmath805 .
the assumption about @xmath974 ensures that @xmath975 contains enough generators so that with lemma [ setupdata ] ( 4 ) we will be able show the surjectivity of @xmath805 . with some needless awkward intuition
, the geodesic representative of @xmath971 based at @xmath800 may look like a long , sharp , closed segment , of which the direction vectors at the endpoints are not extremely close to any direction vector of any element of @xmath975 .
we would rather suggest the reader to understand @xmath971 simply to be an element that guarantees lemma [ setupdata ] ( 3 ) and ( 4 ) .
the statement ( 1 ) is implied by the connection principle ( lemma [ connectionprinciple ] ) .
the statement ( 2 ) follows from the fact that any presentation @xmath976 of a group @xmath977 induces a triangular generating set @xmath978 consisting of all the elements that are represented by subwords of all the relators from @xmath979 .
note that @xmath978 is finite if @xmath976 is a finite presentation .
since @xmath823 is finitely presented , we may choose @xmath965 at least @xmath966 to be sufficiently large so that @xmath967 generates @xmath823 .
then for any finite presentation of @xmath823 over @xmath967 , the induced finite triangular generating set @xmath980 is finite , and hence contained in @xmath981 for some sufficiently large @xmath410 . to prove the statement ( 3 )
, we must find an element @xmath961 such that @xmath982 is disjoint from @xmath981 for all @xmath983 .
note that for any nontrivial @xmath984 , if an open half - space @xmath985 is sufficiently far from the axis of @xmath843 , depending only on the translation distance of @xmath843 , then @xmath986 . because @xmath975 is a finite set , there exists an open half - space @xmath985 so that @xmath986 for all @xmath983 .
let @xmath961 be an element whose axis is contained in @xmath985 .
we may further assume that @xmath987 .
possibly after passing to a sufficiently great power of @xmath971 , we may assume @xmath988 where @xmath989 denotes the complement of @xmath985 .
see figure [ figsetupdata ] . for such @xmath971 , the half - space @xmath990
is contained in @xmath985 , so we have @xmath991 , for all @xmath983 . in other words , @xmath982 is disjoint from @xmath981 for all @xmath983 as desired . moreover , if we took @xmath985 so that @xmath992 , then @xmath971 would be contained in @xmath970 . in fact
, @xmath988 implies that @xmath993 .
the statement ( 4 ) is now obvious since we have chosen @xmath19 , @xmath410 , and @xmath960 in a way depending only on @xmath4 and @xmath14 .
[ sharpck ] with the constants from lemma [ setupdata ] , for any @xmath994 , the initial direction and the terminal direction of @xmath843 at @xmath800 are @xmath103-close to @xmath844 and @xmath845 , respectively . the complement of the half - space @xmath952 is contained in the cone at @xmath949 with axis along @xmath954 of cone angle @xmath9 where @xmath995 , so @xmath9 is clearly much less than @xmath103 when @xmath160 equals @xmath996 .
since @xmath997 , @xmath998 and @xmath999 are both contained in the complement of @xmath952 .
this implies that @xmath843 is @xmath103-sharp with the initial direction and the terminal direction @xmath103-close to @xmath844 and @xmath845 , respectively .
given any positive constant @xmath14 , we fix a collection of setup data @xmath1000 subject to @xmath947 as provided by lemma [ setupdata ] , and obtain a positive constant @xmath791 accordingly . for any constant @xmath15 at least @xmath791 , we will construct the homomorphism @xmath944 in the following .
more precisely , let @xmath1001 and @xmath1002 denote the preimages of @xmath957 and @xmath958 in @xmath814 , respectively .
since @xmath814 is a central extension of @xmath823 by @xmath17 , the conjugation @xmath960 of @xmath823 naturally induces a conjugation of @xmath814 , which will be denoted as @xmath1003 .
we will construct a set - theoretic map @xmath1004 the restriction of @xmath1005 to @xmath1006 , denoted as @xmath1007 will be a partial homomorphism , so the further restriction to @xmath1008 will extend uniquely over @xmath814 to be a homomorphism , which will still be denoted as @xmath1009 . however , it will be verified that @xmath1009 descends to the abelianization , yielding a homomorphism @xmath1010 which is exactly the inverse of @xmath11 ( subsubsection [ subsubsec - verifications ] ) , so eventually we may drop the subscript simply writing @xmath1009 as @xmath805 .
for any element @xmath1011 , by lemmas [ sharpelementlemma ] , [ sharpck ] , we may choose an oriented @xmath13-framed segment @xmath1012 associated to the @xmath103-sharp element @xmath846 . note that @xmath1013 is unique up to @xmath103-small change of the initial and terminal framings , and @xmath1014 can be chosen as the framing flipping @xmath1015 .
we define the claimed set - theoretic map @xmath1016 as follows . for each element @xmath1011 , choose an oriented @xmath13-framed segment @xmath1013 which is associated to @xmath846 as above . for convenience , choose a unit vector @xmath1017 orthogonal to @xmath844 , such that @xmath1018 and @xmath1019 are both @xmath103-close to @xmath1020 . by the connection principle ( lemma [ connectionprinciple ] ) , choose a right - handed nearly regular tripod @xmath379 and an oriented @xmath13-framed segment @xmath363 satisfying the following : * the right - handed tripod @xmath379 is @xmath1021-nearly regular . for each @xmath377 , the terminal endpoint @xmath1022 equals @xmath800 , and the terminal direction @xmath1023 is @xmath699-close to @xmath844 , and
the terminal framing @xmath1024 is @xmath699-close to @xmath1020 . *
the oriented @xmath13-framed segment @xmath363 has length @xmath699-close to @xmath1025 and phase @xmath699-close to @xmath358 .
the initial and terminal directions of @xmath363 are @xmath103-close to @xmath845 and @xmath844 respectively , and the initial and terminal framings of @xmath363 are both @xmath103-close to @xmath1020 . ] in @xmath786 , for any element @xmath1011 , define @xmath1026_{r,\epsilon } + [ \mathfrak{s}_{\hat{g}}\mathfrak{a}_{12}]_{r,\epsilon } + [ \mathfrak{s}_{\hat{g}}\mathfrak{a}_{20}]_{r,\epsilon } -\,[\mathfrak{s}_{\hat{g}}\mathfrak{b}]_{r,\epsilon } -\,[\mathfrak{s}_{\hat{g}}\bar{\mathfrak{b}}]_{r,\epsilon}.\ ] ] for any element @xmath1027 , we define @xmath1028 ( lemma [ setupdata ] ( 3)(4 ) ) .
[ welldefinition ] the set - theoretic map @xmath1016 is well defined . in other words , 1 .
all the reduced cyclic concatenations involved are curves in @xmath1029 ; 2 . for any @xmath1030
, @xmath1031 depends only on @xmath846 ; the statement ( 1 ) follows from straightforward verification using the length and phase formula ( lemma [ lengthandphaseformula ] ) under our fixed choice of setup data ( lemma [ setupdata ] ) . to prove the statement ( 2 ) ,
observe that @xmath1031 is clearly independent of the choice of @xmath1013 and @xmath1020 , since the carrier segment of @xmath1013 is unique and @xmath1020 is unique up to @xmath103-closeness .
suppose that @xmath1032 is another oriented @xmath13-framed segment satisfying the same conditions as of @xmath379 , and similarly for @xmath371 .
we write the new @xmath1033 as @xmath1034 to distinguish .
we must show that @xmath1033 equals @xmath1034 in @xmath786 .
choose an auxiliary left - handed nearly regular tripod @xmath724 satisfying the following .
* the left - handed tripod @xmath724 is @xmath1035-nearly regular . for each @xmath377 , the terminal endpoint @xmath1036 equals @xmath853 , and the terminal direction @xmath1037 is @xmath103-close to @xmath845 , and
the terminal framing @xmath1038 is @xmath103-close to @xmath1020 .
since both @xmath379 and @xmath1032 have the opposite chirality to that of @xmath724 , it follows from rotation ( construction [ rotationconstruction ] ( 1 ) ) that @xmath1039_{r,\epsilon}\,=\,0\ ] ] and @xmath1040_{r,\epsilon}\,=\,0,\ ] ] where @xmath621 means @xmath1041 for @xmath377 , and similarly for the notations @xmath1042 and @xmath1043 . on the other hand , by swapping ( construction [ swapping ] ) , @xmath1044_{r,\epsilon}-[\mathfrak{s}_{\hat{g}}\mathfrak{a}'_{i , i+1}]_{r,\epsilon}= [ \bar{\mathfrak{c}}_{i , i+1}\mathfrak{a}_{i , i+1}]_{r,\epsilon}-[\bar{\mathfrak{c}}_{i , i+1}\mathfrak{a}'_{i , i+1}]_{r,\epsilon},\ ] ] and @xmath1045_{r,\epsilon}-[\mathfrak{s}_{\hat{g}}\mathfrak{b}']_{r,\epsilon}= [ \bar{\mathfrak{s}}_{\hat{g}}\mathfrak{b}]_{r,\epsilon}-[\bar{\mathfrak{s}}_{\hat{g}}\mathfrak{b}']_{r,\epsilon}.\ ] ] for convenience , we write @xmath1046 and @xmath1047 for @xmath363 and @xmath1048 , respectively , and similarly for @xmath1049 and @xmath1050
. then @xmath1051 equals @xmath1052_{r,\epsilon } -[\mathfrak{s}_{\hat{g}}\mathfrak{a}'_{i , i+1}]_{r,\epsilon } ) -\sum_{j\in{{\mathbf z}}_2}\,([\mathfrak{s}_{\hat{g}}\mathfrak{b}_{j , j+1}]_{r,\epsilon } -[\mathfrak{s}_{\hat{g}}\mathfrak{b}'_{j , j+1}]_{r,\epsilon})\\ & = & \sum_{i\in{{\mathbf z}}_3}\,([\bar{\mathfrak{c}}_{i , i+1}\mathfrak{a}_{i , i+1}]_{r,\epsilon}- [ \bar{\mathfrak{c}}_{i , i+1}\mathfrak{a}'_{i , i+1}]_{r,\epsilon } ) -0\\ & = & 0 , \end{aligned}\ ] ] or in other words , @xmath1033 equals @xmath1034 in @xmath786 .
this proves the statement ( 2 ) .
[ triangularrelation ] the set - theoretic map @xmath1009 on @xmath1006 is a partial homomorphism . in other words , 1 . for any @xmath1053 , @xmath1054 2 . for any triple
@xmath1055 satisfying the triagular relation @xmath917 , @xmath1056 [ remarktriangularrelation ] we point out that in the proof of lemma [ triangularrelation ] ( 2 ) , the last equality in step 1 uses antirotation of tripod pairs with opposite charalities ( construction [ antirotationconstruction ] ( 1 ) ) .
the presence of the coefficient @xmath307 there is not only indispensible but also crucial for theorem [ theorempantedcobordism ] to work in the integral coefficient case .
the corresponding fact is that in the paper @xcite , the conclusion of the second rotation lemma ( lemma 8.2 ) should be @xmath1057 if one attempts to state with integral coefficients . to simplify notations ,
we prove the statements with respect to @xmath1016 assuming @xmath817 in @xmath1058 . to prove the statement ( 1 ) , suppose that @xmath846 is an element of @xmath1059 .
observe that @xmath1060 can be chosen as the orientation reversed framing flipping @xmath1061 , and that the defining right - handed tripod and @xmath13-framed segment can be chosen as @xmath383 and @xmath1062 respectively , provided that @xmath1013 , @xmath379 , and @xmath363 have been chosen to define @xmath1031 .
this should be compared to the wrong choice @xmath1063 , @xmath379 , and @xmath363 , which is actually right for defining @xmath1064 by lemma [ setupdata ] ( 4 ) . for convenience , we write @xmath1046 and @xmath1047 for @xmath363 and @xmath1048 , respectively
. then @xmath1065 equals @xmath1066_{r,\epsilon}+[\bar{\mathfrak{s}}^*_{\hat{g}}\mathfrak{a}^*_{-i ,- i-1}]_{r,\epsilon } ) -\sum_{j\in{{\mathbf z}}_2}\ , ( [ \mathfrak{s}_{\hat{g}}\mathfrak{b}_{j , j+1}]_{r,\epsilon } + [ \bar{\mathfrak{s}}^*_{\hat{g}}\bar{\mathfrak{b}}^*_{j , j+1}]_{r,\epsilon})\\ & = & \sum_{i\in{{\mathbf z}}_3}\ , ( [ \mathfrak{s}_{\hat{g}}\mathfrak{a}_{i , i+1}]_{r,\epsilon}+[\bar{\mathfrak{s}}_{\hat{g}}\mathfrak{a}_{-i ,- i-1}]_{r,\epsilon } ) -\sum_{j\in{{\mathbf z}}_2}\ , ( [ \mathfrak{s}_{\hat{g}}\mathfrak{b}_{j ,
j+1}]_{r,\epsilon}+ [ \bar{\mathfrak{s}}_{\hat{g}}\bar{\mathfrak{b}}_{j , j+1}]_{r,\epsilon})\\ & = & \sum_{i\in{{\mathbf z}}_3}\ , ( [ \mathfrak{s}_{\hat{g}}\mathfrak{a}_{i , i+1}]_{r,\epsilon}+[\bar{\mathfrak{s}}_{\hat{g}}\mathfrak{a}_{i+1,i}]_{r,\epsilon } ) -0\\ & = & 0 . \end{aligned}\ ] ] this proves the statement ( 1 ) . to prove the statement ( 2 ) , suppose that @xmath1067 is a triple of elements in @xmath1068 satisfying the triagular relation @xmath917 . from the relation @xmath917 , the carrier segments of @xmath1069 form the boundary cycle of an oriented @xmath307-simplex @xmath570 in @xmath4 , so by simplex subdivision ( definition [ axiomsofconstructions ] ( 2 ) , cf .
subsubsection [ subsubsec - describingaconstruction ] ) and by lemma [ fermatpoint ] , there is a left - handed @xmath715-tame tripod @xmath864 , such that @xmath1070 and @xmath1071 are carried by the same segment for all @xmath746 . by lemma [ sharptripodlemma ]
, there are @xmath866 with @xmath885 and @xmath1072 @xmath103-close to the angle from @xmath869 to @xmath870 with respect to @xmath844 , such that @xmath871 is the same as @xmath1071 up to @xmath103-small change of the initial and terminal framings . in other words
, we may choose @xmath1071 to be @xmath871 instead .
the strategy is to prove @xmath1073 using the defining expression by manipulation and cancellation .
the fundamental case is when @xmath864 is nearly regular with the terminal framing of legs close to each other , or equivalently , when @xmath1074 are close to each other and @xmath873 are all close to @xmath358 .
as we can choose the defining right - handed tripod and the defining @xmath13-framed segment to be the same for all @xmath1075 , we expect the most cancellation in this case and expect the triangular relation @xmath917 to be translated into a cancellation by an antirotation with distinct chiralities
. in general , the difficulty to cancellation lies in that @xmath1074 may not be close to each other and @xmath873 may not be close to @xmath358 .
then we write @xmath1076 as a nearly consecutive tame concatenation @xmath1077 , in which @xmath1078 are close to each other .
apply the connection principle to construct a sequence of nearly consecutive tame concatenations @xmath1079 so that @xmath1080 have the same initial and terminal endpoints respectively , and @xmath1081 equals @xmath1082 .
assume that @xmath1080 and @xmath1083 have length and framings close to each other for @xmath1084 , and that the initial and terminal directions of @xmath1080 are close to those of @xmath1082 for all @xmath1085 .
in a way similar to @xmath864 and @xmath817 , we obtain a sequence of left - handed tripods @xmath1086 , and sharp elements @xmath1087 .
we expect to use swapping to obtain the equations @xmath1088 for @xmath1084 .
once this is done , we can immediately derive the general case by requiring that @xmath1089 falls into the fundamental case .
note that we will use the triangular relation @xmath917 essentially once , and that the number @xmath246 will essentially depend on the geometry of @xmath817 .
* step 1*. suppose in addition that @xmath864 is @xmath1090-nearly regular for some constant @xmath364 with @xmath869 are all @xmath103-close to the unit vector @xmath336 , and that @xmath873 are all @xmath358 .
we prove @xmath1073 in this basic case .
observe that in this case , for each @xmath1091 where @xmath1092 , @xmath1093 may be chosen as @xmath1094 , and the defining tripod can be chosen as @xmath379 , and the defining @xmath13-framed segment @xmath363 may be chosen as @xmath1095 .
then @xmath1096 equals @xmath1097_{r,\epsilon}\right ) -[\mathfrak{t}_{r , r+1}\mathfrak{a}_{r , r+1}]_{r,\epsilon}-[\mathfrak{t}_{r , r+1}\bar{\mathfrak{a}}_{r , r+1}]_{r,\epsilon } \right)\\ & = & \sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\mathfrak{a}_{r+1,r+2}]_{r,\epsilon } + \sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\mathfrak{a}_{r+2,r}]_{r,\epsilon } -\sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\bar{\mathfrak{a}}_{r , r+1}]_{r,\epsilon}\\ & = & \sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\mathfrak{a}_{r+1,r+2}]_{r,\epsilon } + \sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\mathfrak{a}_{r+2,r}]_{r,\epsilon } -\sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\bar{\mathfrak{a}}_{r , r+1}]_{r,\epsilon}\\ & = & 2\sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\mathfrak{a}_{r , r+1}]_{r,\epsilon}-0\\ & = & 2\sum_{r\in{{\mathbf z}}_3}[\mathfrak{t}_{r , r+1}\bar{\mathfrak{a}}_{r+1,r}]_{r,\epsilon}\\ & = & 0 .
\end{aligned}\ ] ] in the third equality , the last summation equals zero by rotation ( construction [ rotationconstruction ] ( 1 ) ) ; the first two summations are both equal to the summation of @xmath1098_{r,\epsilon}$ ] over @xmath1092 , since by swapping ( construction [ swapping ] ) , @xmath1099_{r,\epsilon}+ [ \mathfrak{t}_{12}\mathfrak{a}_{20}]_{r,\epsilon}+ [ \mathfrak{t}_{20}\mathfrak{a}_{01}]_{r,\epsilon}\\ & = & [ \mathfrak{t}_{01}\mathfrak{a}_{12}]_{r,\epsilon}+ [ \mathfrak{t}_{12}\mathfrak{a}_{01}]_{r,\epsilon}+ [ \mathfrak{t}_{20}\mathfrak{a}_{20}]_{r,\epsilon}\\ & = & [ \mathfrak{t}_{01}\mathfrak{a}_{01}]_{r,\epsilon}+ [ \mathfrak{t}_{12}\mathfrak{a}_{12}]_{r,\epsilon}+ [ \mathfrak{t}_{20}\mathfrak{a}_{20}]_{r,\epsilon } , \end{aligned}\ ] ] and similarly for @xmath1100_{r,\epsilon}+ [ \mathfrak{t}_{12}\mathfrak{a}_{01}]_{r,\epsilon}+ [ \mathfrak{t}_{20}\mathfrak{a}_{12}]_{r,\epsilon}$ ] .
the last equality follows from antirotation ( construction [ antirotationconstruction ] ( 1 ) ) .
this proves @xmath1073 in the basic case . *
step 2*. we prove a connecting step which is the following claim .
suppose that @xmath1101 and @xmath1102 are @xmath760-tame left - handed tripods satisfying the following : * the left - handed tripod @xmath724 is @xmath1103-nearly regular .
* for each @xmath377 , the chain @xmath1104 is @xmath699-consecutive and @xmath698-tame .
the terminal direction of @xmath1077 is @xmath103-close to @xmath844 .
the same holds for @xmath1105 . * for each @xmath377 , @xmath1106 is @xmath699-close to @xmath1107 , and @xmath1108 is @xmath699-close to @xmath1109 .
let @xmath866 be the angles guaranteed by lemma [ sharptripodlemma ] with respect to @xmath1101 , which , hence , works for @xmath1102 up to error of @xmath1110 .
let @xmath1111 be the @xmath103-sharp element associated to @xmath1112 , @xmath1113 , respectively .
assuming that @xmath1114 lie in @xmath1115 for all @xmath377 , we claim @xmath1116 to prove the claim , observe that it suffices to prove a simple case that @xmath1082 equals @xmath1117 except for one @xmath377
. then the claim follows by applying the simple case successively to each neighboring pair in the sequence of tripods @xmath1101 , @xmath1118 , @xmath1119 , @xmath1120 . without loss of generality
, we may assume that @xmath1121 except for @xmath746 being @xmath358 .
then @xmath1122 , and we must show @xmath1123 . observe further that for both @xmath1124 and @xmath1125 , we may choose the same defining right - handed tripod @xmath1126 and the same defining @xmath13-framed segment @xmath1127 .
similarly , for @xmath1128 and @xmath1129 we choose the same @xmath1130 and @xmath1131 for both @xmath1128 and @xmath1129 . by swapping the bigon pair @xmath1132 $ ] and @xmath1133 $ ] into @xmath1134 $ ] and @xmath1135 $ ] ( construction [ swapping ] ) , after rearrangement , we have @xmath1136_{r,\epsilon}+[\mathfrak{t}_{20}(\phi_1)\mathfrak{a}^{(2)}_{01}]_{r,\epsilon } \,=\,[\mathfrak{t}'_{01}(\phi_2)\mathfrak{a}^{(1)}_{01}]_{r,\epsilon}+[\mathfrak{t}'_{20}(\phi_1)\mathfrak{a}^{(2)}_{01}]_{r,\epsilon},\ ] ] where @xmath1137 and @xmath1138 ; similarly , @xmath1139_{r,\epsilon}+[\mathfrak{t}_{20}(\phi_1)\mathfrak{a}^{(2)}_{12}]_{r,\epsilon } \,=\,[\mathfrak{t}'_{01}(\phi_2)\mathfrak{a}^{(1)}_{12}]_{r,\epsilon } + [ \mathfrak{t}'_{20}(\phi_1)\mathfrak{a}^{(2)}_{12}]_{r,\epsilon};\ ] ] @xmath1140_{r,\epsilon}+[\mathfrak{t}_{20}(\phi_1)\mathfrak{a}^{(2)}_{20}]_{r,\epsilon } \,=\,[\mathfrak{t}'_{01}(\phi_2)\mathfrak{a}^{(1)}_{20}]_{r,\epsilon } + [ \mathfrak{t}'_{20}(\phi_1)\mathfrak{a}^{(2)}_{20}]_{r,\epsilon};\ ] ] and @xmath1141_{r,\epsilon}-[\mathfrak{t}_{20}(\phi_1)\mathfrak{b}^{(2)}]_{r,\epsilon } \,=\,-[\mathfrak{t}'_{01}(\phi_2)\mathfrak{b}^{(1)}]_{r,\epsilon}-[\mathfrak{t}'_{20}(\phi_1)\mathfrak{b}^{(2)}]_{r,\epsilon};\ ] ] @xmath1142_{r,\epsilon}-[\mathfrak{t}_{20}(\phi_1)\bar{\mathfrak{b}}^{(2)}]_{r,\epsilon } \,=\,-[\mathfrak{t}'_{01}(\phi_2)\bar{\mathfrak{b}}^{(1)}]_{r,\epsilon}-[\mathfrak{t}'_{20}(\phi_1)\bar{\mathfrak{b}}^{(2)}]_{r,\epsilon}.\ ] ] summing up the five equations above yields @xmath1123 .
this finishes the proof the claim of the connecting step . * step 3*. we finish the proof in the general case .
let @xmath864 and @xmath1143 be as before so that @xmath895 is associated to @xmath871 .
we may write @xmath1076 as concatenation of consecutive @xmath13-framed segments @xmath1077 , so that @xmath724 is @xmath1144-nearly regular . by the connection principle ( lemma [ connectionprinciple ] ) , we may interpolate a sequence of tripods @xmath1086 where @xmath1085 runs over @xmath1145 , such that @xmath1146 is the @xmath103-concatenation @xmath1147 , and @xmath1080 satisfies the following .
* for all @xmath377 and @xmath1148 , @xmath1080 have length at least @xmath1149 and phase @xmath699-close to @xmath358 .
* for @xmath377 and @xmath1084 , @xmath1080 and @xmath1083 have length @xmath699-close to each other , and terminal framings @xmath699-close to each other . * for @xmath377
, @xmath1150 equals @xmath1082 . * for @xmath377 , @xmath1081 have length @xmath699-close to each other , and terminal framings @xmath699-close to each other . for each @xmath1086 ,
let @xmath1151 be a triple of angles guaranteed by lemma [ sharptripodlemma ] , and let @xmath1152 be the @xmath103-sharp element associated to @xmath1153 .
it follows that @xmath1154 for @xmath1148 .
moreover , we may assume without loss of generality that @xmath1155 are all @xmath1143 respectively , and that @xmath1156 are all @xmath358 .
note that we may also require @xmath1087 to lie in @xmath1115 , for example , by constructing @xmath1080 as a @xmath103-consecutive @xmath356-tame concatenation @xmath1157 , for @xmath1158 , where @xmath1159 is carried by a segment of length @xmath996 with the initial direction @xmath844 .
the @xmath13-framed segment @xmath1159 can be constructed using simplex subdivision in dimension @xmath358 ( definition [ axiomsofconstructions ] ( 2 ) , cf .
subsubsection [ subsubsec - describingaconstruction ] ) , and @xmath1160 can be constructed by the connection principle ( lemma [ connectionprinciple ] ) .
therefore , step 1 implies that @xmath1161 and step 2 implies that @xmath1162 for @xmath1084 .
it follows that when @xmath1085 equals @xmath358 , we have @xmath1163 this completes the proof .
[ homomorphismextension ] the restriction of @xmath1009 to @xmath1008 extends uniquely to be a homomorphism @xmath1009 from @xmath814 to @xmath786 .
this follows from the facts that @xmath967 contains a generating set of @xmath823 , and that @xmath970 contains @xmath969 which further contains a triangular generating set of @xmath823 ( lemma [ setupdata ] ( 2)(4 ) ) .
in fact , @xmath1164 contains a generating set of @xmath814 which is the preimage of the generating set inside @xmath967 , so the extension is unique ; @xmath1165 contains a triangular generating set of @xmath814 which is the preimage of the triangular generating set inside @xmath969 , so the extension exists by lemma [ triangularrelation ] . by lemma [ homomorphismextension ] , we conclude that the restriction of @xmath1009 to @xmath1164 extends uniquely to be a homomorphism @xmath1166
descending to the abelianization , we denote the induced homomorphism as @xmath1167 the reader should compare the definition of @xmath1009 with the definition of the operator @xmath1168 in ( * ? ? ?
* subsection 7.1 ) .
since @xmath1168 was defined with a coefficient @xmath68 , it does not work in integral coefficients .
in fact , the argument of good correction theorem ( * ? ? ? * theorem 3.2 ) essentially implies that @xmath1168 induces an isomorphism @xmath1169 for any closed oriented hyperbolic surface @xmath792 , which is the inverse of the homomorphism @xmath1170 given by @xmath795_{r,\epsilon})=[\gamma]$ ] . by introducing a right - handed tripod @xmath379 in addition to segment @xmath363 , we may get rid of the coefficient @xmath68 and write down an expression of @xmath1009 with integral coefficients
. however , the ambiguity of the choice of @xmath1020 makes it necessary to pass to @xmath12 rather than to stay in @xmath4 .
it remains to verify that @xmath1171 is the inverse of @xmath11 .
we complete this by proving that @xmath1171 is the pre - inverse of @xmath11 ( lemma [ preinverse ] ) , and that @xmath1009 is onto ( lemma [ spanning ] ) .
[ preinverse ] for any element @xmath846 in @xmath1164 , @xmath1172\ ] ] in @xmath27 .
hence the composition @xmath1173 is the identity transformation of @xmath27 .
it suffices to prove for any @xmath1174 with respect to @xmath1016 .
let @xmath1175 in @xmath814 be the @xmath103-sharp element associated to @xmath621 for @xmath377 , and @xmath1176 be associated to @xmath363 ( definition [ sharpelement ] ) .
hence @xmath1177 is associated to @xmath1048 .
it is clear from the construction of @xmath11 that the the image of @xmath1178_{r,\epsilon}$ ] under @xmath11 is equal to @xmath1179 $ ] .
similarly , @xmath1180_{r,\epsilon})=[\hat{g}\hat{b}]$ ] , and @xmath1181_{r,\epsilon})=[\hat{g}\hat{c}\hat{b}^{-1}]$ ] . then in @xmath27 , @xmath1182 equals @xmath1183+[\hat{g}\hat{a}_{12}]+[\hat{g}\hat{a}_{20 } ] -[\hat{g}\hat{b}]-[\hat{g}\hat{c}\hat{b}^{-1}]\\ & = & [ \hat{g}]+[\hat{a}_{01}]+[\hat{a}_{12}]+ [ \hat{a}_{20}]-[\hat{c}]\\ & = & [ \hat{g}]+[\hat{a}_{01}\hat{a}_{12}\hat{a}_{20}\hat{c}]\\ & = & [ \hat{g } ] , \end{aligned}\ ] ] where @xmath1184=0 $ ] because @xmath379 is right - handed , ( remark [ sharptripodlemmaremark ] ) .
this shows @xmath1185 $ ] .
the ` hence ' part is because @xmath1164 generates @xmath814 ( lemma [ setupdata ] ( 2 ) ) .
[ spanning ] for any curve @xmath112 , the @xmath18-panted cobordism class @xmath796_{r,\epsilon}$ ] is equal to an integral linear combination of elements in the image of @xmath1164 under @xmath1009 .
hence the homomorphism @xmath1171 surjects @xmath786 .
the ` hence ' part follows immediately from the main statement since @xmath1164 generates @xmath814 ( lemma [ setupdata ] ( 2 ) ) .
it remains to prove the main statement . by lemma [ finercurves ]
, it suffices to assume @xmath1186 .
let @xmath1187 denote the subset of @xmath823 consisting of elements @xmath1188 such that the length of @xmath1188 is at most @xmath1189 , and that the initial and terminal directions of @xmath1188 is @xmath699-close to @xmath1190 and @xmath1191 , respectively .
note that @xmath1192 is contained in @xmath1193 .
let @xmath1194 denote the preimage of @xmath1187 in @xmath814 . * step 1*. we find @xmath1195 , such that @xmath1196_{r,\epsilon}=\psi_1(\hat{x}_-)+\psi_1(\hat{x}_+).\ ] ] since @xmath1186 , we may bisect @xmath36 into an @xmath1197-nearly regular bigon @xmath1198 $ ] by interpolating a pair of antipodal points with suitably chosen normal vectors . by lemmas [ setupdata ] ( 3 ) and [ sharpck ] , we can lift @xmath971 to be a @xmath103-sharp element @xmath1199 with an associated @xmath13-framed segment @xmath1200 . by the connection principle ( lemma [ connectionprinciple ] )
there are oriented @xmath13-framed segments @xmath1201 from @xmath1202 to @xmath800 , satisfying the following : * the length and phase of @xmath1201 are @xmath103-close to @xmath1203 and @xmath358 respectively .
the initial direction @xmath1201 is @xmath103-close to @xmath1204 , and the initial framing of @xmath1201 is @xmath103-close to @xmath1205 .
the terminal direction of @xmath1201 is @xmath103-close to @xmath1206 , and the terminal framing of @xmath1201 is @xmath103-close to @xmath1207 . see figure [ figonto1 ] .
let @xmath1208 , where @xmath1209 , be the reduced concatenation .
then @xmath1210 and @xmath1211 are associated to @xmath699-sharp elements @xmath1212 and @xmath1213 in @xmath1214 .
choose a right - handed tripod @xmath379 and a @xmath13-framed segment @xmath363 for defining @xmath1215 , and choose @xmath383 with the indices understood in @xmath384 and @xmath1062 for defining @xmath1216 .
note that for @xmath377 , @xmath1217_{r,\epsilon}+ [ \mathfrak{x}_-^*\mathfrak{a}^*_{-i ,-
i-1}]_{r,\epsilon}\\ & = & [ \mathfrak{x}_+\mathfrak{a}_{i , i+1}]_{r,\epsilon}+ [ \mathfrak{x}_-\overline{\mathfrak{a}_{i , i+1}}]_{r,\epsilon}\\ & = & [ \mathfrak{s}_+(\mathfrak{w}_+\mathfrak{a}_{i , i+1}\bar{\mathfrak{w}}_-)]_{r,\epsilon}+ [ \mathfrak{s}_-\overline{(\mathfrak{w}_+\mathfrak{a}_{i , i+1}\bar{\mathfrak{w}}_-)}]_{r,\epsilon}\\ & = & [ \gamma]_{r,\epsilon } , \end{aligned}\ ] ] where the last equality follows from splitting @xmath1218 ( construction [ splitting ] ) .
thus @xmath1219_{r,\epsilon}+ [ \mathfrak{x}_-^*\mathfrak{a}^*_{02}]_{r,\epsilon}= [ \gamma]_{r,\epsilon};\ ] ] @xmath1220_{r,\epsilon}+ [ \mathfrak{x}_-^*\mathfrak{a}^*_{21}]_{r,\epsilon}= [ \gamma]_{r,\epsilon};\ ] ] @xmath1221_{r,\epsilon}+ [ \mathfrak{x}_-^*\mathfrak{a}^*_{10}]_{r,\epsilon}= [ \gamma]_{r,\epsilon};\ ] ] similarly , @xmath1222_{r,\epsilon } -[\mathfrak{x}_-^*\bar{\mathfrak{b}}^*]_{r,\epsilon}= -[\gamma]_{r,\epsilon};\ ] ] @xmath1223_{r,\epsilon } -[\mathfrak{x}_-^*\mathfrak{b}^*]_{r,\epsilon}= -[\gamma]_{r,\epsilon}.\ ] ] summing up the five equations above shows that @xmath1224_{r,\epsilon}.$ ] * step 2*. for any @xmath1225 of length at least @xmath966 we find @xmath1226 , such that @xmath1227 and that for the images @xmath1228 of @xmath1229 respectively , both @xmath1230 have length less than @xmath1231 .
in fact , we may write the pointed geodesic loop as the concatenation of to geodesic segments @xmath1232 joint at the midpoint of @xmath1233 . by the connection principle ( lemma [ connectionprinciple ] ) , applied to the unframed case simply by ignoring the framings
, there is a path @xmath1234 from the midpoint of @xmath1233 to @xmath800 , satisfying the following : the length of @xmath1234 is @xmath103-close to @xmath1235 ; the initial direction of @xmath1234 is @xmath103-closely perpendicular to @xmath1233 ; and the terminal direction of @xmath1234 is @xmath103-close to @xmath844 .
let @xmath1236 and @xmath1237 in @xmath823 be @xmath1238 and @xmath1239 , respectively . since @xmath1240 in @xmath823 , we may choose lifts @xmath1241 of @xmath1230 in @xmath814 so that @xmath1242 .
it is straightforward to see that @xmath1241 are as desired . *
step 3*. we complete the proof of the main statement . as mentioned above , we may assume that @xmath1186 .
by step 1 , @xmath796_{r,\epsilon}$ ] can be written as a sum of elements in the image of @xmath1194 under @xmath1009 . by iterately applying step 2 , any element in @xmath1243 can be replaced with a sum of elements of the form @xmath1244 where the image of @xmath1245 in @xmath823 has length at most @xmath966 . in particular , @xmath1246 since @xmath965 is assumed to be at least @xmath966 ( lemma [ setupdata ] ( 2 ) ) .
thus @xmath796_{r,\epsilon}$ ] is equal to an integral linear combination of elements in the image of @xmath1247 under @xmath1009 .
this completes the proof . in summary ,
given any oriented closed hyperbolic @xmath0-manifold @xmath4 , and any universally small positive constant @xmath14 , with the positive constant @xmath791 guaranteed by lemma [ setupdata ] ( 4 ) , suppose that @xmath15 is a positive constant greater than @xmath791 .
then the homomorphism @xmath1248 constructed in subsection [ subsec - phi ] is a canonically defined isomorphism ( lemma [ welldefinedphi ] and subsection [ subsec - theinverseofphi ] ) . by lemma [ canonicallydefinedphi ] , for all @xmath24_{r,\epsilon}\in{{\mathbf{\omega}}}_{r,\epsilon}(m)$ ] , the image of @xmath26_{r,\epsilon})$ ] under the bundle projection is the homology class @xmath24\in h_1(m;{{\mathbf z}})$ ] .
this completes the proof of theorem [ theorempantedcobordism ] .
in this section , we show that second homology classes of an oriented closed hyperbolic @xmath0-manifold @xmath4 can be represented by @xmath18-panted surfaces , as precisely stated in theorem [ secondhomologyclass ] .
this will imply the absolute case of theorem [ homologyviapants ] ( 1 ) , namely , when the collection of curves @xmath125 is empty ( subsection [ subsec - proofofhomologyviapants ] ) . roughly speaking ,
theorem [ secondhomologyclass ] follows from inspecting the homology classes of the @xmath18-panted surfaces constructed in the proof of theorem [ theorempantedcobordism ] , so our argument and notations will heavily rely on section [ sec - pantedcobordismgroup ] .
in particular , throughout this section , it will suffice to assume @xmath14 to be a universally small positive constant and @xmath15 to be a positive constant greater than the constant @xmath791 as guaranteed by lemma [ setupdata ] ( 4 ) .
[ secondhomologyclass ] let @xmath4 be an oriented closed hyperbolic @xmath0-manifold . for any small positive @xmath14 and sufficiently large positive @xmath15 depending on @xmath4 and @xmath14 , the following holds . for any homology class @xmath1249 , there exists an ( oriented ) closed @xmath18-panted subsurface @xmath97 so that @xmath1250 $ ] equals @xmath9 .
there is a canonical free resolution of the integral module @xmath786 given by @xmath1251 where @xmath246 denotes the kernel of of the boundary homomorphism .
there is also a natural homomorphism @xmath1252 since the natural homomorphism @xmath1253 uniquely lifts to @xmath1254 restricted to @xmath246 .
therefore , theorem [ secondhomologyclass ] asserts that @xmath246 surjects @xmath1254 . in this sense , it reveals certain finer structure of @xmath18-panted cobordisms in addition to theorem [ theorempantedcobordism ] .
the key idea of the proof of theorem [ secondhomologyclass ] is to apply a process called _ homologous substitution_. to illustrate how it works , suppose that @xmath792 is a connected oriented closed surface and that @xmath1255 is a map so that @xmath1256 $ ] equals @xmath9 . to replace @xmath1255 with a homologous @xmath18-panted subsurface @xmath97 , we endow @xmath792 with a triangulation with a single vertex @xmath800 , and assume that @xmath1257 has been homotoped so that @xmath800 is sent to a chosen basepoint of @xmath4 , and that the @xmath2-simplices of @xmath792 are long geodesic segments , and that the @xmath307-simplices of @xmath792 are totally geodesic in @xmath4 .
following the construction of @xmath805 in subsection [ subsec - theinverseofphi ] , we may replace any ( oriented ) @xmath2-simplex @xmath577 with an @xmath18-multicurve @xmath1258 as in the definition of @xmath805 , without passing to the @xmath18-panted cobordism class . by convention
, we define @xmath1259 to be @xmath1260 .
moreover , we may replace any @xmath307-simplex @xmath570 with an @xmath18-panted surface @xmath1261 , so that if @xmath1262 is a cycle @xmath1263 , @xmath1261 will be bounded by @xmath1264 .
the @xmath18-panted surface @xmath1261 can be obtained explicitly by the constructions in lemmas [ welldefinition ] , [ triangularrelation ] .
thus the @xmath18-panted surface @xmath3 can be obtained by naturally gluing the @xmath18-panted surfaces @xmath1261 along the @xmath18-multicurves @xmath1258 on their boundary , according to the triangulation structure of @xmath792 .
intuitively , it should follow from the spine principle ( lemma [ spineprinciple ] ) that there are natural isomorphisms @xmath1265 . then a mayer vietoris argument
will imply that there is a natural isomorphism @xmath1266 that commutes with the homomorphisms @xmath1267 and @xmath1268 .
in other words , the @xmath18-panted surface @xmath3 is homologous to @xmath792 in @xmath4 , and hence represents @xmath9 as desired . in practice , it is actually more convenient not to specify the homology class @xmath1269 . instead , we consider a triangular presentation complex @xmath1270 ( cf .
subsection [ subsec - triangularpresentationcomplexes ] ) of @xmath823 rather than the triangulated @xmath1271 .
then a similar process of homologous subsitution will yield an @xmath18-panted complex @xmath1272 ( a @xmath307-complex obtained by gluing @xmath18-panted surfaces along @xmath18-multicurves on the boundary , cf .
subsection [ subsec - pantedcomplexes ] ) .
in general , there will be a natural epimorphism @xmath1273 that commutes with @xmath1274 and @xmath1275 .
because of the easy observation that @xmath1276 is onto , @xmath1277 will also be onto .
in other words , any homology class @xmath1269 comes from some @xmath1278 , so it is represented by some @xmath18-panted surface obtained by a composition @xmath1279 .
the rest of this section is devoted to the proof of theorem [ secondhomologyclass ] . in subsections [ subsec - triangularpresentationcomplexes ] , [ subsec - pantedcomplexes ]
, we introduce some notations that we will adopt , namely , triangular presentation complexes and @xmath18-panted complexes ; subsection [ subsec - homologoussubstitution ] is the homologous substitution argument , which is core of the proof ; subsection [ subsec - proofoftheoremsecondhomologyclass ] completes the proof of theorem [ secondhomologyclass ] by a brief summary .
let @xmath4 be an oriented closed @xmath0-manifold .
the bundle of special orthonormal frames of @xmath4 will be denoted as @xmath12 .
fix an orthonormal frame @xmath943 at a fixed basepoint @xmath800 of @xmath4 , so the special orthonormal bundle @xmath12 has a preferred basepoint @xmath835 .
recall that a _ presentation _ of a group @xmath977 is a pair @xmath976 , where @xmath1280 is a set of independent letters and @xmath979 is a set of words in @xmath1281 and @xmath1282 for @xmath1283 , such that the canonical quotient @xmath1284 is isomorphic to @xmath977 . for a presentation of @xmath977 , there is a naturally associated cw @xmath307-complex @xmath410 with a single vertex @xmath800 as the basepoint , called the _ presentation complex _ , such that the @xmath2-cells are in correspondence with the generators , and the @xmath307-cells are in correspondence with the relators .
hence the fundamental group @xmath1285 is naturally isomorphic to @xmath977 .
in fact , if @xmath1286 is a pointed topological space , any homomorphism @xmath1287 can be realized by a map @xmath1288 .
combinatorially , any @xmath307-cell of @xmath410 has polygonal boundary , and the number of edges is equal to the word length of the corresponding relator .
a presentation is said to be _ finite _ if the sets of generators and relators are finite .
[ homologyonto ] if @xmath1289 is a presentation complex of @xmath1290 , then the presentation map @xmath1291 which induces the natural isomorphism on @xmath1 induces an epimorphism @xmath1292 this follows from the fact that @xmath4 is aspherical .
in fact , we may obtain an eilenberg maclane space @xmath1293 by attaching to @xmath410 cells of dimension greater than @xmath307 , and we may extend @xmath1257 to obtain a homotopy equivalence @xmath1294 .
this implies the surjectivity of the induced homomorphism @xmath1274 on the second homology .
[ lifttosom ] if @xmath1289 is a presentation complex of @xmath1290 , then the presentation map @xmath1291 lifts to a map @xmath1295 this follows from the fact that @xmath799 is the splitting extension of @xmath30 by @xmath1296 .
moreover , the homotopy classes of lifts of @xmath1257 are determined by the splittings , which are in bijection with @xmath1297 since @xmath1298 is trivial .
a presentation is said to be _ triangular _ , if the word length of the relators are at most @xmath0 .
the complex @xmath1289 associated to a triangular finite presentation is compact with only monogonal , bigonal , or triangular @xmath307-cells , and we say that @xmath1289 is _ triangular _ and _ finite_. note that any finite presentation gives rise to a triangular finite presentation , by adding a maximal collection of mutually non - intersecting diagonals to subdivide the @xmath307-cells of @xmath410 .
furthermore , if we assume that @xmath410 minimizes the number of generators and the number of relators in the lexicographical order among triangular finite presentations of @xmath977 , it is easy to see that @xmath410 will not contain any monogonal @xmath307-cells , and that any bigonal @xmath307-cells of @xmath410 will be attached to @xmath2-cells representing elements of order @xmath307 . for our application , @xmath823 has no elements of order @xmath307 , and in fact , it is torsion free .
it suffices to consider a specific triangular finite presentation of @xmath823 associated to the triangular generating set guaranteed by lemma [ setupdata ] ( 2 ) .
in particular , the associated triangular finite presentation complex @xmath1289 is a @xmath544-complex , and we denote the presentation map as @xmath1299 which is unique up to homotopy relative to the basepoint . by a ( topological )
_ panted complex _
@xmath1300 we mean a compact cw space obtained from a finite disjoint union of circles by attaching finitely many disjoint pairs of pants via homeomorphisms from cuffs .
recording the pants decomposition of @xmath1300 as part of data , we will refer to the defining circles and pairs of pants as the _ structure curves _ and _ structure pants _ of @xmath1300 .
when a panted complex @xmath1300 is immersed into a closed hyperbolic @xmath0-manifold @xmath4 , we will say that the immersion is _
@xmath18-panted _ , if the restriction to each pair of pants is @xmath18-nearly regular up to homotopy , or ambiguously , we will say that @xmath1300 is an _
@xmath18-panted complex_. note that any connected @xmath18-panted surface is naturally an @xmath18-panted complex .
a _ panted map _ between two panted complexes is a map that sends each structural pair of pants homeomorphically onto a structural pair of pants of the target .
[ pantedrepresentative ] for any nontrivial element @xmath1301 , there is a closed oriented panted surface @xmath3 and a panted map
@xmath1302 , such that the fundamental class of @xmath3 in @xmath1303 is sent to @xmath9 .
let @xmath1304 be the disjoint union of structure pants of @xmath1300 , and @xmath156 be the disjoint union of structure curves of @xmath1300 . the long exact sequence of homology yields @xmath1305 & h_2(\mathcal{k};{{\mathbf z } } ) \ar[r ] & h_2(\mathcal{k},\mathcal{c};{{\mathbf z } } ) \ar[r]^{\partial _ * } & h_1(\mathcal{c})\\ & & h_2(\mathcal{p},\partial\mathcal{p};{{\mathbf z } } ) \ar[r]^{\partial _ * } \ar[u]^{\cong } & h_1(\partial\mathcal{p};{{\mathbf z } } ) \ar[u]^{\chi } } \ ] ] where @xmath1306 means the homomorphism induced by the _
characteristic map _ of the panted complex @xmath1300 that identifies the cuffs of structure pants with the structure curves .
now any element @xmath1307 can be identified as an element @xmath1308 of @xmath1309 in the kernel of @xmath1310 .
since @xmath1309 has a basis formed by the fundamental classes @xmath1311 $ ] of the components @xmath1312 , the element @xmath1308 naturally yields a collection of copies of structure pants with suitable orientations , and @xmath1313)=0 $ ] implies that these copies of pants can be glued up along cuffs , resulting in a closed oriented panted surface @xmath3 .
the naturally induced panted map @xmath1314 is as desired .
let @xmath1289 be the cw complex associated to the triangular generating set guaranteed by lemma [ setupdata ] ( 2 ) , ( cf .
subsection [ subsec - triangularpresentationcomplexes ] ) , and @xmath1270 be the basepoint - preserving map associated to the presentation .
let @xmath960 be the conjugation provided by lemma [ setupdata ] ( 3 ) .
the presentation conjugated by @xmath960 induces an isomorphism @xmath1315 .
we denote the corresponding basepoint - preserving map as @xmath1316 it can be topologically obtained from @xmath1257 by pushing the image of @xmath1317 along the loop corresponding to @xmath1318 .
[ homologoussubstitution ] there exists a cw complex @xmath1319 obtained from @xmath410 by attaching @xmath2-cells , and a compact panted complex @xmath1300 , and there exist maps @xmath1320 satisfying the following .
* the map @xmath1321 restricts to be the basepoint - preserving map @xmath1322 * the map @xmath1323 induces an epimorphism @xmath1324 * the composition @xmath1325 yields an @xmath18-panted complex .
we prove lemma [ homologoussubstitution ] in the rest of this subsection .
we construct @xmath1300 , @xmath1319 and maps @xmath1326 following the construction of @xmath822 in subsection [ subsec - theinverseofphi ] .
fix a lift of @xmath1327 into @xmath12 , denoted as @xmath1328 ( lemma [ lifttosom ] ) .
note that by our assumption , @xmath410 is a @xmath544-complex , so each @xmath2-cell of @xmath410 can be conveniently denoted by the element @xmath1329 that it represents , and each @xmath307-cell of @xmath410 can be conveniently denoted by a triangular relation @xmath1330 of three generators or their inverses . as @xmath410 and @xmath1327 are provided from lemma [ setupdata ] , for each @xmath2-cell @xmath843 of @xmath410 , @xmath1331 is a @xmath103-sharp element .
thus there is an associated oriented @xmath13-framed segment @xmath1332 , or simply written as @xmath1013 in order to be consistent with the notation in subsection [ subsec - theinverseofphi ] . *
step 1*. for each @xmath2-cell @xmath843 of @xmath410 , fix a right - handed nearly regular tripod @xmath379 and an oriented @xmath13-framed segment @xmath363 as in the definition of @xmath805 .
we construct a @xmath307-dimensional @xmath544-complex @xmath1333 and a multcurve @xmath1334 , together with maps @xmath1335 such that the composition yields an @xmath18-multicurve @xmath1336 , which represents the @xmath18-panted cobordism class @xmath1337 , where @xmath842 is the element @xmath1338 .
the construction is as follows .
take the subcomplex @xmath1339 of @xmath410 ; attach @xmath2-cells @xmath1340 for @xmath377 and @xmath1341 , corresponding to the carrier segments of @xmath621 and @xmath363 , respectively ; attach a simplicial @xmath307-cell with boundary the cycle @xmath1342 .
the resulting @xmath307-complex will be denoted as @xmath1333 .
there is a naturally induced map @xmath1343 extending @xmath1344 .
let @xmath1345 \sqcup[\mathfrak{s}_{\hat{g}}\mathfrak{a}_{12 } ] \sqcup[\mathfrak{s}_{\hat{g}}\mathfrak{a}_{20 } ] \sqcup\overline{[\mathfrak{s}_{\hat{g}}\mathfrak{b } ] } \sqcup\overline{[\mathfrak{s}_{\hat{g}}\bar{\mathfrak{b}}]}$ ] be the multicurve of five components , which are the reduced cyclic concatenations of the defining oriented @xmath13-framed segments .
the natural immersion @xmath1336 can be homotoped to factor through @xmath1333 via a composition @xmath1346 in the explicit way as indicated by the construction .
we define @xmath1347 to be @xmath1348 , and @xmath1349 for @xmath1333 .
note that @xmath1347 is the representative of @xmath1350 corresponding to the defining right - handed tripod @xmath1351 and @xmath1352 since the @xmath103-sharp element @xmath860 is associated to @xmath1353 .
thus there are also maps @xmath1354 and the composition is homotopic to an @xmath18-multicurve @xmath1355 , which represents @xmath1356 . *
step 2*. for each simplicial @xmath307-cell @xmath570 of @xmath410 corresponding to a triangular relation @xmath1330 , write @xmath1357 where the wedge is over the basepoint @xmath800 , and @xmath1358 by identifying the copies of @xmath1359 in @xmath1360 and @xmath570 , and @xmath1361 we construct a @xmath307-complex @xmath1362 obtained from @xmath1363 by attaching @xmath2-cells and a panted surface @xmath1261 bounded by @xmath1364 , together with maps @xmath1365 such that the composition is homotopic to an @xmath18-panted surface @xmath1366 ; moreover , the following diagram of maps commutes : @xmath1367 & y(\sigma)\\ l(\partial\sigma ) \ar[r ] \ar[u]^{\cup } & x(\partial\sigma ) \ar[u]^{\cup}. } \ ] ] the construction is as follows .
since @xmath1368_{r,\epsilon}+[l(g_1)]_{r,\epsilon}+[l(g_2)]_{r,\epsilon}=0 $ ] in @xmath786 ( lemma [ triangularrelation ] ) , there exists an @xmath18-panted surface @xmath1261 with @xmath1369 equal to @xmath1364 .
in fact , the construction of @xmath1261 relies on lemmas [ welldefinition ] , [ triangularrelation ] . checking the constructions there
, we see that @xmath1261 can be constructed based only on the @xmath544-complex @xmath1363 and the map @xmath1370 induced from the constructed maps @xmath1371 and the given map @xmath1372 .
formally , we regard @xmath1363 as a partially-@xmath544 space over @xmath4 where the partially-@xmath544 structure is given by the entire @xmath544-complex @xmath1363 , and the map @xmath1370 is as described above ( definition [ partiallydeltaspace ] ) then the construction of @xmath1261 implies that there is a partially-@xmath544 space @xmath559 over @xmath4 which is an extension of @xmath1363 , such that there is a commutative diagram of maps : @xmath1367 & x ' \ar[rd ] \\
l(\partial\sigma ) \ar[r ] \ar[u]^{\cup } & x(\partial\sigma ) \ar[u]^{\cup } \ar[r]&m . } \ ] ] by the spine principle ( lemma [ spineprinciple ] ) , @xmath559 is @xmath2-spined over @xmath1363 , so we may replace @xmath559 with a cw complex @xmath1362 , which is obtained from @xmath1363 by attaching @xmath2-cells
. then @xmath1261 , @xmath1362 and the involved maps from the diagram are as desired .
* step 3*. now we may naturally attach the disjoint union of all @xmath1261 to the disjoint union of all @xmath1334 according to the attaching maps of @xmath410 .
the result is a panted complex @xmath1300 .
similarly , we may attach the disjoint union of @xmath1362 to the disjoint union of all @xmath1333 by naturally identifying the copies of @xmath1333 ( possibly marked by @xmath1373 ) .
the result is a cw @xmath307-complex @xmath1374 containing the subcomplex @xmath410 .
moreover , there are naturally induced maps @xmath1375 the composition is homotopic to an @xmath18-panted complex @xmath1376 ; the restriction of @xmath1377 to @xmath410 is @xmath1327 .
it is clear from the construction that @xmath1374 deformation retracts relative to @xmath410 to a cw subspace @xmath1378 which can be obtained from @xmath410 by attaching @xmath2-cells
. therefore , replacing @xmath1374 with @xmath1319 , we obtain maps @xmath1379 to verify that @xmath1300 , @xmath1319 and the maps @xmath1326 above are as desired , it suffices to prove that @xmath1323 is surjective on the second homology , as the other listed properties are obviously satisfied . by the construction , we may equivalently prove with @xmath1374 instead of @xmath1319 .
write @xmath156 for the disjoint union of all @xmath1334 , and @xmath1380 for the wedge of all @xmath1333 over @xmath800 .
there is a commutative diagram of homomorphisms @xmath1381 & h_2(\mathcal{k};{{\mathbf z}})\ar[r ] \ar[d]^{\phi } & h_2(\mathcal{k},\mathcal{c};{{\mathbf z } } ) \ar[r]^{\partial _ * } \ar[d]^{\phi '' } & h_1(\mathcal{c};{{\mathbf z } } ) \ar[d]^{\phi'}\\ 0 \ar[r ] & h_2(y(k);{{\mathbf z}})\ar[r ] & h_2(y(k),x(k^{(1)});{{\mathbf z } } ) \ar[r]^{\partial _ * } & h_1(x(k^{(1)});{{\mathbf z } } ) \\ 0 \ar[r ] & h_2(k;{{\mathbf z}})\ar[r ] \ar[u]_{\iota}^\cong & h_2(k , k^{(1)};{{\mathbf z } } ) \ar[r]^{\partial _ * } \ar[u]_{\iota''}^\cong & h_1(k^{(1)};{{\mathbf z } } ) \ar[u]_{\iota'}^\cup } \ ] ] where the rows are part of the long exact sequences of homology , and the homomorphisms @xmath1382 are induced from the map @xmath1383 of our construction , and the homomorphisms @xmath1384 are induced from the natural inclusion @xmath1385 .
write the quotient map defining @xmath1300 as @xmath1386 define a homomorphism @xmath1387 by assigning @xmath1388)\,=\,q_*[f(\sigma)]\ ] ] where @xmath1389\in h_2(f(\sigma),l(\partial\sigma);{{\mathbf z}})$ ] is the fundamental class . define a homomorphism @xmath1390 by assigning @xmath1391)\,=\,q_*[l(g)]\ ] ] where @xmath1392\in h_1(l(g);{{\mathbf z}})$ ] is the fundamental class .
there is a commutative diagram @xmath1393^{\partial _ * } & h_1(\mathcal{c};{{\mathbf z } } ) \\
h_2(k , k^{(1)};{{\mathbf z } } ) \ar[r]^{\partial _ * } \ar[u]^{\psi '' } & h_1(k^{(1)};{{\mathbf z } } ) \ar[u]^{\psi'}. } \ ] ] [ phipsi ] with the notations above , @xmath1394 and @xmath1395 .
it is straightforward to check that @xmath1396))=\iota'([g])$ ] for any @xmath2-cell @xmath843 of @xmath410 , by the construction of @xmath1334 .
since @xmath1397 is freely generated by @xmath1398 $ ] where @xmath843 runs over all @xmath2-cells of @xmath410 , we see that @xmath1394 .
we claim that for any @xmath307-cell @xmath570 of @xmath410 , @xmath1399))=\iota''([\sigma])$ ] .
in fact , applying the discussion to the special case when @xmath410 consists of a single @xmath307-simplex @xmath570 together with the @xmath2-skeleton @xmath1262 , we obtain the commutative diagrams @xmath1381 & h_2(f(\sigma),l(\partial\sigma);{{\mathbf z } } ) \ar[r]^{\partial _ * } \ar[d]^{\phi''_\sigma } & h_1(l(\partial\sigma);{{\mathbf z } } ) \ar[d]^{\phi'_\sigma}\\ 0 \ar[r ] & h_2(y(\sigma),x(\partial\sigma);{{\mathbf z } } ) \ar[r]^{\partial _ * } & h_1(x(\partial\sigma);{{\mathbf z } } ) \\ 0 \ar[r ] & h_2(\sigma,\partial\sigma;{{\mathbf z } } ) \ar[r]^{\partial _ * } \ar[u]_{\iota''_\sigma}^\cong & h_1(\partial\sigma;{{\mathbf z } } ) \ar[u]_{\iota'_\sigma}^\cup } \ ] ] where the rows are exact sequences , and @xmath1400^{\partial _ * } & h_1(l(\partial\sigma);{{\mathbf z } } ) \\
h_2(\sigma,\partial\sigma;{{\mathbf z } } ) \ar[r]^{\partial _ * } \ar[u]^{\psi''_\sigma } & h_1(\partial\sigma;{{\mathbf z } } )
\ar[u]^{\psi'_\sigma}. } \ ] ] because @xmath1401))\,=\,\phi'_\sigma\circ\psi'_\sigma(\partial_*([\sigma ] ) ) \,=\,\iota'_\sigma(\partial_*([\sigma]))=\partial_*(\iota''_\sigma[\sigma]),\ ] ] the injectivity of @xmath1402 in this case implies that @xmath1403)\,=\,\iota''_\sigma([\sigma]).\ ] ] by naturality of mayer
vietoris sequences , it follows that @xmath1404)\,=\,\iota''([\sigma]),\ ] ] as claimed .
since @xmath1405 is freely generated by @xmath1406 $ ] where @xmath570 runs over all @xmath307-cells of @xmath410 , we conclude that @xmath1395 .
now an easy diagram chase will show the surjectivity of @xmath1407 in fact , we may identify @xmath1408 and @xmath1409 as kernels of @xmath1402 in @xmath1410 and @xmath1411 , respectively .
if @xmath1412 vanishes under @xmath1402 , @xmath1413 in @xmath1410 also vanishes under @xmath1402 , using the injectivity of @xmath1414 .
moreover , @xmath1415 by lemma [ phipsi ] .
this implies that @xmath1416 is surjective between the kernels of @xmath1402 , or in other words , @xmath154 is surjective .
this completes the verification , and hence completes the proof of lemma [ homologoussubstitution ] .
we summarize the proof of theorem [ secondhomologyclass ] as follows .
let @xmath4 be a closed oriented hyperbolic @xmath0-manifold .
by lemma [ homologoussubstitution ] , there exists a cw complex @xmath1319 obtained from a presentation complex @xmath410 of @xmath823 by attaching @xmath2-cells , and a compact panted complex @xmath1300 , together with maps @xmath1417 satisfying the listed properties .
in particular , the composed map @xmath1325 is an @xmath18-panted complex in @xmath4 .
it is surjective on the second homology by lemmas [ homologoussubstitution ] and [ homologyonto ] . in other words ,
any element @xmath9 of @xmath1254 can be lifted to be an element @xmath1418 of @xmath1408 .
moreover , @xmath1418 can be represented by a panted surface via a panted map @xmath1302 ( lemma [ pantedrepresentative ] ) , which yields an @xmath18-panted surface via the map @xmath1325 .
therefore , the @xmath18-panted surface @xmath941 is an representative of @xmath9 as desired .
this completes the proof of theorem [ secondhomologyclass ] .
in this section , we show that in a closed hyperbolic @xmath0-manifold @xmath4 , the collection of @xmath18-curves @xmath153 are _ @xmath18-panted connected _ in the sense of the following proposition [ pantedconnectedness ] .
this will be applied to show that an ubiquitous measure of @xmath18-pants @xmath98 is irreducible in the proof of theorem [ homologyviapants ] ( subsection [ subsec - proofofhomologyviapants ] ) .
[ pantedconnectedness ] let @xmath4 be a closed hyperbolic @xmath0-manifold .
for any universally small positive @xmath14 and any sufficiently large positive @xmath15 depending on @xmath4 and @xmath14 , the following holds . for any two curves
@xmath1419 , there exists a connected @xmath18-panted subsurface @xmath97 , and @xmath625 contains two components homotopic to @xmath1420 and @xmath1421 , respectively .
the proof of proposition [ pantedconnectedness ] follows from an easy construction using the connection principle ( lemma [ connectionprinciple ] ) .
however , we leave it as a separate section as its statement has certain independent geometric interest .
we say that two curves @xmath1419 are _ @xmath18-panted connected _ if there exists a connected @xmath18-panted subsurface @xmath97 as in the conclusion of the proposition .
this defines an equivalence relation between curves in @xmath153 .
suppose that @xmath14 universally small and @xmath15 is sufficiently large , for instance , as guaranteed by lemma [ setupdata ] . by splitting ( construction [ splitting ] ) , every curve in @xmath153
is @xmath18-panted connected to a curve in @xmath1422 .
it suffices to show that any two curves in @xmath1422 are @xmath18-panted connected .
let @xmath1423 be any two curves in @xmath1422 . by interpolating
a pair of points of distance @xmath1424 on @xmath36 with suitably assigned normal vectors , we may decompose @xmath1420 into a @xmath1425-tame bigon @xmath1426 $ ] with @xmath1427 of length @xmath1424 .
similarly , we decompose @xmath1421 into a @xmath1425-tame bigon @xmath1428 $ ] with @xmath684 of length @xmath1424 . by the connection principle ( lemma [ connectionprinciple ] ) , there
exist oriented @xmath13-framed segments @xmath323 and @xmath1429 of length @xmath1430-close to @xmath1424 and phase @xmath1430-close to @xmath358 , so that @xmath1431 form a @xmath1432-tame cycle .
let @xmath52 be the reduced cyclic concatenation @xmath1433 $ ] , then @xmath1434 by the length and phase formula ( lemma [ lengthandphaseformula ] ) .
furthermore , @xmath1426 $ ] and @xmath1435 $ ] form an @xmath1436-swap pair , and @xmath1428 $ ] and @xmath1437 $ ] form an @xmath1436-swap pair ( definition [ swappair ] ) . by swapping ( construction [ swapping ] )
, we see that @xmath1420 is @xmath18-panted connected with @xmath52 , and that @xmath52 is @xmath18-panted connected with @xmath1421 .
thus @xmath1420 and @xmath1421 are @xmath18-panted connected .
this completes the proof .
an application of proposition [ pantedconnectedness ] is that we can replace any @xmath18-panted surface with a connected one with the same boundary without changing its relative homology class .
[ connectedpantedsurface ] if @xmath4 is an oriented closed hyperbolic @xmath0-manifold and @xmath18 are constants so that @xmath153 is @xmath18-panted connected in the sense of proposition [ pantedconnectedness ] , then for any oriented compact @xmath18-panted subsurface @xmath97 bounded by an @xmath18-multicurve @xmath19 , there exists an oriented compact connected @xmath18-panted subsurface @xmath1438 bounded by @xmath19 such that @xmath1250 $ ] equals @xmath1439 $ ] in @xmath1440 . by induction
, it suffices to prove the case when @xmath3 has only two components @xmath161 and @xmath162 . without loss of generality
, we may assume that each @xmath166 has a nonseparating glued cuff @xmath165 of its pants structure .
in fact , this is automatically true if @xmath166 is closed .
if some @xmath166 is has a nonempty boundary component @xmath158 , there is an @xmath18-pants @xmath200 with a boundary component @xmath238 homotopic to @xmath158 .
take two oppositely oriented copies @xmath1441 of @xmath200 and glue up along the two cuffs other that @xmath1442 .
denote the resulting @xmath18-panted surface as @xmath1443 , so @xmath1444 is @xmath1445 .
if @xmath1446 is has the same orientation as that of @xmath200 , we glue up @xmath1443 and @xmath166 identifying @xmath1447 with @xmath1448 . the resulting @xmath18-panted surface @xmath1449 has the same boundary as that of @xmath166 up to homotopy , and @xmath1449 is homologous to @xmath166 in @xmath4 relative to their boundary .
after replacing @xmath166 with @xmath1449 , each component of @xmath3 has a nonseparating glued cuff @xmath158 as claimed .
now suppose that @xmath257 is a nonseparating glued cuff for each @xmath166 .
let @xmath736 be the @xmath18-panted surface obtained by cutting @xmath166 along @xmath165 , so @xmath749 has two components @xmath1450 and @xmath1451 homotopic to @xmath158 and its orientation reversal , respectively . by proposition [ pantedconnectedness ] , there exists an @xmath18-panted surface @xmath985 with boundary , so that there are two components @xmath1452 and @xmath1453 of @xmath1454 homotopic to @xmath169 and @xmath170 , respectively .
take a copy @xmath1455 of @xmath985 and a copy @xmath1456 of the orientation reversal of @xmath985 .
denote the components of @xmath1457 corresponding to @xmath1458 as @xmath1459 .
we glue up @xmath1460 along the opposite pairs of boundary components other than @xmath1459 , and glue them with @xmath736 by identifying @xmath1459 with @xmath1461 , respectively . the resulting @xmath18-panted surface @xmath167 has the same boundary as that of @xmath3 , and @xmath167 is homologous to @xmath3 in @xmath4 relative to their boundary .
since @xmath167 is connected by the construction , we see that @xmath167 is a connected @xmath18-panted surface as desired .
in this section , we prove theorem [ homologyviapants ] by applying theorem [ theorempantedcobordism ] and propositions [ secondhomologyclass ] , [ pantedconnectedness ] ( subsection [ subsec - proofofhomologyviapants ] ) ; we prove theorem [ main - qfsurface ] by applying theorems [ gluing ] and [ homologyviapants ] , following the methodology of section [ sec - methodology ] ; we prove theorem [ main - pantedsurface ] by applying theorems [ theorempantedcobordism ] , [ secondhomologyclass ] and proposition [ pantedconnectedness ] .
subsection [ subsec - descriptionoftheproblem ] contains some remarks about the formulation of theorem [ main - qfsurface ] generally speaking , the construction problem of geometrically finite surface subgroups in a kleinian group is concerned about the existence of such surface subgroups subject to various conditions .
for instance , one may ask about the existence specifying the boundary of the subsurface , or requiring it to represent some homology class . we will allow ourselves to pass to finite covers of the designated boundary loops or positive multiples of the homology class , and we will only look for immersed subsurfaces rather than embedded ones .
in many motivating applications , these are the usual assumptions under similar circumstances . on the other hand , we will only consider closed hyperbolic @xmath0-manifolds .
this is due to the restriction of our current techniques , and it would certainly be interesting to study the construction problem for cusped hyperbolic @xmath0-manifolds .
let @xmath4 be a closed hyperbolic @xmath0-manifold .
suppose @xmath1462 are @xmath1-injectively immersed loops @xmath1463 , and let @xmath1464 be their disjoint union : @xmath1465 the relative homology of the mapping cone @xmath1466 is well defined , depending only on the free homotopy classes of the loops . without loss of generality , we may hence assume that they are embedded . identifying @xmath19 with its image ,
the relative homology of the cone becomes @xmath1467 by excision .
if @xmath1468 is an immersion of an oriented compact surface @xmath3 , then @xmath3 naturally represents a relative homology class @xmath1469\,\in\ , h_2(m , l;\,{{\mathbf q}}),\ ] ] where the rational coefficient is taken since we are not interested in the torsion .
moreover , it is clear that the restriction of @xmath8 on @xmath625 is a covering if @xmath3 has no disk or sphere component and if @xmath8 is @xmath1-injective on each component of @xmath3 . with the notations above ,
the construction problem that we are concerned about in this paper is the following : for any element @xmath6 , is there a positive multiple of @xmath9 represented by a connected oriented surface @xmath3 and a @xmath1-injective quasi - fuchsian immersion @xmath8 ?
the answer is affirmative as stated in theorem [ main - qfsurface ] .
we summarize the proof of theorem [ homologyviapants ] . note that the ` furthermore ' part follows immediately from the main statements because the boundary homomorphism @xmath13 is integral coefficiented over the natural basis of @xmath129 and @xmath1470 . to prove the first statement
, it suffices to find an @xmath18-panted surface representing any class @xmath1471 up to a scalar multiple .
we may assume without loss of generality that @xmath9 is integral .
under the boundary homomorphism @xmath1472 @xmath1473 can be represented by a multicurve @xmath19 all of whose components are carried by components of @xmath131 . under the composition of the canonical isomorphism @xmath1474 ( theorem [ theorempantedcobordism ] ) and the projection @xmath1475 , @xmath24_{r,\epsilon}$ ]
is sent to @xmath358 since @xmath19 is a boundary .
this means @xmath1476_{r,\epsilon}$ ] is @xmath358 in @xmath786 , so there is an @xmath18-panted surface @xmath1477 with boundary @xmath350 .
let @xmath1478,\ ] ] in @xmath1479 .
since @xmath1480 in @xmath1481 , @xmath1482 can be regarded as an element of @xmath1254 . by theorem [ secondhomologyclass ] , @xmath1482 is represented by an @xmath18-panted surface @xmath1483 .
therefore @xmath1484 is represented by the @xmath18-panted surface @xmath1485 . to prove the second statement , recall that by ( * ? ? ?
* theorem 3.4 , cf .
section 4 ) , there is a measure of @xmath18-nearly regular pants @xmath98 which is @xmath1486-nearly evenly footed ( definition [ gluingconditions ] ) , such that @xmath111 is positive on any curve @xmath112 .
note that the existence of @xmath101 relies on the assumption that @xmath14 is bounded by the injectivity radius of @xmath4 , ( * ? ? ?
* subsection 4.4 ) .
let @xmath1487 be the sum @xmath1488 , where @xmath1489 , so @xmath1490 .
let @xmath1491 be the sum of all @xmath56 for all @xmath99 .
for some sufficiently large positive integer @xmath246 , let @xmath1492 then @xmath1493 is ubiquitous , @xmath18-nearly evenly footed .
it is certainly rich as @xmath1494 vanishes in this case . by proposition [ pantedconnectedness ]
, @xmath143 is irreducible .
in fact , suppose otherwise that @xmath143 were the sum @xmath1495 , such that @xmath116 and @xmath117 have disjoint supports on @xmath153 .
then for @xmath1496 lying in the supports of @xmath116 and @xmath117 respectively , @xmath52 and @xmath1497 can not appear simultaneously on an @xmath18-panted surface @xmath3 whose pants are all from the support of @xmath143 .
however , @xmath143 is ubiquitous so the support of @xmath143 is @xmath611 .
thus we reach a contradiction since @xmath52 and @xmath1497 should be @xmath18-panted connected in the sense of the conclusion of proposition [ pantedconnectedness ] . therefore , @xmath1498 is a measure as claimed in the second statement of theorem [ homologyviapants ] .
it remains to prove that if @xmath1499 , then for some sufficiently large integer @xmath255 , @xmath1500 in @xmath140 is ubiquitous , irreducible , @xmath18-nearly evenly footed , and rich .
technically , one may assume here that @xmath143 is @xmath1486 , and such a @xmath143 can be achieved by using @xmath1501 instead of @xmath14 in the construciton above .
it is clear that @xmath1502 is ubiquitous and irreducible as so is @xmath143 . on the other hand
, @xmath143 being ubiquitous also implies that @xmath1503 is positive on @xmath35 for any curve @xmath36 , so when @xmath255 is sufficiently large , the normalization of the measure @xmath1504 can be @xmath1505-equivalent to the normalization of @xmath1503 restricted to @xmath35 , for all @xmath112 . hence @xmath1502 is @xmath18-nearly evenly footed .
it is also clear that @xmath1502 is rich if @xmath255 is so large that @xmath1506 is less than , for instance , @xmath1507 for all @xmath112 .
this completes the proof of the second statement , and hence the proof of theorem [ homologyviapants ] .
we derive theorem [ main - qfsurface ] from theorems [ homologyviapants ] and [ gluing ] as follows .
let @xmath4 be a closed hyperbolic @xmath0-manifold , and @xmath5 be the union of finitely many mutually disjoint , @xmath1-injectively embedded loops .
[ reductiontoorientable ] if theorem [ main - qfsurface ] is true when @xmath4 is orientable , it holds in the general case as well .
assume that theorem [ main - qfsurface ] is true for the orientable case .
if @xmath4 is not orientable , we consider an orientable @xmath307-fold cover @xmath1508 , and let @xmath1509 be the preimage of @xmath19 . by excision ,
@xmath1510 is isomorphic to @xmath1511 where @xmath246 is the compact @xmath0-manifold with tori boundary obtained from @xmath4 with @xmath19 removed , and similarly , @xmath1512 is isomorphic to @xmath1513 where @xmath1514 is the @xmath307-fold cover of @xmath246 obtained from @xmath1515 with @xmath1509 removed .
because @xmath1513 surjects @xmath1511 under the covering , the same holds for @xmath1512 and @xmath1510 .
therefore , for any element @xmath1516 , we may take an element @xmath1517 which is projected to be @xmath9 .
the orientable case implies that a positive integral multiple of @xmath1418 can be represented by an orientable compact @xmath1-injectively immersed quasi - fuchsian surface @xmath1518 , so @xmath1519 represents a positive integral multiple of @xmath9 as desired .
by lemma [ reductiontoorientable ] , we may assume without loss of generality that @xmath4 is an oriented closed hyperbolic @xmath0-manifold .
[ goodcurvecovers ] with the notations of theorem [ main - qfsurface ] , for any constant @xmath1520 , and for any constant @xmath244 , there exist some @xmath93 , such that every component of @xmath19 has a finite cyclic cover which is homotopic to a curve of @xmath153 .
let @xmath1521 be the components of @xmath19 .
let @xmath1522 be the length of @xmath1523 , and @xmath1524 be the phase of @xmath1523 .
then for any positive integer @xmath1525 , the length and phase of the @xmath1525-fold cyclic cover of @xmath1523 are @xmath1526 and @xmath1527 , respectively .
we must show that for any positive constant @xmath14 and @xmath1528 as above , there exists @xmath93 and positive integers @xmath1529 , such that the complex numbers @xmath1530 are all @xmath14 close to the real number @xmath15 .
consider the complex torus @xmath293 obtained by @xmath1531 modulo the lattice spanned over @xmath1532 by @xmath1533 , and @xmath1534 , for @xmath1535 , where @xmath1536 denotes the standard basis .
the origin of @xmath1531 is projected to be the basepoint @xmath800 of @xmath293 .
the diagonal ray @xmath1537 defined by @xmath1538 is projected to be a ray @xmath1539 start from @xmath1540 . then to find @xmath15 and @xmath1529 as above
, it suffices to show that for any positive constants @xmath14 and @xmath1528 , there exists @xmath93 so that @xmath1541 is @xmath14-close to @xmath800 .
however , this is a well known fact , which can be derived easily from dynamics of geodesic flow on a euclidean torus , so we omit the details .
choose a sufficiently small positive @xmath14 depending on @xmath4 and a sufficiently large @xmath15 depending on @xmath4 and @xmath14 so that lemma [ goodcurvecovers ] , and theorems [ gluing ] and [ homologyviapants ] can be applied .
by lemma [ goodcurvecovers ] , there is a finite cover @xmath1509 of @xmath19 with all components homotopic to curves of @xmath153 .
let @xmath126 be all the unoriented curve classes which are realized by some component of @xmath1509 .
consider the essential case when @xmath1509 does not have two components that are homotopic to each other up to the orientation reversion . in this case
, @xmath131 is in correspondence with components of @xmath1509 .
the relative homology @xmath1479 defined in subsection [ subsec - homologyviapants ] is naturally isomorphic to @xmath1542 ( cf .
subsection [ subsec - descriptionoftheproblem ] ) , and that there is a natural homomorphism @xmath1543 induced by the covering @xmath1544 .
we may realize any homology class @xmath22 as follows . passing to some large multiple of @xmath9 ,
we may lift @xmath9 to be an element @xmath1545 then by theorem [ homologyviapants ] ( 1 ) , there exists a measure @xmath1546 which is mapped onto @xmath1418 , and by theorem [ homologyviapants ] ( 2 ) , we may assume that @xmath101 is ubiquitous , irreducible , @xmath18-nearly evenly footed , and rich .
then there exists an oriented , connected , compact , @xmath1-injectively immersed quasi - fuchsian subsurface : @xmath123 which is @xmath18-nearly regularly panted subordinate to a positive integral multiple of @xmath101 , by theorem [ gluing ] .
it is clear from the construction that ( up to homotopy ) @xmath1547 $ ] can be regarded as an element in @xmath1440 representing a positive multiple of @xmath9 . from the essential case above one
can derive the general case when @xmath1509 may have components that are homotopic to each other up to the orientation reversion .
we will only sketch the argument , as the tricks used below should be easy and less important .
let @xmath1548 be a maximal subunion of components of @xmath1509 so that @xmath1548 does not have two components that are homotopic to each other up to the orientation reversion .
as in the essential case , @xmath131 is in correspondence with components of @xmath1509 , so @xmath1479 can be realized as @xmath1549 .
there is also a natural homomorphism @xmath1550 induced by the inclusion @xmath1551 .
it is easy to verify that @xmath1542 is generated by @xmath1549 together with all @xmath1552 $ ] where @xmath1553 is any annulus between two homotopic components of @xmath1548 .
if the boundary of any such annulus @xmath107 is a curve @xmath112 , we may take a null - homologous @xmath18-panted surface @xmath775 which has a glued cuff @xmath158 homotopic to @xmath36 .
such an @xmath775 can be obtained , for example , by gluing pairs of pants prescribed by an ubiquitous @xmath1498 .
cutting @xmath775 along @xmath158 and homotoping the two boundary components to the two components of @xmath1554 gives rise to an @xmath18-panted surface @xmath1555 such that @xmath1556 $ ] equals @xmath1552 $ ] as in @xmath1542 .
now for any homology class @xmath1516 , we may pass to a positive integral multiple of @xmath9 and lift it as @xmath1557 . from the above
@xmath1558 equals the sum of some @xmath1559 together with some positive rational multiple @xmath1556 $ ] .
possibly after passing to further positive multiple , the essential case implies that @xmath1560 can be represented by an @xmath18-panted surface @xmath1561 with @xmath1562 mapped to @xmath1548 .
we may also represent the difference term @xmath1563 by a union of @xmath18-panted surfaces @xmath1564 as above .
thus the union @xmath1565 is an @xmath18-panted surface representing @xmath1558 in @xmath1566 .
we may assume that @xmath1561 and each @xmath1564 to be obtained by a @xmath18-nearly unit - shearing gluing of a collection of @xmath18-pants prescribed by a ubiquitous , irreducible , @xmath18-nearly evenly footed , and rich measure , then modifying the gluing by a hybriding argument ( cf .
lemma [ hybriding ] ) will yield a connected @xmath1-injectively immersed quasi - fuchsian surface @xmath167 , which still represents @xmath1558 .
this completes the argument of the general case , and hence completes the proof of theorem [ main - qfsurface ] .
we derive theorem [ main - pantedsurface ] from theorems [ theorempantedcobordism ] , [ secondhomologyclass ] and proposition [ pantedconnectedness ] as follows .
we point out that as those results rely only on the constructions of @xmath13-framed segments ( section [ sec - basicconstructions ] ) , the input from dynamics necessary for the proof is only the mixing property of the frame flow on the closed hyperbolic @xmath0-manifold @xmath4 , but not the fact that the mixing rate is exponential .
the invariant @xmath570 can be defined for any null - homologous @xmath18-multicurve as @xmath1567_{r,\epsilon}),\ ] ] where @xmath810 is the canonical isomorphism by theorem [ theorempantedcobordism ] .
note that since @xmath19 is null homologous , @xmath16 lies in the canonical submodule of @xmath27 coming from the center @xmath17 of @xmath799 .
hence @xmath16 has well defined value in @xmath17 .
it follows that @xmath1568 equals @xmath1569 because @xmath11 is a homomorphism .
it also follows that @xmath16 vanishes if and only if @xmath19 is the boundary of an @xmath18-panted subsurface @xmath3 of @xmath4 .
moreover , proposition [ pantedconnectedness ] implies that we may assume @xmath3 to be connected ( lemma [ connectedpantedsurface ] ) .
it remains to show the last statement in the conclusion of theorem [ main - pantedsurface ] .
let @xmath19 be an @xmath18-multicurve with vanishing @xmath16 .
fix an @xmath18-panted surface @xmath1561 bounded by @xmath19 , and denote the relative homology class of @xmath1561 as @xmath1570 . for any homology class @xmath22 with @xmath1571 equal to @xmath24\in h_1(l;{{\mathbf z}})$ ]
, there exists some @xmath1572 such that @xmath1573 . by theorem [ secondhomologyclass ] , @xmath1482
can also be represented by a closed @xmath18-panted subsurface @xmath775 .
thus we may take @xmath3 to be @xmath1574 so that @xmath3 is an oriented compact @xmath18-panted subsurface of @xmath4 representing @xmath9 . by applying lemma [ connectedpantedsurface ] again
, we may substitute @xmath3 with another oriented compact connected @xmath18-panted subsurface @xmath167 representing @xmath9 , as desired .
this completes the proof of theorem [ main - pantedsurface ] .
in conclusion , we are able to construct homologically interesting connected immersed nearly geodesic nearly regularly panted subsurfaces in a closed hyperbolic @xmath0-manifold @xmath4 by knowning a finite presentation of its fundamental group . the existence of plenty of nearly regular pairs of pants in @xmath4 is a consequence of the exponential mixing property of the frame flow , and is the essential reason for the connectedness and the @xmath1-injective quasi - fuchsian property .
even if we did not know the mixing rate , the connection principle can still be deduced from the mixing property , so homologically interesting connected @xmath18-panted subsurfaces can still be constructed .
[ changecoefficient ] is it possible to generalize theorem [ main - qfsurface ] to other coefficients ?
for example , if @xmath1575 is any field , does every homology class @xmath1576 have represented an @xmath1575-oriented compact @xmath1-injectively immersed quasi - fuchsian subsurface ?
it seems that our argument can be modified without difficulty to confirm question [ changecoefficient ] when @xmath1575 is any field of characteristic other than @xmath307 . however , the @xmath17 coefficient case is not clear since the subsurface constructed might be non - orientable , so an unoriented @xmath18-panted cobordism theory needs to be developed .
[ changedimension ] is it possible to generalize theorems [ theorempantedcobordism ] and [ secondhomologyclass ] to other dimensions ?
in particular , can we define and determine the @xmath18-panted cobordism group @xmath10 for any oriented closed hyperbolic manifold @xmath4 ?
we expect that theorem [ theorempantedcobordism ] should hold for all dimensions at least @xmath0 . in dimension
@xmath307 , it seems that @xmath1577 should be a split extension of @xmath1578 by @xmath17 .
this is basically because the special orthonormal frame bundle @xmath1579 should be replaced with the special orthonormal frame bundle of a stabilization @xmath1580 of the tangent bundle @xmath1581 , in order that lemma [ canonicalliftlemma ] holds .
the following two questions are much more difficult but significant . to answer question [ gfsubmanifolds ] ,
we expect a notion of good basic pieces playing the role of nearly regular pairs of pants . to answer question [ cuspcase ] , we need a modified version of the connection principle since the mixing property of frame flow no longer holds .
|
among other things , we prove the following two topologcal statements about closed hyperbolic @xmath0-manifolds . first , every rational second homology class of a closed hyperbolic @xmath0-manifold has a positve integral multiple represented by an oriented connected closed @xmath1-injectively immersed quasi - fuchsian subsurface .
second , every rationally null - homologous , @xmath1-injectively immersed oriented closed @xmath2-submanifold in a closed hyperbolic @xmath0-manifold has an equidegree finite cover which bounds an oriented connected compact @xmath1-injective immersed quasi - fuchsian subsurface . in part , we exploit techniques developed by kahn and markovic in @xcite , but we only distill geometric and topological ingredients from those papers so no hard analysis is involved in this paper .
|
the inflationary paradigm@xcite proposed around @xmath6 was a giant leap towards going beyond the standard model of big - bang cosmology . since then there has been a plethora of inflationary models and it is still increasing @xcite .
however , with the high precision observational cosmology @xcite , any cosmological model building will have to take care of strong observational constraints . by
now inflation has become an integral part of the modern theoretical cosmology .
it is an exponential expansion at the very early stage of our universe .
so far this is the only mechanism which can successfully generate all the suitable initial boundary conditions for the standard big - bang cosmology .
as mentioned before , there exists a large class of models which predict required inflationary observables within the experimental bounds .
particularly , a recently discovered inflationary @xmath7 attractor @xcite gained a significant attentions because of its unification of a large class of models .
in this paper we will introduce new classes of inflationary scenarios extending one of the present authors current work on minimal higgs inflation @xcite . in this class of models we consider a very simple modification of the usual power law potential by incorporating an inflation scale such that it generates a flat plateau for infinite range of inflaton field value .
we may think of these form of the potentials as a small dip over a cosmological constant set by the aforementioned inflationary energy scale .
therefore , it could be very generic in the sense that it does not need extreme fine - tunning to get such a form of the potential . only tunning we have to make
is to model the form of the dip .
however , we believe as well as our subsequent analysis shows that form of the dip will not have significant effect at least on the inflationary observables .
it is the infinite flat plateau which plays the main role to trigger the inflation . depending upon the form of the dip , dynamics after the inflation or more specifically reheating prediction will be different .
therefore , infinite class of models can be constructed by changing the form of the dip . in the recent paper on minimal
higgs inflation @xcite , two different kind of form of the dip has been taken in compatible with the the usual higgs potential and found the models fit extremely well with the cosmological observations . in this paper
we generalize further the form of the dip in compatible with the general power law potential and study in detail their predictions during inflation .
we also study the evolution after the end of inflation , specifically focusing on reheating period . in this paper
we mainly study the perturbative evolution of all the important components of the universe . out of all the components we specifically focus on the heavy dark matter production and constrain the parameter space based on the present dark matter abundance .
the current observational data shows that most of the cold matters in our universe are non - baryonic @xcite , and weakly interacting @xcite .
importantly , direct and indirect probe to detect dark matter in the lower mass range has been inconclusive so far .
hence , it is quite natural to look for dark matter in the heavy mass range , where it is termed as the super - heavy dark matter@xcite .
such a massive particle is very difficult to produce with the current experimental set up because of kinematics .
however , reheating phenomena after the inflation is believed to be an important mechanism which can produce such a massive particle .
therefore , in the cosmological context we can look for indirect evidence of such a heavy dark matter through small scale non - gaussianities in the cmb or even in ultra - high energy cosmic rays @xcite .
the produced mass of the dark matter crucially depends on its production mechanism . for thermal production simple kinematics does not allow the production of dark matter with mass greater than equilibrium
temperature@xmath8 . however , it has long been identified that during reheating , the equilibrium temperature @xmath9 is not the maximum temperature of the universe after inflation @xcite and as a consequence particles of mass larger than the reheating temperature can be produced@xcite .
the mass limit can be further extended if we allow the dark matter production via direct decay of inflaton @xcite . in this paper
we will generalize the existing mechanism of heavy dark matter production for arbitrary equation of state @xmath10 of the oscillating inflaton .
we structured our paper as follows : in section - i , we introduce in detail the form of the potential and their behavior .
we study inflationary dynamics and corresponding conditions for the inflation .
all those conditions , which one generally calls slow roll parameters , will play important roll on the evolution of the cosmological perturbation . by using those we compute three important cosmological observables which are scalar spectral index @xmath11 , the tensor to scalar ratio @xmath12 , and the spectral running @xmath13
. we also considered the behavior of those parameters in terms of model parameter .
after the end of inflation , generically the inflaton starts to have coherent oscillation at the minimum of the potential , which we call re - heating phase of the universe . without explicitly considering the detail mechanism of reheating , one can figure out the possible range of re - heating temperature @xmath8 , and constrain the evolution of scale factor parametrized by efolding number @xmath14 during re - heating period , by observing the background evolution of observable scales and entropy density . in section - ii
, we will consider the aforementioned evolutions of scales and entropy density to constrain further our models .
in section - iii , we study in detail the dynamics of the oscillatory phase of inflaton and study perturbative re - hearting phase . specifically focusing on the heavy dark matter particle production ,
we have solved a set of boltzmann equations for the coupled system of inflaton , radiation and the matter field with different inflaton equation of state during late time reheating phase . assuming the produced heavy particle to be a part of the total dark matter we see today
, we set the upper limit on the value of dark matter annihilation cross - section with respect to the mass and reheating temperature .
as we have mentioned in the introduction , we will model the dip part of the potential by the power law form . however , it has to have smooth connection with the flat infinite plateau as one increases the inflaton field value .
therefore , we suggest the following simple forms of the potential , v ( ) = + , in the above form of the potentials , we have introduced two free parameters @xmath15 .
where , the parameter @xmath16 is defined for @xmath17 , which has been studied as minimal higgs inflation @xcite . for other value of @xmath2
, we can set @xmath18 . as has been advertised
, we can think of above potentials as dip in the background of cosmological constants set by @xmath19 . a simple illustration of the above form of the potentials is shown the fig.[potential ] . through out the paper
, we will refer type - i for first form of above potential and type - ii for the second form .
of course one important generalization of our model can be done by considering the potential to be dependent only upon the modulus of the inflaton field .
therefore , all the odd values of @xmath2 can be included .
however , our analysis through out the paper will be valid for all values of @xmath2 .
for the sake of simplicity we will stick to only even values of @xmath2 .
another simple generalization of our model one can do is by defining @xmath20 as a new potential . , for @xmath21 in this section we will study in detail the background dynamics using the above potential form
. we will start with the following action , s = d^4x [ action ] where @xmath22 is the reduced planck mass . assuming the usual friedmann - robertson - walker(frw ) background ansatz for the space - time ds^2 = dt^2 - a(t)^2 ( dx^2 + dy^2 + dz^2 ) , [ frw ] the system of equations governing the inflationary dynamics in the above background and the dynamics of the scalar field are 3m_p^2 h^2= ^2 + v
( ) + 2m_p^2 = -^2 + + 3 h + v ( ) .
[ friedman ] where , the usual definition of hubble constant is @xmath23 .
as we have seen , our potential has a very flat plateau away from the dip .
therefore , flatness condition which is required for the inflationary dynamics is automatically satisfied for our potential .
the flatness conditions for the potential during inflation are called slow - roll parameters , which are defied as ( ) ^2 & = & + + m_p^2 ( ) & = & + [ eq - slow - roll ] inflation occurs when @xmath24 and @xmath25 .
as one can easily see , the inflation should end near the dip of the potential .
however , the precise location of the end of inflation is usually set by the condition @xmath26 .
let us also define a higher order slow - roll parameter related to the third derivative of the potential for our later purpose .
the expression for the higher order slow - roll parameter is as follows : m_p^4 ( ) = + in addition to provide the successful inflation , we will see how all the aforementioned slow - roll parameters control the dynamics of cosmological perturbation in the inflationary background .
the idea of inflation was first to solve the so - called horizon and flatness problem in the standard big bang cosmology .
an important cosmological parameter which quantify the amount of inflation is called e - folding number @xmath27 . the e - folding number which is defined as the logarithmic ratio of the scale factors at the beginning and at the end of inflation ,
can be written as n = ln ( ) = _ a_in^a_end d lna = _
t_in^t_end h dt [ efold ] in the slow - roll regime one can use hdt = = [ efold2 ] as we have mentioned the inflation ends when @xmath26 , and one can use the eq.([efold ] ) to find the beginning of the inflation .
now by using the eqs.([efold]-[efold2 ] ) , we can express the the efolding number into the following form , @xmath28 & \simeq \frac{\phi_*^2}{n m_p^2 } \frac{1}{(n+2 ) } \tilde{\phi}^{(n+2 ) } \\
\frac{\phi_*^2}{n m_p^2 } \left[\frac{1}{4 } ( \tilde{\phi}^4-\phi_{end}^4 ) + \frac{1}{2 } ( \tilde{\phi}^2-\phi_{end}^2)\right ] & \simeq \frac{\phi_*^2}{4 n m_p^2 } \tilde{\phi}^4 , \end{cases } \label{efold3}\ ] ] where we have defined , @xmath29 . in the right hand side of the above approximate expression for @xmath30 ,
we have ignored the contribution coming from @xmath31 , and also squared term .
we have numerically checked the validity of those expressions for a wide range of value of @xmath32 of cosmological interest . from the observational constraint one needs @xmath33 in order to make the largest scales observed in cmb to be in causal contact before the inflation .
we have all the necessary mathematical expressions to solve for the background inflationary dynamics .
one particular solution has been given in fig.[mi - rhovaplot ] , with a specific value of the efolding number .
once we solve for the background , in the next sections we will do cosmological perturbation analysis , and derive the relevant cosmological parameters as has been described in the introduction . as we have explained in detail in the introduction , the very idea of inflation
was devised to solve some outstanding problems of standard big - bang cosmology .
soon it was realized that inflation also provides seed for large - scale structure formation via quantum fluctuation on the inflationary background .
this remarkable fact gives the inflationary theory a solid basis .
all the cosmologically relevant inflationary observables are essentially identified with various correlation function of those primordial fluctuations calculated in the framework of quantum field theory . in cosmological perturbation theory the most general form the perturbations on the inflationary background eq.([frw ] ) can be expressed as ( x , t ) & = & _ 0(t ) + ( x , t ) , + ds^2 & = & ( 1 + 2 ) dt^2 + 2 a ( _ i b + b_i)dx_i dt + a^2(1 - 2)_ij dx_i dx_j + & + & ( 2 _ i _ j e + 2 _ ( ie_j ) + h_ij)dx_i dx_j , where , the perturbed fields satisfy the following transverse and traceless gauge conditions , _
i h^ij = 0 ; _ ij h^ij = 0 ; ^i b_i = ^i e_i=0 .
therefore , we have ten metric perturbation components , however , only two of them will be dynamical because of general coordinate invariance .
those two are the component of tensor metric perturbation @xmath35 , which are gauge invariant .
in addition we have one scalar degrees of freedom which is the inflaton field . by using the diffeomorphism invariance
, one can identify the gauge invariant scalar perturbation @xcite : - . in the literature ,
we call it as co - moving curvature perturbation
. two point correlation function of this curvature perturbation is known as primordial power spectrum .
this is the fluctuation which turned into density perturbation after their horizon exit , and produce all large scale structure that we see through the gravitational instability after their horizon re - entry . in order to compute the correlation function , necessary quadratic action for curvature perturbation will look like , s_(2 ) = d^4 x a^3 .
evaluating the power spectrum from this action is standard ( see , @xcite for a comprehensive review of cosmological perturbation theory ) , and it is given by = |_k = ah = .
once , we know the power spectrum , cosmological quantity of our interests are the spectral tilt and its running . during inflation
a particular inflaton field value corresponds to a particular momentum mode exiting the horizon .
hence by using the following relation to the leading order in slow - roll parameters , @xmath36 , one obtains the following inflationary observables , n_s-1 = -6 + 2 dn_s^k = -2 + 16 - 24 ^2 . in a similar manner , we can also compute quadratic action for the gauge invariant tenor perturbation @xmath35 and compute its power spectrum @xmath37 .
standard practice is to define a the tensor - to - scalar ratio r = = 16once we have all the expression for cosmological quantities in terms of slow roll parameters , by using eqs.([eq - slow - roll],[efold3 ] ) , and considering @xmath38 in unit of @xmath39 , we express @xmath34 in terms of @xmath2,@xmath30 and @xmath40 , as 1 - n_s & = & + ; dn_s^k = - + - + r & = & 8n^2 ( ) ^ + [ ns - r - r ] we have also numerically checked all the above approximate expressions . from the above analytic expressions for the @xmath34 ,
some general but important comments are as follows . for type - ii class of models
, we see that the value of @xmath41 are insensitive to the value of @xmath2 or on the dimension of the potential . this fact can also bee seen from the fig.([fig - rvlogalpha ] ) , where , no approximation has been done .
in particular one notices that for @xmath42 , value of @xmath43 start to deviate form each other for different value of @xmath2 .
however , in general value of @xmath44 to the leading order .
therefore , form the planck observation considering the upper bound of @xmath45 , we can constrain the upper limit of @xmath2 for a fixed value of @xmath32 . as
one understands from this fact that from the current observation , it is difficult to uniquely fix the form of the potential .
therefore , we need more theoretical inputs to figure out the full form of the potential . for type - i models ,
all quantities are dependent on @xmath2 , which can also be seen from fig.([fig - nsvlogalpha ] ) .
however , an important fact emerges in the limit @xmath46 for these type of models .
if we take the aforementioned limit , the expression of @xmath47 for the type - i model reduce to 1-n_s ; r 8 ( ) ^2 ; dn_s^k = - , which can be identified as a particular model within the class of recently proposed ` @xmath7-attractor ' models @xcite .
therefore , for a wide range of parameter values , we have infinite possibilities namely for @xmath48 , which predict all the cosmological parameters within the planck bound @xcite .
therefore , it is of particular importance to study their uv completion , specifically the supergravity formulation of those form of the potential .
nonetheless , we have numerically checked the dependence of @xmath49 on @xmath2 for a particular value of inflationary scale as shown in fig.([fig - nsvn ] ) . as on observes , from eq.([ns - r - r ] ) , for @xmath50 , @xmath51 as @xmath46 .
this is also clearly seen from the asymptotic behavior of @xmath49 in the fig.([fig - nsvn ] ) +
.the spectral quantities for different values of @xmath2 for 50 efolding .
the two values of @xmath40 are chosen to illustrate that we can have both small field and large field inflation depending on the value of @xmath40 .
the general trend for the variation of these quantities with @xmath40 is illustrated in the figures ( [ fig - nsvlogalpha]-[fig - rvlogalpha ] ) [ cols= " < , < , < , < , < , < , < , < , < , < " , ] + in order to understand the the dynamics of all the component during reheating , we will first numerically solve boltzmann equation for a particular choice of parameter values . as we have seen from the reheating analysis ,
depending upon the equation of state , and within the experimental limit of the scalar spectral index , reheating temperature could be from @xmath52 gev all the way up to @xmath53 gev .
therefore , we assume , @xmath54 , @xmath55 , @xmath56 , which would result in @xmath57 , the value of @xmath58 can be found from the relation ( [ tre ] ) .
we have taken @xmath59 at some arbitrary time during oscillatory period . with these choice of parameters
the results are shown in the figure.([fig - particle1 ] ) . as we noted that the reheating temperature is not the maximum temperature in the radiation era and the maximum temperature
is reached before the radiation domination begins . in our numerical calculation
, we have checked that for reheating temperature @xmath60 the maximum temperature @xmath61 gev and similarly for @xmath62 , @xmath63 .
the value of maximum temperature decreases slightly as we increase the value of @xmath2 .
we have mentioned two possible ways of producing the @xmath64 particle .
we see that changing @xmath65 and keeping @xmath9 fixed , the particle production get enhanced due to the direct production @xmath66 .
this effect is evident from the presence of a second plateau in the curves for the evolution of x - particle energy density ( dotted line for @xmath67 and dot - dashed line for @xmath68 ) .
even , for reheating temperature lower than @xmath69gev , the production via radiation is negligible and the direct production is the predominant one .
we also note that the difference between the two production processes became lesser prominent as we increase @xmath0 .
another important point we must mention is that , in deriving the analytic estimate for @xmath70 , we have assumed @xmath71 , in our numerical calculations we have checked the validity for the assumption for all the values of @xmath0 .
the numerical values of the quantities for different values of @xmath2 are given in the table ( [ xtable ] ) .
it is clear from the table-([xtable ] ) that the production of the dark matter particle decreases as we increase @xmath2 .
it is also observed that for @xmath72 , we will not have radiation dominated universe after the reheating phase for @xmath73 gev .
we arrived at this conclusion for a specific set of parameter values .
therefore , we ask the question of what should be the allowed parameter region for a specific reheating temperature so that our stable heavy x - dark matter particle freezes out at a value such that after evolution it provides us the lower or equal to the dark matter abundance we see today .
therefore , we have scan the parameter space of the dark matter annihilation cross section @xmath4 and its mass @xmath74 for a fixed reheating temperature taking reference of the current dark matter abundance .
to put a bound we will use second form of eq.([abundance ] ) and this will enable us to write a relation between @xmath75 , @xmath74 and @xmath76 .
the full relation is quite complicated , however we can simply write the relation between the three quantities of importance as follows , @xmath77~ \text{for $ m_x \lesssim t_{max}$}\\ \omega_x h^2 & = & d(w ) \left<\sigma|v|\right > m_x^4 exp\left[-\frac{e(w)m_x}{t_{max}}\right]~ \text{for $ m_x \gtrsim t_{max}$ } \label{analytic}\end{aligned}\ ] ] where , @xmath78 and @xmath79 are constants dependent on the equation of state parameter and the reheating parameters such as @xmath80 .
according to the various cosmological observation , the dark matter abundance is measured as @xmath81 @xcite . therefore ,
if we consider this value as an upper limit for our x - particle abundance eq.([analytic ] ) , depending upon different equation of state parameter @xmath0 , we figure out the possible allowed region in fig.([fig - particle1 ] ) . the upper limit on annihilation cross - section @xmath82 is bounded by the red and blue lines coming from unitarity and @xmath83 respectively .
the shaded region is allowed . as one observes from the plots ,
if we consider x - particle is the only dark matter candidate , the parameter space is highly constrained by the unitarity limit .
for example , if we consider reheating temperature @xmath84 gev , production of super - heavy dark matter @xmath85 will not be allowed , mainly because of unitarity limit .
as we increase @xmath9 , because of exponentially suppressed production , very narrow range of @xmath74 is allowed for super - heavy x - particle . for
a fixed value of dark matter mass @xmath74 , as one increases @xmath0 , the upper bound on @xmath86 will decrease . as an example , for @xmath84 gev and @xmath87 gev , one finds @xmath88 for equation of state @xmath89 respectively . from fig.[fig - particle1 ] , as one increases the value of @xmath0 , in the low mass regime slope of the solid and dotted blue curves is decreasing .
this fact widens the range of possible dark matter mass range while narrows down the annihilation cross - section .
in the first part of this paper , we have constructed two new classes of inflationary model with non - polynomial modification of the inflaton potential .
interesting property of these class of potential is that they have infinitely large flat plateau with a dip at the center in the field space .
naturally the large flat plateau drives the inflation . as a result , the predictions of the models are not very much sensitive to the details of the dip and more importantly fit extremely well with latest cosmological observation made by planck .
all the necessary scales in our models are below planck scale , which may imply that our model predictions are robust and any high energy correction to the observables could be planck scale suppressed .
detailed computation on the ultra - violate effect on our model could be important and we left it for our future work . depending upon the choice of scale , our model can have both large field as well as small field inflation .
however , for both the cases with a wide range of scale , the prediction of tensor to scalar ratio @xmath12 turned out to be significantly small .
another important aspect of our model is that we can have significantly low inflationary hubble scale @xmath90 unlike the prediction of usual power law inflationary model
. the value could be as low as @xmath91 gev for different values of @xmath92 for @xmath93 .
as we had discussed , low value fo @xmath90 could be interesting in the context of higgs vacuum instability .
as has been pointed out in @xcite , during as well as after the inflation the quantum fluctuation of higgs field can destabilize the standard model metastable higgs vacuum at around @xmath94 gev .
however , this instability crucially depends upon the value of @xmath90 , and also the height of the higgs potential .
therefore , comparing the naive scale dependence between @xmath95 and @xmath96 , our model have potential to save the higgs vacuum from decaying into the global vacuum .
we leave detail study for our future work . since we have discussed about the possible interesting interplay between our model and the higgs field at high energy scale
, it would be of particular interest to understand the potential application of our model in the particle physics context .
this is what motivates us to study the possible connection and constraints of our models specifically in the context of heavy dark matter production during the reheating period . after the end of inflation
, reheating is the stage during which all the matter fields are produced .
usual strategy is to consider the effective coupling among inflaton and various other matter fields we see today . in this paper
we have specifically concentrated on the perturbative reheating phenomena .
near the origin the inflaton potential behaves like a power law @xmath97 , therefore , coherently oscillating inflaton can be well approximated as an effective fluid with the equation of state parameter , @xmath98 . with the usual pure power law potential , one finds it very difficult to fit the cosmological observation for even @xmath99 .
therefore , detailed study have not been done for general power law potential specifically in the context of reheating . however , as we have advertised through out our paper , with the non - polynomial generalization we can easily have power law form of the inflaton potential specifically at the bottom of the dip where the inflaton coherently oscillates .
this fact leads us to generalize the known analysis of the perturbative reheating @xcite for general effective inflaton equation of state @xmath0 .
hence , we first tried to analytically study the boltzmann equations for the evolution of inflaton , radiation and a massive dark matter field for generic @xmath0 . furthermore , we compared our analytic study with the full numerical solutions for specific form of the potential of our interest .
we have found some interesting behavior such as the existence of minimum extraction of energy from the inflaton field for @xmath100 .
it could be interesting to understand this fact better and constrain the reheating period if any .
however , our main focus was to study the production of heavy dark matter particle with mass @xmath101 .
as there is no observational evidence of light dark matter so far , the heavy dark matter phenomenology could be interesting to explore@xcite .
the reheating mechanism is not fully explored and ill understood because of its non - linearity .
however it is assumed to have huge potential to explain many particle physics as well as cosmological problems starting from baryogenesis , dark matter , neutrino mass and their properties . however , in this paper we concentrated on heavy dark matter production .
we constrained the dark matter annihilation cross - section ( @xmath102 ) as a function of its mass @xmath74 considering the present dark matter abundance @xmath103 as an upper limit .
if we consider single component dark matter , parameter space is tightly constrained .
for example , if one considers low reheating temperature such as @xmath104 gev , we can not have production of super - heavy dark matter of mass @xmath105 gev . as we increase the equation of state @xmath0 ,
the region of parameter space is shrinks .
it would interesting study in details these link between the equation of state of the inflaton and the heavy dark matter particle .
planck collaboration , p. a. r. ade et al.,planck 2013 results .
constraints on inflation , astron .
* 571 * ( 2014 ) a22 , [ arxiv:1303.5082 [ astro-ph.co ] ] , planck 2015 results .
constraints on inflation , astron .
astrophys . * 594 * , a20 ( 2016 ) , [ arxiv:1502.02114 [ astro-ph.co ] ] .
r. kallosh and a. linde , jcap * 1307 * , 002 ( 2013 ) [ arxiv:1306.5220 [ hep - th ] ] ; s. ferrara , et al , phys .
rev . d*88 * , 8 , 085038 ( 2013 ) [ arxiv:1307.7696 [ hep - th ] ] ; r. kallosh , a. linde and d. roest , jhep * 1311 * , 198 ( 2013 ) [ arxiv:1311.0472 [ hep - th ] ] ; s. cecotti and r. kallosh , jhep * 1405 * , 114 ( 2014 ) [ arxiv:1403.2932 [ hep - th ] ] .
j. martin and c. ringeval , jcap * 0608 * , 009 ( 2006 ) [ arxiv : astro - ph/0605367 ] ; l. lorenz , j. martin , and c. ringeval , jcap * 0804 * , 001 ( 2008 ) [ arxiv:0709.3758 ] ; j. martin and c. ringeval , phys.rev .
* d82 * , 023511 ( 2010 ) [ arxiv:1004.5525 ] ; p. adshead _
et al _ , jcap , * 1102 * , 021 ( 2011 ) [ arxiv:1007.3748 ] ; j. mielczarek , phys.rev .
* d83 * , 023502 ( 2011 ) [ arxiv:1009.2359 ] ; r. easther and h. v. peiris , phys.rev . *
d85 * , 103533 ( 2012 ) , [ arxiv:1112.0326 ] ; l. dai , m. kamionkowski , and j. wang , phys .
lett . * 113 * , 041302 ( 2014 ) , [ arxiv:1404.6704 ] .
|
generalizing one of the author s recent paper on minimal higgs inflation , we proposed and analyzed a large class of inflationary models with non - polynomial modification of the potential .
the modification is done by introducing a single scale creating an infinite plateau for large inflaton field value .
one can identify those class of potentials as a small dip at the origin of a constant one dimensional field space . because of this large flat plateau , we find all the predictions are fitting extremely well with the recent observations made by planck .
we have extensively studied perturbative reheating phenomena specifically focusing on the production of dark matter heavier than the reheating temperature .
we generalize the well known analysis of heavy dark matter production for general equation of state @xmath0 of the oscillating inflaton field .
however , at the end we consider effective equation of state @xmath1 emerging form our model .
where , @xmath2 is the index of the power law potential during late time oscillatory phase . considering present value of the dark matter abundance @xmath3 ,
we constrain the parameter space of annihilation cross - section ( @xmath4 ) and mass @xmath5 of a single component dark matter for different reheating temperature and equation of state @xmath0 .
|
non - axisymmetric deformations of rapidly rotating bodies are rather generic phenomena in nature and can appear in a wide class of systems .
particularly interesting within an astrophysical context are those deformations taking place in fluids that are self - gravitating and the literature on this has a long history dating back to the work of @xcite on incompressible newtonian uniformly rotating bodies . since then , the study of these instabilities has continued over the years both in newtonian gravity and in full general relativity .
special attention has traditionally been paid to the study of @xmath3 instabilities , which are characterized by the exponential growth of @xmath3 deformations , where @xmath4 parametrizes the azimuthal dependence @xmath5 in a standard mode decomposition in spherical harmonics .
most of the interest in this type of deformation in compact stars stems from the fact that it has the shortest growth time and leads to the emission of a strong gravitational - wave signal .
the development of non - axisymmetric instabilities is commonly analyzed in terms of the quantity @xmath6 ( _ i.e. _ the ratio between the kinetic rotational energy @xmath7 and the gravitational potential energy @xmath8 ) , that provides a dimensionless measure of the amount of angular momentum that can be tapped to feed the development of the instabilities .
this parameter plays an important role in what is possibly the most celebrated of the non - axisymmetric instabilities : the so - called _ dynamical _ bar - mode instability .
this is an @xmath3 instability which takes place when the parameter @xmath1 is larger than a critical one , @xmath9 . in the case of a newtonian incompressible self - gravitating polytrope , for instance
, the dynamical bar - mode instability develops for @xmath10 @xcite and is only weakly dependent on the considered polytropic index or whether the fluid is compressible .
post - newtonian ( pn ) studies @xcite or fully general - relativistic ones @xcite correct this results only slightly , by reducing the threshold to somewhat lower values of the instability parameter . as an example , for a polytropic relativistic star with polytropic index @xmath11 , the accurate calculations reported in @xcite reveal that the critical value is @xmath12 and that a simple dependence on the stellar compactness allows one to track this threshold from the newtonian limit over to the fully relativistic one @xcite .
the onset and development of the bar - mode instability has been traditionally studied by means of nonlinear 3d simulations of newtonian stars that are either unmagnetized @xcite and , more recently , also magnetized @xcite . in addition ,
pn and fully relativistic simulations have been performed and highlighted , for instance , that the persistence of the bar is strongly dependent on the degree of overcriticality and is generically of the order of the dynamical timescale .
furthermore , generic nonlinear mode - coupling effects between the @xmath13 and the @xmath3 mode appear during the development of the instability and these can severely limit the persistence of the bar deformation and eventually suppress the bar deformation @xcite .
these results have been recently confirmed by the perturbative calculations in @xcite . besides dynamical instabilities , which are purely hydrodynamical , _ secular _ instabilities are also possible in rotating compact stars and these are instead triggered by dissipative processes , such as viscosity or radiation emission . if , in particular , the dissipative mechanism is the emission of gravitational radiation , then the secular instability is also known as chandrasekhar - friedman - schutz or cfs instability @xcite .
contrary to what their name may suggest , secular instabilities do not necessarily develop on secular timescales ( although they normally do ) and are characterized by having a much smaller threshold for the instability . once again , in the case of a newtonian polytrope , the critical secular instability parameter is as small as @xmath14 and thus much more easy to attain in astrophysical circumstances .
although widely observed in numerical simulations , the physical conditions leading to a dynamical bar - mode instability are difficult to be encountered in standard astrophysical scenarios .
such large values of the instability parameter , in fact , can not be easily attained in old and cold neutron stars , which have been brought into uniform rotation and thus to rather small values of @xmath1 .
however , more recently these pessimistic prospects have been changed when a new @xmath3 instability has been discovered in differentially rotating newtonian stars @xcite for values of @xmath15 , therefore well below the expected values for a dynamical bar - mode instability .
the most salient aspect of this new instability is that it appears in stars with a large degree of differential rotation and that it grows on a timescale which is longer but comparable with the dynamical one .
this instability has been referred to as the `` low-@xmath2 instability '' and its dependence on the polytropic index and on the degree of differential rotation has been studied in @xcite . since then , the instability has been observed or discussed in a number of related studies @xcite , all of which have highlighted the possible occurrence of this type of instability during the collapse of a massive stellar core . despite the abundant numerical evidence on the development of this instability ,
the nature of these low-@xmath2 instabilities is still matter of debate and , most importantly , a sufficient criterion for its onset has not been derived yet .
this instability has been studied in great detail by watts and collaborators @xcite , who have made a number of phenomenological predictions either using a toy shell - model first introduced in @xcite , or for a stellar model in newtonian gravity .
overall , the work of watts and collaborators ( but see also @xcite ) recognizes the low-@xmath2 instabilities as the manifestation of a more generic class of instabilities , the _ shear instabilities _
@xcite , that is unstable oscillations that do not exist in uniformly rotating systems and are associated to the existence of a corotation band @xcite .
watts , andersson and jones suggest , in particular , that a necessary condition for the development of the instability is `` corotation '' , that is the presence of a point at which the star rotates at the same pattern speed of the unstable mode @xcite .
an alternative suggestion on the necessary conditions has been made also by ou and tohline @xcite , who instead associate the development of the instability to the presence of a minimum in the vortensity profile of the star .
this minimum can then drive unstable not only the corotating @xmath3-modes but also the odd modes such as the @xmath13 and @xmath16-modes @xcite . in this interpretation
, the growth time of the instability is proportional to distance between the corotation radius , _
i.e. _ the radial position at which the unstable mode corotates with the star , and the minimum of the vortensity .
the purpose of this work is to shed some light on the development of shear instabilities and , in particular , to validate one prediction made , although not explicitly , by watts , andersson and jones .
more specifically , we show that , for sufficient amounts of differential rotation , shear instabilities develop for _ any value _ of the instability parameter @xmath1 and also below the expected critical value for the dynamical bar - mode instability .
we therefore provide evidence that the low-@xmath2 instability is not a new instability but rather the manifestation of a shear instability in a region where the latter is possible only for small values of @xmath1 .
our analysis proceeds via the simulation in full general relativity of sequences of neutron star models having constant rest - mass and constant degrees of differential rotation , but with different amounts of rotation , _
i.e. _ with different values of @xmath1 .
the neutron - star matter is described by a realistic equation of state ( eos ) defined by the unified sly prescription @xcite and we study the development of the non - axisymmetric instabilities from their linear growth up to the fully nonlinear development and suppression .
interestingly , we find that depending on the degree of differential rotation , the shear instability leads either to the growth of a single modes ( for the low-@xmath1 models ) or to the simultaneous presence of up to three unstable modes ( for the high-@xmath1 models ) , which produce beatings in the growth of the overall @xmath3 deformation .
special attention is also paid to the properties of the unstable modes and to their position within the corotation band or the vortensity profiles . in this way we are able to confirm both
the necessary conditions proposed so far for the onset of the instability . in particular , we show that all the unstable modes are within the corotation band of the progenitor axisymmetric model ( _ cf .
_ @xcite ) and that all of the unstable models have vortensity profiles with a local minimum ( _ cf .
_ @xcite ) .
the structure of the paper is as follows : in section [ sec : numerics ] we describe the numerical setting of our simulations , the eos we used and the initial models we generated . in section [ sec : methodology ]
we describe the quantities and tools we used to monitor the evolution of the instability . in section
[ sec : results ] we report the results of the simulations and section [ sec : conclusions ] is dedicated to conclusions and discussion .
we use a spacelike signature @xmath17 and a system of units in which @xmath18 ( or in cgs units whenever more convenient ) .
greek indices are taken to run from @xmath19 to @xmath20 , latin indices from @xmath21 to @xmath20 and we adopt the standard convention for the summation over repeated indices .
in what follows we provide a brief overview of the numerical setup used in the simulations , of the realistic eos adopted and on the procedure followed for the construction of the initial axisymmetric models .
we solve numerically the full set of einstein equations @xmath22 where @xmath23 and @xmath24 are the einstein tensor and the stress - energy tensor , respectively .
the equations are solved within the `` 3 + 1 '' decomposition of spacetime , in which the @xmath25-dimensional metric @xmath26 is decomposed into the spatial metric @xmath27 , the lapse function @xmath28 and the shift vector components @xmath29 .
the field equations , which then also provide an evolution for the extrinsic curvature tensor @xmath30 , are then coupled to those of general relativistic hydrodynamics @xmath31 where , in the case of a perfect - fluid , the stress - energy tensor is given by @xmath32 above @xmath33 is the fluid @xmath25-velocity , @xmath34 is the fluid pressure , @xmath35 the specific internal energy and @xmath36 the rest - mass density , so that @xmath37 is the energy density in the rest frame of the fluid .
the set of hydrodynamics equation is then closed by a prescription for the properties of the matter in the form of a relation between the pressure and other quantities in the fluid , _
e.g. _ @xmath38 , and for which we have chosen a cold and realistic eos which will be discussed in the following section .
the evolution of the spacetime was performed using the ` ccatie ` code , a three - dimensional finite - differencing code providing a solution of a conformal traceless formulation of the einstein equations ( see @xcite for the explicit expressions of the equations solved in the code and also @xcite for a more recent and improved implementation ) .
the relativistic hydrodynamics equations , on the other hand , were solved using the ` whisky ` code , which adopts a flux - conservative formulation of the equations as presented in @xcite and high - resolution shock - capturing schemes or hrsc ( see @xcite for the explicit expressions of the equations solved in the code and also @xcite for a more recent extension of the code to mhd ) .
the ` whisky ` code implements several reconstruction methods , such as total - variation - diminishing ( tvd ) methods , essentially - non - oscillatory ( eno ) methods @xcite and the piecewise parabolic method ( ppm ) @xcite .
also , a variety of approximate riemann solvers can be used , starting from the harten - lax - van leer - einfeldt ( hlle ) solver @xcite , over to the roe solver @xcite and the marquina flux formula @xcite ( see @xcite for a more detailed discussion ) .
all the results reported hereafter have been computed using the marquina flux formula and a ppm reconstruction .
both the einstein and the hydrodynamics equations are solved using the vertex - centered adaptive mesh - refinement ( amr ) approach provided by the ` carpet ` driver @xcite .
our rather basic form of amr consists of box - in - box structures centered on the origin of the coordinate system and with the finest grid covering the whole star at all times .
the simulations reported here make use of @xmath25 levels of refinement , with the finest having a resolution of @xmath39 and the coarsest one a resolution of @xmath40 .
the outer boundary was set relatively close to the star and at a distance of @xmath41 , _
i.e. _ at about @xmath42 times the size of the star . a reflection symmetry across the @xmath43 ( equatorial symmetry ) plane was used to reduce the computational costs , but not a rotational one around the @xmath44-axis ( @xmath45-symmetry ) as it would have artificially prevented the growth of odd-@xmath4 modes ( see discussion in @xcite ) .
as mentioned above , the system of hydrodynamics equations needs to be closed by an eos relating the pressure with the other primitive variables , _
e.g. _ the rest - mass density .
previous studies of the bar - mode instability , both in newtonian gravity and in general relativity , have been focused on the use of ideal fluids and analytic eoss , either in the form a of a `` polytropic '' ( and isentropic ) eos @xmath46 , or of an `` ideal - fluid '' and ( non - isentropic ) eos @xmath47 ( _ cf .
_ discussion in @xcite ) .
while these two descriptions are expected to provide results that are qualitatively correct , a more accurate modelling of these instabilities in compact stars necessarily requires a more physically - motivated description of the neutron - star matter .
it is in this spirit that we have here considered a realistic eos , namely the unified sly eos @xcite , which models high - density and cold ( _ i.e. _ zero temperature ) matter via a skyrme effective potential for the nucleon - nucleon interactions .
the sly eos , which describes via a single effective hamiltonian the neutron star s interior , is supplemented with the hp94 eos @xcite to describe the crustal matter and with the bps eos @xcite for lower density regions .
this prescription results in a one - parameter eos in the form @xmath48 , where the sly eos is used for @xmath49 , the hp94 eos is used for @xmath50 and the bps eos for @xmath51 ( see also figure 1 of @xcite ) .
in addition , at even lower densities the eos becomes temperature dependent ( and thus no longer a simple barotropic eos ) , but because these these regions are well below the threshold for the artificial atmosphere , we do not consider an additional prescription for @xmath52 .
we recall , in fact , that our hrsc methods require the use of a tenuous atmosphere which fills the regions of the computational domain not occupied by the compact star .
the threshold value for the rest - mass density of the atmosphere is chosen to be several orders of magnitude smaller than the maximum value and in our simulation a fluid element is considered to be part of the atmosphere if its rest - mass density @xmath36 satisfies @xmath53 .
when this happens the fluid element is treated as a non - dynamical cold fluid described by a polytropic eos , @xmath54 , with @xmath11 and @xmath55 and its velocity is set to zero ( see @xcite for a more detailed discussion on the use of the atmosphere in the ` whisky ` code ) .
the practical implementation of the realistic eos can take place in a number of different ways .
the simplest is to use standard interpolation techniques , _
e.g. _ based on lagrangian polynomials , on the values of the published tables .
while straightforward , the interpolations in this approach do not guarantee in general that the thermodynamics relations are fulfilled ( see @xcite for a thermodynamical preserving interpolation ) .
in addition , the derivatives of the fields , _ e.g. _ of the pressure to evaluate the sound speed , are typically not available in the tables . furthermore , the use of high - order interpolation and/or finite differences can lead to undesirable spurious oscillations . a second approach that removes
all of these problems , uses analytic fits that have been proposed for the pressure . as an example , ref .
@xcite suggested to fit the specific internal energy of the unified sly eos table with the expression @xmath56 where @xmath57 , @xmath36 and @xmath35 are in cgs units , and the coefficients @xmath58 are @xmath59 ( see table @xmath60 of @xcite ) .
equation ( [ eq : analfit ] ) is obtained from eq .
( 15 ) of @xcite after using @xmath61 , where @xmath62 is the baryon number density and @xmath63 is the mass of the nucleons . as discussed in @xcite , it is then possible to compute the pressure from the value of @xmath35 using the first principle of thermodynamics at @xmath64 @xmath65 and thus to have an evaluation of the pressure which is thermodynamically consistent .
the differences between the fit and the table are typically less then @xmath66 .
unfortunately , although apparently very convenient , the evaluation of the fitting formulas containing several exponential and logarithmic functions , turns out to be computationally rather expensive even if done in a optimized way . as a third approach , which combines the efficiency of a table search with the thermodynamical consistency of an analytic fit ,
consists of performing a simple linear interpolation among the tabulated values constructed from the analytic fit . besides being highly efficient
, a linear interpolation also eliminates the spurious oscillations that arise , for instance , in the derivative of the pressure if high - order interpolation formulas are used . in this case , the interpolation error can be reduced simply by populating the analytically constructed tables with a large number of entries , _ e.g. _
@xmath67 in place of the @xmath68 which are typically available in published tables .
this third approach is the one actually implemented in ` whisky ` and provides a speed up of about @xmath69 with respect to the evaluation of the pressure via the analytic fits and with comparable accuracy .
the initial data for our simulations are prepared as stationary and axisymmetric equilibrium solutions for rapidly rotating relativistic stars @xcite . adopting spherical quasi - isotropic coordinates ,
the line element of the corresponding spacetime is @xmath70 where @xmath71 , @xmath72 , @xmath73 and @xmath74 are functions of @xmath75 and @xmath76 .
moreover we assume the usual relativistic @xmath77-constant law of differential rotation and that amounts to assume an angular - velocity distribution of the form @xmath78 \ , , \label{eq : velocityprofile_gr}\ ] ] where @xmath79 is the coordinate equatorial stellar radius and the coefficient @xmath80 provides a measure of the degree of differential rotation .
expression ( [ eq : velocityprofile_gr ] ) represents the general - relativistic equivalent of the simpler newtonian @xmath77-constant law @xcite
@xmath81 clearly , @xmath82 corresponds to a star in uniform rotation , while @xmath83 corresponds to a star with increasing degree of differential rotation . as a reference
, @xmath84 yields a star with an angular - velocity profile which varies of a factor @xmath85 between the center and the surface of the star ( _ cf .
_ left panel of figure [ fig : ida ] ) . in practice ,
we have computed a very large number of initial models using the sly prescription for the eos for which we have computed _ baryonic _ mass @xmath86 , the gravitational mass @xmath87 , the angular momentum @xmath88 , the rotational kinetic energy @xmath7 and the gravitational binding energy @xmath8 defined as @xmath89 where @xmath90 is the square root of the determinant of three - dimensional metric @xmath27 and @xmath91 is the fluid lorentz factor .
we stress that the definitions ( [ eq : def m])([eq : def w ] ) of quantities such as @xmath88 , @xmath7 , @xmath8 and @xmath1 are meaningful only in the case of stationary axisymmetric configurations and should therefore be treated with care once the rotational symmetry is lost .
plane of the initial models computed with @xmath92 , with the filled circles representing those we have evolved numerically ( a similar behaviour is shown also by the models with @xmath93 .
indicated with solid thin lines are isocontours of constant baryon mass models while indicated with a thick dashed line is the threshold to the dynamical bar - mode instability as computed for a @xmath11 polytrope @xcite .
note that the threshold for the instability tends to increase for smaller rest - masses .
right panel : the same initial models as in the left panel but shown in @xmath94 planes with isocontours of constant @xmath1 ( upper part ) or constant @xmath95 ( lower part ) .
see table [ table : initial ] for a summary of the properties of the initial models.,title="fig:",scaledwidth=49.0% ] plane of the initial models computed with @xmath92 , with the filled circles representing those we have evolved numerically ( a similar behaviour is shown also by the models with @xmath93 . indicated with solid thin lines are isocontours of constant baryon mass models while indicated with a thick dashed line is the threshold to the dynamical bar - mode instability as computed for a @xmath11 polytrope @xcite .
note that the threshold for the instability tends to increase for smaller rest - masses .
right panel : the same initial models as in the left panel but shown in @xmath94 planes with isocontours of constant @xmath1 ( upper part ) or constant @xmath95 ( lower part ) .
see table [ table : initial ] for a summary of the properties of the initial models.,title="fig:",scaledwidth=49.0% ] -.5 cm .main properties of the simulated stellar models .
starting from the left : the name of the model , the differential rotation parameter @xmath80the instability parameter @xmath1 , the compactness @xmath95 , the central rest - mass density @xmath96 , the maximum of the density @xmath97 ( note that for models with @xmath93 , @xmath98 . ) , the ratio between the polar and the equatorial coordinate radii @xmath99 , the proper equatorial radius @xmath100 , the gravitational mass @xmath87 , the total angular momentum @xmath88 divided by the square of the gravitational mass . [ cols="^,^,^,^,^,^,^,^,^,^ " , ] out of this large set , we have then selected for the numerical evolution a number of models so as to build sequences of constant baryonic mass with @xmath101 , degree of differential rotation @xmath102 and values of the instability parameter @xmath1 ranging between @xmath103 and @xmath104 . each model in this sequence is supramassive , namely it has a mass which is larger than the maximum mass allowed for a corresponding nonrotating model , _
@xmath105 , although it is not hypermassive , namely it does not have mass which is above the maximum mass for a uniformly rotating model , _
i.e. _ @xmath106 .
there are two different reasons why such a large - mass model has been chosen .
the first one is that we are interested in the development of a shear instability in the metastable star produced by the merger of a binary system of two neutron stars .
as shown by a number of authors and most recently in @xcite , the product of this merger is either a supramassive or a hypermassive neutron star .
hence , our reference model has a mass which is sufficiently large so as to be a reasonable approximation to the product of a binary neutron star merger .
the second reason is that a sufficiently massive model is necessary in order to reach values of the instability parameter which are above the expected threshold for dynamical bar - mode instabilities as computed in @xcite , _ i.e. _ @xmath107 .
indeed , the maximum possible value for @xmath1 within the computed sequence is around @xmath108 and thus just above the threshold ( _ cf . _ long - dashed line in the left panel of figure [ fig : idb ] ) .
note also that the threshold for the instability tends to increase for smaller rest - masses .
overall , when comparing with equilibrium models generated with a polytropic eos with @xmath55 and @xmath11 ( see table 1 of @xcite ) the realistic eos models reach higher compactness ( models in @xcite typically have @xmath109 ) but lower values of @xmath1 ( _ i.e. _ @xmath110 for the models considered here , while @xmath111 for the polytropic models considered in @xcite ) .
the whole space of parameters is shown in figure [ fig : idb ] , whose left panel reports the position in the @xmath112 plane of the initial models computed with @xmath92 , and where the filled circles represent those we have evolved numerically . indicated with solid thin lines are isocontours of constant baryon mass models while indicated with a thick dashed line is the threshold to the dynamical bar - mode instability as computed for a @xmath11 polytrope @xcite .
the right panel reports the same initial models considered in the left one but shown in @xmath113 planes with isocontours of constant @xmath1 ( upper part ) or constant @xmath95 ( lower part ) .
the main properties of the simulated models are also reported in table [ table : initial ] , where we also introduce our naming convention .
any initial model is indicated as ` m.% .
# ` , with ` % ` being replaced by the value of the differential - rotation parameter @xmath114 and ` # ` by the instability parameter @xmath1 . as an example , ` m.1.200 ` is the star with @xmath92 and @xmath115 . finally , shown in figure [
fig : ida ] are the angular - velocity profiles ( left panel ) and the rest - mass density profiles ( right panel ) of some representative models , namely ` m.1.150 ` , ` m.1.200 ` and ` m.1.250 ` . indicated with different symbols , which match the ones in the left panel of figure [ fig : corotation_band ] ,
are the normalized radial positions of the corotation radii , with the one for model ` m.1.150 ` being shown filled to help distinguish it from the others .
although we postpone to sections [ sec : evolution ] and [ sec : necessary conditions ] the discussion of the implications of these corotation radii , two aspects of the initial data are worth emphasizing .
the first one is that the amount of differential rotation for a given value of @xmath114 effectively decreases when increasing the instability parameter @xmath1 ( _ cf .
_ left panel of figure [ fig : ida ] ) , thus resulting in a smaller corotation band for models with large @xmath1 .
the second one is that all the initial models evolved are axisymmetric but have a `` toroidal - topology '' , namely have the maximum density @xmath97 that is not at the center of the star , and thus @xmath116 .
this toroidal deformation increases with the rotation and thus with @xmath1 . and
@xmath92 ( a similar behaviour is shown also by the models with @xmath93 ) .
indicated with different symbols , which match the ones in the left panel of figure [ fig : corotation_band ] and [ fig : corot_rad ] , are the normalized radial positions of the corotation radii , with the one for model ` m.1.150 ` being shown filled to help distinguish it .
right panel : the same as in the left panel but for the initial rest - mass density.,title="fig:",scaledwidth=49.5% ] and @xmath92 ( a similar behaviour is shown also by the models with @xmath93 ) . indicated with different symbols , which match the ones in the left panel of figure [ fig : corotation_band ] and [ fig : corot_rad ] , are the normalized radial positions of the corotation radii , with the one for model ` m.1.150 ` being shown filled to help distinguish it .
right panel : the same as in the left panel but for the initial rest - mass density.,title="fig:",scaledwidth=49.5% ] -.5 cm
a number of different quantities are calculated during the evolution to monitor the dynamics of the instability . among them
is the quadrupole moment of the matter distribution @xmath117 which we compute in terms of the conserved density @xmath118 rather than of the rest - mass density @xmath36 or of the @xmath119 component of the stress energy momentum tensor .
of course , the use of @xmath120 in place of @xmath36 or of @xmath119 is arbitrary and all the three expressions would have the same newtonian limit , though with different amplitudes for the gravitational waveforms produced ( see for example @xcite ) .
however , we here adopt the form ( [ eq : defquadrupole ] ) because @xmath121 is a quantity whose conservation is guaranteed by the form chosen for the hydrodynamics equations . the quadrupole moment ( [ eq : defquadrupole ] )
can be conveniently used to quantify both the growth time of the @xmath3 instability @xmath122 and the oscillation frequency once the instability is fully developed @xmath123 .
( hereafter we will indicate respectively with @xmath124 and @xmath125 the growth time and frequencies of the @xmath126 unstable modes and we note that , as will be discussed later on , during the simulation a number of different modes appear , thus justifying the use of the upper index `` @xmath127 '' ) . in practice
, we perform a nonlinear least - square fit of the @xmath128 component of the computed quadrupole @xmath129 and we generally use as fitting function a sum of @xmath130 ( usually three ) exponentially modulated cosines @xmath131 where @xmath132 . because we commonly have only about @xmath133 cycles in the time interval considered for the fits , extreme care needs to be applied when computing the growth time , especially when the oscillation frequencies and the growth times are close to each other . in these cases , in fact , variations of the initial phase of the modes @xmath134 can result in large variation of the growth times . in view of this , we will not report them . using three components of the quadrupole moment in the @xmath135 plane
we can define the distortion parameters @xmath136 and @xmath137 , as well as the axisymmetric mode @xmath138 as @xmath139 so that the modulus @xmath140 and the instantaneous orientation of the bar are given by @xmath141 finally , as a useful tool to describe the nonlinear properties of the development and saturation of the instability , the rest - mass density is decomposed into its fourier modes @xmath142 as @xmath143 the phase @xmath144 essentially provides the instantaneous orientation of the @xmath4-th mode when the corresponding mode has a nonzero power . note that despite their denomination , the fourier modes ( [ eq : modes ] ) do not represent proper eigenmodes of oscillation of the star . while , in fact , the latter are well defined only within a perturbative regime ,
the former simply represent a tool to quantify , within the fully nonlinear regime , what are the main components of the rest - mass distribution . as a final comment
we note that while all quantities ( [ eq : defquadrupole])([eq : phimodes ] ) are expressed in terms of the coordinate time @xmath145 and are not invariant measurements , the lengthscale of variation of the lapse function at any given time is always larger than twice the stellar radius at that time , ensuring that events on the same timeslice are also close in proper time .
in what follows we first describe the dynamics of the shear instability as deduced from the numerical simulations and then contrast our results with the phenomenological predictions on the necessary conditions for its development . . for each panel
the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 .
see text for details .
note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] 0.5 cm . for each panel
the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 .
see text for details .
note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] . for each panel
the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 .
see text for details .
note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] 0.5 cm . for each panel
the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 .
see text for details .
note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] . for each panel
the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 .
see text for details .
note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] 0.5 cm . for each panel
the upper part reports the evolution of the bar - distortion parameter @xmath146 , while the lower part shows the evolution of the power in the different modes of the fourier decomposition of the rest - mass density @xmath147 .
see text for details .
note that we here report only some models and with @xmath92 as they are representative also of those with @xmath93.,title="fig:",scaledwidth=42.5% ] -0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area
is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] 0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area
is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area
is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] 0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area
is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area
is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] 0.5 cm for the models described in figure [ fig : evolutiongeneral ] ( black continuous line ) and of the axisymmetric mode @xmath148 ( red dashed line ) . indicated with vertical lines ( long - dashed black , dotted blue and dot - dashed magenta ) are the peak frequencies reported in the left panel of figure [ fig : corotation_band ] within the corotation band and which are used in figure [ fig : corot_rad ] to mark the position of the corotation radii . finally , shown with a shaded rectangular area
is the corotation band for all models.,title="fig:",scaledwidth=42.5% ] -0.5 cm as mentioned when discussing the initial data , we have evolved numerically two sequences of constant baryonic mass , with @xmath149 .
the first sequence has a higher degree of differential rotation ( _ i.e. _ @xmath92 ) and instability parameter ranging from @xmath150 to @xmath151 .
the second sequence , on the other hand , has a smaller degree of differential rotation ( _ i.e. _ @xmath152 ) and instability parameter between @xmath153 and @xmath115 . for both sequences
the highest value of @xmath1 considered is also very close to the highest attainable with our initial data code ( see figure [ fig : idb ] ) . to keep the computational costs to an affordable level we have evolved of all these models for @xmath154 and although this time window is in general insufficient to capture the suppression of the instability , is adequate to measure the frequencies of the unstable modes and provide a first estimate of the growth times . after analyzing the results following the method outlined in the previous section and focusing on the properties of the distortion parameters @xmath146 , @xmath155 and @xmath148
, we find that _ all of the models _ show an @xmath3 instability .
the maximum distortions obtained during the simulation time are @xmath156 ( _ i.e. _ distortions of @xmath157 with respect to the axisymmetric progenitor ) , with the maximum values being reached for the models of the @xmath92 sequence .
the models of the @xmath93 sequence , in fact , have in general much smaller distortions , with @xmath158 over the timescale over which they have been evolved .
we believe this is simply the consequence of the fact that the models in this sequence have smaller growth rates ( see also the discussion in the next section ) . to show the general behaviour of the simulation we focus on the description of the result of the simulation for the @xmath92 sequence as this is representative also of the one with @xmath93 .
in particular , in the upper part of each panel in figure [ fig : evolutiongeneral ] we show the time evolution of the distortion parameter @xmath159 computed with eq .
( [ eq : etas ] ) for six representative models : ` m.1.150 ` , ` m.1.160 ` , ` m.1.200 ` , ` m.1.220 ` , ` m.1.241 ` and ` m.1.250 ` .
similarly , in the lower parts of each panel we show the corresponding evolution of the fourier - modes computed from eq .
( [ eq : modes ] ) for @xmath160
. note that for models with lower values of @xmath1 the bar - mode deformation is very similar to the one already discussed in @xcite , growing exponentially and with only one unstable mode appearing ( _ cf .
_ first row of figure [ fig : evolutiongeneral ] ) .
however , as the rotation is increased , the development of the instability is more complex and at least two unstable modes appear which develop in different parts of the star .
these two modes have very similar but distinct frequencies and growth times , leading to a series of beatings in the evolutions of @xmath159 , whose irregular evolution makes the calculation of the growth times challenging ( _ cf . _
second row ) .
as the rotation is increased towards the maximal values of @xmath1 , up to three distinct modes appear and the evolution of the instability is correspondingly more complex ( _ cf . _ third row ) .
a similar behaviour is shown also by the evolution of the fourier modes .
the two models with lower @xmath1 , in fact , show a clear growth of the @xmath3 mode and of the @xmath13 , with the latter becoming first comparable and then larger when the instability is suppressed . in models ` m.1.250 , m.1.241 ` , on the other hand , the @xmath3 and @xmath16 modes have comparable amplitude for a long period and then the @xmath3 becomes the dominant one ( this is more clear in model ` m.1.241 ` ) .
finally , for larger values of @xmath1 , the @xmath13 mode never attains values comparable with either the @xmath3 or the @xmath16 , which instead control the evolution .
figure [ fig : psd ] reports the power spectral density ( psd ) of @xmath146 and @xmath148 for the same @xmath161 models ( solid black lines ) and allows to appreciate how the spectrum changes as the instability parameter is increased .
more specifically , it is very apparent that at low @xmath1 the spectra have only one peak , whose maximum is marked by the vertical blue dotted lines and that is present at all the values of beta considered ; this is the mode we refer to as @xmath162 ( _ cf . _ table [ table : results ] and figure [ fig : corotation_band ] ) . as @xmath1 increases , the amplitude of this peak decreases and it becomes the weakest for very large rotation rates .
starting from model ` m.1.200 ` ( although a hint of a peak is present already in ` m.1.160 ` ) , a second peak appears at higher frequency and is marked with magenta vertical dot - dashed lines ; this is the mode we refer to as @xmath163 and although it never becomes the largest one , its amplitude increases with @xmath1 .
finally , a third peak appears at low frequency in the spectra starting from model ` m.1.220 ` ( even though a hint is present also for model ` m.1.210 ` ) and is marked with black vertical long - dashed lines .
this mode , which we refer to as @xmath164 , becomes the dominant one at high @xmath1 .
note also that in each panel of figure [ fig : psd ] we report with a dashed blue box the corotation band that will be further discussed in the following section . as a final remark we note that the peaks in the spectrum of @xmath165 ( red dashed lines in figure [ fig : psd ] ) are related to axisymmetric @xmath166 modes and most likely to @xmath167-modes oscillation excited by the development of the instability .
indeed , their frequencies @xmath168 match reasonably well the phenomenological fit for the @xmath167-mode frequencies of nonrotating neutron stars with realistic eoss computed in @xcite .
because our sequences do not contain nonrotating configurations ( the rest mass is larger than the maximum one ) , this association is just qualitative and a more detailed investigation of the frequency spectra of the equilibrium models is necessary to confirm this suggestion .
@cccccccccc model & @xmath169 & @xmath170 & @xmath171 & @xmath172 & @xmath173 & @xmath174 & @xmath175 & @xmath176 & @xmath177 + & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) & ( khz ) + [ 1 mm ] ` m.1.140 ` & @xmath178 & @xmath179 & @xmath180 & @xmath179 & @xmath179 & @xmath181 & @xmath179 & @xmath182 & @xmath183 + ` m.1.150 ` & @xmath184 & @xmath179 & @xmath181 & @xmath179 & @xmath179 & @xmath185 & @xmath179 & @xmath186 & @xmath187 + ` m.1.160 ` & @xmath188 & @xmath179 & @xmath189 & @xmath190 & @xmath179 & @xmath191 & @xmath192 & @xmath193 & @xmath194 + ` m.1.180 ` & @xmath195 & @xmath179 & @xmath196 & @xmath197 & @xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 + ` m.1.200 ` & @xmath203 & @xmath204 & @xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 + ` m.1.210 ` & @xmath212 & @xmath213 & @xmath214 & @xmath215 & @xmath216 & @xmath217 & @xmath218 & @xmath219 & @xmath220 + ` m.1.220 ` & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 & @xmath226 & @xmath227 & @xmath228 & @xmath229 + ` m.1.230 ` & @xmath230 & @xmath231 & @xmath232 & @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 + ` m.1.241 ` & @xmath239 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 & @xmath245 & @xmath246 & @xmath247 + ` m.1.244 ` & @xmath248 & @xmath249 & @xmath250 & @xmath251 & @xmath252 & @xmath253 & @xmath254 & @xmath255 & @xmath256 + ` m.1.247 ` & @xmath257 & @xmath258 & @xmath259 & @xmath260 & @xmath261 & @xmath262 & @xmath263 & @xmath264 & @xmath265 + ` m.1.250 ` & @xmath266 & @xmath267 & @xmath268 & @xmath269 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath274 + ` m.2.125 ` & @xmath275 & @xmath276 & @xmath277 & @xmath179 & @xmath278 & @xmath279 & @xmath179 & @xmath280 & @xmath281 + ` m.2.150 ` & @xmath282 & @xmath283 & @xmath284 & @xmath179 & @xmath285 & @xmath286 & @xmath179 & @xmath287 & @xmath288 + ` m.2.175 ` & @xmath289 & @xmath290 & @xmath291 & @xmath179 & @xmath292 & @xmath293 & @xmath179 & @xmath294 & @xmath295 + ` m.2.200 ` & @xmath296 & @xmath297 & @xmath298 & @xmath179 & @xmath299 & @xmath300 & @xmath179 & @xmath301 & @xmath302 + to support the interpretation of these instabilities as shear instabilities we have considered whether the necessary conditions suggested by @xcite and by @xcite are met .
we recall that watts and collaborators pointed out that an unstable configuration should have the unstable mode with a frequency within the corotation band . in the case of the newtonian expression ( [ eq : velocityprofile_newt ] ) , the corotation band is simpler to compute and for a mode with frequency @xmath303 and azimuthal number @xmath4 , this is simply given by @xmath304 we have therefore checked whether any of the unstable @xmath3 ( bar - modes ) with computed frequency @xmath305 has pattern speed velocity , @xmath306 , within the corotation band .
a careful and rather involved analysis has indeed confirmed the prediction of watts and collaborators : namely , _ all the unstable modes are within the corotation band _ of the progenitor axisymmetric model .
these results for the models with @xmath92 are listed in table [ table : results ] , where we report the edges of the general - relativistic corotation band in terms of angular frequency at the surface @xmath307 and on the axis @xmath177 , as well as the frequencies of the unstable @xmath3 modes obtained from the position of the peaks in the psd ( _ i.e. _ @xmath308 ) or from the nonlinear fitting of eq .
( [ eq : quadrupolefit ] ) ( _ i.e. _ @xmath309 ) . note that depending on the rate of rotation and the degree of differential rotation some frequencies may not be present in the corotation band . the data in table [ table : results ] is also shown in the left panel of figure [ fig : corotation_band ] , where we plot the position of the non - axisymmetric frequencies @xmath123 within the corotation band as a function of @xmath1 ; indicated with a filled symbol , to distinguish it from the others , is the @xmath310 frequency for model ` m.1.150 ` ( _ cf .
_ figure [ fig : ida ] and [ fig : corot_rad ] ) .
the right panel of the same figure shows the corresponding information and on the same scale but for the models with @xmath93 ( the inset shows instead a magnified view ) . clearly , for both sequences all the unstable modes are within the band and , as the stellar rotation rate increases
, more unstable modes appear for the same value of @xmath1 . for models with @xmath92 and shown as a function of @xmath1 ; indicated with thick solid lines are the edges of the corotation band , _
i.e. _ the frequency interval between @xmath177 and @xmath176 . shown with a filled symbol , to help distinguish it from the others , is the @xmath164 frequency for model ` m.1.150 ` ( _ cf .
_ figure [ fig : ida ] and [ fig : corot_rad ] ) .
right panel : the same as in the left one and with the same scale but for models with @xmath93 ; the inset shows instead a magnified view .
clearly , in both panels all the unstable modes are within the band and up to three unstable modes appear for the @xmath92 sequence with increasing @xmath1.,title="fig:",scaledwidth=49.5% ] for models with @xmath92 and shown as a function of @xmath1 ; indicated with thick solid lines are the edges of the corotation band , _
i.e. _ the frequency interval between @xmath177 and @xmath176 . shown with a filled symbol , to help distinguish it from the others , is the @xmath164 frequency for model ` m.1.150 ` ( _ cf .
_ figure [ fig : ida ] and [ fig : corot_rad ] ) .
right panel : the same as in the left one and with the same scale but for models with @xmath93 ; the inset shows instead a magnified view .
clearly , in both panels all the unstable modes are within the band and up to three unstable modes appear for the @xmath92 sequence with increasing @xmath1.,title="fig:",scaledwidth=49.5% ] -.5 cm watts , andersson and jones also give a qualitative description , summarized in their figure 2 of @xcite , of how the unstable and stable models should be distributed in the ( @xmath311 ) plane .
the considerations they make are particularly simple : for high degrees of differential rotation , that is at low values of @xmath80 , the corotation band is rather wide and there will be both an upper value and a lower value of @xmath1 between which the shear instability can develop ( these critical values of @xmath1 correspond to the entrance and exit of the unstable mode in the corotation band ) .
this situation corresponds therefore to the one commonly encountered in numerical simulations , such as those in @xcite , and for which the shear instability takes place only for very small values of the instability parameter and on timescales that are much longer than the dynamical one .
when moving to larger degrees of differential rotation , that is when going to smaller values of @xmath80 , the corotation band becomes larger and larger and the shear instability can develop essentially for all values of @xmath1 , merging with the dynamical bar - mode instability for @xmath312 .
this is exactly what has been found here .
conversely , when moving to smaller degrees of differential rotation , that is when going to higher values of @xmath80 , the corotation band becomes thinner and the shear instability can develop only for a smaller range of @xmath1 . as the differential rotation is further decreased and the star tends to rotate uniformly , the corotation band width vanishes , all the models are stable to the shear instability and subject only to the dynamical bar - mode instability for @xmath313 .
this is indeed the case for the unstable models evolved in @xcite , which were purely ( bar - mode ) dynamically unstable and none of which had the unstable mode within the corresponding corotation band , but above it .
in essence , therefore , there should be an intermediate range of @xmath80 for which the instability is absent at low @xmath1 , appears at intermediate values and then disappears again at high @xmath1 , thus defining an interval of values of @xmath1 for which the models are unstable . to validate also this prediction we simulated a sequence of models with smaller differential rotation and @xmath152
this seconds sequence has the same baryonic mass as that with @xmath314 , but with @xmath1 in a smaller range , namely between @xmath153 and @xmath115 ( note that @xmath115 is also very close to the largest value for which we could build an equilibrium model . ) unfortunately , also all of these models show an @xmath3 shear instability , with the unstable frequencies falling within the corotation band , as it was for the @xmath314 stars .
of course , lack of evidence is not evidence of absence and the fact that we have not found stable models within our range of values of @xmath1 most likely means that these stable models have to be searched either for values of @xmath315 ( for this value of @xmath152 ) , or by moving towards higher values of @xmath80 , where the corotation band is less large .
an obvious consequence of the phenomenological scenario described by @xcite and confirmed by the calculations reported here is that the idea of a low-@xmath2 instability is indeed misleading and it is instead more meaningful to think of a more generic shear instability that , depending on the degree of rotation and of differential rotation , may manifest itself on timescales that are comparable with the dynamical ones ( as in the cases reported here ) or on much longer ones ( as in the cases reported in @xcite ) .
another phenomenological prediction made by watts , andersson and jones in @xcite , is that the growth times should be shorter in the center of the band and increase towards the edges ( the growth times should in fact diverge at the edges ) .
as shown in figure [ fig : corotation_band ] , all of the simulated models do have unstable modes well inside the corotation band and this probably explains why we observe them develop on a timescale which is comparable with the dynamical one and not on much longer timescales as in the original finding of @xcite .
we also find indications that the @xmath152 models , which generically have longer growth times , have unstable modes which occupy regions of the corotation band that are overall more central ( _ cf .
_ right panel of figure [ fig : corotation_band ] ) and thus in contrast with what expected from @xcite .
however , the difficulties mentioned above in computing an accurate estimate of the growth times and the intrinsic difficulties in determining what is the central part of the corotation band , prevent us from providing a more quantitative validation of this prediction .
we now switch to consider our results in the light of the other necessary condition for the onset of the shear instability discussed in @xcite .
as mentioned in the introduction , it has in fact been shown that the presence of a minimum in the profile of the vortensity is a necessary condition for a mode in corotation to be unstable @xcite .
the intuitive description is that the vortensity well can act as a resonant cavity inside the star , amplifying the modes that happen to lay near its minimum @xcite . indeed
, the growth rate of the unstable mode is expected to depend on the location of its corotation radius with respect to the vortensity profile , being proportional to the depth of its corotation radius inside the vortensity well @xcite .
note that the mere existence of a local minimum in the vortensity can not be used as a sufficient condition for the occurrence of a shear instability .
all of the unstable models considered in @xcite , in fact , do show a local minimum in the vortensity but do not have the unstable modes in the corotation band .
for three representative models with small , medium and high values of @xmath1 .
the different symbols match the ones in figure [ fig : ida ] and in the left panel of figure [ fig : corotation_band ] , with the one for model ` m.1.150 ` reported as filled , and show the actual position of the corotation radii . note
that none of these coincides with the minimum of @xmath316 ; for compactness we have reported only the models with @xmath92 .
right panel : normalized corotation radii for the different frequencies @xmath123 presented in the left panel of figure [ fig : corotation_band ] shown as a function of @xmath1 ; the same convention of the previous figures is used for the different symbols.,title="fig:",scaledwidth=49.5% ] for three representative models with small , medium and high values of @xmath1 .
the different symbols match the ones in figure [ fig : ida ] and in the left panel of figure [ fig : corotation_band ] , with the one for model ` m.1.150 ` reported as filled , and show the actual position of the corotation radii . note
that none of these coincides with the minimum of @xmath316 ; for compactness we have reported only the models with @xmath92 .
right panel : normalized corotation radii for the different frequencies @xmath123 presented in the left panel of figure [ fig : corotation_band ] shown as a function of @xmath1 ; the same convention of the previous figures is used for the different symbols.,title="fig:",scaledwidth=49.5% ] -.5 cm to validate whether this newtonian condition holds also for the general - relativistic instabilities simulated here , we have computed for all the models the newtonian vortensity , which is defined as the ratio , along the radial cylindrical coordinate , between the radial vorticity and the density @xmath317 where @xmath318 is the radial cylindrical coordinate ( a fully general - relativistic definition of the vortensity is also possible but more complicated to compute and not significantly different from the newtonian one ) .
when doing this we found that all the unstable models have vortensities with a local minimum in the star and this is shown in the left panel of figure [ fig : corot_rad ] , which reports the radial profiles of the vortensity @xmath316 for three representative models with small , medium and high values of @xmath1 .
the different symbols match the ones in the left panel of figure [ fig : ida ] and [ fig : corotation_band ] , with the one for model ` m.1.150 ` reported as filled , and show the actual positions of the corotation radii .
the right panel of the same figure , on the other hand , shows the normalized corotation radii for the different frequencies @xmath123 presented in the left panel of figure [ fig : corotation_band ] as a function of @xmath1 . note
that none of these coincides systematically with the minimum of @xmath316 , nor with the maximum of the rest - mass density .
all of the corotation radii , however , do move towards larger radial positions as the rotation rate is increased , exactly as does the minimum of the vortensity and the rest - mass maximum . although we can not confirm that modes whose corotation radius is closer to the minimum of the vortensity have systematically shorter growth times ( our data is not sufficiently accurate for this ) , we can compare the models with the higher degree of differential rotation @xmath92 with those having a smaller degree of differential rotation @xmath93 .
in the first case the vortensity well is considerably deeper , with @xmath319 , while for the latter the vortensity has shallower wells with @xmath320 .
since the models with @xmath321 have smaller growth rates , our results indicate therefore that also the existence of a local minimum in the vortensity can be taken as a necessary condition for the development of the shear instability and that the depth of the vortensity well can be used to estimate the growth of the instability .
for many years the properties of rapidly rotating and self - gravitating fluids have been characterized by a complete analytic perturbative theory which provided , for instance , sufficient conditions for the development of instabilities ( see , for instance , @xcite for a collection of results ) . as numerical simulations have become increasingly accurate and stable on longer timescales , these predictions , both in newtonian theory and in full general relativity ,
have been verified , corrected and in some cases extended .
our ability of modelling such configurations has now reached a maturity such that a number of new properties and instabilities have been `` discovered '' numerically , but which do not have behind a fully perturbative description .
the main reason for this is that most of this phenomenology is the result of physical scenarios which are much more complex than the ones investigated perturbatively in the past , _
e.g. _ non - isolated systems with high differential rotation and exchanging mass and angular momentum , which are much more difficult to treat analytically .
a most notable example of these complex and yet ubiquitous instabilities is the so - called `` low-@xmath2 instability '' , which was initially found in @xcite and then reproduced in a number of other different scenarios @xcite .
as the phenomenological description of this instability provided by the simulations has become richer and richer , a full understanding of the mechanisms that lead to its development has lagged behind and it is presently unclear .
we are therefore in a situation in which numerical simulations can probe regimes and conditions which are not yet accessible to perturbative calculations , and can guide the latter by confirming or refuting those scenarios that although possible do not find a realization in practice .
this work has followed this spirit and has used fully general - relativistic calculations of rapidly and differentially rotating neutron stars modeled with a realistic eos to shed some light on the development of the low-@xmath2 instability . in particular we have concentrated our attention on validating an indirect prediction , made by watts , andersson and jones @xcite , who recognized the low-@xmath2 instability as the manifestation of a more generic class of instabilities associated to the existence of a corotation band @xcite , the _ shear instabilities _ , and that should develop for _ any value _ of the instability parameter @xmath1 when sufficient amounts of differential rotation are present .
this is exactly what we have found in our simulations .
more specifically , we have performed simulations of sequences of neutron - star models described by a realistic sly eos @xcite and having constant rest - mass and degrees of differential rotation , but with different amounts of rotation . in all cases
considered we have found the development of a bar - mode instability growing on a dynamical timescale , even when the initial axisymmetric models were well below the critical limit for the dynamical bar - mode instability .
these results , which match well the phenomenological scenario portrayed in @xcite , suggest therefore that the idea of a low-@xmath2 instability is indeed misleading and should be replaced by the more general one of shear instability . depending then on the degree of rotation and of differential rotation
, the instability will develop on timescales that are comparable to the dynamical one ( as reported here ) or on much longer ones ( as reported in the first low-@xmath2 instability studies ) .
special attention has also been paid to the properties of the unstable modes and to their position within the corotation band or the vortensity profiles . in particular , we have shown that all the unstable modes are within the corotation band of the progenitor axisymmetric model ( which is the necessary condition for the development of the instability proposed by @xcite ) and that all of the unstable models have vortensity profiles with a local minimum ( which is the necessary condition suggested by @xcite ) .
finally , by comparing the growth times among models with different degree of differential rotation we have shown that there is a correlation , although not a strong one , between the depth of the vortensity well and growth rate of the instability , with the latter being larger for models with deeper wells . in summary ,
the results presented here shed some light on several aspects of shear instabilities , but they also reveal that more work is required , for instance , to distinguish between the predictions based on the corotation band and the ones based on the vortensity well , or to establish whether in effect they just represent two different ways of expressing the same physical conditions .
clarifying these aspects will requires additional analytical and numerical modelling , and this will be part of our future research .
it is a pleasure to thank anna watts for enlightening discussions and for sharing her notes with us , as well as nils andersson , ian jones , joel tohline for useful comments .
we are grateful to nikolaous stergioulas for the code to compute the initial data used in these calculations and to e. schnetter and all the ` carpet / cactus ` developers .
the computations were mainly performed on the damiana cluster at the aei and , in a part , on the parma s albert2 cluster and on the bcx cluster at cineca .
this work was supported in part by the dfg grant sfb / transregio 7 `` gravitational - wave astronomy '' , by `` compstar '' , a research networking programme of the european science foundation , and by the della riccia foundation .
10 url # 1#1urlprefix[2][]#2 f and haensel p 2001 _ astronomy and astrophysics _ * 380 * 151167
|
we present new results on instabilities in rapidly and differentially rotating neutron stars .
we model the stars in full general relativity and describe the stellar matter adopting a cold realistic equation of state based on the unified sly prescription @xcite .
we provide evidence that rapidly and differentially rotating stars that are below the expected threshold for the dynamical bar - mode instability , @xmath0 , do nevertheless develop a shear instability on a dynamical timescale and for a wide range of values of @xmath1 .
this class of instability , which has so far been found only for small values of @xmath1 and with very small growth rates , is therefore more generic than previously found and potentially more effective in producing strong sources of gravitational waves .
overall , our findings support the phenomenological predictions made by watts , andersson and jones @xcite on the nature of the low-@xmath2 instability as the manifestation of a shear instability in a region where the latter is possible only for small values of the @xmath1 . furthermore ,
our results provide additional insight on shear instabilities and on the necessary conditions for their development .
|
rolling scalar fields are a mainstay of modern cosmology .
this is perhaps best - illustrated by the inflationary paradigm @xcite , in which most implementations involve a scalar field rolling towards the minimum of its potential in such a way that the potential energy of the field is the dominant component of the energy density of the universe .
there are , however , many other cosmological instances in which scalar fields are invoked . during the last few years a new consistent picture of the energy budget of the universe has emerged .
large scale structure studies show that matter ( both luminous and dark ) contributes a fraction of about 0.3 of the critical density , while the position of the first acoustic peak of the cosmic microwave background power spectrum indicates that the total energy density is consistent with criticality .
the discrepancy between these two measurements may be reconciled by invoking a negative pressure component which is termed _
dark energy_. while there are a number of different observational tools to study dark energy number counts of galaxies
@xcite and galaxy clusters @xcite for example the most direct evidence to date comes from the light - curve measurements of intermediate redshift type ia supernovae @xcite . consistency between these observations and others such as weak gravitational lensing @xcite and large scale structure surveys @xcite implies that the dark energy @xmath0 satisfy @xmath1 and that the equation of state be @xcite @xmath2 leading to the acceleration of the universe .
it is of course possible that this mystery component is a cosmological constant @xmath3 , for which @xmath4 .
however , understanding the nature of such an unnaturally small @xmath3 is at least as difficult as undestanding one that is zero .
alternatively , it has been suggested @xcite-@xcite that if the cosmological constant itself is zero , the dark energy component could be due to the dynamics of a rolling scalar field , in a form of late - universe inflation that has become known as _
quintessence_. although there are a number of fine - tuning problems associated with this idea , it does provide a way to ensure the late - time acceleration of the universe , albeit at the expense of introducing a second ( after the inflaton ) cosmologically relevant rolling scalar field . while not addressing the cosmological constant problem itself , and suffering from fine - tuning , quintessence itself
has the advantage of avoiding a future horizon in space - time , and hence makes consistency with what is known about perturbative string theory more likely .
it is natural to wonder whether the inflaton and the quintessence field might be one and the same @xcite .
in fact , specific models for this have been proposed @xcite-@xcite .
clearly such models are attractive because we need only postulate a single rolling scalar , but may be problematic either theoretically or phenomenologically . in this paper
we investigate how we may further limit the proliferation of rolling scalar fields required in modern cosmology by studying how the scalar field responsible for late - time acceleration of the universe might also solve another outstanding cosmological puzzle .
specifically we will be interested in the role that such a field may play in the generation of the baryon asymmetry of the universe .
the spectacular success of primordial nucleosynthesis requires that there exist an asymmetry between baryons and antibaryons in the universe at temperatures lower than an mev .
this is quantified by the requirement @xmath5 where @xmath6 , with @xmath7 the number density of ( anti)baryons and @xmath8 is the entropy density . to generate such an asymmetry ,
the underlying particle physics theory must satisfy three necessary conditions the sakharov conditions @xcite .
these are baryon number @xmath9 violation , the violation of the discrete symmetries @xmath10 and @xmath11 and a departure from thermal equilibrium , this last condition resulting from an application of the @xmath12 theorem . in this paper
we are interested in how these conditions may be met within the context of dark energy models .
the relationship between early - time acceleration inflation and baryogenesis has been explored in some detail ( for example see @xcite-@xcite ) . here
we investigate the opposite regime , that the quintessence field may be associated with the generation of the baryon asymmetry .
naturally , it would be particularly efficient if a single scalar field could be responsible for three fundamental phenomena in cosmology inflation , baryogenesis and dark energy , and indeed we will show that baryogenesis occurs quite generically in models in which a single scalar is responsible for the two periods of cosmic acceleration .
the outline of this paper is as follows . in section [ quintinf ]
we will review some details about quintessence and explain how inflation and quintessence may be unified by generalizing the quintessential inflation model of peebles and vilenkin @xcite . in section [ qbg ]
we will describe how quintessence and quintessential inflation may naturally yield a baryon asymmetry without the introduction of any new fields into the theory .
we term this model _ quintessential baryogenesis _
, borrowing the phrasing from peebles and vilenkin .
this turns out to depend to some extent on the details of quintessential inflation . in section [ sec : constraints ] we will discuss experimental and astrophysical constraints on our models and comment on how we may test the physics involved .
we offer our comments and conclusions in the final section of the paper .
as we have mentioned already , one approach to the dark energy problem is to assume that there is some as yet unknown process that sets the vacuum cosmological constant of the universe to zero , but that there exists a cosmologically - relevant scalar field in the universe , that has yet to reach its global minimum and therefore contributes an effective vacuum energy to the total .
this idea has been termed quintessence and we shall briefly review it here . the einstein equations in cosmology may be written as @xmath13 @xmath14 where we are using the friedmann , robertson - walker ( frw ) ansatz for the metric @xmath15 \ .\ ] ] here the energy density @xmath16 and pressure @xmath17 for a real homogeneous scalar field @xmath18 are given by @xmath19 @xmath20 respectively , with @xmath21 the potential , and where we have defined the planck mass by @xmath22 .
the scalar field itself obeys @xmath23 with a prime denoting a derivative with respect to @xmath18 .
now , to explain the current data indicating an accelerating universe , it is necessary to have the dominant type of matter at late times be such that @xmath24
. if this matter is to be @xmath18 , then ( [ acceleration ] ) implies @xmath25 which , since we conventionally write @xmath26 translates into an equation of state parameter that obeys @xmath27 with an appropriate choice of potential the resulting cosmic acceleration can be arranged to occur at late times , and provides an explanation for the supernova data @xcite .
let us now consider extending these ideas to incorporate inflation @xcite-@xcite .
this further constrains our potential , and we will present a general analysis , indicating where appropriate how the particular results of @xcite are recovered .
consider a generic potential which , for convenience and for comparison with other work , we will express in the form @xmath28 \\ v_2(\phi ) & \ \ \ \ \ , \ \ \ \ \ \ & \phi\in ( 0,\infty ) \ , \end{array}\right.\ ] ] such that @xmath29 , @xmath30 @xmath31 , @xmath32 and @xmath33 .
we shall in general be interested in cases where @xmath34 is monotonically decreasing and concave ( @xmath35 and @xmath36 @xmath31 ) .
this is a generalized form of the potential used by peebles and vilenkin @xcite in which @xmath37 since we require inflation to occur at early times , when the expectation value of @xmath18 is large and negative , we would like @xmath38 to satisfy the slow - roll conditions @xmath39 ^ 2 \frac{m_{\rm p}^2}{16\pi } & \ll & 1 \nonumber \\ { \bar \eta } \equiv \left|\frac{v_1''(\phi)}{v_1(\phi)}\right| \frac{m_{\rm p}^2}{8\pi } & \ll & 1 \ , \end{aligned}\ ] ] for sufficiently large and negative @xmath18 .
inflation ends when the slow - roll conditions are violated and the potential and kinetic energies of the inflaton are comparable with each other .
we denote this epoch by the subscript @xmath40 , following peebles and vilenkin @xcite , so that the above statement reads @xmath41 in the models in this paper we will always have @xmath42 and ( [ friedmann ] ) then implies that the hubble parameter at this epoch is given by @xmath43 in traditional inflationary models the inflaton then rapidly transfers its energy to other fields either through perturbative effects ( reheating ) or parametric resonance ( preheating ) . here
, however , there is no such effect , and it is the kinetic energy of the field @xmath18 that is the dominant component of the energy density of the universe immediately after the end of inflation . following joyce @xcite we term this behavior _ kination_.
since time derivatives of scalar fields scale as @xmath44 we may use ( [ potkinequal ] ) to show that the evolution of the energy density in @xmath18 during the kination era obeys @xmath45 now , assuming spatial flatness ( which the epoch of inflation will ensure in general ) , we may solve the friedmann equation ( [ friedmann ] ) for the cosmic scale factor to obtain @xmath46 ^ 3=3\sqrt{\frac{8\pi v_1(-m_{\rm p})}{3m_{\rm p}^2 } } t \ .\ ] ] since the universe is kinetic energy dominated during this epoch , we may also obtain @xmath47 where we have imposed @xmath48 and @xmath49 is the cosmic time at which inflation ends .
a successful cosmology requires the universe be radiation - dominated at the time of nucleosynthesis , since otherwise the precision predictions of that theory are no longer in agreement with observations .
the lack of conventional reheating , the conversion of the potential energy of the inflaton to particle production , in quintessential inflation means that the requisite radiation must be produced another way .
in fact , the radiation era in these models is due to the subtle behavior of quantum fields in changing geometries . at the end of inflation
, the frw line element undergoes an abrupt change from that associated with cosmic expansion ( exponential or power - law ) to that associated with kination .
massless quantum fields in their vacua in the inflation era are no longer in vacuum in the kination era , corresponding to gravitational particle production .
this effect is analogous to hawking radiation , and has been explored in detail @xcite-@xcite in the cosmological context of interest here .
the radiation density produced in this way is @xmath50 where @xmath51 , and the number density of massless particles produced is given by @xmath52 at such early times in the universe , thermal equilibrium is not yet established due to the rapid pace of cosmic expansion .
the massless particles produced by the effects of quantum fields in our changing space - time only establish thermal equilibrium when the hubble parameter has dropped to a value @xmath53 , where @xmath54 is the particle - antiparticle annihilation cross - section .
since @xmath55 where @xmath56 is a coupling constant , we obtain @xmath57 therefore , thermalization takes place at a temperature @xmath58 this is the highest temperature at which there is thermal equilibrium in the universe .
now that we know how both the scalar field and the radiation evolve during the kinetic energy dominated era , we may easily calculate the scale factor at which radiation - domination occurs .
demanding that @xmath59 we obtain @xmath60 and the temperature @xmath61 at which radiation domination begins is then simply calculated to be @xmath62 so that we have @xmath63 thus , for @xmath64 , since the scale of the potential is no greater than @xmath65 , we see that @xmath66 and so when radiation - domination begins the universe is immediately in thermal equilibrium . clearly , to obtain our standard cosmology
it is necessary to have a period of radiation domination ( followed by matter domination ) before dark energy domination begins .
this means requiring that @xmath67 , so that at @xmath61 the universe becomes dominated by radiation , with the scalar field evolving in the background , its potential and kinetic energies subdominant to the radiation .
in particular , nucleosynthesis takes place at @xmath68mev . to ensure that the universe is radiation dominated at this epoch we should conservatively require @xmath69mev . using ( [ tr ] ) , this allows us to bound the energy scale associated with quintessential inflation by @xmath70 such
a bound does not exist for the standard inflationary paradigm because reheating effects ensure that the universe is radiation - dominated immediately following inflation . in quintessential inflation
however , there is a comparatively small amount of radiation produced through gravitational particle production , so that radiation - domination occurs much later . for a significant time subsequent to this
, cosmic evolution is much the same as in the standard cosmology , with a matter dominated epoch eventually succeeding the radiation era .
although the scalar field is not important during these times , the density fluctuations seeded by quantum fluctuations in @xmath18 during inflation lead to structure formation and temperature fluctuations in the cosmic microwave background radiation . in order to obtain agreement with the cobe anisotropy measurements , we must require @xmath71 where @xmath72 denotes the value of @xmath18 60 efolds before the end of inflation . note that in the peebles - vilenkin model this constraint translates to @xmath73 , similar to that tuning required by standard chaotic inflationary potentials .
finally , let us turn to this extreme future of the universe , in which the scalar field is rolling in the potential @xmath74 and becomes responsible for quintessence at an epoch denoted by a subscript @xmath75 .
it is clear that when quintessence begins , the energy density of the field @xmath18 once again becomes dominated by its potential energy density .
this implies @xmath76 where @xmath77 it is a challenge similar to that for conventional quintessence to ensure that this epoch occurs at the present time and yields the correct ratio of matter to dark energy .
however , such considerations apply far after baryogenesis and we shall refer the reader to other treatments for the details of how this occurs @xcite-@xcite .
in order for the quintessence field @xmath18 to play a role in baryogenesis , we must consider how @xmath18 couples to other fields . in principle , the inflaton and quintessence field may lie in any sector of the theory , the phenomenologically safest of which would be one in which there are only gravitational strength couplings to other particles .
this is presumably the best we can do , since attempts to protect gravitational - strength couplings through global symmetries can be thwarted by wormholes and quantum gravitational effects @xcite-@xcite .
we will adopt a conservative approach and assume that @xmath18 couples to standard model fields with couplings specified by a dimensionless constant and an energy scale which we shall leave as a free parameter for the moment and later constrain by observations and the condition that our model produce a sufficient baryon asymmetry .
we consider terms in the effective lagrangian density of the form @xmath78 where @xmath79 is a coupling constant , @xmath80 is the scale of the cutoff in the effective theory and @xmath81 is the current corresponding to some continuous global symmetry such as baryon number or baryon number minus lepton number .
further , let us assume that @xmath18 is homogeneous .
we then obtain @xmath82 where @xmath83 is the number density corresponding to the global symmetry and we have defined an effective time - dependent `` chemical potential '' @xmath84 .
recall that we need to satisfy the sakharov criteria in order to generate a baryon asymmetry ( for reviews see @xcite-@xcite ) .
the first of these requires baryon number @xmath9 to be violated . at this stage , to maintain generality , we shall leave the mechanism of baryon number violation unspecified , and will address particular cases later .
possible sources are the decay of superheavy grand - unified gauge bosons or anomalous electroweak processes at finite temperature .
further , the standard model is maximally c - violating due to its chiral structure , and the coupling ( [ jmucoupling ] ) is @xmath11-odd . in this sense , no _ explicit _
@xmath11-violation is required in this model .
the third sakharov criterion requires a departure from thermal equilibrium if @xmath12 is a manifest symmetry .
however , the crucial point about baryogenesis in the presence of the rolling scalar field @xmath18 is that @xmath12 is broken spontaneously by the explicit value taken by @xmath85 .
thus , the particular model of baryogenesis that is important here is _ spontaneous baryogenesis _
@xcite , which is effective even in thermal equilibrium
. we will refer to this model , in which the rolling scalar responsible for both inflation and dark energy also provides a source for spontaneous baryogenesis as _
quintessential baryogenesis_. in our model it is the field @xmath18 that plays the role of cohen and kaplan s _ thermion
_ @xcite .
the idea that the inflaton could drive spontaneous baryogenesis was discussed briefly in @xcite , where it was correctly noted that accelerated expansion would reduce the baryon number generated _ during _ inflation to a negligible magnitude , and therefore that barogenesis during the reheating phase was the only possibility .
however , as we shall see , in the context of quintessential inflation there exists a significant range of postinflationary cosmic history during which spontaneous baryogenesis may occur . to understand how spontaneous ( and hence quintessential ) baryogenesis works ,
note that in thermal equilibrium we have @xmath86 \ , \ ] ] where @xmath87 is a parameter and @xmath88 is the phase - space distribution of the particles of the current @xmath81 , which may be fermi - dirac or bose - einstein .
thus , for @xmath89 @xmath90 where @xmath91 is the number of degrees of freedom of the field corresponding to @xmath92 .
therefore , @xmath93 now recall that the entropy density is given by @xmath94 where @xmath95 is the effective number of relativistic degrees of freedom in thermal equilibrium at temperature @xmath96 .
whatever the mechanism of baryon number violation , there will exist a temperature @xmath97 below which baryon number violating processes due to this mechanism become sufficiently rare that they freeze out . for @xmath98
these processes can no longer appreciably change the baryon number of the universe .
computing the freeze - out value of the baryon to entropy ratio we then obtain @xmath99 quintessential baryogenesis is effective at temperatures @xmath100 , with corresponding scale factors @xmath101 .
we have seen that @xmath102 , so that baryon number violating interactions are first in equilibrium during the kination epoch in which the evolution of the scalar field can be written as @xmath103 if we assume that @xmath104 then ( [ baueqn ] ) yields @xmath105 where we have used @xmath106 . writing , for convenience , @xmath107 with @xmath108 ( so that @xmath109 ) and using ( [ athoverax ] ) we then obtain @xmath110 inserting the values for @xmath111 and @xmath112 yields @xmath113 notice that , as expected , this final result is linear in the effective chemical potential , and contains a power of @xmath114 reflecting the appropriate amount of redshifting occurring during the kination epoch between the temperatures @xmath115 and @xmath97 . to make further progress we must calculate @xmath97 and @xmath116 in order to find @xmath114 , and
this requires knowledge of the dynamics of @xmath18 . to do this
correctly it is necessary to consider the possible effects of back - reaction of the coupling to @xmath81 on the dynamics of @xmath18 . taking account of the effective lagrangian into account ,
the equation of motion of @xmath18 becomes @xmath117 using ( [ deltan ] ) we obtain @xmath118\left({\ddot \phi}+3h{\dot \phi}\right ) + v'(\phi)=0 \ .\ ] ] therefore , for @xmath119 , we are justified in neglecting the extra term and we may safely neglect the back - reaction on @xmath18 .
this approximation will typically easily be satisfied , and in particular it is well - justified in the peebles - vilenkin model @xcite in which @xmath120 .
how the baryon excess evolves after this point depends on the value of @xmath97 and on the relevant current in equation ( [ jmucoupling ] ) .
if @xmath121 gev , the critical temperature of the electroweak phase transition , then all baryon number violation ceases at @xmath97 and @xmath122 . however , if @xmath123 , then we must take into account the effects of anomalous electroweak processes at finite temperature .
these can be involved in directly generating the baryon asymmetry ( electroweak baryogenesis ) , in reprocessing an asymmetry in other quantum numbers into one in baryon number ( for example in leptogenesis ) or in diluting the asymmetry created by any baryogenesis mechanism which is effective above the electroweak scale and does not produce a @xmath124 asymmetry .
it is important to realize that , in the context of quintessential inflation , the quantitative effects of these electroweak processes may differ substantially from those in the standard cosmology @xcite since in our case , the electroweak phase transition may occur during kination rather than radiation domination . in the electroweak theory baryon number violating processes at zero temperature are mediated by a saddle - point field configuration known as the sphaleron @xcite .
we shall therefore refer to the finite temperature configurations relevant here as thermal sphalerons .
however , given the constraints we shall present in section [ sec : constraints ] , we shall see that it is necessary to have @xmath125 and no electroweak dilution in order to generate a sufficient baryon asymmetry .
we have now provided quite a general description of quintessential baryogenesis .
while this has allowed us to demonstrate the generic features of our model , we can not calculate the magnitude of the actual baryon asymmetry generated without first specifying a mechanism of baryon number violation ( and hence a value for @xmath97 ) and a value for the dimensionless combination @xmath126 .
let us now turn to some concrete examples .
if there exists baryon number violating physics above the standard model , then this physics will manifest itself in non - renormalizable operators in the standard model .
for the purposes of this section we will actually be interested in operators that violate the anomaly - free combination @xmath124 . in that case
the value @xmath127 calculated via ( [ etaffinal ] ) will be the final baryon to entropy ratio @xmath128 , since anomalous electroweak processes preserve this combination of quantum numbers .
consider the effective 4-fermion operator @xmath129 where @xmath130 denote standard model fermions . here
@xmath131 is a dimensionless coupling , obtained after integrating out the @xmath124 violating effects of a particle of mass @xmath132 .
the rate of baryon number violating processes due to this operator is , as usual , defined by @xmath133 , where @xmath134 is the cross - section for @xmath135 , @xmath136 is the number density of @xmath137 particles , @xmath138 is the relative velocity and @xmath139 denotes a thermal average . for temperatures @xmath140 we have @xmath141 , @xmath142 , and @xmath143 which yields @xmath144 the high power of the temperature dependence in this rate results from the fact that ( [ 4fermion ] ) is an irrelevant operator in the electroweak theory and , as we shall see , is crucial for the success of our mechanism .
these interactions are in thermal equilibrium in the early universe , but because their rate drops off so quickly with the cosmic expansion they will drop out of equilibrium at the temperature @xmath97 defined through @xmath145 thus , @xmath146 as a definite example , let us take @xmath147 . as we shall see in [ sec : constraints ] , this is as large as is allowed by current constraints . in this case ( [ etaffinal ] ) implies that we need @xmath148 in order to obtain the correct bau .
this then implies that we must have @xmath149 for reference , note that in the peebles - vilenkin model @xcite this becomes @xmath150 gev , the intermediate scale that appears in some supersymmetric models . a natural source of baryon number violation to consider
is that arising from gauge - mediated interactions in grand unified theories ( guts ) , in which quarks and leptons lie in the same representation of the gut gauge group .
however , as we shall demonstrate briefly here , this is not a viable source for quintessential baryogenesis .
consider a gut gauge boson @xmath0 with mass @xmath132 .
this particle decays through a renormalizable coupling with decay rate @xmath151 it is then relatively straightforward to show that @xmath0 particles decay in equilibrium ( as is required by our mechanism ) only if @xmath152 since gut gauge bosons are much heavier than this value this source of baryon number violation is not useful for our model .
one might be tempted to consider a particle with the appropriate mass produced by some physics beyond the standard model .
however , it quickly becomes apparent that decays through renormalizable interactions can never work , since these particles then decay and the interactions are never again out of equilibrium ( there is no freeze - out ) .
thus , as we commented earlier , nonrenormalizable operators seem essential for decays of massive particles to work . another important source of baryon number violation in baryogenesis models comes from anomalous electroweak processes at finite temperature .
thermal sphalerons are in equilibrium at temperatures above the critical temperature of the electroweak phase transition @xmath153gev and thus , the final baryon to entropy ratio generated purely through electroweak processes is given by ( [ etaffinal ] ) with @xmath154 .
this yields @xmath155 which , since @xmath156 , is far too low to explain the observed baryon asymmetry ( [ bauobserved ] ) .
the presence of an extremely light scalar field in the universe has the potential to lead to a number of observable consequences in the laboratory and in cosmology . in the case of quintessence
these effects have been analyzed in some detail by carroll @xcite .
particularly strong constraints arise due to couplings of the form @xmath157 where @xmath158 and @xmath159 are the electromagnetic field strength tensor and its dual respectively . if , as in quintessence models , the field @xmath18 is homogeneous and time varying , then it affects the dispersion relation for electromagnetic waves and leads to a rotation in the direction of polarized light from radio sources @xcite .
the bound obtained is @xmath160 where we have assumed that @xmath18 rolls over about a planck mass during the last half a redshift of the universe . to avoid this bound
, quintessence models usually struggle to have such couplings be as small as possible . in our model
however , we are making important use of the coupling ( [ jmucoupling ] ) .
if the relevant current is that for baryon number @xmath161 , then using the anomaly equation we may rewrite our term ( [ jmucoupling ] ) as @xmath162 where @xmath91 and @xmath163 are the gauge couplings of @xmath164 and @xmath165 , respectively , @xmath166 is the number of families and @xmath167 and @xmath168 are the duals of the @xmath164 and @xmath165 field strength tensors respectively .
thus , the constraint ( [ seanbound ] ) should apply to give @xmath169 putting in the appropriate standard model numbers we obtain @xmath170 if the relevant current is @xmath171 then the above argument does not apply formally .
however , we may still generally expect an analogous coupling to the term @xmath172 of a similar order .
we have already demonstrated that , for an appropriate scale @xmath132 , successful quintessential baryogenesis takes place for @xmath173 , and so it is possible to generate the observed bau and to evade existing constraints .
if quintessential baryogenesis is correct , then it may be that the relevant coupling @xmath79 lies just below the existing observational bounds and that future studies of the rotation of polarized light from distant galaxies will reveal the presence of such a term .
one might also worry about the laboratory constraints on the operator corresponding to a coupling to @xmath172 resulting from ( [ anomaly ] ) after electroweak symmetry breaking .
for example , in the standard model these can lead to contributions to the electric dipole moments of the electron and the neutron .
these effects have been considered @xcite in the context of electroweak baryogenesis . applying the experimental bounds @xcite the relevant bound is significantly weaker than the cosmological one quoted above .
we should also comment briefly on the naturalness of the potentials required .
we have taken the quintessence and quintessential inflation paradigms and explored another generic effect in these models .
it should be pointed out however , that the potentials require a great deal of fine - tuning . in particular , the extremely small value of the self - coupling @xmath174 to explain the fluctuations in the cmb is unexplained .
if we consider embedding the model in a supersymmetric theory we will come up against the problems that standard quintessence faces , that the flatness of the potential required at late times in the universe may be ruined by the soft supersymmetry breaking terms required to obtain a realistic phenomenology today .
modern particle cosmology concerns the search for dynamical explanations for the initial conditions required by the standard frw cosmology .
particular attention has been paid recently to one of these initial condition problems , that of the size of the vacuum energy contribution to the total energy density of the universe .
the root of this issue is that vacuum energy , or something approximating it , can lead to the acceleration of the universe .
the best fit cosmology to all current observational data is one in which the universe undergoes two separate epochs of acceleration .
the first of these , inflation , is the only clear way to seed adiabatic , scale - free perturbations in the cosmic microwave background radiation .
the second epoch , that of dark - energy domination , is required to simultaneously understand the power spectrum of the cmb and the expansion of the universe at intermediate redshifts as revealed by type ia supernovae .
these requirements have led cosmologists to introduce a new scalar field to account for the newly required late - time acceleration . in this paper
we have extended the approach of spokoiny @xcite and of peebles and vilenkin @xcite in exploring the extent to which the dynamics of a single scalar field can be responsible for setting multiple parameters required by the standard cosmology .
our contribution has been to generalize the mechanisms by which inflation and dark energy domination may be due to a single scalar and to introduce the idea that this same rolling scalar might be responsible for generating the baryon asymmetry of the universe .
the mechanism that we propose , quintessential baryogenesis , is an application of the spontaneous baryogenesis model of cohen and kaplan@xcite to the quintessential inflation case .
it seems a particularly powerful idea to us that a single rolling scalar field might be responsible for a number of the fundamental initial conditions required to make the standard cosmology work . in the case of the baryon asymmetry , this allows us to associate the existence of an asymmetry with the spontaneous breaking of @xmath12 and the direction of the rolling of the scalar field .
the evolution of the universe we envisage may be summarized as follows ( see figure [ fig : graphic ] ) . at the earliest times in the universe
, inflation occurs due to the potential energy dominance of the field @xmath18 which begins rolling at very large and negative values .
inflation ends when the kinetic energy of the scalar field becomes important and the slow - roll conditions are violated . since our potential does not have a minimum at finite @xmath18 , unlike typical inflationary models , conventional reheating does not occur . instead
, matter is created gravitationally due to the mismatch of vacuum states between the approximately de - sitter state of inflation and that of the kination era . since kinetic energy density redshifts more rapidly that radiation energy density , the universe eventually becomes radiation - dominated . at this stage
the rolling scalar has negligible effect on the expansion rate of the universe .
however , the direction of rolling spontaneously violates @xmath12 . if @xmath18 couples to other fields , as we expect it to generically , then the expectation value of the baryon number operator in this background in thermal equilibrium is nonzero .
thus a baryon excess is generated .
after the electroweak phase transition baryon number violation is no longer effective in the universe and the baryon number existing at that time is frozen in .
the scalar field continues to evolve and in the late universe , after matter - domination has begun , its potential energy can once again become dominant leading to a new period of dark - energy domination .
the couplings required to make quintessential baryogenesis effective may be generated in a number of different ways , for example by gravitational effects coupling the inflaton / dark energy sector to visible sector fields .
we have considered the current experimental constraints on the necessary operator and have found that there exist considerable regions of parameter space in which our mechanism is consistent .
further , it is possible that a restricted region of this parameter space may be accessible to future experiments .
we have left a number of questions unanswered and will return to them in future work .
perhaps the most pressing issue is one that plagues rolling scalar models of dark energy in general , namely the question of technical naturalness of the potentials involved , and their stability to quantum corrections .
we have omitted any discussion of this here , while laying out the general features of the model , but these issues must be addressed to put our mechanism on firmer ground .
for example , it may be most natural to identify the field @xmath18 with a pseudo - goldstone boson @xcite , since its coupling to the current @xmath81 is derivative .
however , this is a general issue for quintessence models , and is not specific to our baryogenesis mechanism .
we have therefore chosen to concern ourselves with this issue separately .
taken at face value , current observations imply that our universe is entering an accelerating phase that may be governed by the rolling of a scalar field . if we are to understand the physics of such a field then it is important that we investigate other ways in which it may impact cosmology and particle physics .
in particular , if , as we have suggested here , the field is responsible for the generation of the baryon asymmetry , then the result will be a more economical and attractive cosmology .
we thank robert brandenberger and alex vilenkin for helpful comments on a first draft of this paper . the work of adf and mt
is supported by the national science foundation ( nsf ) under grant phy-0094122 .
sn is supported by the us department of energy ( doe ) .
a. h. guth , phys .
d * 23 * , 347 ( 1981 ) .
a. d. linde , phys .
b * 108 * , 389 ( 1982 ) .
a. albrecht and p. j. steinhardt , phys .
lett . * 48 * , 1220 ( 1982 ) .
j. a. newman and m. davis , arxiv : astro - ph/9912366 .
z. haiman , j. j. mohr and g. p. holder , arxiv : astro - ph/0002336 .
a. g. riess _ et al .
_ [ supernova search team collaboration ] , astron .
j. * 116 * , 1009 ( 1998 ) [ arxiv : astro - ph/9805201 ] . s. perlmutter _ et al .
_ [ supernova cosmology project collaboration ] , astrophys .
j. * 517 * , 565 ( 1999 ) [ arxiv : astro - ph/9812133 ] .
d. huterer , phys .
d * 65 * , 063001 ( 2002 ) [ arxiv : astro - ph/0106399 ] .
w. hu , d. j. eisenstein , m. tegmark and m. j. white , phys . rev .
d * 59 * , 023512 ( 1999 ) [ arxiv : astro - ph/9806362 ] . s. perlmutter , m. s. turner and m. j. white , phys .
lett . *
83 * , 670 ( 1999 ) [ arxiv : astro - ph/9901052 ] .
l. m. wang , r. r. caldwell , j. p. ostriker and p. j. steinhardt , astrophys . j. * 530 * , 17 ( 2000 ) [ arxiv : astro - ph/9901388 ] . c. wetterich , nucl .
b * 302 * , 668 ( 1988 ) .
b. ratra and p. j. peebles , phys .
d * 37 * , 3406 ( 1988 ) .
r. r. caldwell , r. dave and p. j. steinhardt , phys .
lett . * 80 * , 1582 ( 1998 ) [ arxiv : astro - ph/9708069 ] .
b. spokoiny , phys .
b * 315 * , 40 ( 1993 ) [ arxiv : gr - qc/9306008 ] .
j. peebles and a. vilenkin , phys .
d * 59 * , 063505 ( 1999 ) [ arxiv : astro - ph/9810509 ] .
k. dimopoulos and j. w. valle , arxiv : astro - ph/0111417 .
a. d. sakharov , pisma zh .
fiz . * 5 * , 32 ( 1967 ) [ jetp lett .
* 5 * , 24 ( 1967 sopua,34,392 - 393.1991 ufnaa,161,61 - 64.1991 ) ] .
i. affleck and m. dine , nucl .
b * 249 * , 361 ( 1985 ) .
j. yokoyama , h. kodama , k. sato and n. sato , utap-40 _ in * liege 1986 , proceedings , origin and early history of the universe * 197 - 203 .
( see conference index)_. j. r. ellis , k. enqvist , d. v. nanopoulos and k. a. olive , phys .
b * 191 * , 343 ( 1987 ) .
j. d. barrow , e. j. copeland , e. w. kolb and a. r. liddle , phys .
d * 43 * , 977 ( 1991 ) . j. d. barrow , e. j. copeland , e. w. kolb and a. r. liddle , phys . rev .
d * 43 * , 984 ( 1991 ) .
m. dine , l. randall and s. thomas , nucl .
b * 458 * , 291 ( 1996 ) [ arxiv : hep - ph/9507453 ] .
e. w. kolb , a. d. linde and a. riotto , phys .
lett . *
77 * , 4290 ( 1996 ) [ arxiv : hep - ph/9606260 ] .
e. d. stewart , m. kawasaki and t. yanagida , phys .
d * 54 * , 6032 ( 1996 ) [ arxiv : hep - ph/9603324 ] .
a. dolgov , k. freese , r. rangarajan and m. srednicki , phys . rev .
d * 56 * , 6155 ( 1997 ) [ arxiv : hep - ph/9610405 ] . c. f. kolda and j. march - russell , phys .
d * 60 * , 023504 ( 1999 ) [ arxiv : hep - ph/9802358 ] .
e. w. kolb , a. riotto and i. i. tkachev , phys .
b * 423 * , 348 ( 1998 ) [ arxiv : hep - ph/9801306 ] .
l. m. krauss and m. trodden , phys .
lett . * 83 * , 1502 ( 1999 ) [ arxiv : hep - ph/9902420 ] .
j. garcia - bellido , d. y. grigoriev , a. kusenko and m. e. shaposhnikov , phys .
d * 60 * , 123504 ( 1999 ) [ arxiv : hep - ph/9902449 ] .
e. j. copeland , d. lyth , a. rajantie and m. trodden , phys .
d * 64 * , 043506 ( 2001 ) [ arxiv : hep - ph/0103231 ] .
d. v. nanopoulos and r. rangarajan , arxiv : hep - ph/0103348 .
m. joyce , phys .
d * 55 * , 1875 ( 1997 ) [ arxiv : hep - ph/9606223 ] .
l. h. ford , phys .
d * 35 * , 2955 ( 1987 ) .
t. damour and a. vilenkin , phys .
d * 53 * , 2981 ( 1996 ) [ arxiv : hep - th/9503149 ] .
m. giovannini , phys .
d * 58 * , 083504 ( 1998 ) [ arxiv : hep - ph/9806329 ] .
s. b. giddings and a. strominger , nucl . phys .
b * 306 * , 890 ( 1988 ) .
l. f. abbott and m. b. wise , nucl .
b * 325 * , 687 ( 1989 ) .
s. r. coleman and s. hughes , phys .
b * 309 * , 246 ( 1993 ) [ arxiv : hep - th/9305123 ] .
r. kallosh , a. d. linde , d. a. linde and l. susskind , phys .
d * 52 * , 912 ( 1995 ) [ arxiv : hep - th/9502069 ] .
a. g. cohen , d. b. kaplan and a. e. nelson , ann .
nucl . part .
* 43 * , 27 ( 1993 ) [ arxiv : hep - ph/9302210 ] .
v. a. rubakov and m. e. shaposhnikov , usp .
nauk * 166 * , 493 ( 1996 ) [ phys .
usp . * 39 * , 461 ( 1996 ) ] [ arxiv : hep - ph/9603208 ] .
m. trodden , rev .
phys . * 71 * , 1463 ( 1999 ) [ arxiv : hep - ph/9803479 ] .
a. riotto and m. trodden , ann . rev
. nucl . part .
* 49 * , 35 ( 1999 ) [ arxiv : hep - ph/9901362 ] .
a. g. cohen and d. b. kaplan , nucl .
b * 308 * , 913 ( 1988 ) .
m. joyce and t. prokopec , phys .
d * 57 * , 6022 ( 1998 ) [ arxiv : hep - ph/9709320 ] .
f. r. klinkhamer and n. s. manton , phys .
d * 30 * , 2212 ( 1984 ) .
s. m. carroll , phys .
lett . * 81 * , 3067 ( 1998 ) [ arxiv : astro - ph/9806099 ] .
s. m. carroll , g. b. field and r. jackiw , phys .
d * 41 * , 1231 ( 1990 ) .
s. m. carroll and g. b. field , phys .
d * 43 * ( 1991 ) 3789 .
x. zhang and b. l. young , phys .
d * 49 * , 563 ( 1994 ) [ arxiv : hep - ph/9309269 ] .
a. lue , k. rajagopal and m. trodden , phys .
d * 56 * , 1250 ( 1997 ) [ arxiv : hep - ph/9612282 ] .
e. d. commins , s. b. ross , d. demille and b. c. regan , phys . rev . a * 50 * ( 1994 ) 2960 .
k. f. smith _
_ , _ phys .
lett . _ * b234 * , 191 ( 1990 ) ; + i. s. altarev _ et al .
_ , _ phys .
lett . _ * b276 * , 242 ( 1992 ) .
|
the simplest explanation for early time acceleration ( inflation ) and the late time acceleration indicated by recent data is that they have a common origin .
we investigate another generic cosmological implication of this possiblity , that the baryon asymmetry of the universe may be generated in such models .
we identify several novel features of baryogenesis in such a universe , in which a rolling scalar field is always part of the cosmological energy budget .
we also propose a concrete mechanism by which the baryon asymmetry of the universe may be generated in this context .
we analyze the generic properties of and constraints on these cosmologies , and then demonstrate explicitly how a complete cosmology may develop in some specific classes of models .
|
learning linear separators is one of the central challenges in machine learning .
they are widely used and have been long studied both in the statistical and computational learning theory .
a seminal result of @xcite , using tools due to @xcite , showed that @xmath0-dimensional linear separators can be learned to accuracy @xmath1 with probability @xmath2 in the classic pac model in polynomial time with @xmath3 examples .
the best known lower bound for linear separators is @xmath4 , and this holds even in the case in which the distribution is uniform @xcite . whether the upper bound can be improved to match the lower bound via a polynomial - time algorithm is been long - standing open question , both for general distributions @xcite and for the case of the uniform distribution in the unit ball @xcite . in this work
we resolve this question in the case where the underlying distribution belongs to the class of log - concave and nearly log - concave distributions , a wide class of distributions that includes the gaussian distribution and uniform distribution over any convex set , and which has played an important role in several areas including sampling , optimization , integration , and learning @xcite .
we also consider active learning , a major area of research of modern machine learning , where the algorithm only receives the classifications of examples when it requests them @xcite .
our main result here is a polynomial - time active learning algorithm with label complexity that is exponentially better than the label complexity of any passive learning algorithm in these settings .
this answers an open question in @xcite and it also significantly expands the set of cases for which we can show that active learning provides a clear exponential improvement in @xmath5 ( without increasing the dependence on @xmath0 ) over passive learning .
remarkably , our analysis for passive learning is done via a connection to our analysis for active learning to our knowledge , this is the first paper using this technique .
we also study active and passive learning in the case that the data might not be linearly separable .
we specifically provide new improved bounds for the widely studied tsybakov low - noise condition @xcite , as well as new bounds on the disagreement coefficient , with implications for the agnostic case ( i.e. , arbitrary forms of noise ) .
* passive learning * in the classic passive supervised machine learning setting , the learning algorithm is given a set of labeled examples drawn i.i.d . from some fixed but unknown distribution over the instance space and labeled according to some fixed but unknown target function , and the goal is to output a classifier that does well on new examples coming from the same distribution .
this setting has been long studied in both computational learning theory ( within the pac model @xcite ) and statistical learning theory @xcite , and has played a crucial role in the developments and successes of machine learning .
however , despite remarkable progress , the basic question of providing polynomial - time algorithms with _ tight _
bounds on the sample complexity has remained open .
several milestone results along these lines that are especially related to our work include the following .
the analysis of @xcite , proved using tools from @xcite , implies that linear separators can be learned in polynomial time with @xmath6 labeled examples .
@xcite proved a bound that implies an @xmath4 lower bound for linear separators and explicitly posed the question of providing tight bounds for this class .
@xcite established an upper bound of @xmath7 , which can be achieved in polynomial - time for linear separators .
@xcite achieved polynomial - time learning by finding a consistent hypothesis ( i.e. , a hypothesis which correctly classifies all training examples ) ; this is a special case of erm @xcite .
an intensive line of research in the empirical process and statistical learning theory literature has taken account of `` local complexity '' to prove stronger bounds for erm @xcite . in the context of learning , local complexity takes account of the fact that really bad classifiers can be easily discarded , and the set of `` local '' classifiers that are harder to disqualify is sometimes not as rich . a recent landmark result of @xcite
( see also @xcite ) is the bound for consistent algorithms of @xmath8 where @xmath9 is the alexander capacity , which depends on the distribution @xcite ( see section [ se : dis ] and appendix [ rel - appendix ] for further discussion ) .
however , this bound can be suboptimal for linear separators . in particular , for linear separators in the case
in which the underlying distribution is uniform in the unit ball , the sample complexity is known @xcite to be @xmath10 , when computational considerations are ignored .
@xcite , using the doubling dimension @xcite , another measure of local complexity , proved a bound of @xmath11 for a polynomial - time algorithm . as a lower bound of @xmath12 on @xmath9 for @xmath13 for the case of linear separators and
the uniform distribution is implicit in @xcite , the bound of @xcite given by ( [ e : cap ] ) can not yield a bound better than @xmath14 in this case . in this paper
we provide a _ tight _ bound ( up to constant factors ) on the sample complexity of polynomial - time learning of linear separators with respect to log - concave distributions . specifically , we prove an upper bound of @xmath15 using a polynomial - time algorithm that holds for any zero - mean log - concave distribution .
we also prove an information theoretic lower bound that matches our ( computationally efficient ) upper bound for _ each _ log - concave distribution .
this provides the first bound for a polynomial - time algorithm that is tight for an interesting non - finite class of hypothesis functions under a general class of data - distributions , and also characterizes ( up to a constant factor ) the distribution - specific sample complexity for each distribution in the class . in the special case of the uniform distribution ,
our upper bound closes the existing @xmath16 gap between the upper bounds ( [ e : doubling ] ) and ( [ e : cap.con ] ) and the lower bound of @xcite . *
active learning * we also study learning of linear separators in the active learning model ; here the learning algorithm can access unlabeled ( i.e. , unclassified ) examples and ask for labels of unlabeled examples of its own choice , and the hope is that a good classifier can be learned with significantly fewer labels by actively directing the queries to informative examples .
this has been a major area of machine learning research in the past fifteen years mainly due the availability of large amounts of unannotated or raw data in many modern applications @xcite , with many exciting developments on understanding its underlying principles as well @xcite .
however , with a few exceptions @xcite , most of the theoretical developments have focused on the so called disagreement - based active learning paradigm @xcite ; methods and analyses developed in this context are often suboptimal , as they take a conservative approach and consider strategies that query even points on which there is a small amount of uncertainty ( or disagreement ) among the classifiers still under consideration given the labels queried so far .
the results derived in this manner often show an improvement in the @xmath5 factor in the label complexity of active versus passive learning ; however , unfortunately , the dependence on the @xmath0 term typically gets worse . by analyzing a more aggressive , margin - based active learning algorithm
, we prove that we can efficiently ( in polynomial time ) learn homogeneous linear separators when the underlying distribution is log - concave by using only @xmath17 label requests , answering an open question in @xcite .
this represents an exponential improvement of active learning over passive learning and it significantly broadens the cases for which we can show that the dependence on @xmath5 in passive learning can be improved to only @xmath18 in active learning , but without increasing the dependence on the dimension @xmath0 .
we note that an improvement of this type was known to be possible only for the case when the underlying distributions is ( nearly ) uniform in the unit ball @xcite ; even for this special case , our analysis improves by a multiplicative @xmath19 factor the results of @xcite ; it also provides better dependence on @xmath0 than any other previous analyses implementable in a computationally efficient manner ( both disagreement - based @xcite and more aggressive ones @xcite ) , and over the inefficient splitting index analysis of @xcite . * techniques * at the core of our results is a novel characterization of the region of disagreement of two linear separators under a log - concave measure .
we show that for any two linear separators specified by normal vectors @xmath20 and @xmath21 , for any constant @xmath22 we can pick a margin as small as @xmath23 , where @xmath24 is the angle between @xmath20 and @xmath21 , and still ensure that the probability mass of the region of disagreement outside of band of margin @xmath25 of one of them is @xmath26 ( theorem [ lemma : vectors - sophist ] ) . using this fact ,
we then show how we can use a margin - based active learning technique , where in each round we only query points near the hypothesized decision boundary , to get an exponential improvement over passive learning .
we then show that any passive learning algorithm that outputs a hypothesis consistent with @xmath27 random examples will , with probability at least @xmath28 , output a hypothesis of error at most @xmath29 ( theorem [ t : passive ] ) .
interestingly , our analysis is quite dissimilar to the classic analyses of erm .
it proceeds by conceptually running the algorithm online on progressively larger chunks of examples , and using the intermediate hypotheses to track the progress of the algorithm .
we show , using the same tools as in the active learning analysis , that it is always likely that the algorithm will receive informative examples .
our analysis shows that the algorithm would also achieve @xmath30 accuracy with high probability even if it periodically built preliminary hypotheses using some of the examples , and then only used borderline cases for those preliminary classifiers for further training . to achieve the optimal sample complexity
, we have to carefully distribute the confidence parameter , by allowing higher probability of failure in the later stages , to compensate for the fact that , once the hypothesis is already pretty good , it takes longer to get examples that help to further improve it .
* non - separable case * we also study label - efficient learning in the presence of noise .
we show how our results for the realizable case can be extended to handle ( a variant of ) the tsybakov noise , which has received substantial attention in statistical learning theory , both for passive and active learning @xcite ; this includes the random classification noise commonly studied in computational learning theory @xcite , and the more general bounded ( or massart ) noise @xcite .
our analysis for massart noise leads to optimal bounds ( up to constant factors ) for active and passive learning of linear separators when the marginal distribution on the feature vectors is log - concave , improving the dependence on @xmath0 over previous best known results .
our analysis for tsybakov noise leads to bounds on active learning with improved dependence on @xmath0 over previous known results in this case as well .
we also provide a bound on the alexander s capacity @xcite and the closely related disagreement coefficient notion @xcite , which have been widely used to characterize the sample complexity of various ( active and passive ) algorithms @xcite .
this immediately implies concrete bounds on the labeled data complexity of several algorithms in the literature , including active learning algorithms designed for the purely agnostic case ( i.e. , arbitrary forms of noise ) , e.g. , the @xmath31 algorithm @xcite and the dhm algorithm @xcite .
* nearly log - concave distributions * we also extend our results both for passive and active learning to deal with nearly log - concave distributions ; this is a broader class of distributions introduced by @xcite , which contains mixtures of ( not too separated ) log - concave distributions . in deriving our results ,
we provide new tail bounds and structural results for these distributions , which might be of independent interest and utility , both in learning theory and in other areas including sampling and optimization .
we note that our bounds on the disagreement coefficient improve by a factor of @xmath32 over the bounds of @xcite ( matching what was known for the much less general case of nearly uniform distribution over the unit sphere ) ; furthermore , they apply to the nearly log - concave case where we allow an arbitrary number of discontinuities , a case not captured by the @xcite conditions at all .
we discuss other related papers in appendix [ rel - appendix ] .
we focus on binary classification problems ; that is , we consider the problem of predicting a binary label @xmath33 based on its corresponding input vector @xmath34 . as in the standard machine learning formulation ,
we assume that the data points @xmath35 are drawn from an unknown underlying distribution @xmath36 over @xmath37 ; @xmath38 is called the _ instance space _ and @xmath39 is the _ label space_. in this paper we assume that @xmath40 and @xmath41 ; we also denote the marginal distribution over @xmath38 by @xmath42 .
let @xmath43 be the class of linear separators through the origin , that is @xmath44 . to keep the notation simple
, we sometimes refer to a weight vector and the linear classifier with that weight vector interchangeably .
our goal is to output a hypothesis function @xmath45 of small error , where @xmath46.$ ] we consider two learning protocols : passive learning and active learning . in the passive learning setting
, the learning algorithm is given a set of labeled examples @xmath47 drawn i.i.d . from @xmath36 and
the goal is output a hypothesis of small error by using only a polynomial number of labeled examples . in the ( pool - based )
active learning setting , a set of labeled examples @xmath48 is also drawn i.i.d . from @xmath36 ; the learning algorithm is permitted direct access to the sequence of @xmath49 values ( unlabeled data points ) , but has to make a label request to obtain the label @xmath50 of example @xmath49 .
the hope is that in the active learning setting we can output a classifier of small error by using many fewer label requests than in the passive learning setting by actively directing the queries to informative examples ( while keeping the number of unlabeled examples polynomial ) . for added generality
, we also consider the selective sampling active learning model , where the algorithm visits the unlabeled data points @xmath49 in sequence , and , for each @xmath51 , makes a decision on whether or not to request the label @xmath50 based only on the previously - observed @xmath52 values ( @xmath53 ) and corresponding requested labels , and never changes this decision once made .
both our upper and lower bounds will apply to both selective sampling and pool - based active learning . in the `` realizable case ''
, we assume that the labels are deterministic and generated by a target function that belongs to @xmath43 . in the non - realizable case ( studied in sections [ se : dis ] and [ se : tsy ] ) we do not make this assumption and instead aim to compete with the best function in @xmath43
. given two vectors @xmath20 and @xmath21 and any distribution @xmath54 we denote by @xmath55 ; we also denote by @xmath56 the angle between the vectors @xmath20 and @xmath21 .
throughout this paper we focus on the case where the underlying distribution @xmath42 is log - concave or nearly log - concave .
such distributions have played a key role in the past two decades in several areas including sampling , optimization , and integration algorithms @xcite , and more recently for learning theory as well @xcite . in this section
we first summarize known results about such distributions that are useful for our analysis and then prove a novel structural statement that will be key to our analysis ( theorem [ lemma : vectors - sophist ] ) . in section [ sec : more_distr ] we describe extensions to nearly log - concave distributions as well .
a distribution over @xmath57 is log - concave if @xmath58 is concave , where @xmath59 is its associated density function .
it is isotropic if its mean is the origin and its covariance matrix is the identity .
log - concave distributions form a broad class of distributions : for example , the gaussian , logistic , and uniform distribution over any convex set are log - concave distributions .
the following lemma summarizes known useful facts about isotropic log - concave distributions ( most are from @xcite ; the upper bound on the density is from @xcite ) .
[ isotropic - basic ] assume that @xmath42 is log - concave in @xmath57 and let @xmath59 be its density function . 1 .
if @xmath42 is isotropic then @xmath60 } \leq e^{-\alpha + 1}.$ ] if @xmath61 then : @xmath62 } \leq |b - a|.$ ] 2 .
if @xmath42 is isotropic , then @xmath63 whenever @xmath64 . furthermore , @xmath65 and @xmath66 where @xmath67 is @xmath68 and @xmath69 is @xmath70 , for all @xmath34 of any norm .
3 . all marginals of @xmath42 are log - concave . if @xmath42 is isotropic , its marginals are isotropic as well .
4 . if @xmath71= c^2 $ ] , then @xmath72 } \leq e^{-r + 1}.$ ] 5 . if @xmath42 is isotropic and @xmath61 we have @xmath73 and @xmath74 for all @xmath34 . throughout our paper
we will use the fact that there exists a universal constant @xmath75 such that the probability of disagreement of any two homogeneous linear separators is lower bounded by the c times the angle between their normal vectors .
this follows by projecting the region of disagreement in the space given by the two normal vectors , and then using properties of log - concave distributions in 2-dimensions .
the proof is implicit in earlier works ( e.g. , @xcite ) ; for completeness , we include a proof in appendix [ a : angle ] .
l : angle ] [ l : angle ] assume @xmath42 is an isotropic log - concave in @xmath57 .
then there exists @xmath75 such that for any two unit vectors @xmath20 and @xmath21 in @xmath76 we have @xmath77 to analyze our active and passive learning algorithms we provide a novel characterization of the region of disagreement of two linear separators under a log - concave measure : [ lemma : vectors - sophist ] for any @xmath78 , there is a @xmath79 such that the following holds .
let @xmath20 and @xmath21 be two unit vectors in @xmath57 , and assume that @xmath80 . if @xmath42 is isotropic log - concave in @xmath57 , then : @xmath81 choose @xmath82 .
we will show that , if @xmath83 is large enough relative to @xmath84 , then ( [ e : largemargin ] ) holds .
let @xmath85 .
let @xmath86 be the set whose probability we want to bound .
since the event under consideration only concerns the projection of @xmath34 onto the span of @xmath20 and @xmath21 , lemma [ isotropic - basic](c ) implies we can assume without loss of generality that @xmath87 .
next , we claim that each member @xmath34 of @xmath86 has @xmath88 .
assume without loss of generality that @xmath89 is positive .
( the other case is symmetric . )
then @xmath90 , so the angle of @xmath34 with @xmath20 is obtuse , i.e. @xmath91 . since @xmath92 , this implies that @xmath93 .
but @xmath94 , and @xmath21 is unit length , so @xmath95 , which , since @xmath93 , implies @xmath96 this , since @xmath97 for all @xmath98 $ ] , in turn implies @xmath96 this implies that , if @xmath99 is a ball of radius @xmath100 in @xmath101 , that @xmath102 = \sum_{i=1}^{\infty } { \mathbb p } [ e \cap ( b((i+1 ) c_2 ) - b(i c_2 ) ) ] .\ ] ] to obtain the desired bound , we carefully bound each term in the rhs . choose @xmath103 .
let @xmath104 be the density of @xmath42 .
we have @xmath105 = \int_{(x_1,x_2 ) \in b((i+1 ) c_2 ) - b(i c_2 ) } 1_e(x_1,x_2 ) f(x_1,x_2 ) \ ;
dx_1 dx_2.\ ] ] applying the density upper bound from lemma [ isotropic - basic ] with @xmath87 , there are constants @xmath106 and @xmath107 such that @xmath108 & \leq \int_{(x_1,x_2 ) \in b((i+1 ) c_2 ) - b(i c_2 ) } 1_e(x_1,x_2 ) c_1 \exp(- c_2 c_2 i ) \ ; dx_1 dx_2 \\ & = c_1 \exp(- c_2 c_2 i ) \int_{(x_1,x_2 ) \in b((i+1 ) c_2 ) - b(i c_2 ) } 1_e(x_1,x_2 ) \ ; dx_1 dx_2.\end{aligned}\ ] ] if we include @xmath109 in the integral again , we get @xmath105 \leq c_1 \exp(- c_2 c_2 i ) \int_{(x_1,x_2 ) \in b((i+1 ) c_2 ) } 1_e(x_1,x_2 ) \ ; dx_1 dx_2.\ ] ] now , we exploit the fact that the integral above is a rescaling of a probability with respect to the uniform distribution .
let @xmath110 be the volume of the unit ball in @xmath101 .
then , we have @xmath105 \leq c_1 \exp(- c_2 c_2
i ) c_3 ( i+1)^2 c_2 ^ 2 \alpha/\pi \\ = c_4 c_2 ^ 2 \alpha ( i+1)^2 \exp(-c_2 c_2 i ) , \ ] ] for @xmath111 . returning to ( [ e : shells ] ) , we get @xmath112 = \sum_{i=1}^{\infty } c_4 c_2 ^ 2 \alpha ( i+1)^2 \exp(-c_2 c_2 i ) = c_4 c_2 ^ 2 \times \frac{4 e^{2 c_2 c_2 } - 3 e^{c_2 c_2 } + 1}{\left(e^{c_2 c_2 } - 1 \right)^3 } \times \alpha.\ ] ] since @xmath113 this completes the proof .
we note that a weaker result of this type was proven ( via different techniques ) for the uniform distribution in the unit ball in @xcite .
in addition to being more general , theorem [ lemma : vectors - sophist ] is tighter and more refined even for this specific case this improvement is essential for obtaining tight bounds for polynomial time algorithms for passive learning ( section [ passive ] ) and better bounds for active learning as well .
in this section we analyze a margin - based algorithm for actively learning linear separators under log - concave distributions @xcite ( algorithm [ fig : active - uniform - simple - offline ] ) .
lower bounds proved in section [ s : lower ] show that this algorithm needs exponentially fewer labeled examples than any passive learning algorithm .
this algorithm has been previously proposed and analyzed in @xcite for the special case of the uniform distribution in the unit ball . in this paper
we analyze it for the much more general class of log - concave distributions .
* input * : a sampling oracle for @xmath114 , a labeling oracle , sequences @xmath115 , @xmath116 ( sample sizes ) and @xmath117 , @xmath116 ( cut - off values ) .
* output * : weight vector @xmath118 .
* draw @xmath119 examples from @xmath114 , label them and put them in @xmath120 . * * iterate * @xmath121 * * find a hypothesis @xmath122 with @xmath123 consistent with all labeled examples in @xmath124 .
* * let @xmath125 * * until @xmath126 additional data points are labeled , draw sample @xmath34 from @xmath114 * * * if @xmath127 , then reject @xmath34 , * * * else , ask for label of @xmath34 , and put into @xmath128 .
[ th : agg.margin ] assume @xmath42 is isotropic log - concave in @xmath57 .
there exist constants @xmath129 s.t .
for @xmath130 , and for any @xmath131 , @xmath132 , using algorithm [ fig : active - uniform - simple - offline ] with @xmath133 and @xmath134 , after @xmath135 iterations , we find a separator of error at most @xmath29 with probability @xmath28 .
the total number of labeled examples needed is @xmath17 .
let @xmath75 be the constant from lemma [ l : angle ] .
we will show , using induction , that , for all @xmath136 , with probability at least @xmath137 any @xmath138 consistent with the data in the working set @xmath124 has @xmath139 , so that , in particular , @xmath140 . the case where @xmath141 follows from the standard vc bounds ( see e.g.,@xcite ) .
assume now the claim is true for @xmath142 ( @xmath143 ) , and consider the @xmath144th iteration .
let @xmath145 , and @xmath146 by the induction hypothesis , we know that , with probability at least @xmath147 all @xmath138 consistent with @xmath148 , including @xmath149 , have errors at most @xmath150 .
consider an arbitrary such @xmath138 . by lemma [ l : angle ] we have @xmath151 and @xmath152 , so @xmath153 . applying theorem [ lemma : vectors - sophist ] ,
there is a choice of @xmath106 ( the constant such that @xmath154 ) that satisfies @xmath155 and @xmath156 so @xmath157 now let us treat the case that @xmath158 .
since we are labeling @xmath159 data points in @xmath160 at iteration @xmath142 , classic vapnik - chervonenkis bounds @xcite imply that , if @xmath107 is a large enough absolute constant , then with probability @xmath161 , for all @xmath138 consistent with the data in @xmath124 , @xmath162 finally , since @xmath160 consists of those points that , after projecting onto the direction @xmath149 , fall into an interval of length @xmath163 , lemma [ isotropic - basic ] implies that @xmath164 putting this together with ( [ e : s2.ok ] ) and ( [ e : er.given.s1 ] ) , with probability @xmath165 , we have @xmath166 , completing the proof .
[ passive ] in this section we show how an analysis that was inspired by active learning leads to optimal ( up to constant factors ) bounds for polynomial - time algorithms for passive learning .
[ t : passive ] assume that @xmath42 is zero mean and log - concave in @xmath57 .
there exists an absolute constant @xmath110 s.t . for @xmath130 , and for any @xmath131 , @xmath132 , any algorithm that outputs a hypothesis that correctly classifies @xmath167 examples finds a separator of error at most @xmath29 with probability @xmath168 .
we focus here on the case that @xmath42 is isotropic .
we can treat the non - isotropic case by observing that the two cases are equivalent ; one may pass between them by applying the whitening transform .
( see appendix [ appendix : mainpassiverealizable ] for details . ) while our analysis will ultimately provide a guarantee for any learning algorithm that always outputs a consistent hypothesis , we will use intermediate hypothesis of algorithm [ fig : active - uniform - simple - offline ] in the analysis .
let @xmath75 be the constant from lemma [ l : angle ] . while proving theorem [ th : agg.margin ] , we proved that , if algorithm [ fig : active - uniform - simple - offline ] is run with @xmath133 and @xmath169 , that for all @xmath136 , with probability @xmath170 any @xmath138 consistent with the data in @xmath124 has @xmath171 .
thus , after @xmath172 iterations , with probability at least @xmath173 , any linear classifier consistent with _ all _ the training data has error @xmath174 , since any such classifier is consistent with the examples in @xmath175 .
now , let us analyze the number of examples used , including those examples whose labels were not requested by algorithm [ fig : active - uniform - simple - offline ] .
lemma [ isotropic - basic ] implies that there is a positive constant @xmath176 such that @xmath177 : again , @xmath160 consists of those points that fall into an interval of length @xmath163 after projecting onto @xmath149 .
the density is lower bounded by a constant when @xmath178 , and we can use the bound for @xmath179 when @xmath180 .
the expected number of examples that we need before we find @xmath181 elements of @xmath160 is therefore at most @xmath182 . using a chernoff bound ,
if we draw @xmath183 examples , the probability that we fail to get @xmath181 members of @xmath160 is at most @xmath184 , which is at most @xmath185 if @xmath107 is large enough .
so , the total number of examples needed , @xmath186 , is at most a constant factor more than @xmath187 we can show @xmath188 , completing the proof .
we conclude this section by pointing out several important facts and implications of theorem [ t : passive ] and its proof . 1 . the separator in theorem [ t : passive ] ( and the one in theorem [ th : agg.margin ] ) can be found in _
polynomial time _ , for example by using linear programming .
the analysis of theorem [ t : passive ] also bounds the number of unlabeled examples needed by the active learning algorithm of theorem [ th : agg.margin ] .
this shows that an algorithm can request a nearly optimally small number of labels without increasing the total number of examples required by more than a constant factor .
specifically , in round @xmath144 , we only need @xmath189)$ ] unlabeled examples ( whp ) , where @xmath190 , so the total number of unlabeled examples needed over all rounds is @xmath191
. * theorem [ t : passive ] . * _ assume that @xmath42 is zero mean and log - concave in @xmath57 .
there exists an absolute constant @xmath110 s.t . for @xmath130 , and for any @xmath131 , @xmath132 , any algorithm that outputs a hypothesis that correctly classifies @xmath167 examples finds a separator of error at most @xmath29 with probability @xmath168 .
_ first , let us prove the theorem in the case that @xmath42 is isotropic .
we will then treat the general case at the end of the proof .
while our analysis will ultimately provide a guarantee for any learning algorithm that always outputs a consistent hypothesis , we will use intermediate hypothesis of algorithm [ fig : active - uniform - simple - offline ] in the analysis .
let @xmath75 be the constant from lemma [ l : angle ] . while proving theorem [ th : agg.margin ] , we proved that , if algorithm [ fig : active - uniform - simple - offline ] is run with @xmath133 and @xmath169 , that for all @xmath136 , with probability @xmath170 any @xmath138 consistent with the data in @xmath124 has @xmath171 .
thus , after @xmath172 iterations , with probability at least @xmath173 , any linear classifier consistent with _ all _ the training data has error @xmath174 , since any such classifier is consistent with the examples in @xmath175 .
now , let us analyze the number of examples used , including those examples whose labels were not requested by algorithm [ fig : active - uniform - simple - offline ] .
lemma [ isotropic - basic ] implies that there is a positive constant @xmath176 such that @xmath177 : again , @xmath160 consists of those points that fall into an interval of length @xmath163 after projecting onto @xmath149 .
the density is lower bounded by a constant when @xmath178 , and we can use the bound for @xmath179 when @xmath180 .
the expected number of examples that we need before we find @xmath181 elements of @xmath160 is therefore at most @xmath182 . using a chernoff bound ,
if we draw @xmath183 examples , the probability that we fail to get @xmath181 members of @xmath160 is at most @xmath184 , which is at most @xmath185 if @xmath107 is large enough .
so , the total number of examples needed , @xmath186 , is at most a constant factor more than @xmath320 we claim that @xmath321 .
we have @xmath322 completing the proof in the case that @xmath42 is isotropic .
now let us treat the case in which @xmath42 is not isotropic .
suppose that @xmath323 is the covariance matrix of @xmath42 , so that @xmath324 is the `` whitening transform '' .
suppose , for @xmath325 , an algorithm is given a sample @xmath326 of examples @xmath327 for @xmath328 drawn according to @xmath42 , and @xmath329 labeled by a target hypothesis with weight vector @xmath21 .
note that @xmath282 is consistent with @xmath326 if and only if @xmath330 is consistent with @xmath331 ( so those examples are consistent with @xmath332 ) .
so our analysis of the isotropic case implies that , with probability @xmath2 , for any @xmath282 consistent with @xmath327 , we have @xmath333 which of course means that @xmath334
in this section we consider learning with respect to a more general class of distributions .
we start by providing a general set of conditions on a set @xmath192 of distributions that is sufficient for efficient passive and active learning w.r.t .
distributions in @xmath192 .
we now consider nearly log - concave distributions , an interesting , more general class containing log - concave distributions , considered previously in @xcite and @xcite .
we then prove that isotropic nearly log - concave distributions satisfy our sufficient conditions ; in appendix [ app : more_distr ] , we also show how to remove the assumption that the distribution is isotropic .
a set @xmath193 of distributions is _ admissible _ if it satisfies the following : * there exists @xmath75 such that for any @xmath194 and any two unit vectors @xmath20 and @xmath21 in @xmath76 we have @xmath77 * for any @xmath78 , there is a @xmath79 such that the following holds for all @xmath194 . let @xmath20 and @xmath21 be two unit vectors in @xmath57 s.t . @xmath80 .
then @xmath195 * there are positive constants @xmath196 such that , for any @xmath197 , for any projection @xmath42 of @xmath198 onto a one - dimensional subspace , the density @xmath59 of @xmath42 satisfies @xmath199 for all @xmath34 and @xmath200 for all @xmath34 with @xmath201 .
the proofs of theorem [ th : agg.margin ] and theorem [ t : passive ] can be used without modification to show : [ t : admissible ] if @xmath192 is admissible , then arbitrary @xmath202 can be learned with respect to arbitrary distributions in @xmath192 in polynomial time in the active learning model from @xmath17 labeled examples , and in the passive learning model from @xmath15 examples .
a density function @xmath203 is @xmath204 log - concave if for any @xmath205 $ ] , @xmath206 , @xmath207 , we have @xmath208 . clearly , a density function @xmath59 is log - concave if it is @xmath209-log - concave .
an example of a @xmath210-log - concave distribution is a mixture of two log - concave distributions whose covariance matrices are @xmath211 , and whose means @xmath212 and @xmath213 have @xmath214 . in this section we prove that for any sufficiently small constant @xmath215 , the class of isotropic @xmath204 log - concave distribution in @xmath57 is admissible and has light tails ( this second fact is useful for analyzing the disagreement coefficient in sections [ se : dis ] ) . in doing so we provide several new properties for such distributions , which could be of independent interest .
detailed proofs of our claims appear in appendix [ app : more_distr ] .
we start by showing that for any isotropic @xmath204 log - concave density @xmath59 there exists a log - concave density @xmath216 whose center is within @xmath217 of @xmath59 s center and that satisfies @xmath218 , for @xmath219 as small as @xmath220 .
the fact @xmath219 depends only exponentially in @xmath221 ( as opposed to exponentially in @xmath0 ) is key for being able to argue that such distributions have light tails .
[ basic - betalc ] for any isotropic @xmath204 log - concave density function @xmath59 there exists a log - concave density function @xmath216 that satisfies @xmath218 and @xmath222 , for @xmath223 .
moreover , we have @xmath224 for every unit vector @xmath20 .
note that if the density function @xmath59 is @xmath204 log - concave we have that @xmath225 satisfies that for any @xmath205 $ ] , @xmath226 , @xmath207 , we have @xmath227 .
let @xmath228 be the function whose subgraph is the convex hull of the subgraph of @xmath229 . by using caratheodory s theorem of @xmath57
lies in the convex hull of a set @xmath230 , then there is a subset @xmath231 of @xmath230 consisting of @xmath232 or fewer points such that @xmath34 lies in the convex hull of @xmath231 . ]
we can show that @xmath233 this implies @xmath234 and we can prove by induction on @xmath235 that @xmath236 if we further normalize @xmath237 to make it a density function , we obtain @xmath216 that is log - concave and satisfies @xmath238 where @xmath239 this implies that for any @xmath34 we have @xmath240 . using this fact and concentration properties of @xmath216 ( in particular lemma [ isotropic - basic ] )
, we can show that the center of @xmath216 is close to the center of @xmath59 , as desired .
[ l : angle.beta ] assume @xmath204 is a sufficiently small non - negative constant and let @xmath193 be the set of all isotropic @xmath204 log - concave distributions .
( a ) @xmath193 is admissible .
( b ) any @xmath194 has light tails .
that is : @xmath241 , for @xmath223 .
\(a ) choose @xmath194 .
as in lemma [ l : angle ] , consider the plane determined by @xmath20 and @xmath21 and let @xmath242 denote the projection operator that given @xmath243 , orthogonally projects @xmath34 onto this plane . if @xmath244 then @xmath245 by using the prekopa - leindler inequality @xcite one can show that @xmath246 is @xmath204 log - concave ( see e.g. , @xcite ) . moreover , if @xmath42 is isotropic , than @xmath246 is isotropic as well . by lemma [ basic - betalc ] we know that there exists a @xmath219-isotropic log - concave distribution @xmath247 centered at @xmath248 , @xmath249 , satisfying @xmath218 and @xmath250 for every unit vector @xmath20 , for constants @xmath251 and @xmath252 . for @xmath204 sufficiently small we have @xmath253 .
using this , by applying the whitening transform ( see theorem [ almost isotropic ] in appendix [ app : more_distr ] ) , we can show @xmath254 , for @xmath255 , which implies @xmath256 , for @xmath255 .
using a reasoning as in lemma [ l : angle ] we get @xmath77 the generalization of theorem [ lemma : vectors - sophist ] follows from a similar proof , except using theorem [ almost isotropic ] .
the density bounds in the @xmath257 case also follow from theorem [ almost isotropic ] as well .
\(b ) since @xmath38 is isotropic , we have @xmath258 = d$ ] ( where @xmath59 is its associated density ) . by lemma [ basic - betalc ] , there exists a log - concave density @xmath216 such that @xmath218 , for @xmath223 .
this implies @xmath259 \leq cd$ ] . by lemma [ isotropic - basic ]
we get that that under @xmath216 , @xmath260 , so under @xmath59 we have @xmath241 .
using theorem [ t : admissible ] and theorem [ l : angle.beta](a ) we obtain : [ th : agg.margin : beta ] let @xmath215 be a sufficiently small constant .
assume that @xmath42 is an isotropic @xmath204 log - concave distribution in @xmath57 .
then arbitrary @xmath202 can be learned with respect to @xmath42 in polynomial time in the active learning model from @xmath17 labeled examples , and in the passive learning model from @xmath15 examples .
* lemma [ basic - betalc ] . * _ for any isotropic @xmath204 log - concave density function @xmath59 there exists a log - concave density function @xmath216 that satisfies @xmath218 and @xmath222 , for @xmath223 .
moreover , we have @xmath224 for every unit vector @xmath20 .
_ note that if the density function @xmath59 is @xmath204 log - concave we have that @xmath225 satisfies that for any @xmath205 $ ] , @xmath226 , @xmath207 , we have @xmath227 .
let @xmath228 be the function whose subgraph is the convex hull of the subgraph of @xmath229 .
that is , @xmath335 is the maximum of all values of @xmath336 for any @xmath337 and @xmath338 $ ] such that @xmath339 and @xmath340 .
note that , if the components of @xmath341 are @xmath342 , we can get @xmath335 by starting with @xmath343 taking the convex combination of the members of @xmath344 with mixing coefficients @xmath345 , and then reading off the last component .
caratheodory s theorem of @xmath57 lies in the convex hull of a set @xmath230 , then there is a subset @xmath231 of @xmath230 consisting of @xmath232 or fewer points such that @xmath34 lies in the convex hull of @xmath231 . ]
implies that we can get the same result using a mixture of at most @xmath232 members of @xmath344 . in other words , we can assume without loss of generality that @xmath346 , so that @xmath347 because of the case where @xmath348 concentrates all its weight on one component , we have @xmath234
. we also claim that @xmath349 we will prove this by induction on @xmath235 , treating the case in which @xmath232 is a power of @xmath350 .
( by padding with zeroes if necessary , we may assume without loss of generality that @xmath232 is a power of @xmath350 . )
the base case , in which @xmath351 , follows immediately from the definitions .
let @xmath346 .
assume that @xmath352 , @xmath353 , @xmath354 .
we can write this as : @xmath355 where @xmath356 , for all @xmath51 .
now , by induction we have : @xmath357 the last inequality follows from the fact that @xmath358 .
so , we have proved ( [ e : logbound ] ) .
if we further normalize @xmath237 to make it a density function , we obtain @xmath216 that is log - concave and satisfies @xmath238 where @xmath359 this implies that for any @xmath34 we have @xmath240 .
we now show that the center of @xmath216 is close to the center of @xmath59 .
we have : @xmath360 dr}.\end{aligned}\ ] ] using concentration properties of @xmath216 ( in particular lemma [ isotropic - basic ] ) we get @xmath361 as desired .
[ almost isotropic ] ( i ) let @xmath362 be the density function of a log - concave distribution centered at @xmath248 and with covariance matrix @xmath363 $ ] .
assume @xmath59 satisfies @xmath364 and @xmath250 for every unit vector @xmath20 , for @xmath365 constant close to @xmath366 .
we have : ( a ) assume @xmath367 .
then there exist an universal constant @xmath75 s.t .
we have @xmath368 , for all @xmath34 with @xmath369 . ( b )
assume @xmath370 . there exist universal constants @xmath176 and @xmath83 such that @xmath371 for all @xmath34 .
\(ii ) let @xmath372 be the density function of a log - concave distribution centered at @xmath373 with standard deviation @xmath374 .
then @xmath375 for all @xmath34 .
if furthermore @xmath59 satisfies @xmath376 \leq c$ ] for @xmath365 and @xmath377 , then we have @xmath378 for some universal constant @xmath75 .
\(i ) let @xmath379 .
then @xmath380 is a log - concave distribution in the isotropic position .
moreover , the density function of @xmath381 is given by @xmath382 let @xmath383 $ ] .
we have @xmath384={\mathbb e}[x x^t]- z z^t= m- z z^t.\ ] ] also , the fact @xmath250 for every unit vector @xmath20 is equivalent to @xmath385 u \leq c\ ] ] for every unit vector @xmath20 .
using @xmath386 , @xmath387 , and @xmath388 we get that @xmath389 $ ] , @xmath390 $ ] , and @xmath391 $ ] .
we also have @xmath364 and @xmath392 .
all these imply that @xmath393 \(a ) for @xmath394 we have @xmath395 , where @xmath396 is a unit vector , so @xmath397 . if @xmath255 we have @xmath398 , so by lemma [ isotropic - basic ] we have @xmath399 , so @xmath400 , for some universal constants @xmath401 , as desired .
\(b ) we have @xmath402 . by lemma
[ isotropic - basic ] ( b ) we have @xmath403}.\ ] ] by the triangle inequality we further obtain : @xmath404 } \exp{\left[-c \norm{a^{-1/2}x}\right]}.\ ] ] for @xmath370 , we can show that @xmath405 .
it is enough to show @xmath406 , or that @xmath407 , where @xmath408 ( so @xmath409 ) .
this is equivalent to @xmath410 , which is true since the matrix @xmath411 is positive semi - definite .
\(ii ) define @xmath412 .
we have @xmath413=0 $ ] and @xmath414 = 1 $ ] .
the density @xmath381 of @xmath380 is given by @xmath415 .
now , since @xmath381 is isotropic and log - concave , we can apply lemma [ isotropic - basic](e ) to @xmath381 .
so @xmath416 for all @xmath33 .
so , @xmath417 for all @xmath33 , which implies @xmath375 for all @xmath34 .
the second part follows as in theorem [ almost isotropic ] . in this section , we extend theorem [ th : agg.margin ] to the case of arbitrary covariance matrices .
[ t : any.cov.active ] [ t : any.cov.active ] if all distributions in @xmath192 are zero - mean and log - concave in @xmath57 , then arbitrary @xmath202 be learned in polynomial time from arbitrary distributions in @xmath193 in the active learning model from @xmath17 labeled examples , and in the passive learning model from @xmath15 examples .
our proof is through a series of lemma .
first , @xcite have shown how to reduce to the nearly isotropic case .
[ l : lv.reduce ] for any constant @xmath418 , there is a polynomial time algorithm that , given polynomially many samples from a log - concave distribution @xmath42 , outputs an estimate @xmath323 of the covariance matrix of @xmath42 such that , with probability @xmath2 the distribution @xmath198 obtained by sampling @xmath34 from @xmath42 and producing @xmath419 has @xmath420 for all unit vectors @xmath20 . as a result of lemma [ l : lv.reduce ] , we can assume without loss of generality that the distribution @xmath42 satisfies @xmath421 for an arbitrarily small constant @xmath422 . by theorem [ almost isotropic ] , this implies that , without loss of generality , there are constants @xmath423 such that , for the density @xmath59 of any one or two - dimensional marginal @xmath198 of @xmath42 , we have @xmath424 and for all @xmath34 , @xmath425 we will show that these imply that @xmath193 is admissible .
[ l : angle.anycov ] ( a ) there exists @xmath75 such that for any two unit vectors @xmath20 and @xmath21 in @xmath76 we have @xmath426 \(b ) for any @xmath427 , there is a @xmath428 such that the following holds .
let @xmath20 and @xmath21 be two unit vectors in @xmath57 , and assume that @xmath80 .
then @xmath429 \(a ) projecting @xmath42 onto a subspace can only reduce the norm of its mean , and its variance in any direction .
therefore , as in the proof of lemma [ l : angle ] , we may assume without loss of generality that @xmath430 . here
, let us define @xmath314 to be the region of disagreement between @xmath315 and @xmath316 intersected with the ball @xmath431 of radius @xmath83 in @xmath317
. then we have @xmath432 ( b ) this proof basically amounts to observing that everything that was needed for the proof of theorem [ lemma : vectors - sophist ] is true for @xmath42 , because of ( [ e : dens.low ] ) and ( [ e : dens.up ] ) .
armed with lemma [ l : angle.anycov ] , to prove theorem [ t : any.cov.active ] , we can just apply theorem [ t : admissible ] .
in this section we give lower bounds on the label complexity of passive and active learning of homogeneous linear separators when the underlying distribution is @xmath204 log - concave , for a sufficiently small constant @xmath204 .
these lower bounds are information theoretic , applying to any procedure , that might not be necessarily computationally efficient .
the proof is in appendix [ a : lower ] .
[ t : lower ] for a small enough constant @xmath204 we have : ( 1 ) for any @xmath204 log - concave distribution @xmath42 whose covariance matrix has full rank , the sample complexity of learning origin - centered linear separators under @xmath42 in the passive learning model is @xmath261 ( 2 ) the sample complexity of active learning of linear separators under @xmath204 log - concave distributions is @xmath262 note that , if the covariance matrix of @xmath42 does not have full rank , the number of dimensions is effectively less than @xmath0 , so our lower bound essentially applies for all log - concave distributions .
the proof of our lower bounds ( theorem [ t : lower ] ) relies on a lower bound on the packing numbers @xmath433 .
recall that the @xmath29-packing number , @xmath433 , is the maximal cardinality of an @xmath29-separated set with classifiers from @xmath43 , where we say that @xmath434 are @xmath29-separated w.r.t @xmath192 if @xmath435 for any @xmath436 .
[ packing - logconcave ] there is a positive constant @xmath75 such that , for all @xmath437 , the following holds .
assume that @xmath42 is @xmath204 log - concave in @xmath57 , and that its covariance matrix has full rank .
for all sufficiently small @xmath29 , @xmath438 , we have @xmath439 we first prove the lemma in the case that @xmath42 is isotropic .
the proof in this case follows the outline of a proof for the special case of the uniform distribution in @xcite .
let @xmath440 be the uniform distribution on the surface of the unit ball in @xmath76 .
by theorem [ l : angle.beta ] , there exists @xmath75 such that for any two unit vectors @xmath20 and @xmath21 in @xmath76 we have @xmath441 this implies that for a fixed @xmath20 the probability that a randomly chosen @xmath21 has @xmath442 is upper bounded by the volume of those vectors in the interior of the unit ball whose angle is at most @xmath443 divided by the volume of the unit ball . using known bounds on this ratio ( see @xcite ) we have @xmath444 \leq \frac{1}{\sqrt{d } } \left(\frac{2 \epsilon}{c}\right)^{d-1}$ ] , so @xmath445 \leq \frac{1}{\sqrt{d } } \left(\frac{2 \epsilon}{c}\right)^{d-1}$ ] .
that means that for a fixed @xmath446 if we pick @xmath446 normal vectors at random from the unit ball , then the expected number of pairs of half - spaces that are @xmath29-close according to @xmath42 is at most @xmath447 . removing one element of each pair from @xmath326 yields a set of @xmath448 halfspaces that are @xmath29-separated . setting @xmath449 , leads the desired result . to handle the non - isotropic case ,
suppose that @xmath323 is the covariance matrix of @xmath42 , so that @xmath324 is the whitening transform .
let @xmath198 be the whitened version of @xmath42 , i.e. the distribution obtained by first choosing @xmath34 from @xmath42 , and then producing @xmath419 .
we have @xmath450 ( because @xmath451 iff @xmath452 ) .
so we can use an @xmath29-packing w.r.t .
@xmath198 to construct an @xmath29-packing of the same size w.r.t .
@xmath42 .
now we are ready to prove theorem [ t : lower ]
. * theorem [ t : lower ] . * _ _ for a small enough constant @xmath204
we have : ( 1 ) for any @xmath204 log - concave distribution @xmath42 whose covariance matrix has full rank , the sample complexity of learning origin - centered linear separators under @xmath42 in the passive learning model is @xmath261 ( 2 ) the sample complexity of active learning of linear separators under @xmath204 log - concave distributions is @xmath262 first , let us consider passive pac learning .
it is known @xcite that , for any distribution @xmath42 , the sample complexity of passive pac learning origin - centered linear separators w.r.t .
@xmath42 is at least @xmath453 applying lemma [ packing - logconcave ] gives an @xmath454 lower bound .
it is known @xcite that , if for each @xmath29 , there is a pair of classifier @xmath455 such that @xmath456 , then the sample complexity of pac learning is @xmath457 ; this requirement is satisfied by @xmath42 .
now let us consider the sample complexity of active learning .
as shown in @xcite , in order to output a hypothesis of error at most @xmath29 with probabality at least @xmath28 , where @xmath458 and active learning algorithm that is allowed to make arbitrary yes - no queries must make @xmath459 queries . using this together with lemma [ packing - logconcave ] we get the desired result .
[ se : dis ] we consider two closely related distribution dependent capacity notions : the alexander capacity and the disagreement coefficient ; they have been widely used for analyzing the label complexity of non - aggressive active learning algorithms @xcite .
we begin with the definitions . for @xmath263 , define @xmath264 . for any @xmath265 ,
define the region of disagreement as @xmath266 define the alexander capacity function @xmath267 for @xmath268 w.r.t .
@xmath42 as : @xmath269 define the disagreement coefficients for @xmath268 w.r.t .
@xmath42 as : @xmath270.$ ] the following is our bound in the disagreement coefficient .
its proof is in appendix [ a : dis - coeff : beta ] .
[ th : dis - coeff : beta ] let @xmath215 be a sufficiently small constant .
assume that @xmath42 is an isotropic @xmath204 log - concave distribution in @xmath57 .
for any @xmath271 , for any @xmath29 , @xmath272 is @xmath273 .
thus @xmath274 .
theorem [ th : dis - coeff : beta ] immediately leads to concrete bounds on the label complexity of several algorithms in the literature @xcite .
for example , by composing it with a result of @xcite , we obtain a bound of @xmath275 for agnostic active learning when @xmath42 is isotropic log - concave in @xmath57 ; that is we only need @xmath276 label requests to output a classifier of error at most @xmath277 , where @xmath278
. * theorem [ th : dis - coeff : beta ] . * _ let @xmath215 be a sufficiently small constant . assume that @xmath42 is an isotropic @xmath204 log - concave distribution in @xmath57 . for any @xmath271 , for any @xmath29 , @xmath272 is @xmath273 .
thus @xmath274 . _
roughly , we will show that almost all @xmath34 classified by a large enough margin by @xmath280 are not in @xmath460 , because all hypotheses agree with @xmath280 about how to classify such @xmath34 , and therefore all pairs of hypotheses agree with each other .
consider @xmath282 such that @xmath461 ; by theorem [ l : angle.beta ] we have @xmath462 .
define @xmath463 as in the proof of theorem [ l : angle.beta ] . for any @xmath34
such that @xmath464 we have @xmath465 thus , if @xmath34 also satisfies @xmath466 we have @xmath467 . since this is true for all @xmath282 , any such @xmath34 is not in @xmath468 . by theorem [ l : angle.beta ]
we have , for a constant @xmath83 , that @xmath469 moreover , by theorem [ l : angle.beta ] we also have @xmath470 } \leq r.\ ] ] these both imply @xmath471 .
in this section we consider a variant of the tsybakov noise condition @xcite .
we assume that the classifier @xmath229 that minimizes @xmath279 is a linear classifier , and that , for the weight vector @xmath280 of the optimal classifier , there exist known parameters @xmath281 such that , for all @xmath282 , we have @xmath283 by generalizing theorem [ lemma : vectors - sophist ] so that it provides a stronger bound for larger margins , and combining the result with the other lemmas of this paper and techniques from @xcite , we get the following . [ th : agg.margin : tsybakov ] let @xmath284 .
assume that the distribution @xmath36 satisfies the tsybakov noise condition for constants @xmath285 and @xmath286 , and that the marginal @xmath42 on @xmath287 is isotropic log - concave .
( 1 ) if @xmath288 , we can find a separator with excess error @xmath174 with probability @xmath28 using @xmath289 labeled examples in the active learning model , and @xmath15 labeled examples in the passive learning model .
( 2 ) if @xmath290 , we can find a separator with excess error @xmath174 with probability @xmath28 using @xmath291 labeled examples in the active learning model .
in the case @xmath288 ( that is more general than the massart noise condition ) our analysis leads to optimal bounds for active and passive learning of linear separators under log - concave distributions , improving the dependence on @xmath0 over previous best known results @xcite .
our analysis for tsybakov noise ( @xmath292 ) leads to bounds on active learning with improved dependence on @xmath0 over previous known results @xcite in this case as well .
proofs and further details appear in appendix [ a : massart ] .
the label sample complexity of our active learning algorithm for learning homogeneous linear separators under isotropic logconcave distributions is @xmath17 , while our lower bound for this setting is @xmath293 our upper bound is achieved by an algorithm that uses a polynomial number of unlabeled training examples , and polynomial time . if an unbounded amount of computation time and an unbounded number of unlabeled examples are available , it seems to be easy to learn to accuracy @xmath29 using @xmath294 label requests , no matter what the value of @xmath295 .
( roughly , the algorithm can construct an @xmath29-cover to initialize a set of candidate hypotheses , then repeatedly wait for an unlabeled example that evenly splits the current list of candidates , and ask its label , eliminated roughly half of the candidates . ) it would be interesting to know what is the best label complexity for a polynomial - time algorithm , or even an algorithm that is constrained to use a polynomial number of unlabeled examples .
conceptually , our analysis of erm for passive learning under ( nearly ) log - concave distributions is based on a more aggressive localization than those considered previously in the literature .
it would be very interesting to extend this analysis as well as our analysis for active learning to arbitrary distributions and more general concept spaces .
[ [ acknowledgements ] ] acknowledgements + + + + + + + + + + + + + + + + we thank steve hanneke for a number of useful discussions .
this work was supported in part by nsf grant ccf-0953192 , afosr grant fa9550 - 09 - 1 - 0538 , and a microsoft research faculty fellowship .
56 [ 1]#1 [ 1]`#1 ` urlstyle [ 1]doi : # 1 k.s .
rates of growth and sample moduli for weighted empirical processes indexed by sets . _ probability theory and related fields _ , 1987 .
n. alon . a non - linear lower bound for planar epsilon - nets .
_ focs _ , pages 341346 , 2010 .
d. applegate and r. kannan .
sampling and integration of near log - concave functions . in _
stoc _ , 1991 .
p. assouad .
plongements lipschitziens dans .
france _ , 1110 ( 4):0 429448 , 1983 .
m. f. balcan , a. beygelzimer , and j. langford .
agnostic active learning . in _
icml _ , 2006 .
balcan , a. broder , and t. zhang .
margin based active learning . in _ colt _ , 2007 .
balcan , s. hanneke , and j. wortman . the true sample complexity of active learning . in _
colt _ , 2008 .
p. l. bartlett , o. bousquet , and s. mendelson .
local rademacher complexities . _
annals of statistics _ , 2005 .
a. beygelzimer , d. hsu , j. langford , and t. zhang .
agnostic active learning without constraints . in _ nips _ , 2010 .
a. blumer , a. ehrenfeucht , d. haussler , and m. k. warmuth .
learnability and the vapnik - chervonenkis dimension .
_ jacm _ , 360 ( 4):0 929965 , 1989 .
s. boucheron , o. bousquet , and g. lugosi .
theory of classification : a survey of recent advances .
_ esaim : probability and statistics _ , 2005 .
n. h. bshouty , y. li , and p. m. long . .
_ jcss _ , 2009 . c. caramanis and s. mannor .
an inequality for nearly log - concave distributions with applications to learning .
_ ieee transactions on information theory _ , 2007 . r. castro and r. nowak .
minimax bounds for active learning . in _
colt _ , 2007 .
n. cesa - bianchi , c. gentile , and l. zaniboni .
learning noisy linear classifiers via adaptive and selective sampling .
_ machine learning _ , 2010 .
k. l. clarkson and k. varadarajan .
improved approximation algorithms for geometric set cover .
_ discrete comput .
_ , 370 ( 1):0 4358 , 2007 .
d. cohn , l. atlas , and r. ladner . .
in _ icml _ , 1994 .
s. dasgupta .
coarse sample complexity bounds for active learning . in _ nips _ , volume 18 , 2005 .
s. dasgupta
. active learning . _ encyclopedia of machine learning _
, 2011 .
s. dasgupta , a. kalai , and c. monteleoni .
analysis of perceptron - based active learning . in _ colt _ , 2005 .
s. dasgupta , d.j .
hsu , and c. monteleoni .
a general agnostic active learning algorithm .
_ advances in neural information processing systems _ , 20 , 2007 .
o. dekel , c. gentile , and k. sridharan .
selective sampling and active learning from single and multiple teachers .
_ journal of machine learning research _
, 2012 .
a. ehrenfeucht , d. haussler , m. kearns , and l. g. valiant .
a general lower bound on the number of examples needed for learning . _ information and computation _ , 1989 .
y. freund , h.s .
seung , e. shamir , and n. tishby .
selective sampling using the query by committee algorithm .
_ machine learning _
, 280 ( 2 - 3):0 133168 , 1997 . e. j. friedman .
active learning for smooth problems . in _ colt _ , 2009 .
r. j. gardner . .
soc . _ , 2002 .
e. gin and v. koltchinskii .
concentration inequalities and asymptotic results for ratio type empirical processes . _
the annals of probability _ , 340 ( 3):0 11431216 , 2006 .
a. gonen , s. sabato , and s. shalev - shwartz .
efficient pool - based active learning of halfspaces . in _
icml _ , 2013 .
s. hanneke .
a bound on the label complexity of agnostic active learning . in _ icml _ , 2007 .
s. hanneke .
rates of convergence in active learning .
_ the annals of statistics _ , 390 ( 1):0 333361 , 2011 .
s. hanneke and l. yang .
surrogate losses in passive and active learning , 2012 . http://arxiv.org/abs/1207.3772 .
d. haussler , n. littlestone , and m. k. warmuth .
predicting @xmath296-functions on randomly drawn points . _ information and computation _
, 1150 ( 2):0 129161 , 1994 .
david haussler and emo welzl .
epsilon nets and simplex range queries . _ disc .
comp . geometry _ , 2:0 127151 , 1987 .
a. kalai , a. klivans , y. mansour , and r. servedio .
agnostically learning halfspaces . in _ proceedings of the 46th annual symposium on the foundations of computer science ( focs ) _ , 2005 .
m. kearns and u. vazirani .
_ an introduction to computational learning theory_. mit press , 1994 .
a. r. klivans , p. m. long , and r. a. servedio . learning halfspaces with malicious noise .
_ jmlr _ , 2009 .
a. r. klivans , p. m. long , and a. tang .
s algorithm learns intersections of halfspaces with respect to log - concave distributions . in _
random _ , 2009 .
v. koltchinskii .
rademacher complexities and bounding the excess risk in active learning . _ journal of machine learning research _ , 11:0 24572485 , 2010 .
j. komls , j. pach , and g. woeginger .
almost tight bounds on epsilon - nets . _ discrete and computational geometry _ , 7:0 163173 , 1992 .
s. r. kulkarni , s. k. mitter , and j. n. tsitsiklis . .
_ machine learning _ , 1993 .
p. m. long . on the sample complexity of pac learning halfspaces against the uniform distribution .
_ ieee transactions on neural networks _ , 60 ( 6):0 15561559 , 1995 .
p. m. long .
an upper bound on the sample complexity of pac learning halfspaces with respect to the uniform distribution . _ information processing letters _ , 2003 .
l. lovasz and s. vempala . .
_ random structures and algorithms _ , 2007 .
e. mammen and a.b . tsybakov .
smooth discrimination analysis . _ the annals of statistics _ , 27:0 18081829 , 1999 .
p. massart and e. nedelec .
risk bounds for statistical learning . _
the annals of statistics _ , 2006 .
s. mendelson . estimating the performance of kernel classes .
_ journal of machine learning research _ , 4:0 759771 , 2003 .
r. nowak . .
_ ieee transactions on information theory _ , 2011 .
j. pach and p.k .
agarwal . _ combinatorial geometry_. john wiley and sons , 1995 .
m. raginsky and a. rakhlin .
lower bounds for passive and active learning . in _ nips _ , 2011 .
l.g . valiant .
a theory of the learnable . _ communications of the acm _ , 270 ( 11):0 11341142 , 1984 .
s. van de geer .
_ empirical processes in m - estimation_. cambridge series in statistical and probabilistic methods , 2000 .
a. van der vaart and j. a. wellner .
_ weak convergence and empirical processes with applications to statistics_. springer , 1996 . v. vapnik and a. chervonenkis . on the uniform convergence of relative frequencies of events to their probabilities
_ theory of probability and its applications _ , 160 ( 2):0 264280 , 1971 . v. n. vapnik . _ estimation of dependencies based on empirical data_. springer verlag , 1982 . v. n. vapnik .
_ statistical learning theory_. john wiley and sons , 1998 .
s. vempala .
a random - sampling - based algorithm for learning intersections of halfspaces .
_ jacm _ , 570 ( 6 ) , 2010 .
* learning with noise . alexander capacity and the disagreement coefficient * roughly speaking the alexander capacity @xcite quantifies how fast the region of disagreement of the set of classifiers at distance @xmath100 of the optimal classifier collapses as a function @xmath100 ; of a set of classifiers @xmath43 is the of set of instances @xmath34 s.t . for each @xmath297 there
exist two classifiers @xmath298 that disagree about the label of @xmath34 . ]
the disagreement coefficient @xcite additionally involves the supremum of @xmath100 over a range of values .
@xcite provides guarantees on these quantities ( for sufficiently small @xmath100 ) for general classes of functions in @xmath76 if the underlying data distribution is sufficiently smooth .
our analysis implies much tighter bounds for linear separators under log - concave distributions ( matching what was known for the much less general case of nearly uniform distribution over the unit sphere ) ; furthermore , we also analyze the nearly log - concave case where we allow an arbitrary number of discontinuities , a case not captured by the @xcite conditions at all .
this immediately implies concrete bounds on the labeled data complexity of several algorithms in the literature including the @xmath31 algorithm @xcite and the dhm algorithm @xcite , with implications for the purely agnostic case ( i.e. , arbitrary forms of noise ) , as well as the koltchinskii s algorithm @xcite and the cal algorithm @xcite .
furthermore , in the realizable case and under tsybakov noise , we show even better bounds , by considering aggressive active learning algorithms .
note that as opposed to the realizable case , all existing active learning algorithms analyzed under massart and tsybakov noise conditions using the learning model analyzed in this paper ( including our algorithms in theorem [ th : agg.margin : tsybakov ] ) , as well as those for the agnostic setting , are not known to run in time @xmath299 .
in fact , even ignoring the optimality of sample complexity , there are no known algorithms for passive learning that run in time @xmath299 for general values of @xmath29 , even for the massart noise condition and under log - concave distributions .
existing works on agnostic passive learning under log - concave distributions either provide running times @xmath300 ( e.g. , the work of @xcite ) or can only achieve values of @xmath29 that are significantly larger than the noise rate @xcite . *
other work on active learning * several papers @xcite present efficient online learning algorithms in the selective sampling framework , where labels must be actively queried before they are revealed . under the assumption that the label conditional distribution is linear function determined by a fixed target vector , they provide bounds on the regret of the algorithm and on the number of labels it queries when faced with an adaptive adversarial strategy of generating the instances .
as pointed by @xcite , these results can be converted to a statistical setting when the instances @xmath301 are drawn i.i.d and they further assume a margin condition .
in this setting they obtain exponential improvement in label complexity over passive learning .
while very interesting , these results are incomparable to ours ; their techniques significantly exploit the linear noise condition to get these improvements note that such an improvement would not be possible in the realizable case ( as pointed for example in @xcite ) .
@xcite considers an interesting abstract `` generalized binary search '' problem with applications to active learning ; while these results apply for more general concept spaces , it is not clear how to implement the resulting procedures in polynomial time and by using access to only a polynomial number of unlabeled samples from the underlying distribution ( as required by the active learning model ) .
another interesting recent work is that of @xcite , which study active learning of linear separators via an aggressive algorithm using a margin condition , using a general approximation guarantee on the number of labels requested ; note that while these results work for potentially more general distributions , as opposed to ours , they do not come with explicit ( tight ) bounds on the label complexity .
* @xmath29-nets , learning , and geometry * small @xmath29-nets are useful for many applications , especially in computational geometry ( see @xcite ) .
the same fundamental techniques of @xcite have been applied to establish the existence of small @xmath29-nets @xcite and to bound the sample complexity of learning @xcite , and a number of interesting upper and lower bounds on the smallest possible size of @xmath29-nets have been obtained @xcite .
our analysis implies a @xmath302 upper bound on the size of an @xmath29-net for a set of regions of disagreement between all possible linear classifiers and the target , when the distribution is zero - mean and log - concave . in particular
, since in theorem [ t : passive ] we prove that any hypothesis consistent with the training data has error rate @xmath174 with probability @xmath2 , setting @xmath295 to a constant gives a proof of a @xmath302 bound on the size of an @xmath29-net for the following set : @xmath303
* lemma [ l : angle ] . * _ assume @xmath42 is an isotropic log - concave in @xmath57
. then there exists @xmath75 such that for any two unit vectors @xmath20 and @xmath21 in @xmath76 we have @xmath77 _ consider two unit vectors @xmath20 and @xmath21 .
let @xmath242 denote the projection operator that , given @xmath243 , orthogonally projects @xmath34 onto the plane determined by @xmath20 and @xmath21 .
that is , if we define an orthogonal coordinate system in which coordinates @xmath304 lie in this plane and coordinates @xmath305 are orthogonal to this plane , then @xmath306 . also , given distribution @xmath42 over @xmath57 , define @xmath307 to be the distribution given by first picking @xmath308 and then outputting @xmath309 .
that is , @xmath307 is just the marginal distribution over coordinates @xmath304 in the above coordinate system .
notice that if @xmath309 then @xmath310 where @xmath311 and @xmath312 .
so , if @xmath244 then @xmath313 by lemma [ isotropic - basic](c ) , we have that if @xmath42 is isotropic and log - concave , then @xmath246 is as well . let @xmath314 to be the region of disagreement between @xmath315 and @xmath316 intersected with the ball of radius @xmath179 in @xmath317 .
the probability mass of @xmath314 under @xmath246 is at least the volume of @xmath314 times @xmath318 .
so , using lemma [ isotropic - basic](b ) @xmath319 as desired .
in this section we analyze label complexity for active learning under the popular massart and tsybakov noise conditions , proving theorem [ th : agg.margin : tsybakov ] .
we consider a variant of the tsybakov noise condition @xcite .
we assume that the classifier @xmath229 that minimizes @xmath279 is a linear classifier , and that , for the weight vector @xmath280 of that optimal classifier , there exist known parameters @xmath281 such that , for all @xmath282 , we have @xmath472 by generalizing theorem [ lemma : vectors - sophist ] so that it provides a stronger bound for larger margins , and combining the result with the other lemmas of this paper and techniques from @xcite , we get the following . * theorem [ th : agg.margin : tsybakov ] . * _ let @xmath284 .
assume that the distribution @xmath36 satisfies the tsybakov noise condition for constants @xmath285 and @xmath286 , and that the marginal @xmath42 on @xmath287 is isotropic log - concave .
( 1 ) if @xmath288 , we can find a separator with excess error @xmath174 with probability @xmath28 using @xmath289 labeled examples in the active learning model , and @xmath15 labeled examples in the passive learning model .
( 2 ) if @xmath290 , we can find a separator with excess error @xmath174 with probability @xmath28 using @xmath291 labeled examples in the active learning model .
_ note that the case where @xmath288 is more general than the well - known massart noise condition @xcite . in this case , for active learning , theorem [ th : agg.margin : tsybakov ] improves over the previously best known results @xcite by a ( disagreement coefficient ) @xmath473 factor . for passive learning ,
the bound on the total number of examples needed improves by @xmath474 factor the previously known best bound of @xcite .
it is consistent with recent lower bounds of @xcite that include @xmath474 because those bounds are for a worst - case domain distribution , subject to a constraint on @xmath475 .
when @xmath290 , the previously best result for active learning @xcite is @xmath476 combining this with our new bound on @xmath473 ( theorem [ th : dis - coeff : beta ] ) we get a bound of @xmath477 for log - concave distributions .
so our theorem [ th : agg.margin : tsybakov ] saves roughly a factor of @xmath478 , at the expense of an extra @xmath479 factor . * input * : a sampling oracle for @xmath114 , and a labeling oracle a sequence of sample sizes @xmath115 , @xmath116 ; a sequence of cut - off values @xmath117 , @xmath116 a sequence of hypothesis space radii @xmath480 , @xmath116 ; a sequence of precision values @xmath481 , @xmath116 * pick random @xmath482 : @xmath483 . *
draw @xmath119 examples from @xmath484 , label them and put into @xmath485 . *
* * iterate * @xmath121 * * * find @xmath486 ( @xmath123 ) to approximately minimize training error : @xmath487 .
* * * clear the working set @xmath485 * * * until @xmath126 additional data points are labeled , draw sample @xmath34 from @xmath484 * * * * if @xmath127 , reject @xmath34 * * * * otherwise , ask for label of @xmath34 , and put into @xmath485 + * end iterate * assume now the claim is true for @xmath142 ( @xmath491 ) .
then at the @xmath144-th iteration , we can let @xmath492 @xmath149 has excess errors at most @xmath493 , implying , using ( [ e : angle.up.massart ] ) , that @xmath494 . by assumption , @xmath495 .
taking the sum , we obtain : @xmath498 therefore : @xmath499 by standard vapnik - chervonenkis bounds , we can choose @xmath219 s.t . with @xmath159 samples ,
we obtain @xmath500 with probability @xmath501 .
therefore @xmath502 with probability @xmath490 , as desired . [ lemma : vectors - sophist.bbig ] there is a positive constant @xmath75 such that the following holds .
let @xmath20 and @xmath21 be two unit vectors in @xmath57 , and assume that @xmath503 .
assume that @xmath42 is isotropic log - concave in @xmath57 .
then , for any @xmath504 , we have @xmath505 for absolute constants @xmath506 and @xmath507 .
next , we claim that each member @xmath34 of @xmath86 has @xmath508 .
assume without loss of generality that @xmath89 is positive .
( the other case is symmetric . )
then @xmath90 , so the angle of @xmath34 with @xmath20 is obtuse , i.e. @xmath91 . since @xmath509 , this implies that @xmath510 but @xmath94 , and @xmath21 is unit length , so @xmath95 , which , using ( [ e : theta.big.bbig ] ) , implies @xmath96 which , since @xmath511 for all @xmath512 $ ] , in turn implies @xmath96 this implies that , if @xmath99 is a ball of radius @xmath100 in @xmath101 , that @xmath513 = \sum_{i=1}^{\infty } { \mathbb p } [ e \cap ( b((i+1 ) ( b/\eta ) ) - b(i ( b/\eta ) ) ) ] .\ ] ] let us bound one of the terms in rhs .
choose @xmath103 .
let @xmath515 . applying the density upper bound from lemma [ isotropic - basic ] with @xmath87
, there are constants @xmath106 and @xmath107 such that @xmath514 \\ & & \leq \int_{(x_1,x_2 ) \in r_i } 1_e(x_1,x_2 ) c_1 \exp(- ( b/\eta ) c_2 i ) dx_1 dx_2 \\ & & = c_1 \exp(- ( b/\eta ) c_2 i ) \cdot \\ & & \int_{(x_1,x_2 ) \in r_i } 1_e(x_1,x_2 ) \ ; dx_1 dx_2.\end{aligned}\ ] ] if we include @xmath516 in the integral again , we get @xmath514 \\ & & \leq c_1 \exp(- ( b/\eta ) c_2 i ) \int_{(x_1,x_2 ) \in b((i+1 ) ( b/\eta ) ) } 1_e(x_1,x_2 ) \ ; dx_1 dx_2.\end{aligned}\ ] ] now , we exploit the fact that the integral above is a rescaling of a probability with respect to the uniform distribution .
let @xmath110 be the volume of the unit ball in @xmath101 .
then , we have @xmath514 \\ & & \leq c_1 \exp(- ( b/\eta ) c_2 i ) c_3 ( i+1)^2 ( b/\eta)^2 \eta/\pi \\ & & = c_4 ( b/\eta)^2 \eta ( i+1)^2 \exp(-(b/\eta ) c_2 i ) , \end{aligned}\ ] ] for @xmath111 .
returning to ( [ e : shells.bbig ] ) , we get @xmath517 & = & \sum_{i=1}^{\infty } c_4 ( b/\eta)^2 \eta ( i+1)^2 \exp(-(b/\eta ) c_2 i ) \\ & = & c_4 ( b/\eta)^2 \eta \sum_{i=1}^{\infty } ( i+1)^2 \exp(-(b/\eta ) c_2 i ) \\ & = & c_4 ( b/\eta)^2 \times \frac{4 e^{2 ( b/\eta ) c_2 } - 3 e^{(b/\eta ) c_2 } + 1}{\left(e^{(b/\eta ) c_2 } - 1 \right)^3 } \times \eta.\end{aligned}\ ] ] now , if @xmath518 , we have @xmath517 & \leq & c_4 ( b/\eta)^2 \times \frac{5 e^{2 ( b/\eta ) c_2}}{\left(e^{(b/\eta ) c_2}/2\right)^3 } \times \eta \\ & \leq & c_5 \eta \times ( b/\eta)^2 \exp(-(b/\eta ) c_2 ) \mbox{(where $ c_5 = 40 c_4 $ ) } \\ & = & c_5 \eta \times \exp(-(b/\eta ) c_2 + 2 \ln ( b/\eta ) ) \\ & \leq & c_5 \eta \times \exp(-(b/\eta ) c_2/2),\end{aligned}\ ] ] completing the proof . under the noise condition [ eq : nonsep - tsybakov ] and from the log - concavity assumption , we obtain that there exists @xmath75 such that for all @xmath282 we have : @xmath519 let us denote by @xmath520 . for all @xmath282
, we have : @xmath521 assume now the claim is true for @xmath142 ( @xmath491 ) . then at the @xmath144-th iteration , we can let @xmath524
, @xmath149 has excess errors at most @xmath525 , implying @xmath526 by assumption , @xmath527 . by standard bounds
, we can choose @xmath106 , @xmath107 and @xmath110 s.t . with @xmath159 samples
, we obtain @xmath531 with probability @xmath532 .
therefore @xmath533 with probability @xmath165 , as desired , completing the proof of theorem [ th : agg.margin : tsybakov ] .
|
we provide new results concerning label efficient , polynomial time , passive and active learning of linear separators . we prove that active learning provides an exponential improvement over pac ( passive ) learning of homogeneous linear separators under nearly log - concave distributions .
building on this , we provide a computationally efficient pac algorithm with optimal ( up to a constant factor ) sample complexity for such problems .
this resolves an open question of @xcite concerning the sample complexity of efficient pac algorithms under the uniform distribution in the unit ball .
moreover , it provides the first bound for a polynomial - time pac algorithm that is tight for an interesting infinite class of hypothesis functions under a general and natural class of data - distributions , providing significant progress towards a longstanding open question of @xcite .
we also provide new bounds for active and passive learning in the case that the data might not be linearly separable , both in the agnostic case and and under the tsybakov low - noise condition . to derive our results , we provide new structural results for ( nearly ) log - concave distributions , which might be of independent interest as well . active learning , pac learning , erm , nearly log - concave distributions , tsybakov low - noise condition , agnostic learning .
|
rapidly developing trend in nonperturbative approach to quantum gravity and string theory , which is based on holographic ideas of ads / cft correspondence @xcite , involves vasiliev theory of interacting higher spin fields @xcite and naturally leads to the notion of conformal higher spin fields ( chs ) @xcite .
though these chs fields represent thus far only a playground for rather sophisticated verification of the ads / cft correspondence @xcite , quite interestingly they turn out to be important in recent cosmological applications associated with the problem of initial conditions in the early inflationary universe @xcite .
this is the model of the cft driven cosmology @xcite which incorporates two main ideas a new concept of the cosmological microcanonical density matrix as the initial state of the universe and the implementation of this concept in cosmology with a large number of quantum fields conformally coupled to gravity .
this model plays important role within quantum cosmology and within the cosmological constant and dark energy problems .
in particular , it resolves the issue of infrared catastrophe associated with the observer independent treatment of the no - boundary state @xcite an anti - intuitive conclusion that the origin of an infinitely big universe ( with an insufficient amount of inflation produced at the zero minimum of the inflaton potential rather than its maximum ) is infinitely more probable than that of a finite one .
another property is that its statistical ensemble is bounded to a finite range of values of the effective cosmological constant @xcite and gives rise to a new type of hill - top inflation @xcite .
also this model incorporates a certain version of the ds / cft holographic duality @xcite and is potentially capable of generating the cosmological acceleration phenomenon within the so - called big boost scenario @xcite .
the setting of the initial conditions problem is based on canonical quantization of gravity theory and a natural notion of the microcanonical density matrix as a projector on the space of solutions of the quantum gravitational dirac constraints the system of the wheeler - dewitt equations @xcite .
its statistical sum has a representation of the euclidean quantum gravity ( eqg ) path integral @xcite @xmath8\ ; e^{-s[\,g_{\mu\nu},\varphi\ , ] } , \label{z } \end{aligned}\ ] ] over metric @xmath9 and matter fields @xmath10 which are periodic on the euclidean spacetime with a time compactified to a circle @xmath11 .
as shown in @xcite , this statistical sum has a good predictive power in the model with a primordial cosmological constant @xmath12 and the matter sector of a large number @xmath1 of quantum fields @xmath10 conformally coupled to gravity
conformal field theory ( cft ) with the action @xmath13 $ ] , @xmath14=-\frac{m_p^2}2 \int d^4x\,g^{1/2}\,(r-2\varlambda ) + s_{cft}[\,g_{\mu\nu},\varphi\ , ] .
\label{tree0 } \end{aligned}\ ] ] what allows one to go beyond a usual semiclassical expansion is that the cft modes @xmath10 dominate over conformally non - invariant fields and spatially - inhomogeneous metric modes .
integration over @xmath10 in ( [ z ] ) then leads to the effective action @xmath15=-\frac{m_p^2}2
\int d^4x\,g^{1/2}\,(r-2\varlambda ) + \vargamma_{cft}[\,g_{\mu\nu } ] , \label{eff}\\ & & e^{-\vargamma_{cft}[\,g_{\mu\nu}]}=\int d\varphi\,e^{-s_{cft}[\,g_{\mu\nu},\varphi\,]}. \end{aligned}\ ] ] it differs from ( [ tree0 ] ) by @xmath13 $ ] replaced with @xmath16 $ ] the effective action of the conformal fields on the background of @xmath9 .
on friedmann - robertson - walker ( frw ) background this action is exactly calculable by using the local conformal transformation to the static einstein universe and the well - known gravitational trace anomaly @xmath17 where the coefficients of local curvature invariants
@xmath18 , gauss - bonnet term @xmath19 and weyl tensor squared @xmath20 are determined by the cft particle content .
these coefficients are additive sums of contributions of all conformal fields of different spins @xmath21 . in particular
, the coefficient @xmath0 of the gauss - bonnet term , which is of major interest in what follows , reads @xmath22 where @xmath23 is a partial contribution of spin @xmath21 conformal field and @xmath24 is the number of such fields in the set of all @xmath10 .
the resulting @xmath16 $ ] becomes the sum of the anomaly contribution and a free energy of conformal matter fields on the sphere @xmath25 at the temperature determined by the period of the euclidean time .
then the physics of the cft driven cosmology is entirely determined by the effective action ( [ eff ] ) .
solutions of its equations of motion , which give a saddle point of the statistical sum path integral , are the cosmological instantons of @xmath26 topology with the friedmann - robertson - walker metric @xmath27 where a periodic lapse function @xmath28 and scale factor @xmath29 are the functions of the euclidean time belonging on @xmath11 @xcite .
these instantons serve as initial conditions for the cosmological evolution @xmath30 in the physical lorentzian signature spacetime , which follows from @xmath29 by analytic continuation @xmath31 at the point of the maximum value of the euclidean scale factor @xmath32 .
the coefficient @xmath0 of the gauss - bonnet term in ( [ anomaly ] ) plays especially important role because it imposes an upper bound on the range of @xmath12 within which these instantons exist , @xmath33 the fact that these instantons exist only in the finite range of @xmath12 implies the restriction of the microcanonical ensemble of universes to this range , which from the viewpoint of string theory can , in particular , be interpreted as the solution of the landscape problem for stringy vacua if one assumes that this model is a low energy approximation of the string theory .
as was recently shown in @xcite , this model with the fundamental cosmological constant can be generalized to the case when the role of an effective @xmath12 is simulated by the hill like potential of the inflaton scalar field @xmath34 in the regime of the slow roll approximation , @xmath35 .
then the cft driven cosmology can be regarded as a source of the new type of hill - top inflation scenario . in particular
, it can provide initial conditions for the starobinsky model of @xmath5-inflation @xcite or the model of higgs inflation with the higgs boson playing the role of the inflaton non - minimally coupled to gravity @xcite .
a major difficulty with this scenario is the problem of hierarchy between the sub - planckian energy scale of inflation the inflaton energy density @xmath36 @xcite compatible with current cmb observational data @xcite and the energy scale of the cft driven cosmology ( [ bound ] ) . to match these energy scales
one needs the value of @xmath37 . in the standard model or its gut generalizations containing only three low spin fields , which can be conformally coupled to gravity , @xmath38 , @xmath39 and @xmath40 ,
this is of course impossible .
their contribution to @xmath0 @xmath41 can reach such a magnitude only by the price of unnaturally high numbers of these particle species .
hidden sector of @xmath42 low spin particles sounds too unrealistic to be physically acceptable .
hidden sector of numerous , actually infinitely many , particles is possible in string theory which is believed to underlie the effective quantum field theory and quantum gravity .
however , these particles are massive , of the planckian scale mass , and can not incorporate local conformal invariance which is a corner stone of the cft cosmology . on the other hand , there is a growing believe that string theory can be a broken phase of the vasiliev theory of higher spin gauge fields @xcite , which was recently very actively considered in various tests of ads / cft correspondence including very wide implications of _ conformal higher spin _ ( chs ) field models @xcite .
therefore , it seems reasonable to try as a hidden weakly interacting sector of the cft cosmology the tower of these chs fields , that could provide a large value of @xmath0 .
though very speculative , in view of problems with unitarity for higher spins , this idea is strongly motivated by the recent observation that the partial contributions @xmath23 to ( [ 101 ] ) very rapidly grow with spin as @xmath43 @xcite , so that the needed value can be attained with the finite tower of hsc fields up to @xmath44 containing @xmath45 polarizations @xcite . therefore , the replacement of multiple species of the same spin by a tower of higher spins is a much more efficient mechanism for large @xmath0 , and the goal of this paper is to discuss this mechanism and its possible consequences .
the concept of a hidden sector of chs fields allows one to solve another important problem in the cft cosmology .
it suggests the mechanism protecting the theory from uncontrollable contribution of perturbatively nonrenormalizable graviton loops .
this is achieved by bringing the maximal energy scale of the above cosmological instantons ( [ bound ] ) below the gravitational cutoff . with the definition of this cutoff as the scale under which _ only _
the contribution of the _ graviton _ loops ( loops containing at least one graviton propagator ) is suppressed , there arises a strong distinction between the number of quantum species @xmath1 and the cft central charge @xmath46 which participate in the expressions for a conventional gravitational cutoff @xmath47 @xcite and the inflation scale @xmath48 corresponding to ( [ bound ] ) . in models of multiple quantum species the cutoff is usually defined as the scale below which the _ total _ one - loop contribution to the graviton propagator is smaller than the tree - level part @xcite , and this leads to the cutoff @xmath49 which is not distinguishable from @xmath50 in simple models , but can be very different when @xmath0 and @xmath1 are very different in magnitude .
our definition of the cutoff which involves smallness of only the graviton loop corrections leads to the expression @xmath51 which can be much higher than @xmath52 for @xmath53 , and this is exactly the case of chs fields . in this way we develop the approximation scheme in which the nonrenormalizable graviton sector is subject to effective field theory under the cutoff @xmath3 , whereas the renormalizable sector of multiple conformal species is treated beyond perturbation theory .
there are several other interesting features associated with the contribution of higher spins to the conformal anomaly ( [ anomaly ] ) .
first , unlike for the case of lower spins @xmath23 is negative for fermionic fields with @xmath54 @xcite .
this means that overall @xmath0 in cft cosmology can be negative , and this essentially modifies its scenario .
secondly , irrespective of the sign of @xmath23 one can construct the zeta - function regularized sum over infinite set of spins of chs fields ( described by totally symmetric bosonic tensors or fermionic spin - tensors ) which yields overall zero value of the total @xmath0 in ( [ 101 ] ) @xcite .
this nontrivial manifestation of the ads / cft correspondence is conjectured to underlie hypothetical free of trace anomalies and therefore quantum consistent
fundamental theory that might be based on the fusion of string theory and vasiliev gauge theory of higher spins .
motivated by the prospect of solving the hierarchy problem via chs fields we consider peculiarities of cosmological scenario driven by their finite tower in case of both positive and negative @xmath0 and by the anomaly free chs theory with an infinite set of fields .
the effective action of the cft driven cosmology ( [ eff ] ) on the frw metric ( [ frw ] ) with @xmath26 topology ,
@xmath55\equiv s_{\rm eff}[\,a , n\,]$ ] , was obtained by the conformal transformation to the static einstein universe @xcite with a compactified euclidean time .
it consists of the minisuperspace reduced einstein term , riegert - fradkin - tseytlin action @xcite and the contribution of the einstein static spacetime
free energy of conformal fields and their vacuum casimir energy . in units of the rescaled planck mass @xmath56 it reads @xcite @xmath57=m_p^2\int_{s^1 }
d\tau\,n \left\{-aa'^2
-a+ \frac\varlambda3 a^3+\,b\left(\frac{a'^2}{a } -\frac{a'^4}{6 a}\right ) + \frac{b}{2a}\,\right\ } + f(\eta ) , \label{effaction0 } \end{aligned}\ ] ] where @xmath58 .
the first three terms in curly brackets of ( [ effaction0 ] ) represent the einstein action with a fundamental cosmological constant @xmath59 ( @xmath60 is the corresponding hubble parameter ) .
the constant @xmath61 is a coefficient of the contributions of the conformal anomaly and vacuum ( casimir ) energy @xmath62 on a conformally related static einstein spacetime .
it is proportional to the coefficient @xmath0 of the gauss - bonnet term @xmath63 in the trace anomaly of conformal matter fields ( [ anomaly ] ) @xmath64 the free energy @xmath65 of the set of all conformal fields labelled by their spin @xmath21 also comes from this einstein space , @xmath66 this is a typical boson or fermion sum over field oscillators with energies @xmath67 on a unit 3-sphere , @xmath68 denoting the number of physical polarizations of a spin-@xmath21 field , which is negative for fermions , @xmath69 .
the role of the temperature is played here by the inverse of @xmath70 an overall circumference of @xmath11 in the @xmath26 instanton in units of the conformal time ( [ period ] ) .
the effective action is independent of the anomaly coefficient @xmath71 , because it is assumed that @xmath71 is renormalized to zero by a local counterterm , @xmath72 this guarantees the absence of higher derivative terms in ( [ effaction0 ] ) @xcite non - ghost nature of the scale factor and simultaneously shifts the uv renormalized casimir energy ( which universally expresses in terms of anomaly coefficients @xmath71 and @xmath0 @xcite ) to a partiular value independent of @xmath71 and proportional to @xmath73 @xcite ,- model which also provides the inflaton mode simulating a slowly varying cosmological term which decays at the exit from inflation @xcite . ]
@xmath74 both of these properties are critically important for the instanton solutions of effective equations .
the coefficient @xmath75 of the weyl tensor term @xmath76 also does not enter ( [ effaction0 ] ) because @xmath77 identically vanishes for any frw metric . the statistical sum ( [ z ] )
is dominated by solutions of the effective equation , @xmath78 , which in the gauge @xmath79 reads @xmath80 this is the modification of the euclidean friedmann equation by the anomalous @xmath61-term and the radiation term @xmath81 .
the constant @xmath82 here characterizes the sum of the casimir energy and the energy of thermally excited particles with the inverse temperature @xmath70 given by ( [ period ] ) , @xmath83 it is a nonlocal functional of the history @xmath29 the equation ( [ bootstrap ] ) plays the role of the bootstrap equation for the amount of radiation which is determined by the background on top of which this radiation evolves and produces back reaction .
the quadratic equation ( [ efeq ] ) can be solved for @xmath84 , @xmath85 where the positive sign of the square root is chosen to provide a periodic oscillation of @xmath86 between its maximal and minimal values @xmath87 . in order to guarantee
that at @xmath88 is a turning point with a vanishing @xmath89 the value @xmath90 should satisfy the bound @xmath91 .
this gives the first two restrictions on the range of @xmath92 and @xmath82 @xmath93 whereas the third one follows from the requirement of real turning points @xmath87 , @xmath94 there are two sets of solutions of this integro - differential equation @xcite .
the main set consists of periodic @xmath95 instantons with the oscillating scale factor _ garlands _ that can be regarded as the thermal version of the hartle - hawking instantons .
the scale factor oscillates @xmath96 times ( @xmath97 ) between the maximum and minimum values ( [ apm ] ) , @xmath98 , so that the full period of the conformal time ( [ period ] ) is the @xmath99-multiple of the integral between the two neighboring turning points of @xmath29 , @xmath100 , @xmath101 this value of @xmath70 is finite and determines effective temperature @xmath102 as a function of @xmath103 and @xmath104 .
this is the artifact of a microcanonical ensemble in cosmology @xcite with only two freely specifiable dimensional parameters the gravitational and cosmological constants . according to ( [ hbound ] )
these garland - type instantons exist only in the limited range of the cosmological constant @xmath104 @xcite . in view of ( [ hbound ] ) and ( [ uppercbound ] ) they belong to the domain in the two - dimensional plane of the hubble constant @xmath92 and the amount of radiation constant @xmath82 . in this domain they form an countable , @xmath105 , sequence of one - parameter families curves interpolating between the lower straight line boundary @xmath106 and the upper hyperbolic boundary @xmath107 .
each curve corresponds to a respective @xmath96-fold instantons of the above type .
therefore , the range of admissible values of @xmath12 , @xmath108 has a band structure , each band @xmath109 being a projection of the @xmath96-th curve to the @xmath92 axis .
the sequence of bands of ever narrowing widths with @xmath110 accumulates at the upper bound of this range @xmath111 .
the lower bound @xmath112 the lowest point of @xmath113 family can be obtained numerically for any field content of the model .
another set of solutions follows from rewriting the effective equation ( [ efeq ] ) in the form ( retaining in contrast to ( [ maineq ] ) both signs of the square root ) @xmath114 and noting that for @xmath115 it reduces to @xcite ( without radiation contribution in lorentzian signature spacetime this solution was derived in @xcite ) @xmath116 obviously the solutions to these two equations , @xmath117 , represent spherical euclidean instantons @xmath118 of the radii @xmath119 respectively , or the strings of such spheres touching each other at their poles and forming a _ necklace _ " with any number of such spherical beads @xcite .
note that the value of @xmath115 is consistent with the bootstrap equation ( [ bootstrap ] ) , because the time period for such a necklace consisting of @xmath96 beads , @xmath120 diverges at the poles of spherical beads , where they touch each other the range of integration over @xmath86 in contrast to eq.([period1 ] ) is a multiple of the range between @xmath121 at the pole of the 4-sphere @xmath118 and its value @xmath119 at the equator of @xmath118 .
therefore both @xmath65 and @xmath122 vanish and give in view of ( [ bootstrap ] ) the value of @xmath115 .
, width=264 ] these vacuum ( or zero temperature , @xmath123 ) necklace instantons existing for all values of @xmath124 are , however , uninteresting because their contribution to the statistical sum is suppressed to zero by their infinite _ positive _ euclidean action .
for @xmath125 the on - shell value of the action ( [ effaction0 ] ) , @xmath126 diverges to @xmath127 at the poles of necklace beads with @xmath121 , where @xmath128 and @xmath129 .
thus the cft cosmology scenario is free from the infrared catastrophe of vacuum no - boundary instantons , which would otherwise have a _ negative _ tree - level euclidean action ( proportional to @xmath130 at @xmath131 ) and which would imply that the origin of an infinitely big universe is infinitely more probable than that of a finite one .
elimination of this infrared catastrophe is the quantum effect of the trace anomaly which flips the sign of the effective action and sends it to @xmath127 .
goes over into two different points of @xmath132 , @xmath133 , with equal field values .
[ fig.2],width=264 ] the explanation why the trace anomaly action does not produce the same effect for the no - boundary prescription of hartle and hawking consists in the observation @xcite that the density matrix prescription in the cft cosmology , despite the same @xmath132-geometry of the cosmological instanton , has boundary conditions other than those of the no - boundary case
. graphical representation of the origin of the periodic garland instanton from the density matrix , whose two arguments are associated with spatial hypersurfaces @xmath134 and @xmath135 , is demonstrated on fig.1 for a single - fold case .
analogous single - bead necklace instanton originates by a similar procedure shown on fig.2 .
it implies that prior to transition to the statistical sum the values of the fields on two hemispheres should coincide at the point where these hemispheres touch . after tracing out the fields at the identified surfaces @xmath136 the hemispheres
get glued into a complete @xmath132 instanton , but its antipodal pole points should still be identified as well as their field values @xmath137 .
this is the additional boundary condition which is absent in the conventional no - boundary prescription . and
, as shown in @xcite , path integration over the quantum field on a generic euclidean manifold with the identification of values of this field at two spacetime points leads to the suppression of the result to zero by its one - loop prefactor .
this result strongly relies on positivity of the classical euclidean action or unitarity , which is apparently related to the positivity of @xmath138 in the above derivation .
the main motivation for introducing chs fields in cosmology is an attempt of solving the hierarchy problem .
realistic cft driven cosmology capable of generating a finite inflationary stage arises by the generalization of the model ( [ tree0 ] ) to the case of effective cosmological term with a slowly varying @xmath12 .
@xmath12 is simulated by a potential @xmath139 of a dynamical scalar field @xmath34 the inflaton in the regime of the slow roll from the slope of @xmath139 .
such a model of a new hill - top inflation scenario was recently built in @xcite .
it turns out that the slow roll conditions smallness of inflationary parameters @xmath140 and @xmath70 , which determine the properties of the primordial cmb power spectrum @xcite , are satisfied for those models which are very close to the upper boundary of the effective cosmological constant range ( [ lambda_range ] ) .
this leads to a major difficulty in a realistic inflationary scenario the problem of hierarchy between the planckian scale of this bound and the estimates for the scale of inflation based on the cmb data in the models which provide a good fit of these data @xcite . here
we show that using chs fields suggests a solution to this hierarchy problem and , moreover , provides a mechanism protecting the obtained results from the contribution of nonrenormalizable graviton loops .
two models which perhaps give the best fit are the starobinsky @xmath5-inflation model and quantitatevely very close to it higgs inflation model with a large non - minimal coupling of the higgs inflaton to curvature @xcite . in particular , the latter model establishes a very convincing relation between the observable cmb spectral parameter @xmath141 and the value of the higgs mass very close to the one discovered at lhc @xcite . for higgs inflation with a large non - minimal coupling the energy density ( in the einstein frame of fields ) at the start of inflation should be essentially sub - planckian , @xmath142 @xcite . matching with the upper bound ( [ lambda_range ]
) @xmath143 implies that the total beta must be of the order of magnitude @xmath144 in order to reach this value with the conventional low spin particle phenomenology characteristic of the standard model one would need unrealistically high numbers of conformal invariant scalar bosons @xmath145 , dirac fermions @xmath146 and vector bosons @xmath147 in the expression ( [ 100 ] ) for @xmath0 .
hidden sector of so numerous low spin weakly interacting particles does not seem to be realistic . however , unification of interactions inspired by the ideas of string theory , holographic duality @xcite and higher spin gauge theory @xcite suggests that this hidden sector might contain conformal higher spin ( chs ) fields described by totally symmetric tensors and spin - tensors @xcite , and the total value of @xmath0 consists of the additive sum of all their partial contributions ( [ 101 ] ) .
recently there was essential progress in the theory of these fields .
in particular , it was advocated in @xcite that the values of @xmath23 can be explicitly calculated for both bosons and dirac fermions of a generic spin @xmath21 .
they read @xmath148 where @xmath68 is their respective number of dynamical degrees of freedom
polarizations ( negative for fermions ) .
violates the rule ( [ boson ] ) which is valid for @xmath149 . ]
though these fields serve now basically as a playground for holographic ads / cft duality issues and suffer from the problems of perturbative unitarity , which is anticipated to be restored only at the non - perturbative level , . ]
it is worth trying to exploit them as a possible solution of hierarchy problem in the cft driven cosmology .
strong motivation for this is that partial contributions of individual higher spins rapidly grow with the spin as @xmath43 , so that the tower of spins up to some large @xmath150 generates the total value of @xmath0 @xmath151 ( for simplicity we consider only bosons and assume that every higher spin species is taken only once , @xmath152 ) .
therefore , in order to provide the hierarchy bound ( [ beta_bound ] ) the maximal spin should be @xmath153 , which corresponds to the following estimate of the total number of particle modes ( polarizations ) in the hidden sector of the theory @xmath154 when , instead of a tower of spins , the cosmological model is driven by a conformal field of an individual spin @xmath155 with @xmath156 and @xmath157 , then the needed value of spin and the number of polarizations are @xmath158 quite interestingly , in the case of ( [ beta])-([nu ] ) this coincides with the estimate for the average value of @xmath0 per one conformal degree of freedom , @xmath159 at which the thermal correction to the spectral parameter of cmb , @xmath160 , depending on the properties of the hill like inflaton potential might appear in its third decimal order @xcite the precision anticipated to be reachable in the next generation of cmb observations following planck . for the case of the individual spin ( [ ind_spin ] )
the thermal correction is even stronger and can appear in the second or even the first decimal order .
this means that a potential resolution of the hierarchy problem in the cft scenario via chs simultaneously would make measurable the thermal contribution to the cmb red tilt , which is complementary to the conventional tilt caused by the deviation of the slow roll evolution from the exact de sitter evolution @xcite .
chs fields provide a mechanism which does not only solve the phenomenological problem of hierarchy between the planckian and inflation scales , but is also likely to justify the approximation underlying the predictions of the above type .
as mentioned in introduction this approximation goes beyond semiclassical expansion , because its subleading one - loop order is critically important for the construction of the model . in particular
, it contributes a large one - loop contribution whose balance against the tree - level part establishes the upper bound ( [ bound ] ) on the energy scale of inflation .
therefore , the inflation scale @xmath161 is deeply below the planck scale for a large @xmath0 generated by large numbers of quantum species @xmath24,@xmath162 , @xmath163 .
this seems to imply validity of the semiclassical expansion in which all graviton loops can be disregarded .
however , there is a problem associated with the fact that quantum gravity with a large number @xmath1 of quantum species has the effective field theory cutoff @xmath164 which decreases for growing @xmath1 along with the scale ( [ lambdai ] ) and can be even higher than the latter .
this is a well known statement based on perturbation theory arguments @xcite or implications of the hawking radiation from a semiclassical black hole @xcite . within a conventional perturbation theory
this completely breaks the predictions in our model , because we can not guarantee smallness of quantum corrections due to nonrenormalizable graviton loops
the theory is not protected from uncontrollable radiative corrections .
the origin of the gravitational cutoff ( [ cutoff ] ) , below which a nonrenormalizable contribution of graviton loops can be disregarded , is based on the following reasoning .
suppose we have @xmath1 quantum species which generate in the external gravitational field a full set of feynman diagrams .
inclusion of quantum graviton loops consists in the insertion into these diagrams graviton propagators with the relevant vertices of gravity - matter interaction .
each graviton propagator carries extra factor of @xmath165 which suppresses a relevant background quantity of the scale @xmath166 a spacetime curvature or a second order spacetime derivative . with a small @xmath167 number of quantum fields the suppression factor for this contribution would be @xmath168 which implies a standard planck scale cutoff @xmath169
however , with many quantum fields this contribution gets enhanced . for noninteracting multiple species the above contribution
is multiplied by the first power of @xmath1 , since the graviton propagator connects two vertices of one and the same field .
this is because another matter field propagator , connecting these vertices , for free ( linear ) fields can connect only the vertices of one and the same species , whereas another graviton propagator which could have connected vertices of different fields would give ( together with the first added propagator ) a subdominant contribution @xmath170 .
therefore , the overall suppression factor for the insertion of any single graviton propagator becomes @xmath171 , and the cutoff equals ( [ cutoff ] ) .
matrix propagator of quantum species can be non - diagonal ) , so that the suppression factor equals @xmath172 and the cutoff seemingly reduces to @xmath173 .
but off - diagonal elements should carry the coupling constant which should scale as @xmath174 to guarantee perturbative regime in the matter field sector , and this raises @xmath175 back to the non - interacting case of ( [ cutoff ] ) .
i am grateful to s.sibiryakov for the discussion of this point . ]
thus , a hidden sector of multiple quantum species does not help to protect the model from unrenormalizable graviton corrections .
its energy scale @xmath2 with @xmath176 either coincides with the gravitational cutoff ( [ cutoff ] ) or even exceeds it .
this is the consequence of the universality of the gravitational interaction no matter how small is the coupling of the hidden sector to observable sector , universal gravitational interaction can not be reduced below the gravitational cutoff .
it should be emphasized here that in the above derivation of the gravitational cutoff we did not require smallness of the _ full _ quantum ( say one - loop ) correction relative to the tree - level part as it was done in @xcite . in the perturbation theory approach of @xcite
this would have led instead of @xmath1 to the central charge of the theory
the relevant coefficient of the conformal anomaly @xmath46 , and the cutoff would be @xmath2 .
rather , in our definition of the cutoff we demanded smallness of only the graviton corrections all nonrenormalizable contributions containing any single graviton propagator . on the contrary , the contribution of purely quantum species loops can be large and treated beyond perturbation theory , as it was done in the construction of cosmological instantons above . with this definition
the cutoff is inverse proportional to @xmath177 rather than @xmath178 .
for conventional low spin theories these two quantities are of the same order of magnitude and are usually not distinguished from one another in the cutoff expression .
remarkable property of chs fields is that they qualitatively change this situation .
this is because for this theory either for an individual sufficiently high spin @xmath21 or for a tower of spins up to some @xmath179
the parameter @xmath0 grows with the spin much faster than the number of fields . for an individual spin
@xmath21 the role of the number of quantum species is played now by the number of polarizations @xmath157 , while for a tower of spins ( for bosonic case with every spin taken once ) it is @xmath180 correspondingly , the total value of @xmath0 equals @xmath181 for an individual spin and @xmath182 , see eq.([beta ] ) , for such a tower of spins .
therefore the ratio of the inflation scale ( [ lambdai ] ) to the cutoff ( [ cutoff ] ) is decreasing with the growth of the individual spin @xmath21 or the height of the spin tower @xmath150 respectively as which should equally enter the expressions for the inflation scale ( [ lambdai ] ) , cf . eq.([bound ] ) , and the gravitational cutoff ( [ cutoff ] ) and , therefore , cancel out in their ratio . ]
@xmath183 where the inflation scales and cutoffs labeled by @xmath21 and @xmath150 obviously denote the cases of the individual spin or a tower of those . in both cases of the individual spin ( [ ind_spin ] ) , @xmath184 , and the tower of spins ( [ beta])-([nu ] ) , @xmath153 , these ratios are very small and range within the limits @xmath185 .
therefore , the model is in the quantum state three or four orders of magnitude below its gravitational cutoff , and an uncontrollable contribution of nonrenormalizable graviton loops is negligible . for strongly coupled species with the cutoff @xmath173 mentioned above ( see footnote 4 )
these ratios are higher @xmath186 , @xmath187 , and amount to @xmath188 .
so for a tower of spins , @xmath153 , the system becomes too close to the gravitational cutoff .
this case is , however , not so important because ( non - diagonal ) propagator mixing between the chs species is possible only on the nonzero background of these fields , whereas we assume that in conformal cosmology it is vanishing
conformal species have vanishing expectation values and contribute only via their quantum fluctuations .
thus the model is deeply below the gravitational cutoff , and we remain with the dominant quantum contribution of only the chs fields in the external gravitational field .
this contribution is big , because it is weighted by @xmath53 , and it is treated beyond perturbation theory . for linear fields
it is exhausted by the one - loop order , which is exactly calculable for the frw metric by the trace anomaly method described above .
the generalization to a nonlinear case also looks straightforward .
it is important that this sector is perturbatively renormalizable , which is obvious for low spin fields @xmath189 , while for higher order spins it directly follows from their higher - derivative nature
the conformal field of spin @xmath21 has a kinetic operator of order @xmath190 @xcite . due to renormalizability gravitational multi - loop counterterms
are exhausted by the same three tensor invariants as in the one - loop order , @xmath20 , @xmath63 and @xmath18 ( @xmath5 is absent due to conformal invariance ) which generate the three - parameter trace anomaly ( [ anomaly ] ) , and the principal effect of nonlinearity is a slow logarithmic rg running of @xmath71 , @xmath0 and @xmath75 @xcite . moreover , for the frw background this sector is actually free from _ logarithmic _ uv divergences , because on this closed @xmath26 instanton the counterterms @xmath191 , the euler number ( contributed by the gauss - bonnet invariant @xmath63 ) as well as @xmath192 are all vanishing .
thus , this sector of the model is free from uv divergences , and its power divergences are absorbed by the renormalization of the cosmological @xmath12 and gravitational @xmath193 coupling constants .
it is in terms of these two _ renormalized _ constants and without any renormalization ambiguity that the frw cosmological instantons were built within the dynamically suppressed energy scale ( [ bound ] ) .
note that the @xmath5-term of the starobinsky model , considered above ( [ alpha_renorm ] ) and used for a _
finite _ renormalization of the casimir energy ( [ casimir_renorm ] ) and simulation of the effective cosmological term @xcite , belongs to uv finite sector .
therefore , elimination of higher - derivative ghosts in the gravitational sector by this finite renormalization can not be broken by leading radiative corrections , which justifies the criterion of naturalness " in this model .
eq.([fermion ] ) shows that when the model is dominated by chs fermions the total @xmath0 and @xmath61 can be negative , because fermionic @xmath23 are negative starting with @xmath194 .
apparently , this is associated with the fact that hsc fields are not free from ghosts their kinetic operator contains higher derivatives of order @xmath190 .
there exist a hope that at the nonperturbative level chs theories can be rendered unitarity , so that there effect can still be interesting within the cft cosmology scenario .
the picture of this scenario with a negative @xmath61 is somewhat different from the case of @xmath125 and looks as follows .
first consider the garland instantons .
for a negative @xmath195 the equation ( [ maineq ] ) should be chosen in the form with the negative sign of the square root @xmath196 if we want periodic oscillations between @xmath88 and @xmath197 . here
the argument of the square root equals @xmath198 and is positive definite , and also no restrictions on the range of @xmath92 and @xmath82 similar to ( [ hbound ] ) follow .
apriori , in the two dimensional plane of @xmath199 and @xmath200 garland instanton solutions can occupy all points below the hyperbola @xmath107 .
however , the bootstrap equation ( [ bootstrap ] ) with a negative @xmath61 imposes the upper bound on the value of the conformal time period of garland instantons , and this leads to a reduction of this domain .
the constant @xmath82 can not be negative because @xmath90 should be positive for garlands , so that the range @xmath201 gives , in view of @xmath202 being a monotonically decreasing function of @xmath70 , a restriction on the domain of possible
@xmath70 @xmath203 this immediately rules out solutions belonging to @xmath92 and @xmath82 axes of the curvilinear triangle , because both for @xmath204 ( @xmath205 ) and @xmath206 ( @xmath207 ) the time period is infinite , which contradicts @xmath208 .
similarly , the asymptotics @xmath209 can not be reached along the upper hyperbolic boundary , @xmath107 , because the bootstrap equation would imply that @xmath210 and @xmath211 ( in view of bosonic contributions to @xmath202 tending to infinity for vanishing @xmath70 ) .
but this is impossible because at this boundary the conformal time period is exactly calculable the turning points @xmath87 approach each other and the integral for @xmath70 becomes exactly calculable @xcite . ]
@xcite and equals @xmath212 the values of @xmath92 are also bounded from above because for @xmath213 the conformal time period @xmath214 , and large @xmath60 are also ruled out by the maximal value of @xmath70 .
in contrast to the case of positive @xmath61 , however , the upper bound on the cosmological constant is not universal and can be obtained only numerically .
thus , negative @xmath61 rules out infrared catastrophe in the distribution of garland type cosmological models with @xmath209 , because there are no corresponding garland instantons in this limit .
this restricts the range of the cosmological constant from above .
one would think that the upper limit on @xmath70 might also bound the value of @xmath92 and solve the hierarchy problem .
say , on the hyperbolic boundary with @xmath215 it reads @xmath216 .
but this bound turns out to be to weak to provide @xmath217 , unless the value of @xmath218 in @xmath219 is very big , which again implies a high tower of fermionic chs .
the domain of negative @xmath82 and @xmath61 in ( [ bootstrap ] ) corresponds to _ necklace _
instantons described above in sect.2 . for @xmath220
the value of @xmath90 is also negative , so that a sensible minimal value of the scale factor is actually zero .
this does not lead to a contradiction with ( [ negativeb ] ) , because the conformal time @xmath70 in view of eq.([period2 ] ) diverges to infinity at the lower limit and @xmath221 , so that the geometry smoothly closes at @xmath121 without a conical singularity to form a round sphere @xmath222 ( or a necklace the string of round spheres ) for all positive @xmath92 .
however , in contrast to necklace instantons with a positive @xmath61 they are not ruled out by their on - shell effective action ( [ action - instanton ] ) because @xmath223 and the action is infinitely negative @xmath224 . note that this divergence occurs not at a particular value of @xmath92 the statistical sum is badly defined for all values of @xmath92 , and the situation is qualitatively different from the no - boundary state .
therefore , the cosmology with @xmath225 is in principle inconsistent , because its full statistical sum contains this divergent contribution of necklace instantons .
the additional ensemble of garland instantons , discussed above , contributes to the statistical sum a finite part , but it is hard to interpret it on top of the divergent part of the necklace distribution .
another interesting case of chs theory is the infinitely high tower of spins @xmath226 . in this case
the total @xmath0 and other characteristics of the chs tower are formally divergent , but in the case when all spin particles are taken only once , @xmath152 , the use of @xmath227-function regularization for the divergent spin series gives a remarkable result an overall @xmath61 becomes zero @xcite , @xmath228 this result holds independently for boson , @xmath229 , and fermion , @xmath230 @xcite , chs towers and underlies a non - trivial check of the @xmath231-correspondence in which the vanishing result on the 5-dimensional ads side simply follows from the absence of conformal anomaly in odd - dimensional spacetime . explicit summation over spins is possible not only for trace anomaly coefficients but also for the thermal free energies of chs particles @xcite .
this can be done by expanding the logarithm in ( [ freeenergy ] ) in the sum over one - particle " statistical sums @xmath232 with growing inverse temperature @xmath233 and interchanging the order of summation over spins and equidistant levels of the inverse temperature , @xmath234 explicit summation of @xmath235 over spin @xmath21 gives the answer @xcite @xmath236 which gives a good approximation for the sum over @xmath96 in the limit of large @xmath70 .
irrespective of this approximation , the answer for the total free energy , given by this convergent sum over @xmath96 , has important property
unexpectedly it is negative .
similarly for fermions @xmath237 therefore , both bosonic and fermionic theories of chs have negative thermal energy densities @xmath202 .
obviously , this unnatural conclusion is the result of the oversubtraction performed by the @xmath227-function renormalization of @xcite which annihilates all uv divergences in the infinite sum over spins .
this is clearly seen in the special cutoff regularization by the parameter
@xmath238 @xmath239 this regularization retains power divergences which give a dominant positive contribution , while the @xmath227-function regularization subtracts these divergences and leaves us with the negative remnant .
potential justification for this property could be a conjectured nonperturbative treatment of such theories which solves the problem of their unitarity .
we will see now that outside of this nonperturbative approach these models recover the situation of the vacuum no - boundary instantons with the infrared catastrophe of @xmath131 . indeed , for @xmath240 and @xmath241 eqs .
( [ efeq])-([bootstrap ] ) lead to a negative @xmath220 , so that the turning points ( [ apm ] ) can not be both positive . similarly
to the situation a negative @xmath82 above , the negative value of @xmath90 implies that the minimal value of the scale factor is zero , which does not lead to a conical singularity of the solution at @xmath205 , because the conformal time @xmath70 in view of ( [ period1 ] ) diverges to infinity and @xmath242 .
the solution becomes @xmath243 , which represents the hartle - hawking instanton @xmath132 . in view of the overall @xmath61 equal to zero ,
the on - shell action ( [ action - instanton ] ) does not diverge
its integral term is finite for all @xmath92 .
it coincides with the tree - level action of the no - boundary instanton the 4-sphere of the radius @xmath244 and diverges to @xmath245 at @xmath209 , @xmath246 .
our main conclusion is that cosmology with numerous chs fields can solve the problem of hierarchy between the planck scale and the subplanckian inflationary scale compatible with cmb observations within a number of inflation models including @xmath5-gravity and higgs inflation model @xcite .
for that a finite height tower of bosonic and fermionic chs fields should generate a large positive coefficient @xmath0 of the gauss - bonnet invariant in the overall conformal anomaly of the theory . in chs models
the coefficient @xmath0 is much higher than the number of field polarizations @xmath1 which means that the energy scale of the model @xmath2 is essentially ( by three or four orders of magnitude for @xmath5-gravity and higgs inflation model ) below the gravitational cutoff @xmath3 , the latter being determined from the requirement of smallness of graviton loop corrections .
this justifies a special approximation scheme in which these nonrenormalizable corrections are treated within the effective field theory below this cutoff , while the dominant quantum contribution of chs fields is treated beyond perturbation theory and dynamically puts the bound on the energy scale of the model .
though it fits presently very popular ideas of string theory , higher spin gauge theory @xcite and holographic duality @xcite , this mechanism of a large @xmath0 is still vulnerable to criticism regarding the problem of perturbative unitarity for chs fields due to their higher - derivative nature .
how important is the lack of unitarity in the hidden sector of our model ?
free or weakly selfinteracting chs fields are coupled to the observable sector only gravitationally , but below the gravitational cutoff this coupling is apparently suppressed similarly to the contribution of graviton loops and the loops containing any single graviton propagator .
this effective decoupling essentially reduces the effects of non - unitarity .
of course , this is not a fundamental solution of the problem , which is expected to be achieved only at the nonperturbative level under a better understanding of the gauge theory of interacting higher spin fields .
the problem of unitarity can manifest itself and even can be destructive for a negative @xmath0 , which is possible in the case of the dominant contribution of higher spin conformal fermions with @xmath54 .
as we saw , cosmology with @xmath225 is inconsistent , because its statistical sum contains a divergent contribution of necklace instantons .
the source of this inconsistency is apparently related to the breakdown of perturbative unitarity .
suppression of vacuum necklace instantons , mentioned in sect.2 , is based on the gaussian integration with a positive definite quadratic form @xcite .
the boundary condition @xmath247 , induced by the microcanonical density matrix as depicted in fig.2 , can be enforced with the lagrange multiplier @xmath248 in the path integral for a generic theory with the quadratic action @xmath249=\frac12\int dx\,\phi f\phi$ ] .
therefore , the modification due to this boundary condition reduces to a single gaussian integral over @xmath248 @xmath250 + i\lambda\big(\phi(x_+)-\phi(x_-)\big)\big)\nonumber\\ & & \qquad\qquad=\big({\rm det } \,f\big)^{-1/2}\int d\lambda\,\exp\left(-\frac12\int dx\,dy\,j_\lambda(x)g(x , y ) j_\lambda(y)\right ) \label{1000 } \end{aligned}\ ] ] with the local source @xmath251 peaked at @xmath252 , @xmath253 , and the kernel @xmath254 the green s function of @xmath255 .
therefore , the statistical sum reduces to zero , @xmath256 , in view of the divergent coincidence limits for @xmath254 , @xmath257 .
violation of unitarity is associated with the indefiniteness of the operator @xmath255 which makes @xmath258 divergent , and this is consistent with the divergence to @xmath245 of the integral part of ( [ action - instanton ] ) for @xmath223 .
the situation with the anomaly free theory of the infinite chain of chs fields is different .
its statistical sum is finite , but it suffers from the infrared catastrophe at @xmath131 similar to the problem with the no - boundary prescription .
therefore , it is hardly acceptable as a source of cosmological initial conditions , because it suggests that the origin of infinitely big universe is infinitely more probable than that of the finite one .
this means that the anomaly free chs model , as a potential basis of a consistent theory of everything " , does not give any advantages in context of cosmological applications . what ruins these advantages is the oversubtraction of power divergences by @xmath227-regularization of the sum over spins in ( [ bosonf1 ] ) , which leads to a negative value of the thermal energy the other side of breakdown of perturbative unitarity
apparently , the disregard of power divergences in the sum over spins physically is not such a harmless procedure . their recovery in the regularization ( [ reg ] ) with
the parameter @xmath238 , which is equivalent to the cutoff in the height of the spin tower at @xmath259 , brings us back to the predictions of the model with a finite tower of chs fields .
this spin cutoff becomes a phenomenological parameter related to the inflation energy scale detectable via cmb observations .
thus , the chs models dominated by fermions with @xmath223 and anomaly free models with @xmath240 remain a playground of holographic methods in ads / cft correspondence , but fail to generate interesting cft driven cosmology .
this matches with the fact established in @xcite that the cft driven scenario implies another type of holographic duality the dgp / cft correspondence . for positive @xmath260
this is the duality between the 4-dimensional cft driven cosmology of @xcite and the tree - level dynamics of the 4d brane which is embedded into the 5-dimensional schwarzschild - de sitter bulk and carries the 4d einstein - hilbert term the generalization of the dgp model @xcite .
i would like to express my gratitude to a.tseytlin , m.vasiliev and m.beccaria for fruitful and thought - provoking correspondence and discussions .
i also greatly benefitted from helpful discussions with f.bezrukov , d.blas , c.burgess , c. deffayet , j.garriga , j.b.hartle , a.yu.kamenshchik , d.marolf , r.metsaev , d.v.nesterov , m.shaposhnikov , s.sibiryakov , n.tsamis , w.unruh , a.vilenkin and r.woodard .
i am grateful for hospitality of theory division , cern , and pacific institute of theoretical physics , ubc , where this work was initiated .
this work was also supported by the rfbr grant no .
14 - 01 - 00489 and by the tomsk state university competitiveness improvement program .
99 j. m. maldacena , _ the large n limit of superconformal field theories and supergravity _ , adv .
phys . 2 ( 1998 ) 231252 [ hep - th/9711200 ] ; + s. s. gubser , i. r. klebanov , and a. m. polyakov , _ gauge theory correlators from non - critical string theory _ ,
lett . b428
( 1998 ) 105114 [ hep - th/9802109 ] ; + e. witten , _ anti - de sitter space and holography _ , adv .
phys . 2 ( 1998 ) 253291 [ hep - th/9802150 ] .
m. a. vasiliev , _ consistent equation for interacting gauge fields of all spins in ( 3 + 1)-dimensions _ , phys .
b243 ( 1990 ) 378 - 382 ; + m. a. vasiliev , _ more on equations of motion for interacting massless fields of all spins in ( 3 + 1)-dimensions _ , phys .
b285 ( 1992 ) 225 - 234 ; + m. a. vasiliev , _ nonlinear equations for symmetric massless higher spin fields in ( a)ds(d ) _ , phys . lett .
b567 ( 2003 ) 139 - 151 [ hep - th/0304049 ] .
a. o. barvinsky and a. yu .
kamenshchik , _ cosmological landscape from nothing : some like it hot , jcap _ * 09 * ( 2006 ) 014 [ hep - th/0605132 ] ; + a. o. barvinsky and a. yu .
kamenshchik , _ thermodynamics via creation from nothing : limiting the cosmological constant landscape , phys . rev . _ * d 74 * ( 2006 ) 121502 [ hep - th/0611206 ]
. f. l. bezrukov , a. magnin and m. shaposhnikov , _ standard model higgs boson mass from inflation , phys .
* b 675 * ( 2009 ) 88 [ arxiv:0812.4950 ] ; + a. de simone , m. p. hertzberg and f. wilczek , _ running inflation in the standard model , phys .
* b 678 * ( 2009 ) 1 [ arxiv:0812.4946 ] ; + f. bezrukov and m. shaposhnikov , _ standard model higgs boson mass from inflation : two loop analysis , jhep _ * 07 * ( 2009 ) 089 [ arxiv:0904.1537 ] ; + a. o. barvinsky , a. yu . kamenshchik , c. kiefer , a.a .
starobinsky and c. steinwachs , _ asymptotic freedom in inflationary cosmology with a non - minimally coupled higgs field , jcap _ * 12 * ( 2009 ) 003 [ arxiv:0904.1698 ] .
g.hinshaw et al , _ nine - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological parameter results , astrophys .
j. suppl . _
* 180 * ( 2009 ) 225 [ arxiv:1212.5226 ] ; + e. komatsu et al . , _ five - year wilkinson microwave anisotropy probe ( wmap ) observations : cosmological interpretation , astrophys . j. suppl . _ * 180 * ( 2009 ) 330 [ arxiv:0803.0547 ]
. p. ade et al .
( planck collaboration ) , _ planck 2013 results .
cosmological parameters , astron .
astrophys . _
* 571 * ( 2014 ) a16 [ arxiv:1303.5076 ] ; + p. ade et al .
( planck collaboration ) , _ planck 2013 results .
constraints on inflation , astron .
astrophys . _
* 571 * ( 2014 ) a22 [ arxiv:1303.5082 ] .
r. j. riegert , _ a nonlocal action for the trace anomaly _ , phys .
b134 ( 1984 ) 56 ; + e. s. fradkin and a. a. tseytlin , _ conformal anomaly in weyl theory and anomaly free superconformal theories _ , phys .
b134 ( 1984 ) 187 ; + i. antoniadis , p. o. mazur and e. mottola , _ scaling behavior of quantum four - geometries _ , phys .
b323 ( 1994 ) 284 ; + a. o. barvinsky , a. g. mirzabekian and v. v. zhytnikov , _ conformal decomposition of the effective action and covariant curvature expansion _ [ gr - qc/9510037 ] .
l. s. brown and j. p. cassidy , _ stress tensors and their trace anomalies in conformally flat space - times , phys .
* d 16 * ( 1977 ) 1712 ; + a. cappelli and a. coste , _ on the stress tensor of conformal field theories in higher dimensions , nucl .
* b 314 * ( 1989 ) 707 ; + i. antoniadis , p. o. mazur and e. mottola , _ physical states of the quantum conformal factor , phys .
_ * d 55 * ( 1997 ) 4770 [ hep - th/9509169 ] ; + m. beccaria and a.a .
tseytlin , _ higher spins in @xmath262 at one loop : vacuum energy , boundary conformal anomalies and ads / cft , jhep _ * 11 * ( 2014 ) 114 [ arxiv:1410.3273 ] .
a. m. pelinson , i. l. shapiro and f. i. takakura , _ on the stability of the anomaly - induced inflation _ ,
phys . b648 ( 2003 ) 417 [ hep - ph/0208184 ] ; + t. p. netto , a. m. pelinson , i. l. shapiro and a. a. starobinsky , _ from stable to unstable anomaly - induced inflation _ [ arxiv:1509.08882 ] v. f. mukhanov and g. v. chibisov , _ quantum fluctuation and nonsingular universe , j. exp .
lett . _ * 33 * ( 1981 ) 532 ; + v. mukhanov , h. feldman and r. brandenberger , _ theory of cosmological perturbations .
classical perturbations .
quantum theory of perturbations .
extensions , phys .
* 215 * ( 1992 ) 203 .
g. r. dvali , g. gabadadze , m. kolanovic and f. nitti , _ scales of gravity _ , phys .
d 65 , 024031 ( 2002 ) [ arxiv : hep - th/0106058 ] ; + g. dvali and m. redi , _ black hole bound on the number of species and quantum gravity at lhc _ , phys .
d77 ( 2008 ) 045027 [ arxiv:0710.4344 ] ; + g. dvali , _ black holes and large n species solution to the hierarchy problem _ , fortsch .
58:528 - 536,2010 [ arxiv:0706.2050 ] ; + g. dvali and s. n. solodukhin , _ black hole entropy and gravity cutoff _
, cern - ph - th-2008 - 141 [ arxiv:0806.3976 ] . z.komargodski and
a.schwimmer , _ on renormalization group flows in four dimensions _ , jhep 1112 ( 2011 ) 099 [ arxiv:1107.3987 ]
; + z.komargodski , _ the constraints of conformal symmetry on rg flows _ , jhep 1207 ( 2012 ) 069 [ arxiv:1112.4538 ] .
|
conformal higher spin ( chs ) field theory , which is a solid part of recent advanced checks of ads / cft correspondence , finds applications in cosmology .
hidden sector of weakly interacting chs fields suggests a resolution of the hierarchy problem in the model of initial conditions for inflationary cosmology driven by a conformal field theory .
these initial conditions are set by thermal garland type cosmological instantons in the sub - planckian energy range for the model of chs fields with a large positive coefficient @xmath0 of the gauss - bonnet term in their total conformal anomaly and a large number of their polarizations @xmath1 .
the upper bound of this range @xmath2 is shown to be much lower than the gravitational cutoff @xmath3 which is defined by the requirement of smallness of the perturbatively nonrenormalizable graviton loop contributions . in this way we justify the approximation scheme in which the nonrenormalizable graviton sector is subject to effective field theory under this cutoff ,
whereas the renormalizable sector of multiple chs fields is treated beyond perturbation theory and dynamically generates the bound on the inflation scale of the cft cosmology @xmath4 .
this confirms recent predictions for the origin of the starobinsky @xmath5 and higgs inflation models from the chs cosmology , which occurs at the energy scale three or four orders of magnitude below the gravitational cutoff , @xmath6 .
we also consider cosmological models dominated by fermionic chs fields with a negative @xmath0 and anomaly free models of infinite towers of chs fields with @xmath7 and briefly discuss the status of unitarity in chs models .
_ theory department , lebedev physics institute , leninsky prospect 53 , moscow 119991 , russia _ + _ department of physics , tomsk state university , lenin ave .
36 , tomsk 634050 , russia _ + _ and _ + _ pacific institue for theoretical physics , department of physics and astronomy , university
_ + _ of british columbia , 6224 agricultural road , vancouver , bc v6 t 1z1 , canada _
|
the most interesting idea associated with composite quantum systems is quantum entanglement .
a pair of particles is said to be entangled in quantum mechanics if its state can not be expressed as a product of the states of its individual constituents .
einstein , podolsky and rosen@xcite were the first to point out certain nontrivial consequences of entanglement on the ontology of quantum theory .
the preparation and manipulation of these entangled states lead to a better understanding of basic quantum phenomena .
for example , complex entangled states , such as the greenberger , horne and zeilinger@xcite triplets of particles are used for tests of quantum nonlocality@xcite . beyond these fundamental aspects
, entanglement has become a fundamental resource in quantum information processing , and there has been a rapid development of this subject in recent years@xcite .
cavity - qed has been a favourite tool to test the foundations of quantum mechanics including entanglement .
many beautiful experiments have been carried out , and in recent years , entangled states have been created and verified@xcite .
maximally entangled states between two modes in a single cavity have been generated using a rydberg atom coherently interacting with each mode in turn@xcite .
practical realization of various features of quantum entanglement are obtained in atom - photon interactions in optical and microwave cavities .
several studies have been performed to quantify the entanglement generated in atom - photon interactions in cavities@xcite . the above cavity - qed related investigations involved mostly the absorption or emission of a single photon in an atomic transition .
however , involvement of more than one photon , in particular , two photons in the transition between two atomic levels via a non - resonant intermediate level has been known for a long time @xcite .
the output radiation from such interactions exhibits novel non - classical properties such as sub - poissonian photon statistics .
needless to say , the idea of squeezed light has originated from a two - photon process@xcite .
two - photon processes have also been studied in cavity - qed @xcite .
it showed compact and regular quantum revivals in the atomic population in the single atom two - photon cavity - qed @xcite .
haroche and co - workers have demonstrated experimentally the two - photon maser action in a micromaser cavity@xcite . in total , the two - photon process mostly exhibits non - classical properties compared to the one - photon process@xcite .
thus , it would be interesting to study the properties of atom - atom entanglement in the framework of a two - photon process .
the two - photon atomic transition process also introduces a dynamic stark shift in the atomic transition which is related to the magnitude of the electric field of the radiation inside the cavity . this non - trivial effect which is naturally present in actual experiments involving two - photon transitions
has to be properly acounted for in the its theoretical analysis , or example in the two - photon micromaser@xcite .
various possibilities of exploiting the stark effect in quantum optical applications have been noticed in recent years . to name a few , schemes for applying the dc as well as the ac stark shifts towards implementation of quantum logic gates and algorithms@xcite , and in the improvement of photon sources for interferometry@xcite have been suggested .
it is thus tempting to study if the stark shift can be utilized to enhance atom - atom entanglement , as well .
the purpose of the present paper is to investigate the possibility of controlling atomic entanglement by the stark shift generated in atomic transitions inside cavities . in section
ii , we derive an effective hamiltonian which efficiently describes the atom - field two - photon interaction .
this hamiltonian is quadratic in the field operators , which is at the root of nonclasssical properties that this process exhibits .
the properties of the atom - atom entanglement generated through this interaction is studied in section iii .
we find that the atomic entanglement can be generally sustained , and for certain interaction times enhanced too , by making the stark shift compensate for the two - photon detuning . in section iv we show that similar trends are also obtained for a thermal cavity field .
we conclude the paper with a summary of our results in section v.
we consider a ladder system of a three - level rydberg atom interacting with a single mode of a microwave cavity field . the middle level may be a group of closely spaced levels removed far away from one - photon resonance .
thus the interaction involves simultaneous absorption ( or emission ) of two photons between the two atomic levels via a group of ( or one ) intermediate levels .
this is called a degenerate process since the two photons are from the same mode of the radiation field .
let us label the lower and the the upper level as @xmath1 and @xmath2 respectively and the intermediate levels are labelled as @xmath3 ( see figure @xmath4 ) .
0.02 cm ( upper level ) , @xmath5 ( middle level ) , @xmath1 ( lower level ) respectively.,width=226 ] [ fc3 ] the atom initially in the lower state @xmath1 absorbs a photon and jumps to one of the intermediate levels @xmath3 and from there moves to the upper level @xmath2 by absorbing another photon .
the return path of the atom from @xmath2 to @xmath1 is again via @xmath3 by the emission of two photons to the same mode . to make the equations look simpler
let us consider only one intermediate level @xmath5 .
a microscopically correct hamiltonian describing the above process can be written as @xmath6 where the interaction hamiltonian @xmath7 is of the order of one - photon dipole interaction strength .
the @xmath8 and @xmath7 can be written in operator form as @xmath9 and @xmath10 respectively . @xmath11 and @xmath12 are coupling constants ( in dipole approximation ) for the one - photon interactions responsible for the transitions between the level @xmath1 and @xmath5 and that between @xmath2 and @xmath5 respectively .
the atomic operators are given by @xmath13 and @xmath14 and similar definitions go for the operators involving the upper level @xmath2 . in the basis of the states @xmath15 , @xmath16 and @xmath17 , allowed by the rotating - wave approximation ,
we can write the hamiltonian in the form of a @xmath18 matrix . the density matrix of the atom - field system obeys the equation @xmath19.\end{aligned}\ ] ] the equation of motion can be solved by one of the usual known methods , but , the derivations get more and more tedious as the number of intermediate levels increases .
however , as these levels are removed far away from one - photon resonance , we can use this to obtain an effective hamiltonian which efficiently describes the two - photon process . in doing so ,
we follow the method outlined in refs.@xcite .
we start by making a canonical transformation @xcite @xmath20 where @xmath21 is time - independent and hermitian .
@xmath22 obeys the equation of motion @xmath23\end{aligned}\ ] ] where @xmath24 - 1/2[k,[k , h]]+ ........ \label{ec2}\end{aligned}\ ] ] we know that the probability of one - photon transition is inversly proportional to the one photon detuning @xmath25 ( @xmath26 ) . since this detuning is large , the one - photon transition probabilities are very small . in this situation ,
it is safe to retain terms upto second order in one - photon coupling constants . then , retaining terms of the order of square of the coupling constants in the above expression , we have @xmath27-\frac{1}{2}[k,[k , h_0 ] ] .
\label{ec3}\end{aligned}\ ] ] since @xmath21 is arbitrary , we choose that @xmath28=-ih_1 , \label{ec4}\end{aligned}\ ] ] from which we determine all the elements of @xmath21 .
this reduces the effective hamiltonian to @xmath29 .
\label{ec5}\end{aligned}\ ] ] written back in operator form , @xmath30 takes the form @xmath31s_z + \frac{1}{2}(\beta_e-\beta_g)a^\dagger a\nonumber\\ & + & g(s^+a^2+s^-{a^\dagger}^2 ) \label{ec6}\end{aligned}\ ] ] where the spin operators are @xmath32 , @xmath33 and @xmath34 .
@xmath35 is the two - photon detuning and is given by @xmath36 and @xmath37 is the two - photon coupling constant having the form @xmath38\end{aligned}\ ] ] where @xmath39 and @xmath40 .
the stark shifts associated with the levels @xmath41 and @xmath42 are , respectively , @xmath43 and @xmath44 we now have a hamiltonian describing the interaction between an an effective two - level system or an effective spin-@xmath45 system and a single mode radiation field of frequency @xmath46 .
if we take @xmath47 the hamiltonian reduces to @xmath48s_z+g(s^+a^2+s^-{a^\dagger}^2 ) \label{ec7}\end{aligned}\ ] ] it may be noted here that @xmath30 is a function of the two - photon detuning @xmath35 which is an outcome of the procedure followed here . in other words , we need not assume the resonance condition @xmath49 unlike in other methods in literature@xcite , but this method gives the two - photon detuning @xmath35 as an independent parameter for the analysis . the effective hamiltonian can easily be contrasted when compared with a hamiltonian describing one - photon process @xcite .
first , the effective hamiltonian is now cavity photon number dependent and is thus dynamic .
secondly , the hamiltonian is quadratic in annihilation and creation operators .
as mentioned earlier , they are at the root of all the nonclassical behaviours in two - photon processes . using such a hamiltonian
, we now study the entanglement of two such atoms passing through the cavity one after the other .
the effective hamiltonian derived above can be written in the matrix form in the basis of @xmath50 , @xmath51 states as @xmath52 the eigenvalues of @xmath30 are @xmath53 ^ 2+g^2(n+1)(n+2)},\\ { \lambda}_2=-\beta-\sqrt{[\frac{\delta}{2}+\beta ( n+1)]^2+g^2(n+1)(n+2)}.\end{aligned}\ ] ] the corresponding eigenstates can be written as @xmath54 where @xmath55 and @xmath56 we envisage a process in which two atoms pass through an ideal cavity ( @xmath57 ) such that there is no overlap of their flights there @xcite .
the interaction of each atom in the cavity is described by @xmath30 in eq.([ec8 ] ) .
we assume that the two atoms are in their respective upper states @xmath2 before they enter the cavity empty of photons .
after passage of the first atom through the cavity , the joint atom - field state is given at any time @xmath58 is @xmath59 where @xmath60,\\ s_1&=&[c_1 ^ 2\sin{({\lambda}_1t)}+c_2 ^ 2\sin{({\lambda}_2t)}],\\ r_2&=&c_1c_2[\cos{({\lambda}_1t)}-\cos{({\lambda}_2t)}],\\ s_2&=&c_1c_2[\sin{({\lambda}_1t)}-\sin{({\lambda}_2t)}].\end{aligned}\ ] ] the second atom then interacts with the cavity field modified by the passage of the first atom .
assuming the flight time of the two atoms through the cavity to be the same , the joint state of the two atoms and the cavity after the second leaves the cavity is given by @xmath61 where @xmath62 , @xmath63 , @xmath64 and @xmath65 .
next we calculate the two - atom mixed state taking trace over the field variables . the joint two - atom mixed state density matrix in the basis of @xmath66 , @xmath67 , @xmath68 and @xmath69 states
is given by @xmath70 where @xmath71 @xmath72 @xmath73 @xmath74 and @xmath75 we compute the two - atom entanglement using the well - known measure of the entanglement of formation@xcite given by @xmath76 where @xmath77 is called the concurrence defined by the formula @xmath78 where the @xmath79 are the eigenvalues of @xmath80 in descending order , and @xmath81 is the binary entropy function .
the entanglement of formation is monotone of concurrence .
the eigenvalues of @xmath82 in this case are given by @xmath83 , @xmath83 , @xmath84 and @xmath85 respectively .
is plotted vs rabi angle @xmath0 for ( i ) @xmath86 and @xmath87 ( solid line ) ( ii ) @xmath88 and @xmath89 ( dotted line).,width=245 ] we compute numerically the entanglement of formation @xmath90 for the two atoms , and plot it versus the rabi angle @xmath0 for different combinations of the two - photon detuning @xmath35 and the stark shift @xmath91 in the figures 2 and 3 .
we find that entanglement between the two atoms is controlled by the two - photon detuning and the stark - shift parameters .
we first plot @xmath90 in fig.2 for the resonant condition ( @xmath49 and @xmath92 ) .
now , if one includes a non - vanishing detuning and the resultant stark shift , one sees from fig.2 that the two - atom entanglement is diminished in general over a range of values of the rabi angle . it can be verified that as long as @xmath35 and @xmath91 are of the same sign , the magnitude of entanglement between the two atoms decreases with the increase of @xmath35 or @xmath91 .
is plotted vs rabi angle @xmath0 for ( i ) @xmath86 and @xmath87 ( solid line ) ( ii ) @xmath93 and @xmath94 ( dotted line).,width=245 ] the more interesting case is obtained when the detuning and the stark shifts are of opposite signs .
this is revealed in fig.3 where the two - atom entanglement @xmath90 versus @xmath0 for @xmath95 is compared with @xmath90 when @xmath96 .
we see that the magnitude of the entanglement between the two atoms increases when @xmath96 with respect to the case when @xmath95 for certain values of the rabi angle , e.g. , @xmath97 .
such enhancement of entanglement is observed in varying measures for other values of the rabi angle also .
its origin lies in the presence of the photon number operator @xmath98 in the effective hamiltonian in eq.(14 ) which makes the `` effective '' two - photon detuning @xmath99 dynamic as the cavity photon number oscillates in time .
one sees over a range of interaction times that the atomic entanglement is in general sustained with variations in oscillatory behaviour with respect to @xmath0 , if the two - photon detuning @xmath35 is compensated by the stark shift . or in other words , the maximum entanglement that can be obtained by varying @xmath0 over a range of interaction times remains similar to that in the resonant case .
this is in striking contrast to the case displayed in fig.2 where the entanglement reduces substantially with the increase of @xmath35 and @xmath91 , both having the same sign .
it is interesting to note that the entanglement between two atoms is also preserved if we interchange the sign of @xmath35 and @xmath91 but with the condition @xmath100 .
overall , we find that the stark shift acts as a control parameter for the atom - atom entanglement .
the thermal field is the most easily available radiation field , and so , its influence on the entanglement of spins is of much interest . the atomic entanglement mediated by the thermal field through single photon processes
have been studied earlier@xcite . since the thermal field is related to the temperature of the medium , photons of this field
are naturally present inside the cavity .
so it is not out of place to include the bose statistics for the thermal field in our analysis .
the field at thermal equilibrium obeying bose - einstein statistics has an average photon number at temperature @xmath101 , given by @xmath102 the photon statistics is governed by the distribution @xmath103 given by @xmath104 this distribution function always peaks at zero , i.e. , @xmath105 . for a field in a thermal state , the joint two - atom - cavity state is obtained by summing over all @xmath106 , and is given by @xmath107,\end{aligned}\ ] ] where @xmath108 is the photon distribution function of the thermal field .
the reduced mixed density matrix of two atoms after passing through the the thermal cavity field in the basis of @xmath66 , @xmath67 , @xmath68 and @xmath69 states is given by @xmath109 where @xmath110 @xmath111 @xmath112 @xmath113 and @xmath114 is plotted vs rabi angle @xmath0 for ( i ) @xmath86 and @xmath87 ( solid line ) ( ii ) @xmath88 and @xmath89 ( dotted line ) . the average thermal photon number @xmath115.,width=245 ] we again compute the entanglement of formation of the joint two - atom state after it emerges from the cavity .
similar to the case of the vacuum cavity field considered in section iii , the thermal field mediates entanglement between the two atoms even though there is no direct interaction between them .
this feature was also observed earlier in context of the one - photon atomic transition process@xcite .
the variation of the magnitude of the two - photon entanglement versus the rabi angle is displayed in the figs .
4 , 5 and 6 . is plotted vs rabi angle @xmath0 for ( i ) @xmath86 and @xmath87 ( solid line ) ( ii ) @xmath93 and @xmath94 ( dotted line ) . the average thermal photon number @xmath115.,width=245 ] our purpose here is to investigate the effect of stark shift on atomic entanglement , and to this end we plot in figs . 4 and 5 the entanglement of formation as a function of the rabi angle @xmath0 for different combinations of the two - photon detuning @xmath35 and the stark shift parameter @xmath91 when the thermal field has an average photon number @xmath115 .
we notice that the variation in @xmath90 as a function of @xmath0 is similar ( barring differences in magnitudes ) to the case of vacuum cavity field . from fig.4 , we note again , that as long as @xmath35 and @xmath91 are of same sign , the magnitude of entanglement between the two atoms decreases compared to the resonant case with the increase of @xmath35 or @xmath91 .
but , as seen from fig.5 , atomic entanglement can be increased for particular values of @xmath0 by choosing @xmath91 to be of opposite sign as @xmath35 .
again , similar to the case of the vacuum cavity field , we find that if the two - photon detuning is compensated by the stark shift ( fig .
@xmath116 ) , the atomic entanglement can be sustained on average over a range of values of the rabi angle .
these characteristics are still noticed for higher values of the average thermal photon number @xmath117 .
as seen from fig.6 , only the magnitude of @xmath90 is reduced with increase of average thermal photons .
thus , the stark shift can be used to control the atomic entanglement mediated by the thermal field by preserving the maximum magnitude of entanglement obtained in a large range of atom - cavity interaction times .
is plotted vs rabi angle @xmath0 for @xmath118 and @xmath89.,width=245 ]
in this paper we have investigated the possibility of the control of atomic entanglement by the stark shift generated in atomic transitions inside cavities . to this end
we have considered a degenerate two - photon process in a ladder system and obtained an effective hamiltonian which describes the interaction efficiently@xcite .
the two - atom entanglement is shown to be mediated by the cavity field through which the two atoms pass successively without any spatial overlap between them@xcite .
we are able to use the two - photon detuning which comes out naturally in the method presented in section ii , to compensate the dynamic stark shift to get the atom - atom entanglement . through this method
we have shown the possibility of using the dynamical stark shift in controlling atomic entanglement mediated both by the vacuum cavity field as well as by the thermal cavity field .
we have shown that the entanglement between two atoms depends on the two - photon detuning and the stark shift parameter .
the magnitude of atomic entanglement quantified by the entanglement of formation diminishes with the increase of the detuning and the stark shift .
however , interestingly , we have found that such a trend could be reversed if the values of the detuning and the stark shift are made to compensate each other . in the latter case the entanglement could be even enhanced compared to the resonant situation for particular values of the atom - photon interaction time .
more generally , it has been shown that the maximum magnitude of entanglement generated over a range of values of the rabi angle is nearly sustained if we set the values of the two - photon detuning and the stark - shift to be equal and opposite in sign .
the effects of photon statistics of the thermal field on the mediated entanglement@xcite has also been studied .
we have shown that the general characteristics of the atomic entanglement as a function of the photon detuning and the stark shift parameter are maintained for the case of the thermal field inside the cavity .
the use of stark shifts in some quantum information protocols has been suggested recently@xcite , and an experiment to demonstrate the enhancement of rydberg atom interactions has actually been perfomed@xcite .
our present study should motivate further investigations on the feasability of using stark shifts in the practical manipulation of quantum entanglement .
99 a. einstein , b. podolsky and n. rosen , phys . rev . * 47 * , 777 ( 1935 ) .
d. m. greenberger , m. a. horne , and a. zeilinger , am .
* 58 * , 1131 ( 1990 ) .
j. w. pan et al . ,
nature ( london ) * 403 * , 515 ( 2000 ) . c. h. bennett et al .
lett . * 70 * , 1895 ( 1993 ) ; m. zukowski , a. zeilinger , m. a. horne and a. k. ekert , phys , rev . lett . * 71 * , 4287 ( 1993 ) . j. m. raimond , m. brune and s. haroche , rev .
phys . * 73 * , 565 ( 2001 ) .
a. rauschenbeutel , p. bertet , s. osnaghi , g. nogues , m. brune , j. m. raimond and s. haroche , phys .
a * 64 * , 050301 ( 2001 ) .
masiak , phys .
a * 66 * , 023804 ( 2002 ) ; t. tessier , a. delgado - hilgado , i. fuentes - guridi , and i. h. deutsch , phys . rev .
a * 68 * , 062316 ( 2003 ) .
m. s. kim , j. lee , d. ahn and p. l. knight , phys .
a * 65 * , 040101(r ) ( 2002 ) ; l. zhou , h. s. song and c. li , j. opt .
b : quantum semiclass . opt . * 4 * , 425 ( 2002 ) .
a. datta , b. ghosh , a. s. majumdar and n. nayak , europhys . lett . * 67 * , 934 ( 2004 ) ; b. ghosh , a. s. majumdar , and n. nayak , int .
* 4 * , 665 ( 2006 ) ; b. ghosh , a. s. majumdar , and n. nayak , phys .
a * 74 * , 052315 ( 2006 ) .
b. ghosh , a. s. majumdar and n. nayak , int .
. inf . * 5 * , 169 ( 2007 ) .
m. lipeles , r. novick , and n. tolk , phys .
* 15 * , 690 ( 1965 ) ; k. j. mcneil and d. f. walls , j. phys .
a * 7 * , 617 ( 1974 ) ; n. nayak and b. k. mohanty , phys .
a * 19 * , 1204 ( 1979 ) ; b. nikolaus , d. z. zhang , and p. e. toschek , phys .
* 47 * , 171 , ( 1981 ) ; z. c. wang and h. haken , z. phys .
b * 55 * , 361 ( 1984 ) .
h. p. yuen , phys .
a * 13 * , 2226 ( 1976 ) ; c. m. caves , phys .
d * 23 * , 1693 ( 1981 ) .
r. r. puri and r. k. bullough , j. opt .
b * 5 * 2021 ( 1988 ) . v. bartzis and n. nayak , j. opt .
b * 8 * 1779 ( 1991 ) .
m. brune , j. m. raimond , p. goy , l. davidovich , and s. haroche , phys .
* 59 * , 1899 ( 1987 ) .
yoo and j. h. eberly , phys .
rep . * 118 * , 239 ( 1985 ) .
e. t. jaynes , f. w. cummings , proc . ieee * 51 * , 89 ( 1963 ) .
a. biswas and g. s. agarwal , phys .
rev . a*69 * , 062306 ( 2004 ) ; s. dasgupta , a. biswas and g. s. agarwal , phys . rev .
a*71 * , 012333 ( 2005 ) ; x
zou , y .- f .
xiao , s .- b . li and g .- c .
guo , phys .
* , 064301 ( 2007 ) .
m. j. fernee , h. r .- dunlop , and g. j. milburn , phys .
rev . a*75 * , 043815 ( 2007 ) .
m. takatsuji , phys .
a * 11 * , 619 ( 1975 ) .
s. hill and w. k. wootters , phys . rev .
* 78 * , 5022 ( 1997 ) ; w. k. wootters , phys .
lett . * 80 * , 2245 ( 1998 ) .
zanjani , j. a. petrus and j. d. d. martin , phys .
lett . * 98 * , 203005 ( 2007 ) .
|
we study the entanglement properties of two three - level rydberg atoms passing through a single - mode cavity .
the interaction of an atom with the cavity field allows the atom to make a transition from the upper most ( lower most ) to the lower most ( upper most ) level by emission ( absoprtion ) of two photons via the middle level .
we employ an effective hamiltonian that describes the system with a stark shifted two - photon atomic transition .
we compute the entanglement of formation of the joint two - atom state as a function of rabi angle @xmath0 .
it is shown that the stark shift can be used to enhance the magnitude of atomic entanglement over that obtained in the resonant condition for certain parameter values .
we find that though the two - atom entanglement generally diminishes with the increase of the two - photon detuning and the stark shift , it is possible to sustain the entanglement over a range of interaction times by making the detuning and the stark shift compensate each other .
similar characteristics are obtained for a thermal state cavity field too
. 0.5 cm
|
in this section , we outline the derivation of the tunneling current in eq .
( 2 ) of the main text .
this standard calculation is included to make the presentation self contained and closely follows the derivation for @xmath107 green functions presented in ref .
we describe the tunneling contact by the hamiltonian @xmath108 , where the three terms refer to the lead ( tip ) , the sample , and the tunnel coupling . the superconducting tip with chemical potential @xmath109 and gap @xmath6 is described by the bcs hamiltonian @xmath110,\end{aligned}\ ] ] where @xmath111 and @xmath112 ( @xmath113 ) annihilates ( creates ) an electron in the tip with momentum @xmath114 and spin @xmath115 .
the sample hamiltonian generally takes the form @xmath116 where @xmath117 is the hamiltonian in first quantization and @xmath118 annihilates an electron with spin @xmath115 at position @xmath45 in the sample .
we choose the superconducting order parameters in tip and sample to be real such that the superconducting phase difference @xmath119 enters the tunneling hamiltonian @xmath120,\ ] ] where @xmath121 the time argument , @xmath15 is the hopping strength , and @xmath122 annihilates an electron in the tip at the tunneling contact , which is located at the origin .
the sample is contacted at position @xmath45 and we suppress the position arguments in the following for simplicity . the time - dependent phase difference between the tip and the sample , @xmath123 , depends on the voltage @xmath124 applied to the junction .
we evaluate the current from the heisenberg equation of motion @xmath125 $ ] , where @xmath126 is the electron - number operator of the tip .
taking the expectation value , we obtain @xmath127\right\},\ ] ] where @xmath128 is a pauli matrix acting in nambu space , @xmath129 , and we have introduced the lesser green function in nambu and spin space with matrix elements given by @xmath130 .
here the @xmath131 are components of the nambu spinor @xmath132 with @xmath133 .
we are interested in tunneling processes to lowest order in the tunneling amplitude and therefore neglect andreev reflections in the tip , which give rise to multiple andreev reflections . as these processes involve several single - particle tunneling events they enter only at higher orders in the tunneling amplitude .
this approximation is exact when the sample is spin polarized ( e.g. , a proximity - coupled semiconductor nanowire in a strong magnetic field ) and one of the two spin components is fully normal reflected . in this case
, spin - flipping andreev reflections in the tip do not contribute to transport .
we follow the nonequilibrium green function approach described in ref . @xcite setting the offdiagonal elements of the tip green function in nambu space to zero and denoting the diagonal elements by @xmath134 .
we obtain the @xmath135 current @xmath136 where @xmath137g_r(\tau_1,\tau_2)$ ] is the sample green function in presence of the tip , @xmath138 are indices in nambu space denoting particle and hole components , and the trace is taken in spin space .
this expression has been derived in ref .
@xcite for the case when tip and sample are both spin - conserving @xmath78-wave superconductors .
according to the langreth rule , the lesser green function of the sample can be written as @xmath139 similar expressions also exist for the greater green function .
the first term involving @xmath140 gives rise to a single - particle current which at subgap energies requires inelastic processes in the sample .
the second term leads to the current carried by andreev reflections .
the sample self energy due to the presence of the tip can be written as @xmath141 we now focus on the andreev current and defer the discussion of the single - particle current to sec . [
sec : single - particle - current ] . thus neglecting the first term in eq .
( [ eq : gf_l ] ) we find eq .
( 2 ) of the main text , @xmath142 \label{eq : current}\end{aligned}\ ] ] where @xmath143 is the electron - hole block of the ( retarded ) green function of the sample and @xmath144 denotes the frobenius norm of matrix @xmath145 . here
we have used the relations @xmath146 and @xmath147 , where @xmath148 is the fermi distribution function , @xmath149 and @xmath150 is the normal density of state at the fermi energy in the tip . due to the step functions in @xmath151 the integration interval
is restricted to @xmath152 .
note that in this interval the self energy is purely imaginary .
in this section , we calculate the conductance for isolated majorana or andreev states at zero energy , cf . eqs . ( 6 ) and ( 9 ) of the main text , neglecting the contributions of all other states in the sample . while the main text focuses on andreev states in @xmath78-wave superconductors , we also consider more general spin structures here .
we first consider a topological superconductor substrate with a single zero - energy majorana state to the tip .
the majorana wavefunction has the form @xmath153 which maps onto itself under a particle - hole transformation . neglecting contributions from other states , we can approximate the sample green function by @xmath154 .
note that we suppress position arguments throughout this section . including the coupling to the tip the full green function can be written as @xmath155 where @xmath156 is the self energy projected onto the majorana bound state .
we obtain @xmath157}\left(\begin{array}{cc } u_{\uparrow}u_{\downarrow}^ { * } & -u_{\uparrow}^{2}\\ u_{\downarrow}^{*2 } & -u_{\downarrow}^{*}u_{\uparrow}. \end{array}\right)\ ] ] where we introduced @xmath158 .
thus we find @xmath159^{2}}\ ] ] and using eq .
( [ eq : current ] ) we obtain the current @xmath160,\label{eq : majorana_current}\ ] ] with @xmath161 .
we now evaluate the current near the threshold @xmath162 , @xmath163 the current at the opposite threshold @xmath164 follows from @xmath165 . at low temperatures , @xmath166
, we can set @xmath167 and the majorana current reads @xmath168 to lowest order in @xmath169 , one can approximate @xmath170 this yields @xmath171 as given in eq . ( 6 ) of the main text , where we have defined @xmath172 at the threshold we find @xmath173 which yields eq .
( 1 ) of the main text . at large voltages ,
@xmath174 , we instead find a negative differential conductance @xmath175 in agreement with the lineshape shown in fig .
( 1 ) of the main text .
a non - topological zero - energy andreev bound state is characterized by two nambu spinors @xmath176 where @xmath177 , @xmath178 are functions of space and @xmath179 is the time - reversal operator with @xmath180 the complex conjugation .
the lehmann representation of the real space green function is thus a @xmath181 matrix in nambu and spin space @xmath182 when spin is a good quantum number the green function may be reduced to a @xmath52 matrix in particle - hole space only . in the case of pure @xmath78-wave pairing , where cooper pairs are formed from electrons with opposite spin , we can set @xmath183 .
the spinors @xmath184 and @xmath185 then belong to the orthogonal subspaces spanned by @xmath186 and @xmath187 .
the green function decomposes into two @xmath107 blocks , which are related by particle - hole symmetry and have equal contributions to the current . in the opposite case of a spin
polarized @xmath54-wave superconductor we can set @xmath188 . now @xmath184 and @xmath185 belong to the same subspace spanned by @xmath189 and the green function is a single @xmath107 matrix .
it is therefore useful to first discuss a general @xmath107 nambu green function @xmath190 we will return to the @xmath181 case in sec . [ sec : spin - orbit ] when discussing superconductors with spin - orbit coupling .
the coupling to the tip can be included through the self - energy @xmath191 .
the green function of the coupled system is obtained from the dyson equation @xmath192 .
we find @xmath193 where @xmath194 .
using eq .
( [ eq : current ] ) , we obtain the current @xmath195 ^ 2+\pi^2\omega^2 t^4\left[(g_{ee}(\omega)\rho(\omega_-)+g_{hh}(\omega)\rho(\omega_+)\right]^2 } \label{eq : current_2by2}.\end{aligned}\ ] ] we now calculate the conductance for a zero - energy andreev state in the limiting cases of pure @xmath78-wave and spinless @xmath54-wave pairing and for a general spin - structure of the order parameter in the presence of spin - orbit coupling . the result for the three cases
are compared in fig .
[ fig : generic_conductance ] .
[ [ sec : s - wave - pairing ] ] when @xmath196 the green function in eq .
( [ eq:2abs_gf ] ) is block diagonal .
the @xmath52 block in the basis @xmath197 reads @xmath198 we find @xmath199 and thus @xmath62 vanishes . using eq .
( [ eq : current_2by2 ] ) , we arrive at the current as given in eq . ( 9 ) of the main text , @xmath200 where @xmath201 , @xmath202 .
we have included an extra factor of two to account for the second @xmath52 block of the green function which yields an equal contribution as a consequence of particle - hole symmetry . near the threshold
when @xmath162 , @xmath203 we find @xmath204 which can be obtained by a similar analysis as for the majorana bound state in the previous section .
the dimensionless function @xmath205 takes on values between 0 and 1 .
thus , the threshold differential conductance is @xmath206 [ [ sec : p - wave - pairing ] ] in a spin - polarized @xmath54-wave superconductor cooper pairs we can set @xmath207 .
the green function of an andreev state reduces to a single @xmath52 matrix @xmath208 we have introduced @xmath209 and using eq .
( [ eq : current_2by2 ] ) we obtain @xmath210^{2}+\pi^{2}t^{4}({\left| u_{\uparrow } \right|}^{2}+{\left| v_{\uparrow } \right|}^{2})^{2}[\rho(\omega_{+})+\rho(\omega_{-})]^{2}}\ ] ] with @xmath162 .
close to the threshold @xmath211 the denominator is dominated by @xmath212 and we generically find @xmath213 .
hence the conductance @xmath214 is continuous in contrast to the case of @xmath78-wave pairing . at certain points in space
it may be possible that @xmath215 in which case the conductance still jumps at the threshold .
slightly moving away from such points should restore the smooth onset of the conductance at the threshold .
the conductance exhibits a peak at the characteristic scale @xmath216 with @xmath217 as shown in fig .
[ fig : generic_conductance ] . realistic proposals of topological superconductors typically involve a mixture of @xmath78-wave and @xmath54-wave pairing .
in particular , such pairing arises in any superconductor with spin - orbit coupling . in this case
all components of @xmath218 and @xmath219 are generically nonzero . the full sample green function including the coupling to the tip due to the self - energy in eq .
( [ eq : self - energy ] ) can be written in terms of dyson series @xmath220\left(\begin{array}{c } \phi_{+}^{\dagger}\\ \phi_{-}^{\dagger } \end{array}\right ) .\end{aligned}\ ] ] a straightforward calculation reveals @xmath221\mathbf{v}^{\dagger}-\left[(\omega-\tilde{\sigma}_{++}^{r}){\cal c}\mathbf{v}+\tilde{\sigma}_{+-}^{r}\mathbf{u}\right]({\cal c}\mathbf{u})^{\dagger}}{\omega^{2}-\omega(\tilde{\sigma}_{++}^{r}+\tilde{\sigma}_{--}^{r})+\tilde{\sigma}_{++}^{r}\tilde{\sigma}_{--}^{r}-\tilde{\sigma}_{+-}^{r}\tilde{\sigma}_{-+}^{r}}\ ] ] where the projected self - energies are @xmath222 \\
\tilde{\sigma}^r_{--}(\omega)=-i \pi t^{2}\left[\|\mathbf{v}\|^{2}\rho(\omega_{-})+\|\mathbf{u}\|^{2}\rho(\omega_{+})\right ] \\
\tilde{\sigma}^r_{+-}(\omega)=-i \pi t^{2}\langle\mathbf{u},\theta\mathbf{v}\rangle\left[\rho(\omega_{-})+\rho(\omega_{+})\right ] \\
\tilde{\sigma}^r_{-+}(\omega)=-i \pi t^{2}\langle \theta\mathbf{v},\mathbf{u}\rangle\left[\rho(\omega_{-})+\rho(\omega_{+})\right]=-\tilde{\sigma}^r_{+-}(\omega)^{*},\end{gathered}\ ] ] and @xmath223 is the inner product . in the above derivations , we have used the anti - unitarity of the time - reversal operator and that @xmath224 , namely @xmath225 .
then the norm can be written as @xmath226z+4\rho(\omega_{+})\rho(\omega_{-})y\right\}}{\left\ { \omega^{2}-w\pi^{2}t^{4}\left[(\rho(\omega_{+})^{2}+\rho(\omega_{-})^{2})y+z\rho(\omega_{+})\rho(\omega_{-})\right]\right\ } ^{2}+w\omega^{2}\pi^{2}t^{4}(z+2)[\rho(\omega_{-})+\rho(\omega_{+})]^{2}}\ ] ] where @xmath227 the differential conductance is then a function of @xmath228 , @xmath229 and @xmath230 , where @xmath231 .
the parameter @xmath228 interpolates between @xmath78-wave pairing ( @xmath232 , sec .
[ sec : s - wave - pairing ] ) , where the threshold conductance is maximal , @xmath54-wave pairing ( @xmath233 , sec .
[ sec : p - wave - pairing ] ) , where the threshold conductance is zero .
-wave pairing case are @xmath232 , @xmath234 . for the @xmath54-wave pairing case , @xmath233 , @xmath235 . and in the case with spin - orbit coupling , @xmath236 , @xmath237.,scaledwidth=60.0% ] in fig .
[ fig : generic_conductance ] we show the conductance for different values of this parameter .
in this section , we provide details on evaluating the contribution of the quasiparticle continuum to the conductance near the threshold .
we identify two different effects of the continuum states , namely , ( a ) a possible smoothening of the step in the bound - state conductance [ see fig .
1(a ) and eq .
( 10 ) of the main text ] due to interference between resonant and nonresonant andreev reflections and ( b ) additonal contributions to the conductance from nonresonant andreev reflections .
the latter effect becomes important when probing the conductance away from the bound state , where resonant and nonresonant andreev reflections compete as discussed in the main text . as follows from sec .
[ sec : bound_state_conductance ] , the lineshape of a zero - energy andreev state in an @xmath78-wave superconductor closely resembles that of a majorana state .
in particular , the conductance exhibits a step at the threshold @xmath102 in both cases .
we now revisit these two cases and discuss whether this step is robust when the quasiparticle continuum is taken into account . as discussed in the main text , we find a suppression of the conductance for the andreev state close to the threshold , while the step remains robust for the majorana .
this behavior is reminiscent of tunneling from a normal metal tip , where majorana states appear as a robust zero - bias conductance peaks while zero - energy andreev states generically exhibit zero conductivity . for simplicity ,
we focus our analysis on samples in which spin is a good quantum number and that can be described in @xmath107 nambu space .
an example of a more general model is discussed in sec .
[ sec : qshi_edge ] .
in particular , we consider an @xmath78-wave superconductor with a zero - energy andreev state and a spinless @xmath54-wave superconductor with a majorana state . in these cases
the lehmann representation of the green function reads @xmath238 where the spinor @xmath239 describes a single zero - energy bound state and @xmath240 are continuum states above the gap @xmath11 with energy @xmath241 and index @xmath242 labeling the degeneracy . using this expression in the green function in the presence of a tunnel coupling in eq .
( [ green_func_2by2 ] ) and expanding to lowest order in energy , we obtain @xmath243 to determine the effect of the extra term @xmath244 we analyze the energy dependence of @xmath245 in more detail . the function @xmath246 has a simple pole at @xmath247 and branch cuts on the real axis for @xmath248 but is analytic elsewhere .
the residue of @xmath246 at @xmath64 is @xmath249 which has zero determinant and we can thus write @xmath250 and expand @xmath245 as a laurent series around @xmath64 @xmath251 furthermore @xmath246 is hermitian at nonzero subgap energies and thus its determinant is real .
we now determine the lowest - order cotribution to @xmath252 for majorana and andreev states .
the green function of a topological superconductor satisfies particle - hole symmetry @xmath253 with a unitary operator @xmath254 .
this yields @xmath255 and reality further requires @xmath245 to be an even function of @xmath58 .
hence , we obtain @xmath256 and @xmath257 , which approaches a constant in the limit @xmath258 with @xmath259 .
the majorana contribution to the current reads @xmath260 the rates @xmath261 diverge at the threshold and the continuum term @xmath62 becomes negligible .
the threshold conductance of majorana states thus remains unaffected by the continuum states .
in contrast to majoranas , the @xmath52 green function describing zero - energy andreev states generically does not satisfy particle - hole symmetry and thus @xmath262 . as an example of a system with trivial zero - energy andreev state
, we consider a magnetic impurity in an @xmath78-wave superconductor , which induces a bound state localized at the impurity , known as a shiba state .
the hamiltonian describing a shiba state localized at the origin due to a magnetic impurity can be written in first quantization as @xmath263 where @xmath264 and @xmath265 are pauli matrices in nambu and spin space , where @xmath124 and @xmath266 are the potential scattering and exchange coupling strength .
since @xmath267 is block diagonal in spin space , we only need to deal with one of the blocks , say @xmath268 . the other block related to
the first by particle - hole symmetry contributes a second ( equivalent ) zero - energy state as mentioned above .
the sample green function @xmath269 at the impurity position has the form @xcite @xmath270 where @xmath271 and @xmath11 are the normal density of states and gap of the sample and we introduced the dimensionless parameters @xmath272 and @xmath273 .
the shiba bound state energy is given by @xmath274 thus , since we focus on zero - energy bound states we set @xmath275 .
after a straightforward calculation , we obtain @xmath276 for small @xmath58 . from the green function
we also obtain ( for @xmath30 ) @xmath277 near the threshold , for small @xmath38 , the term @xmath278 in the denominator of eq .
( [ eq : current_2by2 ] ) dominates and we find @xmath279 thus the conductance @xmath280 is zero at @xmath102 and then rises linearly as described by eq .
( 10 ) of the main text .
this originates from the interference between resonant and nonresonant andreev reflection , which suppresses the differential conductance at @xmath102 .
in order to estimate the width of this suppression we determine the value of @xmath169 when the term @xmath281 in the denominator of eq .
( [ eq : current_2by2 ] ) becomes of the same order of magnitude as the tunneling broadening @xmath282 $ ] , i.e. , @xmath283 which yields @xmath284 we now determine the conductance due to nonresonant andreev reflections from the quasiparticle continuum in the substrate . to evaluate the conductance contributions from the continuum , we focus on a specific model of a dirac hamiltonian with a domain wall , where the mass changes sign .
we first calculate the real space green function and then evaluate the conductance as a function of separation from the impurity . as an example of a topological superconductor hosting a majorana state we calculate the green function of a dirac hamiltonian with a domain wall @xmath285\tau_z\label{dirac_hamiltonian}\ ] ] in the spinless nambu basis @xmath286 . here
@xmath287 is the velocity of the dirac fermion and @xmath288 is the effective mass .
the system satisfies particle - hole symmetry @xmath289 with the charge conjugation operator @xmath290 , where @xmath291 is the complex conjugation .
this is a generic model for the low - energy behavior of a topological superconductor close to the phase transition .
specifically , a spinless @xmath54-wave superconductor with a chemical potential close to the bottom of the band can be approximated by eq .
( [ dirac_hamiltonian ] ) . to compute the eigenstates of this hamiltonian
it is convenient to solve for the rotated hamiltonian @xmath292 .
notice that the charge conjugation operator is now given by @xmath293 .
one readily shows that the domain wall supports a single majorana state at zero energy with the wavefunction @xmath294 all other states in the system are extended with energies above the gap @xmath288 .
the wavefunctions of the continuum states can be determined using the ansatz @xmath295
+ \left ( \begin{array}{c } m{\rm sgn}\{x\ } \\
( v_fp\pm \epsilon ) \end{array } \right)e^{-ipx } \left[c\theta(-x)+d\theta(x ) \right ] , \end{aligned}\ ] ] where @xmath296 and @xmath297 is the positive eigenvalue .
the matching condition at @xmath298 leads to @xmath299 for positive - energy solutions , we get @xmath300 for incoming waves from @xmath301 we require @xmath302 and obtain @xmath303 similarly , for outgoing waves to @xmath301 we have @xmath304 and thus @xmath305 this yields the scattering states @xmath306 one can easily check that incoming and outgoing states are orthogonal @xmath307 . from the normalization conditions
@xmath308 we obtain the coefficients @xmath309^{-1/2}$ ] and @xmath310^{-1/2}$ ] for a system of size @xmath311 .
the green function can now be obtained from its spectral decomposition @xmath312 \label{eq : gf_rotated}\ ] ] where we have used the charge conjugation operator @xmath313 to relate the negative solution to the positive ones . focusing on @xmath314 for simplicity
we find the wavefunctions of the continuum states @xmath315 using eq .
( [ eq : gf_rotated ] ) , we obtain the green function @xmath316 here we have used @xmath317 notice that the matrix commutes with @xmath318 , and thus commutes with the rotation operation introduced at the beginning .
we find the green function for the original hamiltonian @xmath319 @xmath320 where @xmath321 is the normal density of states .
away from the mass domain wall , the sample green function at @xmath322 can be computed similarly from eq .
( [ eq.in_out_state ] ) and ( [ eq : gf_rotated ] ) , which yields @xmath323 where @xmath324^{-1}=v_f
/ m\sqrt{1-(\omega / m)^{2}}$ ] is the coherence length that characterizes the exponential decay of the majorana wavefunction .
the green function has a pole at zero energy whose amplitude decays exponentially away from the domain wall reflecting the probability density of the majorana bound state .
this singular part fully determines the threshold conductance and yields the quantized value @xmath325 .
notice that @xmath326 guarantees the robustness of this quantization consistent with the general argument based on particle - hole symmetry given in sec .
[ sec : continuum_majorana ] .
the conductance at voltages above the threshold depends on the competition between the singular part and a nonsingular contribution which originates from the quasiparticle continuum .
the continuum contribution can be readily evaluated for @xmath327 , where the green function describes a homogeneous @xmath54-wave superconductor , @xmath328 using this expression in the current formula in eq .
( [ eq : current_2by2 ] ) we find @xmath329 where we introduced the transmission probability of the junction @xmath330 and assumed @xmath331 .
this yields the conductance @xmath332 the conductance has a maximum at @xmath333 with a magnitude @xmath334 and a peak width @xmath335 . in fig .
[ fig.position_conductance ] we show the differential conductance _ vs _ voltage for different separations between tip and domain wall .
as the tip is moved away from the domain wall , nonresonant andreev reflections become more important and the conductance peak becomes narrower and slightly higher .
while the quantized conductance step for the majorana state derived in the main text is model independent , its robustness to continuum effects was only shown for spinless models described by a @xmath107 hamiltonian . as an example of a more realistic model , we consider the edge of a quantum spin hall insulator ( qshi ) .
we calculate the threshold conductance including the continuum contribution for the case when the sample is contacted at the domain wall .
we find the universal step at the threshold to be robust in line with the results for the spinless case .
this suggests that the conductance step at the threshold may be robust even for more general models of topological superconductors .
the edge of a qshi , with a domain wall at the origin , can be described by the first quantized hamiltonian @xmath336 where @xmath337.\ ] ] applying the unitary transformation @xmath338 , we have @xmath339 let @xmath340 , and focus on the inner block , which decribes a dirac field with a domain wall created by a mass jump , as discussed in sec .
[ sec : majorana_dirac ] . according to eq .
( [ eq : gf_dirac ] ) , the green function at the domain wall is @xmath341 we rotate it back to the original basis @xmath342 and obtain the full green function @xmath343 where the self energy is given by @xmath344 calculation reveals @xmath345 + 2i\pi^{2}\nu_smt^{2}\sqrt{1-\frac{\omega^{2}}{m^{2}}}\left[\rho(\omega_{-})+\rho(\omega_{+})\right]}\\ & \times\left(\begin{array}{cc } i & \sqrt{1-\frac{\omega^{2}}{m^{2}}}-i\pi\left(\frac{\pi\nu_s\omega}{m}\right)t^{2}(\rho(\omega_{-})+\rho(\omega_{+}))\\ \sqrt{1-\frac{\omega^{2}}{m^{2}}}-i\pi\left(\frac{\pi\nu_s\omega}{m}\right)t^{2}(\rho(\omega_{-})+\rho(\omega_{+ } ) ) & -i \end{array}\right).\end{aligned}\ ] ] denoting @xmath346 we find @xmath347}{\omega^{2}\left[1-(\gamma_e+\gamma_h)^{2}/16m^{2}\right]^2+(\gamma_e+\gamma_h)^{2}/4}.\ ] ] we readily verify that the denominator is dominated by the second term at the threshold .
in particular , we see the continuum contribution in the denominator appears proportional to @xmath348 as for the spinless case .
hence , the conductance quantization at @xmath102 is unaffected by the continuum contribution in this model .
in this section we analyze the single - particle tunneling current that can flow in addition to the andreev current discussed above .
a single - particle current is possible if the quasiparticle occupying the bound state can relax to the quasiparticle continuum .
such relaxation can occur , e.g. , due to inelastic transitions assisted by phonons or photons .
we concentrate on the majorana state , although a similar analysis is possible for a trivial andreev state .
we neglect the quasiparticle continuum and include relaxation as a phenomenological parameter , without specifying its microscopic origin .
we first derive the single - particle current [ given in eq . (
11 ) of the main text ] , from which we then calculate the threshold conductance from single - particle tunneling [ eq . ( 12 ) ] .
let us assume that the local environment of the majorana state introduces relaxation processes , induced by phonons or photons , via the self - energy @xmath349 .
the substrate green function can be determined from the dyson series @xmath350 where @xmath57 is the green function without relaxation processes .
we approximate the bare substrate green function @xmath57 by the majorana contribution , @xmath351 and project the self - energy onto the majorana subspace introducing @xmath352 where we approximate the self - energy by its value at zero energy .
the invariance of the majorana state under particle - hole transformation guarantees that the expectation value @xmath353 is purely imaginary .
thus , the retarded and advanced green functions of the majorana state read @xmath354 in quasi - equilibrium , we can express the greater and lesser green function in terms of the retarded and advanced green functions , @xmath355 where @xmath356 is the quasi - equilibrium distribution function and we used the relations @xmath357 these terms are the rates for emptying and filling of the delocalized fermion formed by the majorana at the contact and a second one far away .
note that local transitions can change the occupation of this state . since this fermion has zero energy
the two rates are equal @xmath358 according to detailed balance .
we can now evaluate the current in eq .
( [ current_formula_green_function ] ) . besides the andreev current in eq .
( [ eq : current ] ) , we find the single - particle current from the first term in eq .
( [ eq : gf_l ] ) @xmath359-\gamma_{\rm qp}[\gamma_{e}(\omega)(1-n_f(\omega_-))-\gamma_{h}(\omega)(1-n_f(\omega_+))]}{\omega^{2}+\left(\gamma_e(\omega)+\gamma_h(\omega)+\gamma_{\rm qp}\right)^2/4},\label{is_app}\ ] ] which gives eq . ( 11 ) in the main text if we take @xmath360 and @xmath361 assuming @xmath362 .
we compute the single - particle current near the threshold at @xmath162 with @xmath363 @xmath364 to lowest order in @xmath169 .
we thus obtain the differential conductance @xmath365^{2 } } , \quad x=\frac{ev-\delta}{\omega_t}.\ ] ] now we focus on the conductance at the threshold , namely @xmath298 . for weak tip - substrate tunneling , @xmath366 , we find @xmath367^{2}}\nonumber \\ & = \frac{2\pi e^2}{h } \frac { \omega_t^{3/2}}{\gamma_{\rm qp}^{3/2 } } , \end{aligned}\ ] ] where the @xmath45-integration is elementary . in the opposite limit of strong tip - substrate tunneling , @xmath368 , we can neglect the contribution of @xmath96 in the denominator . in this limit
, we find the peak conductance @xmath369^{2}}\nonumber \\ & = \frac{2\pi e^2}{9h } \frac{\gamma_{\rm qp}}{\omega_t}. \end{aligned}\ ] ] the parametric dependence of @xmath370 is summarized in eq .
( 12 ) of the main text .
|
experimental evidence for majorana bound states largely relies on measurements of the tunneling conductance .
while the conductance into a majorana state is in principle quantized to @xmath0 , observation of this quantization has been elusive , presumably due to temperature broadening in the normal - metal lead . here , we propose to use a superconducting lead instead , whose gap strongly suppresses thermal excitations . for a wide range of tunneling strengths and temperatures ,
a majorana state is then signaled by symmetric conductance peaks at @xmath1 of a universal height @xmath2 . for a superconducting scanning
tunneling microscope tip , majorana states appear as spatial conductance plateaus while the conductance varies with the local wavefunction for trivial andreev bound states .
we discuss effects of nonresonant ( bulk ) andreev reflections and quasiparticle poisoning .
_ introduction._motivated by possible applications in quantum information processing @xcite , topological superconductors hosting majorana bound states are currently under intense investigation @xcite . based on the superconducting proximity effect ,
various realistic platforms have been proposed to support majorana states including topological insulators @xcite , semiconductor nanowires @xcite , and atomic chains @xcite .
although these systems are available in the laboratory , the experimental observation of unique majorana signatures remains challenging .
a widely employed diagnostic tool is the tunneling conductance of normal metal
superconductor junctions , in which majoranas manifest themselves as characteristic zero - bias peaks @xcite .
experimental signatures consistent with theoretical predictions have been observed in quantum wires @xcite and atomic chains @xcite .
however , it is a major challenge in these experiments to uniquely distinguish majoranas from conventional fermionic subgap states .
spin - polarized subgap states such as shiba states bound to magnetic impurities @xcite or andreev bound states in a magnetic field can exhibit a zero - energy crossing as a function of exchange interaction or zeeman energy @xcite .
thus , such fermionic states may accidentally occur at zero energy and give rise to similar conductance features .
as magnetic impurities or external magnetic fields are also required for the most relevant realizations of topological superconductors , such trivial conductance peaks can generally not be disregarded .
in contrast to fermionic subgap states , majoranas exhibit a celebrated quantized zero - bias conductance of @xmath0 @xcite .
unfortunately , this has so far proved difficult to observe in experiment . the fermi distribution in the metal lead is smooth on the scale of the temperature @xmath3 , which strongly limits the experimental energy resolution . when temperature is larger than the tunnel coupling , the majorana peak is broadened and the zero - bias conductance is reduced . even at low temperatures ( e.g. , @xmath4mk in ref .
@xcite ) , it may be difficult to observe the quantized peak height as multichannel effects limit the relevant tunneling strength @xcite .
quasiparticle poisoning may also lead to deviations from quantization .
a fermion - parity breaking rate exceeding the tunnel coupling broadens the peak and reduces its height .
this requires one to work at temperatures below the lowest fermionic excitations in the topological superconductor . in this paper
, we show how a robust conductance signature of majorana bound states can be obtained by employing superconducting leads . in striking contrast to normal - state contacts ,
effects of thermal broadening are strongly suppressed for a superconducting lead because quasiparticle excitations are exponentially suppressed @xmath5 by its superconducting gap @xmath6 .
bound states no longer appear as zero - bias anomalies but rather as two symmetric peaks in the differential conductance @xmath7 which occur when the bcs singularity of the superconducting gap lines up with the majorana bound state , i.e. , at the thresholds @xmath8 .
these peaks have a universal height @xmath9 which persists over a wide range of tunnel couplings .
this yields particularly striking evidence when employing a scanning tunneling microscope ( stm ) with a superconducting tip which allows for spatially resolved measurements .
this has previously been used to map out bound state wavefunctions in conventional and unconventional superconductors @xcite .
here we propose that such maps can clearly distinguish between majoranas and trivial zero - energy bound states .
indeed , the peak conductance is uniform in the vicinity of majorana states and a conductance map exhibits a characteristic mesa or plateau structure . in contrast
, the conductance of trivial subgap states exhibits a spatial pattern which is governed by the bound - state wavefunction .
in addition , stm measurements allow for systematic studies as a function of tunneling strength by varying the tip height .
it was recently demonstrated @xcite that this can be exploited to probe quasiparticle relaxation processes . in the present context ,
varying the tunneling strength may help to identify majorana signatures despite competing effects such as nonresonant andreev reflections or quasiparticle poisoning .
_ subgap conductance for majorana bound state._at subgap voltages @xmath10 and zero temperature , the tunneling current between superconducting tip or lead and substrate ( with gap @xmath11 ) flows by multiple andreev reflections . near the threshold @xmath12 , the differential conductance @xmath13 is dominated by single andreev reflections from the sample . for tip locations
far from the zero - energy bound state in the sample , this yields the familiar peak in @xmath13 due to the singular densities of states of incoming electrons and outgoing holes . in the vicinity of the bound state ,
tunneling is further enhanced by the zero - energy resonance @xcite .
formally , the subgap current due to single andreev reflections from the sample can be expressed as @xcite @xmath14 \nonumber\\ & \times \rho(\omega_-)\rho(\omega_+)[n_f(\omega_-)-n_f(\omega_+)],\label{current}\end{aligned}\ ] ] where @xmath15 is the amplitude for tip - substrate tunneling , @xmath16 , @xmath17 denotes the fermi function , and the superconducting tip enters through its bcs density of states @xmath18 with @xmath19 the normal density of states at the fermi energy .
spin or subband degrees of freedom are accounted for by a possible matrix structure of the anomalous retarded green function @xmath20 of the substrate at the tip position @xmath21 . in terms of its lehmann representation , @xmath20 has contributions from both the bound state and the above - gap continuum . in the following ,
we first consider the resonantly enhanced andreev current from a majorana bound state and subsequently discuss the contribution of the quasiparticle continuum . for @xmath22 , we can approximate @xmath23 in eq .
( [ current ] ) , up to corrections of order @xmath24 .
this insensitivity to temperature is a key advantage of superconducting leads .
the bound - state contribution to the substrate green function is @xmath25 here , @xmath26^{\rm t}$ ] denotes the local bogoliubov de gennes wavefunction of the majorana bound state with @xmath27 the time - reversal operator .
the broadening @xmath28 of the bound state is induced by the tunnel coupling to the lead .
the corresponding self energy @xmath29 $ ] is diagonal as andreev reflections in the lead can be neglected near @xmath12 .
inserting eq .
( [ dirac_green_function ] ) into ( [ current ] ) yields ( for @xmath30 ) @xcite @xmath31 ^ 2/4 } \label{diraccurrent}\end{aligned}\ ] ] in terms of the electron and hole tunneling rates @xmath32 . while the integrand in eq .
( [ diraccurrent ] ) has a resonance denominator , its behavior is peculiar due to the strong energy dependence of the tunneling rates .
specifically , the square - root singularity of the bcs density of states implies that the integrand involves a characteristic energy scale @xmath33 which depends on a fractional power of the tunneling rate from a normal tip @xmath34 . in the weak - tunneling regime @xmath35 , we can write @xmath36 for @xmath37 , where @xmath38 measures the voltage from the threshold @xmath6 . in the vicinity of the threshold , @xmath39 ,
the resonance denominator is dominated by the second term and we obtain @xmath40 and thus eq . ( [ majorana_conductance ] ) .
the entire peak lineshape @xmath41 involves the function @xmath42 which vanishes for @xmath43 , jumps to @xmath44 , and falls off with a small negative differential conductance tail at large @xmath45 , cp . fig .
[ fig : lineshape ] . for majorana ( solid line ) and andreev state with @xmath46 ( dashed line ) .
for a majorana , the conductance exhibits a step of height @xmath47 at the threshold . for an andreev state
, the conductance has a smooth onset , cf .
eq .
( [ dirac_conductance_bulk ] ) .
both peaks have a negative - differential conductance dip at high voltages .
inset : graph of @xmath48 as defined in the main text .
( b ) spatial conductance maps for majorana ( left ) and andreev state ( right ) for @xmath49 .
the majorana gives rise to a conductance plateau whereas the andreev state exhibits a pattern reflecting the spatial dependence of the ratio @xmath50 .
the majorana conductance drops far from the bound state when the broadening exceeds @xmath51.,title="fig:",scaledwidth=45.0% ] + for majorana ( solid line ) and andreev state with @xmath46 ( dashed line ) . for a majorana ,
the conductance exhibits a step of height @xmath47 at the threshold . for an andreev state
, the conductance has a smooth onset , cf .
eq .
( [ dirac_conductance_bulk ] ) .
both peaks have a negative - differential conductance dip at high voltages .
inset : graph of @xmath48 as defined in the main text .
( b ) spatial conductance maps for majorana ( left ) and andreev state ( right ) for @xmath49 .
the majorana gives rise to a conductance plateau whereas the andreev state exhibits a pattern reflecting the spatial dependence of the ratio @xmath50 .
the majorana conductance drops far from the bound state when the broadening exceeds @xmath51.,title="fig : " ] for majorana ( solid line ) and andreev state with @xmath46 ( dashed line ) . for a majorana ,
the conductance exhibits a step of height @xmath47 at the threshold . for an andreev state
, the conductance has a smooth onset , cf .
eq .
( [ dirac_conductance_bulk ] ) .
both peaks have a negative - differential conductance dip at high voltages .
inset : graph of @xmath48 as defined in the main text .
( b ) spatial conductance maps for majorana ( left ) and andreev state ( right ) for @xmath49 .
the majorana gives rise to a conductance plateau whereas the andreev state exhibits a pattern reflecting the spatial dependence of the ratio @xmath50 .
the majorana conductance drops far from the bound state when the broadening exceeds @xmath51.,title="fig : " ] thus , the differential conductance between a conventional superconductor and a majorana state exhibits a peak which is independent of tunneling strength and majorana wavefunction .
while the peak height is close to the quantized majorana peak height @xmath0 for a normal - metal lead , there are several differences : ( a ) there are two symmetric , finite - bias majorana peaks at @xmath1 rather than a single zero - bias peak , ( b ) the conductance peak is strongly asymmetric with a discontinuous step at the threshold , and ( c ) the width of the peak is set by @xmath51 with its sublinear dependence on junction transparency .
the threshold discontinuity in the conductance persists even when including the contributions of the quasiparticle continuum in the substrate green function . to see this
, we model the substrate superconductor by a @xmath52 nambu green function @xmath53 . for a topological substrate , this is appropriate for perfect spin polarization ( spinless @xmath54-wave superconductor ) . including the tunnel coupling to the tip through the self energy @xmath55 as given above , the substrate green function becomes @xmath56^{-1}$ ] .
we first focus on the vicinity of the bound state where the conductance is dominated by andreev reflections from the bound state . by straightforward calculation and expansion of @xmath57 in @xmath58 @xcite , we find @xmath59 this differs from the pure bound - state contribution by the additional term @xmath60 in the denominator which involves the determinant ( in particle - hole space ) of the bare substrate green function . while the determinant of the bound - state contribution to the green function vanishes , this is no longer the case when including the quasiparticle continuum . at subgap energies away from bound states ,
the green function @xmath53 is a hermitian @xmath52 matrix , so that @xmath61 and hence @xmath62 are real .
thus , we find @xmath63 for a majorana state , the real part of the resonance denominator must vanish exactly at @xmath64 . indeed , particle - hole symmetry further constrains @xmath61 to be an even function of @xmath58 which can be approximated as a constant at small @xmath58 ( see @xcite , where this conclusion is confirmed by model calculations ) .
then , we find @xmath65 near the threshold . even with this term , the denominator in eq .
( [ threshold_current_bulk ] ) remains dominated by the divergent tunnel broadenings @xmath66 and the discontinuous conductance step as well as the universal value of the threshold conductance in eq .
( [ majorana_conductance ] ) persist . in experiment ,
the square - root singularity of the bcs density of states of the tip may be broadened intrinsically due to higher - order processes or effectively due to experimental resolution .
the universal threshold conductance persists as long as @xmath51 exceeds this broadening .
this condition also determines the spatial extent of the conductance plateau , @xmath67/3 $ ] , where @xmath68 is the majorana localization length , @xmath69 denotes the value of @xmath51 at the center of the majorana bound state , and @xmath70 is the broadening of the tip density of state , cf .
fig .
[ fig : lineshape](b ) .
of course , a well - resolved majorana peak also requires @xmath71 , i.e. , the tunnel broadening needs to be small compared to the induced gap .
if the peak is not fully resolved , it is suppressed below the universal value and its height may vary as a function of space . for tip locations far from the bound state , the tunneling conductance is dominated by conventional ( nonresonant " ) andreev reflections . these still yield a threshold peak due to the singular tip density of states in @xmath72 and @xmath73 , but are not enhanced by a bound - state resonance . for a one - dimensional @xmath54-wave superconductor ,
this conductance peak has height @xmath74 and width @xmath75 quadratic in the junction transparency @xmath76 @xcite . observing the conventional andreev peak
thus requires that the broadening of the tip density of states is small compared to @xmath75 .
this is a much more stringent condition than for the resonant andreev peak as the width of the bound - state peak @xmath77 involves a lower power of @xmath15 .
we note that in a typical stm experiment @xcite , conventional andreev peaks can be resolved only for small tip - sample distances , while bound - state signatures persist to much weaker tunnel couplings .
_ subgap conductance for andreev bound state._these results should be contrasted with those for trivial zero - energy andreev bound states . for concreteness , consider an @xmath78-wave superconductor with conserved spin @xcite , whose bogoliubov de gennes description decomposes into two independent spin sectors that interchange under particle - hole transformations .
a zero - energy andreev state corresponds to two bogoliubov de gennes wavefunctions , @xmath79^{\rm t}$ ] and @xmath80^{\rm t}$ ] , one in each sector .
an analogous calculation @xcite yields the threshold current @xmath81 reflecting the two zero - energy wavefunctions , the maximal threshold conductance is twice that in the majorana case , @xmath82 , and realized for the particle - hole symmetric case @xmath46 . in general
, the peak conductance depends on the ratio of electron and hole wavefunction at the tip position .
this dependence is captured by the dimensionless function @xmath83 which takes on values between 0 and 1 and is plotted in fig .
[ fig : lineshape](a ) .
the function satisfies @xmath84 as the two spin sectors contribute equally . in the limit of large particle - hole asymmetry , @xmath85 .
the lineshape of the conductance peak is similar to the majorana peak , with a width of order @xmath51 upon replacing @xmath86 by @xmath87 .
our results imply that the height of the conductance peak allows for a clear distinction between a conventional andreev bound state and a majorana state .
even when @xmath88 for one location of the stm tip , moving the tip to another location modifies the conductance peak height for a conventional bound state , tracking the ratio of electron and hole wavefunctions .
in contrast , the conductance map exhibits a characteristic mesa structure for a majorana state , see fig .
[ fig : lineshape](b ) . in non - stm
tunneling experiments , changes of parameters ( e.g. , gate voltages ) which affect the majorana wavefunction should leave the peak height unchanged for a majorana but not for a conventional andreev bound state . as there is no locking of the bound state to zero energy ,
also the continuum contribution is distinctly different for conventional andreev states .
the two spin sectors are described by separate @xmath52 nambu green functions which map into one another under particle - hole transformations .
this is quite unlike the majorana green function which maps onto itself . for each sector
, @xmath61 is therefore no longer an even function of @xmath58 and will generally have a singular contribution @xmath89 at the threshold so that @xmath90 .
these general arguments can be confirmed explicitly for shiba states in @xmath78-wave superconductors @xcite . near the threshold , the resonance denominator in the expression for the current
is now dominated by @xmath62 . as illustrated in fig .
[ fig : lineshape ] by a numerical evaluation of the current , this suppresses the conductance step . analytically , we find that just above the threshold , the conductance increases linearly , @xmath91 and matches with the conductance obtained from eq .
( [ dirac_current ] ) for @xmath92 .
we note that this suppression of the conductance step depends on @xmath93 and can thus be probed by varying the tip - sample distance in an stm experiment .
this may serve as an additional signature to distinguish between majorana and conventional andreev bound states . .
the single - particle contribution affects the conductance only a in a window of transmission values , where @xmath94 .
while the maximum is of order @xmath95 , the position of the maximum in tunneling strength depends sensitively on temperature ( through @xmath96 ) .
inset : line shape of the total conductance as a function of voltage away from the threshold , for different ratios of @xmath97.,scaledwidth=45.0% ] _ effects of quasiparticle poisoning._so far , we only included bound - state broadening by the tunneling contact . at finite temperatures
, the bound - state occupation also changes by inelastic transitions to other subgap states or the quasiparticle continuum in the sample @xcite .
we account for these processes by an additional contribution @xmath98 to the self energy of the bound - state green function eq .
( [ bulk_green_function ] ) .
this does not affect the andreev current at the threshold , where the denominator is dominated by the diverging tunnel coupling .
however , the overall weight of the peak is reduced by a _
narrowing _ of the linewidth by a factor @xmath99 once @xmath100 , see fig .
[ fig : peakheight ] ( inset ) .
in addition , quasiparticle poisoning generates a single - electron current @xmath101 which involves tunneling of single particles followed by inelastic transitions from the zero - energy bound state to other bound states or the quasiparticle continuum @xcite . for a majorana state
, we find near the threshold @xmath102 ( with analogous results applying for andreev bound states ) @xcite @xmath103}{\omega^2+[\gamma_{\rm qp } + \gamma_e(\omega)+\gamma_h(\omega)]^2/4}.\end{aligned}\ ] ] for weak and strong tunneling , this yields @xcite @xmath104 figure [ fig : peakheight ] shows that this single - particle contribution assumes a maximum of @xmath105 when @xmath106 .
however , it can be easily made negligible by tuning the system away from this maximum through varying temperature or tunneling strength .
_ conclusions._we show that conductance measurements with superconducting leads constitute a promising technique to identify majorana states .
the presence of majoranas is signaled by conductance peaks of universal height which are largely unaffected by thermal broadening , a key obstacle in previous experiments with normal - metal contacts .
we discuss strategies to systematically rule out parasitic effects such as quasiparticle poisoning or trivial subgap states .
the proposed setup is readily available in the laboratory and , in fact , has already been realized in previous experiments @xcite .
( notice , however , that temperature was comparable to the induced gap in the stm experiments performed to date , precluding observation of the universal conductance , and that the nanowire experiments focused on zero - bias peaks . ) our results also imply that quasiparticle poisoning rates can be extracted from systematic measurements as a function of tip height and temperature . _
acknowledgments._we thank p. brouwer , k. franke , b. heinrich , j. meyer , y. oreg , and m .-
t .
rieder for stimulating discussions .
we acknowledge financial support by the helmholtz virtual institute new states of matter and their excitations , " sfb 658 , spp1285 and spp1666 of the deutsche forschungsgemeinschaft , the humboldt foundation , the minerva stiftung , as well as doe contract de - fg02 - 08er46482 at yale university .
we are grateful to the aspen center for physics , supported by nsf grant no .
phys-106629 , for hospitality while this line of work was initiated
. 99 a. kitaev , ann .
phys . * 303 * , 2 ( 2003 ) . c. nayak , s.h .
simon , a. stern , m. freedman , and s. das sarma , rev .
mod .
phys . * 80 * , 1083 ( 2008 ) .
j. alicea , rep .
prog .
phys .
* 75 * , 076501 ( 2012 ) .
c.w.j .
beenakker , annu .
rev .
con . mat .
phys . * 4 * , 113 ( 2013 ) .
s.r .
elliott and m. franz , rev .
mod .
phys .
* 87 * , 137 ( 2015 ) .
l. fu and c.l .
kane , phys .
rev .
lett . * 100 * , 096407 ( 2008 ) .
l. fu and c. l. kane , phys .
rev .
b * 79 * , 161408(r ) ( 2009 ) . r.m .
lutchyn , j.d .
sau , and s. das sarma , phys .
rev .
lett . * 105 * , 077001 ( 2010 ) .
y. oreg , g. refael , and f. von oppen , phys .
rev .
lett . *
105 * , 177002 ( 2010 ) .
s. nadj - perge , i.k .
drozdov , b.a .
bernevig , and a. yazdani , phys .
rev .
b * 88 * , 020407(r ) ( 2013 ) .
b. braunecker and p. simon , phys .
rev .
lett . * 111 * , 147202 ( 2013 ) .
m.m .
vazifeh and m. franz , phys .
rev .
lett . * 111 * , 206802 ( 2013 ) .
j. klinovaja , p. stano , a. yazdani , and d. loss , phys .
rev .
lett . * 111 * , 186805 ( 2013 ) .
f. pientka , l.i .
glazman , and f. von oppen , phys .
rev .
b * 88 * , 155420 ( 2013 ) .
y. kim , m. cheng , b. bauer , r.m .
lutchyn , and s. das sarma , phys .
rev .
b * 90 * , 060401(r ) ( 2014 ) .
y. peng , f. pientka , l.i .
glazman , and f. von oppen , phys .
rev .
lett . * 114 * , 106801 ( 2015 ) .
k.t .
law , p.a .
lee , t.k .
ng , phys .
rev .
lett . * 103 * , 237001 ( 2009 ) .
k. flensberg , phys .
rev .
b * 82 * , 180516 ( 2010 ) .
m. wimmer , a.r .
akhmerov , j.p .
dahlhaus , c.w.j .
beenakker , new j. phys .
* 13 * , 053016 ( 2011 ) .
v. mourik , k. zuo , s.m .
frolov , s.r .
plissard , e.p.a.m .
bakkers , and l.p .
kouwenhoven , science * 336 * , 1003 ( 2012 ) .
a. das , y. ronen , y. most , y. oreg , m. heiblum , and h. shtrikman , nature phys .
* 8 * , 887 ( 2012 ) .
h.o.h .
churchill , v. fatemi , k. grove - rasmussen , m.t .
deng , p. caroff , h.q .
xu , and c.m .
marcus , phys .
rev .
b * 87 * , 241401(r ) ( 2013 ) .
s. nadj - perge , i.k .
drozdov , j. li , h. chen , s. jeon , j. seo , a.h .
macdonald , b.a .
bernevig , a. yazdani , science * 346 * , 602 ( 2014 ) .
m. ruby , f. pientka , y. peng , f. von oppen , b.w .
heinrich , k.j .
franke , phys .
rev .
lett . * 115 * , 197204 ( 2015 ) .
l. yu , acta phys . sin .
* 21 * , 75 ( 1965 ) .
h. shiba , prog .
theor .
phys . * 40 * , 435 ( 1968 ) .
a.i .
rusinov , zh . eksp .
teor .
fiz .
pisma red .
* 9 * , 146 ( 1968 ) [ jetp lett . *
9 * , 85 ( 1969 ) ] .
a.v .
balatsky , i. vekhter , and j .- x .
zhu , rev .
mod .
phys . * 78 * , 373 ( 2006 ) .
k.j .
franke , g. schulze , and j.i .
pascual , science * 332 * , 940 ( 2011 ) .
r.s .
deacon , y. tanaka , a. oiwa , r. sakano , k. yoshida , k. shibata , k. hirakawa , and s. tarucha , phys .
rev .
lett . * 104 * , 076805 ( 2010 ) .
e.j.h .
lee , x. jiang , m. houzet , r. aguado , c.m .
lieber and s. de franceschi , nat .
nano .
* 9 * , 79 ( 2014 ) .
f. pientka , g. kells , a. romito , p.w .
brouwer , f. von oppen , phys .
rev .
lett . * 109 * , 227006 ( 2012 ) .
a. yazdani , b.a .
jones , c.p .
lutz , m.f .
crommie , and d.m .
eigler , science * 275 * , 1767 ( 1997 ) .
a. yazdani , c.m .
howald , c.p .
lutz , a. kapitulnik , and d.m .
eigler , phys .
rev .
lett . * 83 * , 176 ( 1999 ) .
e.w .
hudson , k.m .
lang , v. madhavan , s.h .
pan , h. eisaki , s. uchida , and j.c .
davis , nature ( london ) * 411 * , 920 ( 2001 ) .
s .- h .
ji , t. zhang , y .- s .
fu , x. chen , x .- c .
ma , j. li , w .- h .
duan , j .- f .
jia , and q .- k .
xue , phys .
rev .
lett . * 100 * , 226801 ( 2008 ) .
m. ruby , f. pientka , y. peng , f. von oppen , b.w .
heinrich , k.j .
franke , phys .
rev .
lett . *
115 * , 087001 ( 2015 ) d. badiane , m. houzet , j.s .
meyer , phys .
rev .
lett . * 107 * 177002 ( 2011 ) p. san - jose , j. cayao , e. prada , r. aguado , new .
j. phys .
* 15 * , 075019 ( 2013 ) supplementary material j.c .
cuevas , a. martn rodero , and a. levy yeyati , phys .
rev .
b , * 54 * , 7366 ( 1996 ) .
i. martin and d. mozyrsky , phys .
rev .
b * 90 * , 100508 ( 2014 ) .
a. levy yeyati , j.c .
cuevas , a. lpez - dvalos , and a. martn - rodero , phys .
rev .
b * 55 * , r6137(r ) ( 1997 ) the conductance of andreev states in @xmath54-wave or spin - orbit coupled superconductors is discussed in @xcite .
the importance of quasiparticle poisoning depends strongly on temperature .
recent experiments on proximity - coupled quantum wires [ a.p .
higginbotham _ et al .
_ , nat
. phys . , doi:10.1038/nphys3461 ( 2015 ) ] find time scales as long as 10ms .
m.t .
deng , c.l .
yu , g.y . huang , m. larsson , p. caroff , and h.q .
xu , nano lett .
* 12 * , 6414 ( 2012 ) .
a.d.k .
finck , d.j .
van harlingen , p.k .
mohseni , k. jung , and x. li , phys .
rev .
lett .
* 110 * , 126406 ( 2013 ) .
|
the software projects for high energy physics experiments are international with often several hundreds programmers from institutions world - wide involved .
the coordination of distributed software development is an important factor of success .
the automated nightly builds become a major component in collaborative software organization and management . in multi - person , multi - platform environment they provide a fast feedback to developers on new code submissions and facilitate collective work on different or same parts of software .
nicos @xcite was originally created for the atlas experiment @xcite and evolved in a versatile nightly builds system .
it operates on unix - like platforms and works with known release management tools , such as scram @xcite and cmt @xcite .
the nicos nightly builds become a media in which advanced programmers perform code development and make this development more effective and high - quality .
currently the nicos tool is also used in the lhc computing grid project @xcite .
the task of nicos is to provide a nightly build system that * works with known software release tools , * provides options for version management and number of releases in a cycle , * performs testing of the newly built software , * informs programmers about results with dynamic web pages and personal e - mails . the design strategy is based on a goal to make the system flexible , portable , stable , and easy to use . flexibility is attained by organizing nicos in a modular way so that each module is responsible for a certain step and independently described in the nicos configuration file . the modular structure allows to create the nightly builds framework that allows to plug in build , test , validation and other external tools . for the task of software management nicos
relies on external release tools @xcite .
nicos is based on perl scripts that makes possible porting to unix and windows systems .
nicos creates web pages that can be viewed with major web browsers , such as netscape , internet explorer , and opera .
the nicos system include the peer process that controls the execution of modules .
it detects problems and makes an attempt to rerun failed modules .
this feature allows to compensate for temporary technical problems and achieve maximum stability of the nightly builds .
nicos is easy to use for both administrators and software developers .
the nicos project configuration is stored in one unique text file named nicos_cache .
the versions of packages for the nightly builds can be specified in the nicos database files or supplied by an external tool .
a fast feedback to developers is one of the most important goals of nightly builds .
nicos automatically posts the information about the progress of nightly builds , identifies problems , and creates the web pages with build results .
the authors of failed packages can receive automatic e - mail notifications if desired .
nicos organization is shown in figure [ organiz ]
. the main process , nicos controller , can be run at scheduled times by system tools , such as cron .
it runs the nicos job modules and directs the status information to the nicos project web page . if desired , the previous nightly release is preserved and new release is built in a separate area .
the number of releases in a cycle is determined by the nicos administrator .
nicos supports multi - platform builds ( starting version 0.2 ) sharing code sources .
the main build process checks out code and triggers the start of builds on other platforms . in case of failure
the nicos controller tries to make a restart from the point of failure . if the nicos job runs overtime ( typically more than 24 hours ) it is automatically destroyed .
there are eight major modules in the nicos job .
they are able to serve as the interfaces to external tools , such as cvs , test , and release management tools .
* nicos configuration . at this step
the project parameters are read from the nicos_cache file .
* release tool setup .
the release tool ( such as scram , cmt ) is setup with commands specified in nicos_cache . * code checkout from the cvs repository specified in nicos_cache .
the packages tags are retrieved from the nicos database file or supplied by an external tool . * project setup . at this step
the project setup commands can be specified .
* project build .
the release tool should insert separators between the packages builds outputs with the packages names .
the unique string from the separators should be specified for this step .
after the build nicos cuts out and writes the logfiles with the build outputs for individual packages to the log area of the project . * unit tests and integrated tests .
the test tools , such as oval @xcite or cppunit @xcite are plugged in , as desired .
the outputs are forwarded to the testlog area of the project .
* error analysis . in the build output nicos searches for critical patterns indicating make problems , such as no rule to make target or symbol referencing errors. for tests , success is determined by the value returned by test tools . *
creation of the web pages with the summary of build results and e - mail notification about problems .
the progress and results of nicos builds are reflected on the dynamically created web pages . its structure is shown in figure [ web ] .
the nicos project web page shows the major parameters , such as project area , compiler options , and list of releases available for supported platforms . for building releases
, the currently performed step is displayed .
the nicos project web page contains links to the pages with detailed results for individual releases and pages showing the nicos database and configuration parameters .
optionally the collection of php scripts , `` nicos webmaster '' , can be used for handling the administration of nicos over the www ( available in 0.2 version of nicos ) .
this tool can modify the project configuration ( `` project configurator '' ) and nicos database of package versions ( `` version administrator '' ) .
in addition it is possible to schedule or stop builds from the job controller page .
the configuration parameters for the nicos job is collected in the xml - like nicos_cache files .
the steps of building process is associated with its markup tag .
every tag consist of a tag name , sometimes followed by an optional list of tag attributes that are placed between brackets ( @xmath0 and @xmath1 ) .
the markup tag is followed by the commands , including environment definitions , needed to be executed at the step .
the example of a markup tag below is for the step of build .
the attribute defines the directory name ( relative to the project area ) where the build should be performed .
the tag is followed by command scram b that builds a scram based project . ....
< project build dir = src > scram b .... nicos checks out versions of packages as specified in the nicos database file .
each line of this text file contains the name of package , tag of package , and e - mail addresses of the developers separated by spaces .
the exact tag of package or option for tag calculation can be indicated .
the options include the selection of recent submission , recent tagged submission , recent version in the `` official '' format : -[0 - 9][0 - 9]-[0 - 9][0 - 9]-[0 - 9][0 - 9 ] .
the nicos database file can be indicated in nicos_cache or dynamically created by a plug - in script .
the latter option is useful for large collaborations where special version management tools are used @xcite .
the 0.1 version of nicos is available since march 2003 and provides basic functionalities .
new 0.2 version ( scheduled for september 2003 ) will contain important improvements ( e.g. multi - platform support ) and additions such as the php script collection for administration over the www .
the long term plans center around improving simplicity of use and portability ( including the support of nicos for windows os ) .
nicos ( nightly control system ) is currently successfully used by several software projects at cern ( e.g. atlas , lcg ) .
the nicos features such as the instantaneous error analysis , dynamic informative web pages , and e - mail notifications about problems are proved to be useful for collaborative software organization and management . as experience gains
, nicos undergoes continuous improvements and evolutions .
|
nicos , nightly control system , is a flexible tool for coordination of software development in large - scale projects .
it manages the multi - platform nightly builds based on the recent versions of software packages , tries to compensate for technical failures , tests the newly built software , identifies possible problems , and makes results immediately available to developers spread over different institutions and countries .
the nicos nightly build services ensure that new software submissions are consistent and provide expected results .
the nicos tool was developed to coordinate the efforts of more than 100 developers from 34 countries for the atlas project at cern and can be easily adapted for other large software projects .
|
recently , @xcite reported the discovery of g2 , an extremely red object with spatially resolved br-@xmath1 emission .
the object was interpreted as a @xmath9 earth mass gas cloud based upon an inferred low dust temperature , observed elongation of the br-@xmath1 emission along the object s direction of motion , and a claimed tail along the same orbital trajectory as g2 .
this interpretation is particularly interesting because g2 is on a highly eccentric orbit with closest approach to our galaxy s central black hole within the next year , potentially allowing us to observe an unprecedented accretion event onto a supermassive black hole and offer insight into the region surrounding the black hole ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) . the interpretation of g2 as a gas cloud , however , is not definitive . one challenge for the pure gas cloud scenario is that , given the strong tidal fields in this region and g2 s low self - gravity , g2 needs to have formed quite recently ( @xmath101995 , just prior to the initial observations ; @xcite ) .
since such a gas cloud would be tidally disrupted during its periapse passage in the upcoming year , the gas cloud model implies g2 will be observed over almost the exact extent of its existence .
therefore several alternative scenarios invoking an underlying star have been proposed @xcite . in these scenarios ,
the observed heated gas is posited to be circumstellar , either intrinsic or a consequence of interaction of the star and surrounding ambient gas . in these stellar scenarios ,
g2 not only has existed for timescales much longer than the observed time baseline , but will also survive periapse passage . regardless of its nature , g2 s properties and possible origin depend critically on its orbital parameters .
these parameters have been estimated from observations with a short time baseline compared to the orbital period ( @xmath1010 vs.@xmath10200 years ; * ? ? ?
* ) and in a very crowded region , making the orbital solution susceptible to biases @xcite .
we therefore present new measurements and analysis of g2 that minimize the effects of source confusion on estimates of g2 s orbital parameters and examine the temporal evolution of g2 s properties .
two types of new data were collected for this study using the laser - guide - star adaptive optics ( lgs ao ) systems at the w.m .
keck observatory @xcite .
spectra were obtained using the osiris integral field spectrograph @xcite through the narrow - band kn3 filter , which is centered on the br-@xmath1 hydrogen line ( @xmath11 ) , at a spectral resolution of @xmath0 .
imaging data were obtained in the @xmath2-band filter ( @xmath12 ) and the @xmath3-band filter ( @xmath13 ) using the keck ii near - infrared camera ( nirc2 , pi : k. matthews ) .
these data were obtained and reduced in a similar manner as in our previous publications @xcite and the specific spectroscopic and @xmath3 observations utilized in this paper are described in table [ tab : obs ] .
lccccccccccc + 2006 jun 18,30 ; jul 1 & 2006.495 & keck ii lgs & 35 & 28 & 27 & 74 & 3 & @xmath14 & @xmath15 & @xmath16 & @xmath17 + 2008 may 16 ; jul 25 & 2008.487 & keck ii lgs & 35 & 22 & 21 & 78 & 4,0 & & & @xmath18 & @xmath19 + 2009 may 5,6 & 2009.344 & keck ii lgs & 35 & 24 & 19 & 79 & 0 & @xmath20 & @xmath21 & @xmath22 & @xmath23 + 2010 may 5,8 & 2010.349 & keck ii lgs & 35 & 17 & 16 & 82 & 0 & @xmath24 & @xmath25 & @xmath26 & @xmath27 + 2012 jun 9,11 ; aug 11,12 & 2012.613 & keck
i lgs & 20 & 27 & 21 & 68 & 0 & @xmath28 & @xmath29 & @xmath30 & @xmath31 + & & & & & & & & + + 2003 jun 10 & 2003.440 & keck ii ngs & 10 & 12 & 12 & 86 & 0 & & & @xmath32 & @xmath33 + 2004 jul 26 & 2004.567 & keck ii lgs & 10 & 11 & 11 & 80 & 1 & & & @xmath34 & @xmath35 + 2005 jul 30 & 2005.580 & keck ii lgs & 10 & 62 & 56 & 87 & 2 & & & @xmath36 & @xmath37 + 2006 may 21 & 2006.385 & keck ii lgs & 10 & 19 & 19 & 81 & 0 & & & @xmath38 & @xmath39 + 2009 jul 22 & 2009.561 & keck ii lgs & 10 & 4 & 4 & 85 & 0 & & & @xmath40 & @xmath41 + 2012 jul 20 - jul 23 & 2012.562 & keck ii lgs & 10 & 1316 & 1314 & 93 & 0 & & & @xmath42 & @xmath43 +
because g2 has no detectable @xmath44 continuum and is fainter than any object we have previously extracted , some analysis steps differ from the spectral extraction presented in @xcite and @xcite . for all epochs
, we created a combined data cube before extracting g2 s spectrum rather than extracting spectra from individual cubes .
the oh sky lines in the data are subtracted using sky frames scaled to the strength of families of oh lines in the observed frames to account for temporal variations . in 2012 ,
the br-@xmath1 emission line from g2 is coincident with the 2.180 oh line , so to minimize the systematic effects associated with oh line subtraction , we scale the sky only to this line .
an iterative process was required to estimate the position and spectral properties of g2 in the osiris data cubes ( see figures [ fig : g2 ] and [ fig : images ] ) since g2 s position is needed to place the aperture for spectral extraction , and g2 s spectral properties are needed to determine which osiris channels should be used to measure its position .
g2 s position was first obtained by visual inspection of the data cube .
then , an initial spectrum was extracted at this position using an aperture with a radius of 35 mas .
emission from local ambient gas was subtracted using a region free of stellar halos within @xmath45 .
a gaussian was fit to the resulting emission line . in order to refine the position , the three - dimensional data cube was median collapsed over the wavelength range corresponding to twice the standard deviation of the gaussian fit to the emission line , centered on the line peak .
continuum emission was subtracted by averaging the median of the 25 nearest channels on either side of the line .
g2 s position was further refined with a two - dimensional gaussian fit to the continuum - subtracted image .
the iterative extraction process was repeated again to obtain a final position and spectrum for g2 .
further iterations produced no significant changes . measuring g2
s position on a continuum - subtracted frame ( see figure [ fig : brgamma ] ) removes the effect of source confusion , thereby avoiding astrometric biases .
g2 s spectral properties were obtained using the final gaussian fits to the br-@xmath1 emission line . the radial velocities ( rvs ) for g2
were calculated from the offset of the line from the rest wavelength ( @xmath46 ) and corrected to the local standard of rest .
the reported fwhm measurements were corrected for instrumental broadening ( fwhm=85@xmath47 ) .
line flux measurements were made by comparing the gaussian fit to the non - variable star s0 - 2 between 2.17 - 2.18 @xmath48 m .
statistical uncertainties were taken from the root mean square ( rms ) error of gaussian fits to three independent data subsets ; however , we note that systematic errors ( e.g. oh sky lines ) are not included .
the final br-@xmath1 line maps were used to evaluate g2 s physical size and absolute astrometric position .
the size of g2 s br-@xmath1 emission was estimated by comparing the fwhm of g2 to that of nearby stars . to track the motions of g2 in the plane of the sky , its osiris position was transformed into an absolute coordinate system . to construct an absolute reference frame
, we used the precise locations of well characterized stars in the field ( see figure [ fig : kn3 ] ) , as measured with _
the position of g2 from the two - dimensional gaussian fit was added to the list of positions for each epoch .
then , these positions were matched to those from lgs ao @xmath2 observations of the same region at a nearby epoch using a second - order polynomial transformation to account for translation , rotation , and pixel scale differences between the images .
once the positions in the osiris frames were transformed into a @xmath2 epoch , they were transformed into an absolute reference frame in which sgra*-radio is at rest at the origin , as originally described in @xcite and updated in yelda et al . _ in prep_. we investigated possible systematic astrometric effects in osiris by cross - checking the osiris astrometry of s0 - 2 to our standard nirc2 analysis .
the results are consistent , validating the use of osiris astrometry .
the @xmath3 position of g2 was estimated from calibrated near - infrared images in a similar fashion to that described above for the br-@xmath1 line maps .
one difference was that we deconvolved the individual images using the lucy - richardson algorithm @xcite in order to help isolate the point sources from the extended @xmath3 dust emission .
both the large - scale background and the psf were estimated using _
starfinder_. after beam - restoring the individual frames with a gaussian having fwhm of half the theoretical resolution limit ( 40 mas ) , the frames were averaged to create a final image for each epoch ( see figure [ fig : lp ] ) .
the results of this analysis do not vary if the image is restored using a wider gaussian beam . to locate stars used as astrometric reference sources ,
_ starfinder _ was run on the combined ( deconvolved ) frame using the beam - restoring gaussian as the reference psf .
g2 s position was determined from an elliptical two - dimensional gaussian fit .
the fitting of an elliptical gaussian allows for extended structure and is analogous to the method used for the br-@xmath1 detection .
the positions for g2 from this method are consistent with those produced via _
starfinder_. the size of g2 was measured by comparing to the psf of stars with a similar flux .
the observed @xmath3 positions were transformed into an absolute coordinate system as described for the br-@xmath1 detections .
one complication for measuring the @xmath3 position of g2 in the plane of the sky is that g2 was superimposed on a filament of dust emission during the early- to mid- 2000 s , which caused significant astrometric bias ( see figure [ fig : lp ] ) . to quantify this effect
, we performed a series of star - planting simulations using a 2012 may @xmath3 image , in which g2 is well off the filament . in each of the four simulations , an extra point source ( @xmath49 )
was planted at the expected g2 position at the time of earlier @xmath3 observations ( 2003.440 , 2004.567 , 2005.580 , and 2006.385 ) based upon the orbit fit to the br-@xmath1 measurements .
the background dust emission does not have noticeable motions in the plane of the sky , thus the astrometric bias at these locations in the 2012 image should be representative of the bias at earlier epochs .
the position of the artificial point source was extracted using _
starfinder_. to constrain the spectral energy distribution of the object , we measured the integrated flux of g2 in the @xmath3 frame ( e.g. * ? ? ? * ) .
in contrast with @xcite , but in agreement with @xcite , we detect no @xmath2 counterpart for g2 ; therefore , we performed star planting simulations to determine the @xmath2 upper limit .
we used the br-@xmath1 astrometry of g2 to predict the positions of g2 in the @xmath2 data .
then , an artificial point source was planted at g2 s predicted position in the may 2010 image , when g2 is most isolated ( see figure [ fig : kp ] ) .
using a modified version of the _ starfinder _ algorithm ( boehle et al . _ in prep _ ) , we searched for a point source within a 3 pixel box centered on the planted star , given the psf originally extracted from the may 2010 image .
we fit an orbit to g2 s astrometric positions and rv measurements .
@xmath3-astrometric and rv measurements reported by @xcite were also included ; however , we fit the br-@xmath1 and @xmath3 astrometry separately .
we assumed g2 follows a purely keplerian orbit around the central black hole .
the orbit inference problem aims to find the most likely orbital parameters ( eccentricity , period , time of periapse passage , inclination , position angle of the ascending node , and the longitude of periapse ) given the data .
in addition to g2 s orbital parameters , the parameters of the gravitational potential need to be inferred ( mass and distance to black hole , and its perceived 2d position and 3d velocity ) .
since the observed motion of g2 does not entail enough information to constrain all these parameters , the orbit of s0 - 2 , which has undergone a full orbit , was fit simultaneously and effectively determines the gravitational potential ( see * ? ? ?
compared to earlier publications , we amended our keplerian orbit fitting code to use the bayesian sampler multinest @xcite .
we verified the results are equivalent with our previous monte carlo approach ( e.g. * ? ? ?
six key results emerge from our analysis of g2 : * _ apparent positional offset between the br-@xmath1 line emission and @xmath3 continuum emission of g2 : _ the @xmath3 astrometry is not consistent with our br-@xmath1 astrometry , which is reported here for the first time ( see figure [ fig : orb ] ) .
the positional difference can be explained by astrometric bias in the @xmath3 astrometry . in our star planting analysis
, we find the extracted position of an artificial @xmath3 source on the dust filament where g2 is located at early epochs differs from the input positions by an average of @xmath101.2 pix ( 12 mas ) .
this is almost an order of magnitude larger than typical centroiding uncertainties at @xmath3 and the astrometric bias found in a region isolated from extended background flux .
we therefore exclude @xmath3 astrometry from our primary orbital fits . *
_ revised orbital parameters : _ the keplerian orbit derived using br-@xmath1 astrometry differs from the orbit derived using @xmath3 astrometry ( e.g. * ? ?
* ) by more than can be explained by formal measurement uncertainties ( see figure [ fig : orb ] ) .
we list the most likely orbital parameters from br-@xmath1 in table [ tab : orb ] .
this new orbit for g2 has pushed the closest approach date to march 2014 , has a closer periastron ( @xmath50 , @xmath6 ) and a higher eccentricity ( @xmath51 ) . * _
g2 s compact size : _
g2 s observed size is comparable to the keck angular resolution in almost all our observations ; however , it exhibits marginal spatial extent in our highest quality spectroscopic data set . at this time ( 2006 ) , the observed half - width at half - maximum is @xmath52 after accounting for the observed psf of s0 - 2 ( @xmath53 ) .
g2 also shows a velocity gradient along its orbit , with the most highly redshifted portion of the object closest to the black hole , consistent with what is expected for a tidal interaction event .
the observed spatial extent is consistent with that presented in @xcite , thus the br-@xmath1 spatial extent of g2 along the direction of motion is relatively constant and corresponds to a size on the order of @xmath10100 au . * _ spectral evolution : _ we find an increasing br-@xmath1 line width for g2 , with fwhm measurements ranging from @xmath54 in 2006 to @xmath55 in 2012 ( see table [ tab : obs ] ) .
these measurements are consistent with those presented by @xcite . * _ constant brightness : _ both at @xmath3 and br-@xmath1 the flux of g2 is constant within uncertainties .
the integrated flux of g2 at @xmath3 has an average value of @xmath56 .
the br-@xmath1 line fluxes in 2006 , 2009 , 2010 , and 2012 are @xmath57 , @xmath58 , @xmath59 , @xmath60 of s0 - 2 s continuum flux between 2.17 - 2.18 , respectively .
* _ source color : _ our star planting simulations indicate that the upper @xmath2 magnitude limit of g2 is @xmath61 and is @xmath62 magnitudes deeper than the earlier upper @xmath2 limit imposed on g2 @xcite .
the deeper @xmath2 limit yields a dust temperature below @xmath10500k if g2 is a pure gas cloud . if g2 has an underlying stellar source , a @xmath2 magnitude limit places distinct limits on the luminosity of the star .
this limit is similar to the expected @xmath2 magnitude for a low - mass t tauri star in the galactic center ( e.g. * ? ? ? * ) , and eliminates a more massive star unless it is shrouded by dust which self - extincts the star at @xmath2 .
+ lc time of closest approach ( @xmath63 ) & @xmath64 + eccentricity ( @xmath65 ) & @xmath66 + period ( @xmath67 ) & @xmath68 + angle to periapse ( @xmath69 ) & @xmath70 + inclination ( @xmath71 ) & @xmath72 + position angle of the ascending node ( @xmath73 ) & @xmath74 +
the two orbits derived by @xcite differ from each other by more than can be explained with formal measurement uncertainties .
systematic errors in the l astrometry naturally explain the discrepancy . because the offset becomes less significant and the rv measurements play an increasingly dominant role in the orbital modeling , the measurements seem to converge on the solution obtained with the less biased br-@xmath1 astrometry and rv measurements .
there are several implications of the new orbital model parameters .
first , the revised periapse date of march 2014 and the closer periastron are important for the design and interpretation of sgra * monitoring programs designed to test for an increased accretion flow . with a special nrao call for proposals and tremendous attention called to this object , which was originally anticipated to experience periapse in june 2013 ,
more than 30 programs have been approved for the summer of 2013 covering radio to @xmath1-ray wavelengths . while some models predict that a cloud would generate enhanced radio emission from sgra * well in advance of periapse passage ( e.g. * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) , none has yet been detected @xcite , although this may not yet be expected with the revised periapse passage .
second , the geometry of the new orbit may call two previously suggested associations into question . most importantly , the previous claimed tail " of low surface brightness emission , which was central to the original claim that g2 is a pure gas cloud , is no longer securely affiliated with the compact head " , which has been analyzed in this work .
the new orbit also falls 3.5@xmath75 off the plane of the disk of young stars orbiting the central black hole at @xmath100.05 - 2 pc ( @xcite ; yelda et al . _ in prep _ ) which has been suggested as the possible origin of the gas cloud through colliding stellar winds of young stars @xcite or a central stellar source ( e.g. * ? ? ?
third , the even higher eccentricity provides strong constraints on g2 s origin .
it is notable that a high eccentricity is consistent with the outcome of three body exchanges between a binary system and the central black hole which may explain the dense concentration of b stars ( the s - stars ) in the innermost regions of the galaxy ( e.g. * ? ? ?
if a triple star system is invoked rather than a binary , g2 could possibly be the result of a recent merger between two components to produce the observed physical properties ( red , compact , and at times marginally resolved ) and an exchange with the central black hole with the remaining component to produce the observed orbital parameters ( highly eccentric , short period ) .
while we favor a stellar model , this question will be soon addressed , as g2 should remain intact through its periapse if this is indeed correct .
the authors would like to acknowledge the invaluable feedback from conference attendees at workshops where this work has been presented ( aas 221st meeting , keck 20th anniversary celebration , galactic nuclei ringberg workshop ) .
support for this work was provided by nsf grant ast-0909218 .
data presented herein were taken at the w. m. keck observatory .
the w. m. keck observatory is operated as a scientific partnership among the california institute of technology , the university of california , and the national aeronautics and space administration .
the observatory was made possible by the generous financial support of the w. m. keck foundation . , h. , martins , f. , fritz , t. k. , genzel , r. , levin , y. , perets , h. b. , paumard , t. , nayakshin , s. , gerhard , o. , alexander , t. , dodds - eden , k. , eisenhauer , f. , gillessen , s. , mascetti , l. , ott , t. , perrin , g. , pfuhl , o. , reid , m. j. , rouan , d. , sternberg , a. , & trippe , s. 2009 , , 697 , 1741 , j. c. y. , wizinowich , p. , campbell , r. , chock , l. , cooper , a. , james , e. , lyke , j. , mastromarino , j. , martin , o. , medeiros , d. , morrison , d. , neyman , c. , panteleev , s. , stalcup , t. , tucker , p. , wetherell , e. , & van dam , m. 2012 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol .
8447 , society of photo - optical instrumentation engineers ( spie ) conference series , a. , mui , k. , yazici , s. , sabha , n. , shahzamanian , b. , witzel , g. , moser , l. , garcia - marin , m. , valencia - s .
, m. , jalali , b. , bremer , m. , straubmeier , c. , rauch , c. , buchholz , r. , kunneriath , d. , & moultaka , j. 2013 , , 551 , a18 , a. m. , hornstein , s. d. , lu , j. r. , bouchez , a. , le mignant , d. , van dam , m. a. , wizinowich , p. , matthews , k. , morris , m. , becklin , e. e. , campbell , r. d. , chin , j. c. y. , hartman , s. k. , johansson , e. m. , lafon , r. e. , stomski , p. j. , & summers , d. m. 2005 , , 635 , 1087 , j. , barczys , m. , krabbe , a. , adkins , s. , aliado , t. , amico , p. , brims , g. , campbell , r. , canfield , j. , gasaway , t. , honey , a. , iserlohe , c. , johnson , c. , kress , e. , lafreniere , d. , magnone , k. , magnone , n. , mcelwain , m. , moon , j. , quirrenbach , a. , skulason , g. , song , i. , spencer , m. , weiss , j. , & wright , s. 2006 , , 50 , 362 , t. , genzel , r. , martins , f. , nayakshin , s. , beloborodov , a. m. , levin , y. , trippe , s. , eisenhauer , f. , ott , t. , gillessen , s. , abuter , r. , cuadra , j. , alexander , t. , & sternberg , a. 2006 , , 643 , 1011 , g. , eckart , a. , bremer , m. , zamaninasab , m. , shahzamanian , b. , valencia - s . ,
m. , schdel , r. , karas , v. , lenzen , r. , marchili , n. , sabha , n. , garcia - marin , m. , buchholz , r. m. , kunneriath , d. , & straubmeier , c. 2012 , , 203 , 18 , p. l. , le mignant , d. , bouchez , a. h. , campbell , r. d. , chin , j. c. y. , contos , a. r. , van dam , m. a. , hartman , s. k. , johansson , e. m. , lafon , r. e. , lewis , h. , stomski , p. j. , summers , d. m. , brown , c. g. , danforth , p. m. , max , c. e. , & pennington , d. m. 2006 , , 118 , 297
|
we present new observations and analysis of g2 the intriguing red emission - line object which is quickly approaching the galaxy s central black hole .
the observations were obtained with the laser guide star adaptive optics systems on the w. m. keck i and ii telescopes and include spectroscopy ( @xmath0 ) centered on the hydrogen br-@xmath1 line as well as @xmath2 ( 2.1 ) and @xmath3 ( 3.8 ) imaging .
analysis of these observations shows the br-@xmath1 line emission has a positional offset from the @xmath3 continuum .
this offset is likely due to background source confusion at @xmath3 .
we therefore present the first orbital solution derived from br-@xmath1 line astrometry , which when coupled with radial velocity measurements , results in a later time of closest approach ( @xmath4 ) , closer periastron ( @xmath5 , @xmath6 ) , and higher eccentricity ( @xmath7 ) compared to a solution using @xmath3 astrometry .
the new orbit casts doubt on previous associations of g2 and a low surface brightness `` tail . ''
it is shown that g2 has no @xmath2 counterpart down to @xmath8 .
g2 s @xmath3 continuum and the br-@xmath1 line - emission is unresolved in almost all epochs ; however it is marginally extended in our highest quality br-@xmath1 data set from 2006 and exhibits a clear velocity gradient at that time . while the observations altogether suggest that g2 has a gaseous component which is tidally interacting with the central black hole , there is likely a central star providing the self - gravity necessary to sustain the compact nature of this object .
|
microlensing describes gravitational lensing of a source whose multiple images are not resolved .
two fundamental microlensing observables are the total magnification ( photometry ) and image centroid shift ( astrometry ) of images of a lensed source .
these observables have important astrophysical applications such as determining the mass and distance to the lens , angular radius of the source , etc .
( see , e.g. , @xcite , @xcite , @xcite , @xcite , @xcite , and references therein ) . a natural issue to explore
is how are the photometry and astrometry of a source being lensed by a point mass changed when the point mass is replaced by a black hole lens .
this could have important implications for the testability of general relativity s predictions about how the gravitational field of a black hole affects light rays .
indeed , the standard theoretical framework for point mass microlensing is based on relativistic calculations to first - order in @xmath4 about a schwarzschild black hole , where @xmath7 where @xmath8 is the black hole s mass and @xmath9 the gravitational radius .
@xcite found that to second - order in @xmath4 , the relativistic corrections appear in the position and magnification of images due to a point mass lens , while no such correction appears in the total magnification .
@xcite also showed that no relativistic correction to second - order in @xmath4 occurs for the associated image centroid shift . in this paper , we extend the work of the previous authors by determining an analytical expression under assumptions * a1**a3 * for the full schwarzschild black hole relativistic correction of the image centroid shift , which includes the total magnification , for point and extended sources with arbitrary surface brightness .
the full relativistic corrections will then be applied to the case of the massive black hole at the center of our galaxy .
microlensing by the galactic black hole has been studied by several authors in the weak - field limit of the black hole ( e.g , @xcite , @xcite , @xcite , @xcite ) .
in addition , the precession of star orbits in the strong - field regime of the black hole was considered by , e.g. , @xcite , @xcite .
@xcite gave a numerical treatment of the magnifications of several relativistic images for a point source being lensed by a schwarzschild black hole lens at the galactic center .
they considered sources within our galaxy that are far away from the black hole ( about @xmath10 ) , while we shall consider sources as close as @xmath5 to the black hole . analytical work on magnification due to a schwarzschild lens
was also done by several authors for point and/or extended sources with uniform brightness profiles ( e.g. , @xcite , @xcite , @xcite , @xcite ) . as noted above
, our microlensing treatment will apply not only to the magnification , but the image centroid of extended sources with arbitrary surface brightness .
we shall also show that a single factor approximates the relativistic corrections to the weak - field total magnification and image centroid due to a schwarzchild black hole lens at the galactic center .
the same factor applies to either a point or extended source .
estimates of this factor will be given for lens - sources distances ranging from @xmath5 to @xmath11 . in principle
, the magnification and image centroid of sources closer to the galactic black hole should provide stronger relativistic microlensing signatures .
we shall show that even for sources about @xmath5 behind the black hole , the full relativistic correction is still negligible .
section [ sec - image - mag ] reviews some basic results about lensing by a schwarzschild black hole . in section [ sec - image - centroid ] , we compute explicitly the image centroid due to a schwarschild black hole acting on point and extended sources with arbitrary brightness profiles .
our image centroid formula expresses the relativistic image centroid in terms of the weak - field image centroid due to a point mass lens .
section [ sec - applications ] estimates the relativistic corrections to microlensing by the galactic black hole .
we review some basic analytical results about the image positions and magnification due to a schwarzschild black hole lens .
a schwarzschild black hole of mass @xmath12 is the unique static , spherically symmetric , vacuum ( i.e. , ricci flat ) spacetime that is minkowski at infinity ( e.g. , @xcite , p. 119 ) : @xmath13 where @xmath14 is the schwarzschild radius @xmath15 and @xmath16 the standard metric on a @xmath17-sphere of radius 1 .
relative to the optical axis through the observer and center of the black hole , let @xmath18 denote the angular position of a point source behind the black hole .
note that a source between the observer and the black hole is also lensed ; in principle , even the observer is lensed .
such situations will not be treated in this paper .
we shall always assume the standard lensing configuration of a source behind the black hole the capture cross section @xmath19 of a light ray in a schwarzschild spacetime is @xmath20 where @xmath21 is the critical apparent impact parameter ( e.g. , @xcite , p.144 ) .
consequently , any light ray that makes it from the source to the observer will have an apparent impact parameter @xmath22 greater than @xmath23 .
the classification of null geodesics in the schwarzschild geometry divides such source - to - observer light rays into those that loop around the black hole at least once before arriving at the observer and those that travel directly to the observer without looping ( cf .
@xcite , p. 674 ) .
let @xmath24 , @xmath25 , and @xmath26 be the distances from lens to source , observer to lens , and observer to source , resp .
denote the angular position of a lensed image relative to the line of sight by @xmath27 . a negative angle @xmath28 will correspond to an angular position on the side of the black hole opposite to the angular position @xmath27 .
a trigonometric argument yields the lens equation , which relates @xmath18 and @xmath27 as follows ( @xcite ) : [ eq - lebh ] = - ( + ( - ) ) , where @xmath29 is the bending angle of the lensed ray .
lensed images _ of a light source at @xmath18 are the solutions @xmath27 of the lens equation .
the _ magnification _
@xmath30 of a lensed image @xmath27 is the ratio of the solid angle subtended at the distance to the source by the lensed image to the solid angle of the unlensed source . explicitly , [ eq - mag ] ( ) = | ( )
the lens equation can not be solved analytically in closed form , so approximations appropriate to the lensing scenario of interest to us will be employed .
our working assumptions are : * the source is assumed to be close to the line of sight , i.e. , @xmath31 , and lie in the asymptotically flat region outside the black hole , i.e. , @xmath32 . * in anticipation of later applications to the galactic black hole , we suppose that @xmath33 and @xmath34 , where @xmath22 is the apparent impact parameter of _ any _ light ray from the source to the observer . * for any light ray from the source to the observer that do not wind around the black hole , we shall assume that its distance of closest approach @xmath3 lies in the weak - field regime regime , i.e. , @xmath35 .
we shall now use * a1 * - * a3 * to simplify the tangent terms in the lens equation .
relative to the observer , the caustic due to the black hole coincides with the optical axis passing the observer and singularity at @xmath36 . by * a1 *
, the source is near to the optical axis ( i.e. , @xmath31 ) , so the source is highly magnified and its angular position satisfies [ eq - tan - approx - a1 ] . in * a2 * , it is important to add that the apparent impact parameter @xmath22 and distance @xmath3 of closest approach are not equal in general and approximate each other in the weak - field limit .
in fact , for our lensing assumptions , the quantities @xmath22 and @xmath3 are related as follows ( e.g. , @xcite , p. 144 ) @xmath37 consequently , we have @xmath38 and @xmath39 for the nearly flat regime @xmath35 .
assumption * a2 * yields that @xmath40 , so [ eq - tan - approx - a2 ] . before discussing how our assumptions affect the term @xmath41 in ( [ eq - lebh ] ) , we consider the bending angle @xmath29 under assumptions * a1 * and * a2*. the bending angle , which is not a priori assumed to be small , is given by ( e.g. , @xcite , p. 189 ; @xcite , p. 145 ) : [ eq - bh - bending - angle ] ( x_0 ) = 2 _ x_0^ - , where @xmath42 , \qquad x_0 = r_0/\aprbh.\ ] ] the integral in ( [ eq - bh - bending - angle ] ) is an elliptic integral and can be expressed in terms of an elliptic integral of the first kind . in fact , set @xmath43 with @xmath44 the zeros of the quartic @xmath45 are & & r_1 = x_0 , r_2 = , + & & r_2 = , r_4 = 0 . note that @xmath46 .
it can be shown ( e.g. , @xcite , p. 47 ) that [ eq - intx ] _
x_0^ = f ( , k ) , where @xmath47 is an elliptic integral of the first kind with arguments @xmath48 using ( [ eq - intx ] ) , we plot the bending angle @xmath29 in figure [ fig-1 ] .
in addition , taylor expanding the integrand of in ( [ eq - bh - bending - angle ] ) yields the following expansion for the bending angle ( @xcite , @xcite ) : [ eq - expansion - bending ] ( x_0 ) = + ( -1 ) + ( ) .
( in radians ) as a function of @xmath49 , where @xmath3 is the distance of closest approach to the black hole . in the top graph , we see that @xmath29 diverges as @xmath50 .
the bottom graph shows that @xmath29 is significantly less than unity for @xmath51 . in the case of the galactic black hole
, the range of @xmath52 in the bottom graph corresponds to a distance of closest approach in the interval @xmath53 .
this yields a lens - source distance ranging from about @xmath54 to @xmath11 see figure [ fig2 ] .
, title="fig:",width=288,height=211 ] ( in radians ) as a function of @xmath49 , where @xmath3 is the distance of closest approach to the black hole . in the top graph
, we see that @xmath29 diverges as @xmath50 .
the bottom graph shows that @xmath29 is significantly less than unity for @xmath51 . in the case of the galactic black hole ,
the range of @xmath52 in the bottom graph corresponds to a distance of closest approach in the interval @xmath53 .
this yields a lens - source distance ranging from about @xmath54 to @xmath11 see figure [ fig2 ] .
, title="fig:",width=288,height=211 ] for light rays that travel directly from the source to the observe without looping around the black hole , assumption * a3 * yields that @xmath55 . by ( [ eq - expansion - bending ] )
, the bending angle takes the standard weak - field form : @xmath56 because @xmath57 in this case and @xmath58 by ( [ eq - tan - approx - a2 ] ) , we obtain @xmath59 and , hence , [ eq - tan - approx - a3 ] ( - ) - . for the source - to - observer rays that do not loop ,
we have @xmath60 . combined with equations ( [ eq - tan - approx - a1 ] ) , ( [ eq - tan - approx - a2 ] ) , and ( [ eq - tan - approx - a3 ] ) we see that the lens equation reduces to the usual _ weak - field lens equation _ : [ eq - wfle ] = - ( ) = - , where [ eq - radii - er ] = . for rays that loop around the black hole at least once before arriving at the observer ,
the bending angle @xmath29 is near a multiple of @xmath61 ( cf .
figure [ fig-1 ] ) .
such light rays are close to the unstable photon orbit at radius @xmath62 ( e.g. , @xcite , p. 132 ) . in this case
, the bending angle has the form @xmath63 , where @xmath64 is the number of times the light ray loops around the black hole and @xmath65 ( representing a small angular deviation from a multiple of @xmath66 ) .
consequently , [ eq - tan - approx - sf ] ( - ) _ n - . equation ( [ eq - tan - approx - sf ] ) , along with * a1 * and * a2 * , yields the following _ strong - field lens equation _ for rays circling near @xmath62 ( @xcite ) : [ eq - sfle ] = - _
n. we now turn to the solutions of the weak- and strong - field lens equations .
the solutions of the weak - field lens equation ( [ eq - wfle ] ) are well known ( e.g. , see @xcite , pp .
187 - 191 , for a detailed treatment ) .
if @xmath67 ( source is on the optical axis ) , then the source is lensed into a formally infinitely magnified circle , called an _ einstein ring _ , which has an angular radius @xmath68 . if @xmath69 ( source is off the optical axis ) , then on opposite sides of the black hole there are a primary image @xmath70 and secondary image @xmath71 of the source : [ eq - imageswf ] _
the primary and secondary images have magnification [ eq - magwf ] _ = .
note that assumption * a1 * was used for the approximations in ( [ eq - imageswf ] ) and ( [ eq - magwf ] ) .
the total magnification is @xmath72 to simplify expressions like ( [ eq - imageswf])and ( [ eq - magwf ] ) that involve @xmath68 , we shall scale all angles as follows : @xmath73 the weak - field image positions and magnifications now become [ eq - images - mag - wf - s ] _ = 1 ,
_ = , = .
note that the angular image positions and magnifications in ( [ eq - images - mag - wf - s ] ) for the weak field regime coincide with that obtained from lensing by a point mass . if @xmath74 , then the source is lensed into the formally infinitely magnified einstein ring mentioned in section [ subsec - einsteinr - psi ] , along with an infinite sequence of relativistic einstein rings near @xmath62 .
the angular radii @xmath75 of the relativistic einstein rings are given by the solutions of the strong - field lens equation ( [ eq - sfle ] ) for @xmath67 ( @xcite ) : [ eq - radiin - er ] = ( a_0 e^-(2n + 1 ) + b_0 ) , where @xmath76 note that @xmath77 is approximately the angle spanned by the schwarzschild radius .
for simplicity , set @xmath78 and rewrite @xmath75 as @xmath79 since @xmath80 we have @xmath81 ( because @xmath82 ) .
consequently , each term in the strictly decreasing sequence @xmath83 obeys [ eq - dn-1 ] _ 1 < 1 , n=1,2, . if @xmath69 , then we know from section [ subsec - einsteinr - psi ] that on opposite sides of the black hole there are a primary image @xmath70 and secondary image @xmath71 from light rays that travel directly ( no looping around the black hole ) to the observer .
in addition , there are two infinite sequences ( on opposite sides of the black hole ) of relativistic images @xmath84 that are due to light rays looping @xmath85 times around the black hole near @xmath62 .
the following expressions for the solutions @xmath84 of the strong - field lens equation ( [ eq - sfle ] ) are convenient ( @xcite ) : [ eq - thrn ] ^_n = ^e_n + .
n = 1,2, .
the images @xmath86 ( resp . ,
@xmath87 ) result from @xmath85 clockwise ( resp .
, counter - clockwise ) windings around the black hole . using ( [ eq - mag ] ) , the magnifications of the relativistic images can be shown to be approximately as follows ( @xcite , @xcite ) : [ eq - magrn ] ^_n = . the total magnification is then @xmath88 equations ( [ eq - radiin - er ] ) and ( [ eq - magrn ] ) imply that [ eq - mag - r-1 ] = . by ( [ eq - dn-1 ] ) , both series in ( [ eq - mag - r-1 ] ) are geometric series .
hence , [ eq - mag - r-2 ] = , where [ eq - c0 ] c_0 = a_0 . equations
( [ eq - thrn ] ) , ( [ eq - magrn ] ) , and ( [ eq - mag - r-2 ] ) simplify to [ eq - thrn - s ] ^_n & = & ^e_n + u , ^e_n = , + [ eq - magrn - s ] ^_n & = & , + [ eq - mag - r - s ] & = & . by ( [ eq - mag - r - s ] ) , the relativistic and total magnifications for a point source are then given as follows ( @xcite , @xcite ) : [ eq - magbht ] = , = + _ r = ( 1 + ) , where [ eq - cr ] = = 4 c_0 ^3 , = .
equation ( [ eq - magbht ] ) shows that @xmath89 determines the relativistic correction to the total weak - field magnification .
we also refer interested readers to the elegant generalization by @xcite of equations ( [ eq - thrn - s])([eq - mag - r - s ] ) to the case of a reissner - nordstrm black hole lens .
in this section , we derive a formula using * a1**a3 * that expresses the image centroid of a schwarzschild black hole as a relativistic correction to the weak - field image centroid of a point mass lens .
the center - of - light or centroid of all the images of a source at @xmath90 is [ eq - cent - def ] = , where @xmath91 is the total magnification . since the centroid shift is @xmath92 it suffices to study @xmath93 .
we now express the image centroid as [ eq - cent ] = + , where @xmath94 and @xmath95 are respectively the centroids of the weak - field and relativistic images .
explicitly , @xmath96 and @xmath97 by ( [ eq - images - mag - wf - s ] ) and ( [ eq - mag - r - s ] ) , the weak - field image centroid is [ eq - cent - wf ] = .
recall that ( [ eq - cent - wf ] ) is the weak - field image centroid for source positions near the caustic .
indeed , the full shifted image centroid due to a point mass lens is an ellipse ( e.g. , @xcite,@xcite,@xcite ) . using ( [ eq - thrn - s ] ) and
( [ eq - magrn - s ] ) , we get @xmath98 now , observe that [ eq - thndn ] _
n=1^ ^e ( ) ^k = , k = 1,2 , , where [ eq - c0k ] c_0 ( k ) = a^k_0 with @xmath99 note that @xmath100 . equation ( [ eq - thndn ] ) yields @xmath101 equations ( [ eq - magbht ] ) , ( [ eq - cr ] ) , and ( [ eq - cent - wf ] ) imply [ eq - cent - r ] = , where [ eq - dr ] = 4 ^2 .
hence , the image centroid ( [ eq - cent ] ) simplifies to [ eq - cent - main ] = , where [ eq - ccent ] = .
it follows that the constants @xmath89 and @xmath102 determine the full relativistic correction to the image centroid due to a point mass lensing a source near the caustic .
the total magnification of an extended source @xmath103 with surface brightness @xmath104 is given by @xmath105 where @xmath106 with @xmath107 , the unit angular vector position of the source .
the spherical symmetry of the schwarzschild black hole implies that @xmath108 .
consequently , equation ( [ eq - magbht ] ) yields [ eq - magtot - ext ] ^ext = ( 1 + ) , where @xmath109 is the weak - field extended source total magnification .
readers are referred to @xcite for a discussion of @xmath110 for a point mass lens .
the image centroid of the extended source is @xmath111 equation ( [ eq - cent - main ] ) gives [ eq - cent - main - ext ] = , where @xmath112 is the weak - field extended image centroid see @xcite for a treatment of @xmath113 with a point mass lens .
thus , as in the point source case , the quantities @xmath89 and @xmath102 determine the relativistic corrections to the extended source image centroid , as well as the total magnification , for a point mass lens .
in this section , we estimate the relativistic corrections to the weak - field total magnification and image centroid for the case of the galactic black hole under * a1**a3*. we begin with estimates of some needed quantities .
the constants @xmath114 and @xmath115 are given approximately as follows : @xmath116 for the massive black hole at the galactic center , we have ( @xcite ) : @xmath117 the linear and angular schwarzschild radii of the black hole are [ eq - gbh - r - vth ] 5.23 10 ^ -2 au , = 6.46 as .
since the relativistic images are near @xmath118 , those on opposite sides of the black hole have angular spacings of order @xmath119 , which is outside the resolving capabilities of near - future instruments .
the expected resolution of the space interferometry mission to be launched in 2009 is about @xmath120 .
hence , the source is microlensed by the strong - field region near the black hole .
the ( weak - field ) angular einstein radius as a function of the lens - source distance @xmath121 is given by [ eq - the ] ( x ) = ( ) ^1/2 1.65 as .
the function @xmath122 has a positive derivative and so is strictly increasing see figure [ fig2 ] .
the angular diameter of the einstein ring is approximately the angular spacing between the primary and secondary images .
for @xmath123 the range of the angular spacing is @xmath124 which is at the boundary of the resolving capabilities of present day instruments .
we also have @xmath125 and by ( [ eq - dr ] ) we get [ eq - dr - est ] ( x ) 2 10 ^ -13 ( ) .
for these distances , the linear einstein ring radius @xmath126 approximates the distance @xmath3 of closest approach as well as the impact parameter @xmath22 ( since @xmath127 ) .
the value of @xmath126 is approximately @xmath128 for @xmath129 .
, title="fig:",width=288,height=211 ] for these distances , the linear einstein ring radius @xmath126 approximates the distance @xmath3 of closest approach as well as the impact parameter @xmath22 ( since @xmath127 ) .
the value of @xmath126 is approximately @xmath128 for @xmath129 .
, title="fig:",width=288,height=211 ] let us now verify that assumptions * a1**a3 * hold .
we are supposing that @xmath31 and @xmath130 , which by ( [ eq - gbh - r - vth ] ) yields @xmath131 so assumption * a1 * is satisfied .
now , the largest value @xmath132 of the impact parameter of light rays from the source to the observed occurs for the rays that travel to the observer without looping around the black hole .
these rays passing through the nearly flat region outside the black hole , which means that the associated impact parameter and distance of closest approach are approximately equal . equation ( [ eq - imageswf ] ) shows that the angular positions of the closest approach of the light rays are given approximately by @xmath133 ( since @xmath31 ) .
it follows that @xmath132 can be approximated by the linear einstein ring radius , @xmath134 ( figure [ fig2 ] ) .
the function @xmath135 is strictly increasing with a minimum value of approximately @xmath136 for @xmath137 see figure [ fig3 ] .
in addition , since @xmath138 is also strictly increasing with maximum @xmath139 for @xmath140 ( see ( [ eq - the ] ) ) , it follows that @xmath141 consequently , assumption * a2 * holds .
now , since @xmath142 is strictly decreasing with a maximum value of order @xmath143 for @xmath144 , the primary and secondary images lie in the nearly flat region outside the black hole . in particular , assumption * a3 * is also satisfied .
we add that since @xmath145 is at most of order @xmath146 for @xmath147 , a source at distance @xmath148 lies in a region of space that is `` flatter '' than the region near the linear einstein radius @xmath138 ( since @xmath149 ) .
furthermore , for a source star of mass @xmath150 and radius @xmath151 , the closest it can get to the galactic black hole without being thorn apart is given roughly by the tidal disruption radius of the source ( e.g. , @xcite ) : @xmath152 where @xmath153 is a parameter of order unity that depends on the model of the source .
for @xmath154 and @xmath155 , the tidal radius is then of order @xmath156 hence , a source at @xmath157 is significantly outside its tidal disruption radius . for @xmath130 , while the bottom illustrates that @xmath158 .
the latter implies that the primary and secondary images lie in the nearly flat region outside the galactic black hole . , title="fig:",width=288,height=211 ] for @xmath130 , while the bottom illustrates that @xmath158 .
the latter implies that the primary and secondary images lie in the nearly flat region outside the galactic black hole .
, title="fig:",width=288,height=211 ] for the galactic black hole , equation ( [ eq - cr ] ) shows that the relativistic correction to the weak - field total magnification is [ eq - cr - est ] ( x ) 2.1 10 ^ -18 ( ) ^3/2 .
since the derivative of @xmath159 is always negative , the correction @xmath159 is a strictly decreasing function of the lens - source distance @xmath148 see figure [ fig4 ] .
the figure shows that the relativistic correction @xmath89 to the magnification is at most of order @xmath160 for sources with @xmath161 .
turning to the relativistic image centroid , equations ( [ eq - dr - est ] ) and ( [ eq - cr - est ] ) yield that the derivative of @xmath162 is always negative .
consequently , the function @xmath163 is strictly decreasing with increasing lens - source distance figure [ fig4 ] .
furthermore , the figure yields that @xmath159 is at most of order @xmath6 and @xmath164 at most of order @xmath165 for @xmath157 , so the relativistic image centroid factor ( [ eq - ccent ] ) can be approximated as follows : [ eq - ccent - est ] ( x ) 1 - ( x ) x 10 pc .
it follows that the relativistic correction @xmath89 to the weak - field total magnification also serves as a relativistic correction to the weak - field image centroid .
the value @xmath166 corresponds to the weak - field case , which clearly holds approximately for @xmath157 . in light of figure [ fig4 ] , measurable relativistic corrections to the total magnification and
image centroid due to standard gravitational lensing by the galactic black hole would most likely require sources very close to the black hole . and
@xmath167 as a function of the distance @xmath24 of the source from the galactic black hole , where @xmath168 .
note that these distances are significantly outside the tidal disruption radius , @xmath169 , for a sun - like source .
the bottom figure guarantees that @xmath89 describes the relativistic correction to both the weak - field image centroid and total magnification ( see discussion in text ) .
the correction is clearly negligible for sources @xmath161 .
, title="fig:",width=288,height=211 ] and @xmath167 as a function of the distance @xmath24 of the source from the galactic black hole , where @xmath168 .
note that these distances are significantly outside the tidal disruption radius , @xmath169 , for a sun - like source .
the bottom figure guarantees that @xmath89 describes the relativistic correction to both the weak - field image centroid and total magnification ( see discussion in text ) .
the correction is clearly negligible for sources @xmath161 .
, title="fig:",width=288,height=211 ]
previous work on gravitational lensing by the black hole at the galactic center investigated the relativistic corrections to the weak - field total magnification and image centroid to second order in @xmath170 , where @xmath3 is the distance of closest approach of the light ray to the black hole .
it was shown recently that for a point mass lens the total magnification and image centroid shift of a point source remain unchanged by relativistic corrections of second order in @xmath4 .
we computed the relativistic corrections for a schwarzschild black hole lens under assumptions * a1**a3*. these corrections were applied to the case of the massive black hole at the galactic center .
we found that the weak - field magnification and image centroid have approximately the same relativistic correction .
this correction is a strictly decreasing function of the lens - source distance @xmath24 . for @xmath130 ,
the relativistic correction is of order at most @xmath171 a minuscule correction .
hence , for standard lensing configurations , a nontrivial relativistic correction to microlensing by the galactic black hole would likely have to come from sources deep inside the black hole s potential well .
many thanks to scott gaudi for stimulating discussions and the referee for a helpful clarifying comment . this work was supported in part by an alfred p. sloan research fellowship and nsf career grant dms-98 - 96274 .
|
the standard treatment of gravitational lensing by a point mass lens @xmath0 is based on a weak - field deflection angle @xmath1 , where @xmath2 with @xmath3 the distance of closest approach to the mass of a lensed light ray .
it was shown that for a point mass lens , the total magnification and image centroid shift of a point source remain unchanged by relativistic corrections of second order in @xmath4 .
this paper considers these issues analytically taking into account the relativistic images , under three assumptions * a1**a3 * , for a schwarzschild black hole lens with background point and extended sources having arbitrary surface brightness profiles .
the assumptions are * a1 : * the source is close to the line of sight and lies in the asymptotically flat region outside the black hole lens ; * a2 : * the observer - lens and lens - source distances are significantly greater than the impact parameters of the lensed light rays ; and * a3 : * the distance of closest approach of any light ray that does not wind around the black hole on its travel from the source to the observer , lies in the weak - field regime outside the black hole .
we apply our results to the galactic black hole for lensing scenarios where * a1**a3 * hold .
we show that a single factor characterizes the full relativistic correction to the weak - field image centroid and magnification . as the lens - source distance increases , the relativistic correction factor strictly decreases .
in particular , we find that for point and extended sources about @xmath5 behind the black hole , which is a distance significantly outside the tidal disruption radius of a sun - like source , the relativistic correction factor is minuscule , of order @xmath6 .
therefore , for standard lensing configurations , any detectable relativistic corrections to microlensing by the galactic black hole will most likely have to come from sources significantly closer to the black hole .
[ firstpage ] black holes : strong field regime gravitational lensing : photometry and astrometry .
|
our galaxy contains over two hundred known supernova remnants ( snrs ) , which are an important source of energy and heavy element release into the interstellar medium ( ism ) , and are also thought to be the sites of the acceleration of cosmic rays . over the last twenty years i have produced several versions of a catalogue of galactic snrs , the most recent revised in 2004 january ( see appendix [ s : appendix ] ) .
since the first version of the catalogue was published in green ( 1984 ) , the number of identified galactic snrs has increased considerably , from 145 to 231 , and here i review some of the statistical properties of galactic remnants based on the most recent version of the catalogue . in section [
s : catalogue ] the catalogue is described , while the selection effects applicable to the identification of galactic snrs are discussed in section [ s : selection ] .
some simple statistical properties of the remnants are presented in section [ s : simple ] , with more detailed discussions of distance - dependant statistical studies of galactic snrs ( including a brief discussion of some aspects of extragalactic remnants ) and the galactic distribution of snrs given in sections [ s : distance ] and [ s : distribution ] respectively .
the summary parameters of the 231 remnants from the 2004 january version of the catalogue of galactic snrs are presented in appendix [ s : appendix ] .
the catalogue of galactic snrs contains : ( i ) basic parameters ( galactic and equatorial coordinates , size , type , radio flux density , spectral index , and other names ) ; ( ii ) short descriptions of the observed structure at radio , x - ray and optical wavelengths , as applicable ; ( iii ) other notes on distance determinations , pulsars or point sources nearby ; and ( iv ) references .
appendix [ s : appendix ] gives the basic parameters of all 231 remnants in the 2004 january version of the catalogue , and describes these parameters in more details .
the detailed version of the catalogue is available on the world - wide - web from : + http://www.mrao.cam.ac.uk/surveys/snrs/ + which includes the descriptions , additional notes and references .
the detailed version is available as postscript or pdf for downloading and printing , or as html web pages for each individual remnant .
the web pages include links to the ` nasa astrophysics data system ' for each of the nearly one thousand references .
notes both on those objects no longer thought to be snrs , and on many possible and probable remnants that have been reported , are also included in the detailed version of the catalogue .
in addition to the observational selection effects that are discussed further in section [ s : selection ] , it should be noted that the catalogue is far from homogeneous .
is is particularly difficult to be uniform in terms of which objects are considered as definite remnants , and are included in the catalogue , rather than listed as possible or probable remnants which require further observations to clarify their nature .
although many remnants , or possible remnants , were first identified from wide area surveys , many others have been observed with a far from uniform set of observational parameters , making uniform criteria for inclusion in the main catalogue difficult .
also , some of the parameters included in the catalogue are themselves of quite variable quality .
for example , the radio flux density of each remnant at 1 ghz .
this is generally of good quality , being obtained from several radio observations over a range of frequencies , both above and below 1 ghz . however , for a small number of remnants often those which have been identified at other than radio wavelengths no reliable radio flux density , or only a limit is available ( which applies to 14 remnants in the current catalogue ) . , width=415 ] are indicated by arrows at the left and right edges of the plot.[f : sigmab],width=415 ]
although several galactic snrs have been identified at other than radio wavelengths , in practice the dominant selection effects are those that are applicable at radio wavelengths .
simplistically , two selection effects apply to the identification of galactic snrs ( e.g. green 1991 ) , due to the difficulty in identifying ( i ) faint remnants and ( ii ) small angular size remnants .
( in the case of extragalactic snrs , the selection effects are different , and these are discussed briefly particularly in the context of snrs identified in m82 in section [ s : extragalactic ] . ) clearly , snrs need to have a high enough surface brightness for them to be distinguished from the background galactic emission .
this selection effect is _ not _ uniform across the sky , both because the galactic background varies with position , and because the sensitivities of available wide area surveys covering different portions of the galactic plane vary .
the most recent large - scale radio surveys that have covered much of the galactic plane are : ( i ) the effelsberg survey at 2.7 ghz ( reich et al .
1990 ; frst et al . 1990 ) , which covered @xmath1 and @xmath2 ; and ( ii ) the most survey at 843 mhz ( whiteoak & green 1996 ; green et al .
1999 ) , which covered @xmath3 , but only to @xmath4 . figs .
[ f : sigmal ] and [ f : sigmab ] show the distribution of surface brightness for known snrs against galactic longitude and latitude .
these show that in the anti - centre and away from @xmath5 , where the galactic background is lower , fainter remnants are relatively more common than brighter remnants , as expected .
also , there are fewer faint remnants identified in the 4th quadrant , which is due to the narrower range of the latitude coverage of the most survey compared with that of the effelsberg survey .
( there are similar numbers of remnants with surface brightnesses less than @xmath6 with @xmath7 , in the 1st and 4th quadrant 9 and 7 respectively but at higher galactic latitude many more snrs have been identified in the 1st quadrant than in the 4th 10 compared to 5 due to the wider latitude coverage of the effelsberg survey . ) , @xmath2 ( i.e. the region covered by the effelsberg 2.7-ghz survey , reich et al .
1990 ; frst et al . 1990 ) identified since 1991 ( cf .
[ f : sigma ] for the distribution for all galactic remnants).[f : post92],width=415 ] the effelsberg survey detected new snrs down to surface brightnesses corresponding to @xmath8 at 1 ghz ( reich et al .
1988 ) , although the completeness limit for regions of brighter galactic emission is higher .
this is not only because of the difficulty in identifying remnants in the presence of extended galactic emission , but is also due to confusion with bright regions ( which is relatively more of a problem at higher frequencies ) .
since the new snrs identified from the effelsberg survey were included in the version of the snr catalogue published in green ( 1991 ) , consideration of the surface brightness of other remnants in the survey region that have subsequently been identified is useful for estimating the completeness limit for this survey . since 1991
an additional 24 remnants within @xmath1 and @xmath2 have been included in the catalogue , most in the first quadrant .
these remnants have been identified from a variety of observations , usually covering small regions of the galactic plane , rather than from large area surveys . of these
, five do not have good radio observations available , and a histogram of the surface brightnesses of the remaining 19 is shown in fig .
[ f : post92 ] . of these remnants ,
three have surface brightnesses above @xmath9 , two of which ( ( 0.3)+(0.0 ) and ( 1.0)-(0.1 ) ) are close to the galactic centre , where the background is particularly bright . regarding these three bright remnants as somewhat special cases , the surface brightnesses of the other more recently identified snrs suggest a completeness limit of @xmath10 for the effelsberg survey .
it is difficult to estimate the completeness limit of the most survey in a similar way , as only three new remnants have been identified in the most survey region since the remnants identified in this survey were included in the 1996 version of the catalogue .
this is due to the limited number of telescopes able to observe this part of the galactic plane . of these
more recently identified , only two have surface brightnesses ( of @xmath11 and @xmath12 at 1 ghz ) .
however , a comparison of the distribution of the brighter snrs in galactic longitude suggests that the completeness limit in the most survey region is not very different from that in the effelsberg survey region .
there are 32 remnants in the 1st quadrant ( i.e. covered by the effelsberg survey ) with surface brightnesses above @xmath9 and 27 in the 4th quadrant ( i.e. covered by the most survey ) .
although the molonglo survey covers a smaller range in galactic latitude than the effelsberg survey , this difference is not important , as only one of the bright remnants in the 1st quadrant has @xmath13 , .[f : sigma],width=415 ] so , the surface brightness limit for completeness of the current catalogue of galactic snrs is approximately @xmath9 .
[ f : sigma ] shows a histogram of the surface brightnesses of the 217 galactic snrs , of which 64 are above this nominal surface brightness limit .
as noted above , many snrs with surface brightnesses below this limit have been identified , both from surveys and from other observations .
these fainter remnants are predominantly in regions of the galaxy where the background is low , i.e. in the 2nd and 3rd quadrants , and away from @xmath5 , as shown in fig .
[ f : lb ] .
it is noticeable in fig .
[ f : sigmal ] that there are more remnants in the 2nd quadrant than the 3rd ( 21 compared with 11 ) .
it seems likely that this is due to the fact that the 2nd quadrant is more accessible to the wider range of northern hemisphere radio telescopes than is the 3rd quadrant . above the nominal surface brightness
limit given above , there are only very few remnants in the 2nd and 3rd quadrants ( 3 and 2 respectively ) , i.e. for these bright remnants there is no indication of any obvious deficit of remnants in the 3rd quadrant . at first sight , fig . [
f : sigmab ] appears to show that there are more remnants identified away from the galactic plane at positive latitudes than at negative latitudes .
there is an asymmetry in the number of snrs at high galactic latitudes .
there are 11 remnants with @xmath14 , but only 4 remnants with @xmath15 .
most of these high positive latitude remnants are in the 1st and 2nd quadrants , which suggests this asymmetry is related to gould s belt ( e.g. stothers & frogel 1974 ) , which is predominantly at positive latitudes in these quadrants .
however , it is not clear that these high latitude remnants are close enough to be associated with gould s belt .
there is no evidence for any asymmetry in the number of remnants at low galactic latitudes ; there are 49 with @xmath16 and 44 with @xmath17 , which are not statistically different . .
( note that the latitude and longitude axes are not to scale.)[f : lb],width=510 ] + .
( note that the latitude and longitude axes are not to scale.)[f : lb],width=510 ] ongoing and future observations will no doubt continue to detect more galactic snrs , although it seems very likely that most of these objects will be faint , and hence difficult to study in detail .
currently there are several large scale radio surveys underway that will cover much of the galactic plane including : ( i ) the canadian galactic plane survey ( cgps , see : taylor et al .
2003 ) which covers much of the northern galactic plane , from @xmath18 to @xmath19 ; ( ii ) the vla galactic plane survey ( vgps , see : lockman & stil 2004 ) which covers @xmath20 to @xmath21 , ( iii ) the southern galactic plane survey ( sgps , see mcclure - griffiths et al .
2001 ; mcclure - griffiths 2002 ) , and ( iv ) the second - epoch molonglo galactic plane survey ( mgps2 , see green 2002 ) , which covers @xmath22 , @xmath23 .
examples of recently identified snrs include two faint remnants with surface brightnesses at 1 ghz less than a few time @xmath24 , which were found from cgps observations by kothes et al .
however , confusion with the galactic background particularly in regions near the galactic centre and near @xmath5 will continue to be a limiting factor in the identification of even moderately bright remnants .
comparison of radio observations over a wide range of frequencies ( so spectral index information can be used ) , or observations at other than radio frequencies , may help to avoid some of the limits caused by this confusion .
recent discoveries include three new , as their identification was published after that catalogue was updated in 2004 january . ] , faint snrs near @xmath25 from brogan et al .
( 2004 ) , which have surface brightness at 1 ghz of @xmath26 .
these new remnants were identified because of the wide range of frequencies , and the relatively high resolution of the radio observations . ,
width=415 ] as discussed below in section [ s : sigmad ] , there is a general trend that fainter remnants tend to be larger , and hence on average older , than brighter remnants .
however , because of the wide range of properties of galactic snrs with known distances , the surface brightness selection effect applies not just to old remnants , but also to young remnants . in particular , note that the remnant of the snr of 1006 ( see table [ t : historical ] , below ) is fainter than the surface brightness completeness limit discussed above .
it should be noted that the study of galactic snrs by filipovi et al.(2002 ) , which used the pmn southern survey images to extract flux densities of snrs at 5 ghz , is seriously limited by observational constraints . since the pmn observations were not processed to image extended objects ( see condon , griffith & wright 1993 ) , the derived flux densities of many snrs are in serious error .
small angular size snrs are likely to be missing from current catalogues
. if they are too small , then their structure is not well resolved by the available galactic plane surveys , and they would not be recognised as likely snrs . fig .
[ f : theta ] is shows the histogram of the angular sizes of known remnants , which peaks at around 10 arcmin .
( note that for elongated remnants , which have angular sizes given as @xmath27 arcmin@xmath28 in the catalogue , a single diameter of @xmath29 has been used in this histogram , and in other figures in this paper concerned with angular size . )
the limiting angular size varies for the different available wide area surveys .
as discussed above , the radio survey that covers most of the galactic plane is the effelsberg 2.7-ghz survey , which has a resolution of @xmath30 arcmin .
so , for this survey , any remnants less than about 13 arcmin in diameter ( i.e. 3 beamwidths ) are not likely to be recognised from their structures ( although , as discussed in section [ s : missing ] , some searches have been made for small remnants among the unresolved and barely resolved sources in the effelsberg 2.7-ghz survey ) .
the most 843-mhz survey has a much better resolution , @xmath31 arcmin , which implies that in the region of the galactic plane covered by this survey only remnants smaller than about @xmath32 arcmin ( i.e. 3 beamwidths ) might be expected to be missed .
however , although the most survey detected 18 new snrs ( whiteoak & green 1996 ) , the smallest new remnant is ( 345.7)-(0.2 ) , which is @xmath33 arcmin@xmath28 in extent , i.e. several times larger than the nominal limit of @xmath32 arcmin .
thus it is difficult to quote a single angular size selection limit for current snr catalogues , although it is clear that it is difficult to identify small angular size remnants from existing wide area surveys .
this selection effect is likely to be more important for filled - centre type remnants than for shell type remnants .
even if filled - centre remnants are large enough to be resolved in a survey at the level of several beamwidths , their centrally brightened structures may not be striking enough to be able to recognise them as filled - centre remnants . using only radio continuum observations ,
it is also easy to confuse the flat - spectrum synchrotron emission from filled - centre remnants with thermal emission . additional observations
e.g. radio polarisation ( as used to identify the small angular size filled - centre remnant ( 54.1)-(0.3 ) , see reich et al . 1985 ) , radio recombination line non - detections ( see , for example , misanovic , cram & green 2002 ) , relatively low infra - red to radio ratios ( e.g. cohen & green 2001 ) , x - ray emission ( e.g. schaudel et al . 2002 ) are useful to distinguish filled - centre remnants from thermal sources .
the lack of small angular size remnants i.e. young but distant remnants is particularly clear when the remnants of known ` historical ' galactic supernovae ( see stephenson & green 2002 ) are considered .
these remnants are relatively close - by as is expected , since their parent sne were seen historically and therefore sample only a small fraction of the galactic disc .
consequently we expect many more similar , but more distant remnants in our galaxy ( e.g. green 1985 ) , but these are not present in current catalogues .
cccdcdccdccd & & & & & & & + date & name or & distance & & @xmath34 & & & size & & & size & + & remnant & /kpc & & / & & & /arcmin & & & /arcmin & + & cas a & 3.4 & 5 & @xmath35 & 2720 & & 2.0 & 435 & & 1.0 & 109 + 1604 & kepler s & 2.9 & 3 & @xmath36 & 19 & & 1.0 & 2.2 & & 0.5 & 0.55 + 1572 & tycho s & 2.3 & 8 & @xmath37 & 56 & & 2.3 & 4.1 & & 1.1 & 1.0 + 1181 & 3c58 & 3.2 & 7 & @xmath38 & 33 & & 2.6 & 4.7 & & 1.3 & 1.2 + 1054 & crab nebula & 1.9 & 6 & @xmath39 & 1040 & & 1.4 & 52 & & 0.7 & 13 + 1006 & ( 327.6)+(14.6 ) & 2.2 & 30 & @xmath40 & 19 & & 7.7 & 1.3 & & 3.9 & 0.31 + table [ t : historical ] gives the distances , angular sizes , flux densities and surface brightnesses at 1 ghz , for the remnants of known historical supernovae from the last thousand years , plus cas a ( which although its progenitor was not seen so it is not strictly a historical remnant is known to be only about 300 years old ) .
the distances used for these remnants are those given in section [ s : distances ] .
this table also lists the parameters of these remnants when scaled to larger distances of 8.5 and 17 kpc , i.e. to represent how they would appear if they if they were at the other side of the galaxy ( from the galactic centre , to the far point on the solar circle ) .
the number of other young ( i.e. less than a thousand year old ) snrs expected in the galaxy can be estimated in two simple ways : ( i ) from the expected supernova rate of one every 45 to 70 years ( cappellaro 2003 ) , 15 to 22 young remnants are expected in total ; ( ii ) considering the fraction of the galactic disc sampled by the historical supernovae ( 6 in a thousand years ) , which are within @xmath41 kpc , i.e.about 16 per cent of the galactic disc modelled as being uniform and having a radius of @xmath42 kpc , implies there should be @xmath43 young remnants ( see also the discussions in strom 1994 ) . of these , @xmath44 ( see cappellaro 2003 ) are expected to be the remnants of massive supernovae
i.e.from type ib / ic / ii sne and therefore be close to the galactic plane . from table [ t : historical ] , any young snrs in the galaxy similar to the known historical remnants , but in the far half of the galaxy , would generally be expected to have angular sizes less than a few arcmin , high surface brightness , greater than @xmath45 ( although remnants similar to the remnant of the sn of 1006 would be much fainter ) .
these remnants would also be expected to lie close to the galactic plane , with @xmath46 .
although the above estimates are rather uncertain due to intrinsic poisson uncertainties from the small numbers of known historical remnants , they imply that about a dozen or more young but distant remnants might be expected . however , there are very few such remnants in the current galactic snr catalogue .
in fact there are only 3 known remnants with angular sizes of 2 arcmin or less : ( 1.9)+(0.3 ) , ( 54.1)+(0.3 ) and ( 337.0)-(0.1 ) . of these ,
( 1.9)+(0.3 ) is the smallest , with an angular size of only 1.2 arcmin .
no distance measurement is available for this remnant which being close to @xmath47 makes kinematic methods unreliable but even if it were at the far side of the galaxy , at say 17 kpc , its physical size would only be 6 pc .
this is comparable to the sizes of the known historical remnants in table [ t : historical ] ( which have physical diameters of 5 , 3 , 5 , 7 , 3 and 19 pc respectively ) .
another indication that this is indeed a young remnant is that it shows a circularly symmetric limb - brightened shell of radio emission ( see fig .
[ f : g1.9 ] for a previously unpublished image of this remnant at 1.5 ghz , from 1985 observations made with the nrao s vla ) .
the known young shell remnants all show highly circular structures , whereas older remnants tend to be less circular , as is expected as they expand into differing regions of the interstellar medium .
the distance to the filled - centre remnant ( 54.1)+(0.3 ) is uncertain ( e.g. camilo et al . 2002 ) , although its small angular size of 1.5 arcmin again suggests it is physically small , and therefore young , even if it were to be situated at the edge of the galaxy .
a reasonably accurate distance estimate is available for ( 337.0)-(0.1 ) ( see section [ s : distances ] ) , and this is @xmath48 kpc , which is consistent with this being a young but distant snr ( with a diameter of @xmath49 pc , from its angular size of 1.5 arcmin ) . the deficit of small remnants is also illustrated in fig . [
f : sigmatheta ] , which shows there are very few known galactic remnants with high surface brightnesses and small angular sizes .
arcsec@xmath28 at a position angle of @xmath50 .
the contour levels are at @xmath51 mjy beam@xmath52 for @xmath53 ( with the negative contour dashed ) .
the coordinates are j2000.0.[f : g1.9],width=226 ] arcmin .
the five historical remnants from table [ t : historical ] included in this plot are indicated by additional crosses .
( for all but the smallest few remnants in the catalogue , the angular size is given to the nearest arcmin.)[f : sigmatheta],width=415 ] this deficit of young but distant remnants has long been recognised , but searches for remnants of this type ( e.g.green & gull 1984 ; helfand et al . 1984 ; green 1985 , 1989 ; sramek et al .
1992 ; misanovic et al.2002 ; see also saikia et al .
2004 ) have had only limited success ( identifying the small remnants ( 1.9)+(0.3 ) and ( 54.1)+(0.3 ) noted above ) . since
the missing young but distant remnants are expected to have angular sizes of a few arcmin or less , they will not have been resolved sufficiently by single - dish radio surveys with resolutions of several arcminutes , so will not have been recognised as snrs .
searching for these remnants is not easy because there are very many candidate sources in single - dish surveys to choose from ( e.g. in the 1st quadrant , there are over a thousand compact sources in the effelsberg survey with @xmath54 ) , and only a fraction of these have been observed with high enough resolution to be able to identify them if they were small angular size snrs . moreover , the missing young but distant remnants are likely to be in the most complex , and hence most confused , regions of the galactic plane , as being distant they will be close to @xmath5 .
the use of additional observational indicators ( e.g. radio spectral index , infra - red to radio ratios ) has yet not proved efficient for improving such searches , nor have the most survey ( see whiteoak & green 1996 ) and the nvss ( see condon et al.1998 ) , which cover the galactic plane with slightly higher resolution than available single - dish surveys .
( note , however , that the nvss does contain previously unrecognised snrs , for example ( 353.9)-(2.0 ) , with an angular size of 13 arcmin , see green 2001a . ) the fact that such missing , small remnants are likely to be in complex regions of the plane may mean that confusion is a very significant problem , and not just at radio wavelengths .
further searches for these missing young but distant remnants are required .
it should be noted that there are unlikely to be other luminous remnants in the galaxy like cas a and the crab nebula .
any such remnants , even on the far side of the galaxy , would have relatively high flux densities , and the nature of all such sources in the galactic plane is known . on the other hand ,
the remnant of the sn of 1006 is faint possibly because it is far from the galactic plane , in a low density region and distant remnants similar to this would be particularly difficult to detect , as they would have both a small angular size and low surface brightness .
however , any remnants similar to the other three historical remnants are detectable .
in the current version of the catalogue , 77% of remnants are classed as shell ( or possible shell ) , 12% are composite ( or possible composite ) , and 4% are filled - centre ( or possible filled centre ) remnants .
the remaining 7% have not yet been observed well enough to be sure of their type , or else are objects which are conventionally regarded as snrs although they do not fit well into any of the conventional types ( e.g. ctb80 ( @xmath55(69.0)+(2.7 ) ) , msh 17@xmath563_9 _ ( @xmath55(357.7)-(0.1 ) ) ) . since the 1991 version of the catalogue ( green 1991 ) , the proportion of shell remnants in the catalogue has stayed very similar , with the proportion of composite remnants increasing from 8% , and the proportion of filled centre remnants has decreasing from 7% .
the increase in the proportion of composite remnants is because more recent , improved observations have continued to identify many more shell remnants , but have also identified faint , pulsar powered nebulae in what until then had been identified as pure shell remnants ( e.g. w44 ( @xmath55(24.7)-(0.4 ) ) ) , and also that faint shells have been detected around some filled - centre remnants ( e.g.(21.5)-(0.9 ) ) .
arcmin.[f : stheta],width=415 ] there are 14 galactic snrs that are either not detected at radio wavelengths , or are poorly defined by current radio observations , so that their flux density at 1 ghz can not be determined with any confidence : i.e. 94% have a flux density at 1 ghz included in the catalogue . of the catalogued remnants ,
36% are detected in x - ray , and 23% in the optical . at both these wavelengths ,
galactic absorption hampers the detection of distant remnants .
some of the properties of the galactic snrs in the catalogue , which are not shown in other sections of this paper , are shown in fig .
[ f : stheta ] .
this shows the flux density at 1 ghz versus angular size for snrs less than 30 arcmin in extent .
this is of interest in terms of which remnants may appear as bright , relatively compact galactic plane sources ( e.g. in future planck surveys ) .
the most prominent sources are , not surprisingly , the very bright snrs cas a and the crab nebula ( see table [ t : historical ] ) , which have similar angular sizes , and very high flux densities .
cas a has the higher flux density at 1 ghz by a factor of about 2.6 , but because the crab nebula has a much flatter spectrum ( with @xmath57 compared with @xmath58 for cas a , e.g. baars et al . 1977 ) , the crab nebula has the higher flux density at frequencies above about 8 ghz .
the statistics of the radio spectral indices of galactic snrs are not discussed here , although there is a short discussion of these in green ( 2001b ) .
many studies of galactic snrs require knowledge of the distances to remnants ( or equivalently their physical sizes , since their angular sizes are known ) . however , accurate distances are not available for many known snrs .
the distances that are available are obtained from a wide variety of methods e.g. optical expansion and proper motion studies , 21-cm absorption spectra , column density ( see foster & routledge 2003 ) , association with or co features in the surrounding interstellar medium , or association with other objects each of which is subject to their own uncertainties , and some of which are subjective .
table [ t : distances ] presents distances for 47 galactic snrs available in the literature .
( in a few cases distances estimates to snrs are also available from the distances to associated pulsar , derived from the observed pulsar dispersion measure and a model of the galactic electron density distribution .
however , these have not been used in table [ t : distances ] . ) in several cases the distances derived from absorption measurements have been recalculated using a modern ` flat ' rotation curve with a galactocentric radius of 8.5 kpc and a constant rotation speed of 220 km s@xmath52 . additionally , in a few cases the distances given depend on re - interpretation of the published observations .
for ( 11.2)-(0.3 ) and ( 21.5)-(0.9 ) , the distances given correspond to the near distances of the last strong absorption seen ( see also safi - harb et al .
2001 for ( 21.5)-(0.9 ) ) .
rdllr remnant & & method & reference & notes + & & & & + ( 4.5)+(6.8 ) & 2.9 & optical proper motion / velocity & blair et al .
( 1991 ) & + ( 6.4)-(0.1 ) & 1.9 & absorption & velzquez et al .
( 2002 ) & + ( 11.2)-(0.3 ) & 4.4 & absorption & becker et al .
( 1985 ) & a , b + ( 18.8)+(0.3 ) & 14.0 & association with co & dubner et al .
( 2004 ) & + ( 21.5)-(0.9 ) & 4.6 & absorption & davelaar et al .
( 1986 ) & a , b + ( 27.4)+(0.0 ) & 6.8 & absorption & sanbonmatsu & helfand ( 1992 ) & + ( 33.6)+(0.1 ) & 7.8 & absorption & frail & clifton ( 1989 ) & a + ( 34.7)-(0.4 ) & 2.8 & absorption & caswell et al . ( 1975 ) & a + ( 39.7)-(2.0 ) & 3.0 & association with & dubner et al . ( 1998 ) & + ( 43.3)-(0.2 ) & 10.0 & association with & lockhart & goss ( 1978 ) & a + ( 49.2)-(0.7 ) & 6.0 & association with co & koo et al . ( 1995 ) & + ( 53.6)-(2.2 ) & 2.8 & association with & giacani et al .
( 1998 ) & + ( 55.0)+(0.3 ) & 14.0 & association with & matthews et al .
( 1998 ) & + ( 74.0)-(8.5 ) & 0.4 & optical proper motion / velocity & blair et al .
( 1999 ) & + ( 74.9)+(1.2 ) & 6.1 & column density & kothes et al .
( 2003 ) & + ( 84.2)-(0.8 ) & 4.5 & association with co & feldt & green ( 1993 ) & + ( 89.0)+(4.7 ) & 0.8 & association with co and & tatematsu et al .
( 1990 ) & + ( 93.3)+(6.9 ) & 2.2 & column density & foster & routledge ( 2003 ) & + ( 93.7)-(0.2 ) & 1.5 & association with & uyaniker et al .
( 2002 ) & + ( 109.1)-(1.0 ) & 3.0 & association with region & kothes et al .
( 2002 ) & + ( 111.7)-(2.1 ) & 3.4 & optical proper motion / velocity & reed et al .
( 1995 ) & + ( 114.3)+(0.3 ) & 0.7 & association with & yar - uyaniker et al ( 2004 ) & + ( 116.5)+(1.1 ) & 1.6 & association with & yar - uyaniker et al ( 2004 ) & + ( 116.9)+(0.2 ) & 1.6 & association with & yar - uyaniker et al ( 2004 ) & + ( 119.5)+(10.2 ) & 1.4 & association with & pineault et al .
( 1993 ) & + ( 120.1)+(1.4 ) & 2.3 & optical proper motion / velocity & chevalier et al .
( 1980 ) & + ( 130.7)+(3.1 ) & 3.2 & absorption & roberts et al .
( 1993 ) & + ( 132.7)+(1.3 ) & 2.2 & association with co & routledge et al .
( 1991 ) & + ( 166.0)+(4.3 ) & 4.5 & association with & landecker et al .
( 1989 ) & + ( 166.2)+(2.5 ) & 8.0 & association with & routledge et al .
( 1986 ) & + ( 184.6)-(5.8 ) & 1.9 & various & trimble ( 1973 ) & + ( 189.1)+(3.0 ) & 1.5 & optical absorption & welsh & sallmen ( 2003 ) & + ( 205.5)+(0.5 ) & 1.6 & various & odegard ( 1986 ) & + ( 260.4)-(3.4 ) & 2.2 & association with & reynoso et al .
( 1995 ) & + ( 263.9)-(3.3 ) & 0.3 & pulsar parallax & caraveo et al .
( 2001 ) & + ( 292.0)+(1.8 ) & 6.0 & various & gaensler & wallace ( 2003 ) & + ( 292.2)-(0.5 ) & 8.4 & absorption & caswell et al .
( 2004 ) & + ( 296.8)-(0.3 ) & 9.6 & association with & gaensler et al .
( 1998a ) & + ( 315.4)-(2.3 ) & 2.3 & optical velocity & sollerman et al .
( 2003 ) & + ( 320.4)-(1.2 ) & 5.2 & absorption & gaensler et al .
( 1999 ) & + ( 327.4)+(0.4 ) & 4.8 & absorption & mcclure - griffiths et al .
( 2001 ) & + ( 327.6)+(14.6 ) & 2.2 & optical proper motion / velocity & winkler et al .
( 2003 ) & + ( 332.4)-(0.4 ) & 3.1 & absorption & caswell et al .
( 1975 ) & a + ( 337.0)-(0.1 ) & 11.0 & various & sarma et al .
( 1997 ) & + ( 348.5)+(0.1 ) & 8.0 & absorption & caswell et al .
( 1975 ) & a + ( 348.7)+(0.4 ) & 8.0 & absorption & caswell et al .
( 1975 ) & a + ( 349.7)+(0.2 ) & 14.8 & absorption & caswell et al .
( 1975 ) & a + + notes : a ) distance recalculated using modern rotation curve ; b ) see text for further discussion . ) .
the dashed line indicates the surface brightness completeness limit discussed in section [ s : surface].[f : withd],width=415 ] the uncertainties in these distances are far from uniform . for kinematic distances which are a large majority of the distances given in table [ t : distances ] there are always some uncertainties in deriving distances from observed velocities , due to deviations from circular motion ( especially an issue for nearby remnants , and for those near @xmath47 and @xmath59 where the observed velocity does not depend strongly on distance ) and ambiguities inside the solar circle .
generally , the published errors in kinematic distances are less than @xmath60 , although there are additional possible larger uncertainties depending on whether the ( often subjective ) association of a particular feature with a remnant is actually correct .
a potential bias for statistical studies is that many distance methods are more easily applied to brighter remnants than to fainter ones .
this is particularly the case for 21-cm absorption studies , which depend on a radio continuum from the remnant being bright , otherwise any absorption could not be studied in reasonable detail .
but this also applies to some of the other methods , e.g.association with other or co features in the ism , where fainter remnants will not be well defined , so that clear morphological association with other features will be more difficult . indeed , fig .
[ f : withd ] shows a histogram of the surface brightness of galactic snrs with known distances , which shows these tend to be the brighter galactic snrs overall ( see fig . [
f : sigma ] ) .
thus , it is likely that , at a given diameter , snrs with known distances are biased to brighter remnants .
the number of snrs with available distances is sufficiently large that statistical studies see section [ s : sigmad ] show that the range of intrinsic luminosities of galactic snrs is large . )
relation for 47 galactic snrs with known distances ( see table [ t : distances ] ) , shown as filled circles .
the open circle shows the parameters of rx j@xmath61 ( @xmath55(266.2)-(1.2 ) ) , if it is at a distance of 200 pc ( see text for discussion ) .
note that the lower left part of this diagram is likely to be seriously affected by selection effects.[f : sigmad],width=340 ] ) relation for 47 galactic snrs with known distances ( see table [ t : distances]).[f : ld],width=340 ] since distances are not available for all snrs , many statistical studies of galactic snrs have relied on the surface - brightness / diameter , or ` @xmath0 ' relation to derive distances for individual snrs from their observed flux densities and angular sizes . for remnants with known distances ( @xmath63 ) , and hence known diameters ( @xmath64 ) , physically large snrs are fainter ( i.e. they have a lower surface brightness ) than small remnants . using this correlation between @xmath65 and @xmath64 for remnants with known distances ,
a physical diameter is deduced from the distance - independent _ observed _ surface brightness of any remnant .
then a distance to the remnant can be deduced from this diameter and the observed angular size of the remnant .
the @xmath0 relation for galactic snrs with known distances is shown in fig .
[ f : sigmad ] . as discussed above ( section [ s : distances ] ) , the distances to individual remnants
are not homogeneous in quality , and many depend on subjective interpretation of data . of the snrs included in this figure ,
three are ` filled - centre ' remnants ( the crab nebula ( @xmath55(184.6)-(5.8 ) ) , 3c58 ( @xmath55(130.7)+(3.1 ) ) and ( 74.9)+(1.2 ) ) .
nevertheless , fig .
[ f : sigmad ] clearly shows a wide range of diameters for a given surface brightness , which is a severe limitation in the usefulness of the @xmath0 relation for deriving the diameters , and hence distances , to individual remnants . for a particular surface brightness , the diameters of snrs vary by up to about an order of magnitude , or conversely , for a particular diameter , the range of observed surface brightnesses seen varies by more than two orders of magnitude . the correlation shown between surface brightness and diameter in fig .
[ f : sigmad ] is , however , largely a consequence of the fact that it is a plot of surface - brightness rather than luminosity against diameter , @xmath64 .
surface brightness is plotted , because it is the distance - independent observable that is available for ( almost ) all snrs , including those for which distances are not available . for remnants
whose distances are known , we can instead consider the radio luminosity of the remnants . since @xmath65 and luminosity , @xmath66 , depends on the flux density @xmath67 , angular size @xmath68 , distance @xmath63 and diameter @xmath64 , as @xmath69 then @xmath70 thus , much of the correlation shown in the @xmath0 relation in fig . [
f : sigmad ] is due to the @xmath71 bias that is inherent when plotting @xmath65 against @xmath64 , instead of @xmath66 against @xmath64 .
the @xmath62 relation for galactic snrs with known distances in fig .
[ f : ld ] shows that there is wide range of luminosities for snrs of all diameters .
cas a is the most luminous galactic snr , but it appears to be at the edge of a wide distribution of luminosities .
the wide range of luminosities is perhaps not surprising , given that the remnants are produced for a variety of types of supernovae , and that they evolve in regions of ism with a range of properties ( e.g. density ) , which may well effect the efficiency of the radio emission mechanism at work .
for example , some snrs may initially evolve inside a low - density , wind - blown cavity , and then collide with the much denser regions of the surrounding ism .
furthermore , the full range of intrinsic properties of snrs may be even wider than that shown in figs [ f : sigmad ] and [ f : ld ] , as the selection effects discussed above mean that it is difficult to identify small and/or faint snrs .
one specific example is the recently identified snr rx j@xmath61 ( @xmath55(266.2)-(1.2 ) , see aschenbach 1998 ) , which may extend the range of properties of snrs considerably ( see duncan & green 2000 ) .
the surface brightness of rx j@xmath61 at 1 ghz is @xmath72 , which places it among the faintest 20 per cent of catalogued remnants .
if it is at a distance as small as 200 pc , as suggested by aschenbach ( 1998 ) see also redman et al .
( 2002 ) then its diameter would be only 7 pc , see fig .
[ f : sigmad ] . on the other hand , the deficit of bright , large snrs in fig .
[ f : sigmad ] can not be explained by any selection effect , and so represents some real limit in the luminosity of remnants at a particular diameter , related to the physics of the underlying radio emission mechanism at work .
this upper bound in the @xmath0 plane can be used to derive an _
upper limit _ for the diameter of any snr from its observed surface brightness , and hence an upper limit on its distance .
another upper bound on the distance to any remnant which may be as useful is to assume it lies within the galactic disc .
case & bhattacharya ( 1998 ) derived a @xmath0 relation , based on the distances available for 36 remnants not including filled - centre remnants from the 1996 version of the catalogue .
they argue that it is useful for deriving distances for individual snrs , with a fractional error of only 0.33 , which is very much smaller than the wide range in diameters for a given surface brightness shown in fig .
[ f : sigmad ] .
this optimistic result was only obtained after excluding cas a from the remnants with known distances used to derived the @xmath0 relation ( as it was deemed to be sufficiently different from other shell snrs ) , and also after excluding 7 remnants with high @xmath73-values ( as three of these showed the largest deviation from the best - fit @xmath0 relation ) .
it is , however , difficult to decide _ a priori _ whether a remnant is or is not to be included in the subset of remnants for which case & bhattacharya derived distances with relatively small uncertainties . also , it is not clear that any best - fit @xmath0 relation not withstanding selection effect problems actually represents the evolutionary track of individual snrs ( see berkhuijsen 1986 ) .
the distribution of snrs with known distances is a snapshot in time of a population of remnants , and individual remnants may evolve in the @xmath0 plane in directions quite different from the overall power law fitted to the overall distribution of snrs ( or to the upper limit of the distribution ) . as a simple example , consider the situation where snrs have a range of intrinsic luminosities , expand with a constant luminosity up to some particular diameter which varies for different snrs depending on their environment ( e.g. the surrounding ism density , which influences their expansion speed , which may affect the efficiency of radio emission mechanism )
after which their radio luminosity fades rapidly . in this case
, the locus of the upper bound to the highest surface brightness remnants for a particular diameter is related to where the luminosities of different snrs begin to decrease , and does not represent the evolutionary track of any individual remnant .
the current direction of the evolutionary track of only one galactic snr , cas a , can be estimated from available observations .
the flux density of cas a is decreasing at approximately 0.8% year@xmath52 ( baars et al .
1977 ; rees 1990 ) , and its bulk expansion is 0.22% year@xmath52 ( ageros & green 1999 ; but see delaney et al .
2004 for alternative expansion timescales ) .
these observations suggest cas a is following a track with @xmath74 , although the uncertainty in this slope is large ( nominally @xmath75 , if the uncertainty in the secular flux density decrease is taken as 0.2% year@xmath52 , and the expansion timescale of the remnant is between 400 to 500 years ) .
as noted above , a major problem with statistical studies of galactic snrs is the difficulty of obtaining reliable distances for remnants .
studies of samples of remnants in external galaxies are more straightforward in this respect , as all the remnants are at a very similar distance .
given this , then it is arguably more appropriate to consider the @xmath62 relation , rather than the @xmath0 , for extragalactic snrs as the latter is not needed to determine distances for individual remnants with the appropriate selection effects .
then , for unresolved sources , in most radio studies the dominant selection effect is a flux density limit , which corresponds to a fixed luminosity limit for a particular galaxy . in a recent study of the statistical properties of extragalactic snrs in several galaxies in addition to those in the milky way arbutina et al.(2004 )
concluded that only in the case of m82 was there a good @xmath62 correlation , and hence a useful @xmath0 relation also .
in other cases there was a poor correlation between the luminosity and diameter of identified snrs ( as noted in section [ s : sigmad ] above for galactic snrs ) .
however , in their study arbutina et al . have not correctly appreciated the observational selection effects applicable to the sample of snrs in m82 .
this sample of snrs was identified by huang et al .
( 1994 ) from observations at 8.4 ghz . in fig .
1 of arbutina et al . a sensitivity limit corresponding to a luminosity of @xmath76 erg s@xmath52 hz@xmath52
is plotted , which although appropriate at 8.4 ghz , is not appropriate for the luminosities of the m82 sample of snrs plotted in this figure , which are at 1 ghz .
the true luminosity limit from huang et al.s observations should be moved upwards on arbutina et al.s fig . 1 by @xmath77 ( i.e. @xmath78 for a typical spectral index of 0.5 to correct from 8.4 to 1 ghz ) .
moreover , the actual observational selection effects provide a more stringent limit on the detectable luminosities of the larger snrs in the sample .
any remnants with a diameter of larger than @xmath77 pc were resolved by huang et al.s observations , and a surface brightness limit ( equivalent to luminosity scaling as @xmath79 ) , not a constant luminosity limit is appropriate ( see fig . 2 of huang et al . , who show these limits on a @xmath0 rather than a @xmath62 plot ) .
consequently , the apparent range of luminosities of snrs in m82 is strongly limited by selection effects , particularly for the larger remnants .
indeed , it is noticeable that the range of luminosities shown for the larger remnants in m82 appears smaller than that of the smaller remnants , which seems unlikely to be real .
thus the range of luminosities of snrs in m82 is likely to extend to lower values , but these objects have not been identified due to selection effects . in this case
the correlation between luminosity and diameter for remnants here is not strong , as is the case in other galaxies , including our own .
consequently , the @xmath0 relation for m82 is also affected by selection effects , and is therefore of limited use .
the distribution of snrs in the galaxy is of interest for many astrophysical studies , particularly in relation to their energy input into the ism and for comparison with the distributions of possible progenitor populations .
such studies are , however , not straightforward , due to observational selection effects and the lack of reliable distance estimates available for most identified remnants . in particular , all snrs in the anti - centre ( i.e. 2nd and 3rd galactic quadrants ) are outside the solar circle , at large galactocentric radii , in regions where the background galactic emission is low , so that low surface brightness remnants are relatively easy to identify ( see section [ s : selection ] ) . without taking selection effects into account , the larger number of fainter snrs in the anti - centre leads to an apparently broad distribution of galactic snrs in galactocentric radius ( e.g. the very broad distribution of snrs derived by li et al .
( 1991 ) , who included all snrs in their analyses ) . a more complicated method to derive the radial distribution of galactic snrs
is that used by case & bhattacharya ( 1996 , 1998 ) , following a method used by narayan ( 1987 ) for pulsars .
this relies on ( i ) assuming catalogues of galactic snrs are complete for snrs within a distance of 3 kpc ; ( ii ) using @xmath0 derived distances for the snrs , and ( iii ) attempts to correct for observational selection effects using a scaling factor that varies in many bins across the disc of the galaxy .
however , this is difficult given the uncertainties in the usefulness of @xmath0 relation discussed above , and the necessity of deconvolving selection effects from the observed distribution of snrs .
an alternative approach is to investigate the distribution of snrs in galactic coordinates , restricting the studies to relatively bright remnants , for which current catalogues are thought to be complete .
van den bergh ( 1988a , b ) discussed the distribution of observed snrs and noted that high surface brightness remnants ( in this case taken to be @xmath80 ) are concentrated in a thin nuclear disc when plotted in galactic coordinates .
as noted by frst s comments to van den bergh ( 1988b ) , this conclusion is strengthened by a more realistic surface brightness completeness limit .
+ + + more quantitatively following the method of li et al . , but using an appropriate selection brightness cut - off the observed distribution of bright snrs with galactic longitude can be compared with that expected from various models .
a major advantage of this method is that it avoids the problem that we lack accurate distances to individual snrs , although on the other hand it uses only a sub - set of the known galactic snrs .
[ f : all - bright ] shows the observed distributions with galactic longitude of all galactic snrs , and of the 64 remnants which have @xmath81 ( this is a similar number to the 71 remnants used in a similar study presented in green 1996a , which used a slightly lower surface brightness cut - off applied to the snr catalogue of green 1996b ) . by applying the surface brightness cut - off , so that the surface brightness selection effect is not important , it is clear that the distribution of galactic remnants is actually much more concentrated towards @xmath47 than if all remnants are considered ( cf .
[ f : lb ] ) .
[ f : all - bright ] shows evidence for a deficit of snrs near @xmath82 , which may be a true deficit if there is a decrease in the space density of sn progenitors towards the galactic centre .
however , it may also be , in part at least , due to the difficulty of finding remnants in this region of the galactic plane , due to the very complex background emission and confusing galactic sources ( e.g. regions ) .
any remaining incompleteness in current catalogues , both for the surface brightness and angular diameter selection effects , are expected to be worse closer to @xmath5 ( because of the increased confusion in the case of the surface brightness selection effect , and the longer line - of - sight through the galaxy for missing small , i.e. young but distant remnants ) .
thus , the true distribution in @xmath83 is likely to be somewhat _ narrower _ than is indicated in fig .
[ f : all - bright ] .
for comparison with the observed distributions in galactic longitude , simple monte carlo models of the distribution of snrs in the disc of the galaxy were constructed assuming a simple , circularly symmetric , gaussian distribution , where the probability distribution varies with galactocentric radius , @xmath84 , as @xmath85 ( where @xmath86 is the gaussian galactocentric scale length , assuming the distance to the galactic centre is 8.5 kpc ) .
[ f : model ] shows plots of the expected distribution of snrs in galactic longitude of three such models for different scale lengths . as noted above
, the true distribution is likely to be somewhat narrower than that derived from the observations , due to residual selection effects , so that this scale length is an upper limit .
a @xmath87 comparison of the observed and model cumulative distributions indicates that for this simple model , a scale length of @xmath88 kpc best matches the observed distribution of high brightness snrs . the model distribution of snrs derived above should , however , be interpreted cautiously , as not only is it a simplistic model without spiral arm structures , but also it is a model of the distribution of _ observed _ remnants .
it is far from clear what factors affect the brightness and lifetime of appreciable radio emission from snrs
i.e. their observability at radio wavelengths and hence how close the distribution of observed snrs is to the parent supernovae distribution .
the distribution of snrs could reflect the distribution of , for example , the density of the ism , or the galactic magnetic field , if these are important factors in determining the brightness and lifetime of radio emission from snrs .
here i have presented a recent catalogue of 231 galactic snrs , and have discussed the selection effects that apply to the identification of remnants . both surface brightness and angular size selection
effects are important , and these need to be borne in mind when statistical studies of galactic snrs are made .
one consequence of the current angular size selection effect is that few young but distant remnants have yet been identified in the galaxy .
these objects are likely to be in complex regions of the galactic plane , and further observations using a wide range of radio wavelengths and/or the combination of radio and other wavelengths are required to identify these missing objects . for remnants with known distances ,
the intrinsic range of luminosity of galactic snrs is large , which combined with selection effects , means that the @xmath0 relation is of limited use for determining distances to individual remnants , or for statistical studies .
i am grateful to many colleagues for numerous comments on , and corrections to , the various versions of the galactic snr catalogue , and for comments on early versions of this work .
i also thank the referee , chris salter , for his useful and detailed report .
this paper was finalised during an extended visit to the national radio astronomy observatory , socorro , nm , usa in 2004 september , and i am very grateful for their hospitality at that time .
this research has made use of nasa s astrophysics data system bibliographic services .
10 ageros m. , green d. a. , 1999 , _ mnras _ , * 305 * , 957 .
arbutina b. , uroevi d. , stankovi m. , tei lj .
, 2004 , _ mnras _ , * 350 * , 346 .
arendt r. g. , 1989 , _ apjs _ , * 70 * , 181 .
aschenbach b. , 1998 , _ nature _ , * 396 * , 141 .
baars j. w. m. , genzel r. , pauliny - toth i. i. k. , witzel , a. , 1977 , _ a&a _ , * 61 * , 99
. becker r. h. , markert t. , donahue m. , 1985 , _ apj _ , * 296 * , 461 .
berkhuijsen e. m. , 1973 , _
* 24 * , 143 . berkhuijsen e. m. , 1986 , _
a&a _ , * 166 * , 257 .
blair w. p. , long k. s. , vancura o. , 1991 , _ apj _ , * 366
* , 484 .
blair w. p. , sankrit r. , raymond j. c. , long k. s. , 1999 , _ aj _ , * 118 * , 942 .
braje t. m. , romani r. w. , roberts m. s. e. , kwai n. , 2002 , _ apj _ , * 565 * , l91 .
brogan c. l. , devine k. e. , lazio t. j. , kassim n. e. , tam c. r. , brisken w. f. , dyer k. k. , roberts m. s. e. , 2004 , _ aj _ , * 127 * , 355 camilo f. , lorimer d. r. , bhat , n. d. r. , gotthelf e. v. , halpern j. p. , wang q. d. , lu f. j. , mirabal n. , 2002 , _ apj _ , * 574 * , l71 . cappellaro m. , 2003 , in _ supernovae and gamma - ray bursters _ , lecture notes in physics , volume 598 , ed k. w. weiler , springer , berlin , p. 21 .
caraveo p. a. , de luca a. , mignani r. p. , bignami g. f. , 2001 , _ apj _ , * 561 * , 930 .
case g. l. , bhattacharya d. , 1996 , _ a&as _ , * 120 * , 437
. case g. l. , bhattacharya d. , 1998 , _ apj _ , * 504 * , 761 .
caswell j. l. , murray j. d. , roger r. s. , cole d. j. , cooke d. j. , 1975 , _
a&a _ , * 45 * , 239 .
caswell j. l. , mcclure - griffiths n. m. , cheung m. c. m. , 2004 , _ mnras _ , * 352 * , 1405 .
chevalier r. a. , kirshner r. p. , raymond j. c. , 1980 , _ apj _ , * 235 * , 186 .
cohen m. , green a. j. , 2001 , _ mnras _ , * 325 * , 531 .
combi j. a. , testori j. c. , romero g. e. , colomb f. r. , 1995 , _ a&a _ , * 296 * , 514 .
condon j. j. , griffith m. r. , wright a. e. , 1993 , _ aj _ , * 106 * , 1095 .
condon j. j. , cotton w. d. , greisen e. w. , yin q. f. , perley r. a. , taylor g. b. , broderick j. j. , 1998 , _ aj _ , * 115 * , 1693 .
davelaar j. , smith a. , becker r. h. , 1986 , _ apj _ , * 300 * , l59 .
delaney t. , rudnick l. , fesen r. a. , jones t. w. , petre r. , morse j. a. , 2004 , _ apj _ , * 613 * , 343 .
dubner g. m. , holdaway m. , goss w. m. , mirabel i. f. , 1998 , _ aj _ , * 116 * , 1842 . dubner g. , giacani e. , reynoso e. , parn s. , 2004 , _ a&a _ , * 426 * , 201
. duncan a. r. , green d. a. , 2000 , _
a&a _ , * 364 * , 732 .
duncan a. r. , stewart r. t. , haynes r. f. , jones k. l. , 1995 , _ mnras _ , * 277 * , 36 . duncan a. r. , stewart r. t. , haynes r. f. , jones k. l. , 1997 , _ mnras _ , * 287 * , 722 .
feldt c. , green d. a. , 1993 , _
a&a _ , * 274 * , 421 .
fesen r. a. , hurford a. p. , 1996 , _ apjs _ , * 106 * , 563 .
filipovi m. d. , stupar m. , jones p. a. , white g. l. , 2002 , in _ neutron stars in supernova remnants _ , asp conference series , volume 271 , eds p. o. slane & b. m. gaensler , asp , san francisco , p. 387 .
foster t. , routledge d. , 2003 , _ apj _ , * 598 * , 1005 .
frail d. a. , clifton t. r. , 1989 , _ apj _ , * 336 * , 854 .
frst e. , reich w. , reich p. , reif k. , 1990 , _ a&as _ , * 85 * , 691 .
gaensler b. m. , wallace b. j. , 2003 , _ apj _ , * 594 * , 326 .
gaensler b. m. , manchester r. n. , green a. j. , 1998a , _ mnras _ , * 296 * , 813 .
gaensler b. m. , stappers b. w. , frail d. a. , johnston s. , 1998b , _
apj _ , * 499 * , l69 .
gaensler b. m. , brazier k. t. s. , manchester r. n. , johnston s. , green a. j. , 1999 , _ mnras _ , * 305 * , 724 . gaensler b. m. , schulz n. s. , kaspi v. m. , pivavaroff m. j. , becker w. e. , 2003 , _ apj _ , * 588 * , 441 .
giacani e. b. , dubner g. , cappa c. , testori j. , 1998 , _ a&as _ , * 133 * , 61 .
giacani e. b. , frail d. a. , goss w. m. , vieytes m. , 2001 , _ aj _ , * 121 * , 3133 .
green a. j. , 2002 , in _ the universe at low radio frequencies _
, iau symposium 199 , eds a. pramesh rao , g. swarup & gopal krishna , asp , san francisco , p. 259 .
green a. j. , cram l. e. , large m. i. , ye t. s. , 1999 , _ apjs _ , * 122 * , 207 .
green d. a. , 1984 , _ mnras _ , * 209 * , 449 .
green d. a. , 1985 , _ mnras _ , * 216 * , 691 .
green d. a. , 1988 , _ ap&ss _ , * 148 * , 3 .
green d. a. , 1989 , _ aj _ , * 98 * , 1358 .
green d. a. , 1991 , _ pasp _ , * 103 * , 209 .
green d. a. , 1996a , in _ supernovae and supernova remnants _ ,
iau colloquium 145 , eds r. mccray & z. wang , cambridge university press , p. 341 .
green d. a. , 1996b , in _ supernovae and supernova remnants _ , iau colloquium 145 , eds r. mccray & z. wang , cambridge university press , p. 419 .
green d. a. , 2001a , in _ high energy gamma - ray astronomy _ , eds f. a. aharonian & h. j. vlk , american institute of physics , melville , new york , p. 59 .
green d. a. , 2001b , _ mnras _ , * 326 * , 283 .
green d. a. , gull s. f. , 1984 , _ nature _ , * 312 * , 527 .
helfand d. j. , chance d. , becker r. h. , white r. l. , 1984 , _ aj _ , * 89 * , 819 .
huang z. p. , thuan t. x. , chevalier r. a. , condon j. j. , yin q. f. , 1994 , _ apj _ , * 424 * , 114 .
jones d. h. , stappers b. w. , gaensler b. m. , 2002 , _ a&a _ , * 389 * , l1 . kassim n. e. , 1989 , _ apjs _ , * 71 * , 799
. kim k .-
, koo b .- c . , 2000 , _ apj _ , * 529 * , 229 .
koo b .- c .
, heiles c. , 1991 , _ apj _ , * 382 * , 204 .
koo b. , kim k. , seward f. d. , 1995 , _ apj _ , * 447 * , 211 . kothes r. , landecker t. l. , foster t. , leahy d. a. , 2001 , _ a&a _
, * 376 * , 641 .
kothes r. , uyaniker b. , yar a. , 2002 , _ apj _ , * 576 * , 169 .
kothes r. , reich w. , foster t. , byun d. , 2003 , _ apj _ , * 588 * , 852 .
kovalenko a. v. , pynzar a. v. , udaltsov v. a. , 1994 , _ astronomy reports _ , * 38 * , 95 .
landecker t. l. , pineault s. , routledge d. , vaneldik j. f. , 1989 , _ mnras _ , * 237 * , 277 . langston g. , minter a. , daddario l. , eberhart k. , koski k. , zuber j. , 2000 , _ aj _ , * 119 * , 2801 .
li z .- w . ,
wheeler j. c. , bash f. n. , jefferys w. h. , 1991 , _ apj _ , * 378 * , 93 lockhart i. a. , goss w. m. , 1978 , _ a&a _ , * 67 * , 355 . lockman f. j. , stil j. m. , 2004 , in _ milky way surveys : the structure and evolution of our galaxy _ ,
asp conference series 317 , eds d. clemens , r. shah & t. brainerd , in press ( astro - ph/0310762 ) .
lum k. s. k. , canizares c. r. , clark s. w. , coyne j. m. , markert t. h. , saez p. j. , schattenburg m. l. , winkler p. f. , 1992 , _ apjs _ , * 78 * , 423 .
mcclure - griffiths n. m. , 2002 , in _ seeing through the dust : interstellar medium in galaxies _ , asp conference series 276 , eds a. r. taylor , t. l. landecker & a. willis , asp , san francisco , p. 58
mcclure - griffiths n. m. , green a. j. , dickey j. m. , gaensler b. m. , haynes r. f. , wieringa m. h. , 2001 , _ apj _ , * 551 * , 394 .
maciejewski w. , murphy e. m. , lockman f. j. , savage b. d. , 1996 , _ apj _ , * 469 * , 238 .
matthews b. c. , wallace b. j. , taylor a. r. , 1998 , _ apj _ , * 493 * , 312 .
misanovic z. , cram l. , green a. , 2002 , _ mnras _ , * 335 * , 114 .
narayan r. , 1987 , _ apj _ , * 319 * , 162 .
normandeau m. , taylor a. r. , dewdney p. e. , basu s. , 2000 , _ aj _ , * 119 * , 2982 .
odegard n. , 1986 , _ apj _ , * 301 * , 813 . pineault s. , landecker t. l. , madore b. , gaumont - guay s. , 1993 , _ aj _ , * 105 * , 1060 .
redman m. p. , meaburn j. , bryce m. , harman d. j. , obrien t. j. , 2002 , _ mnras _ , * 336 * , 1039 .
reed j. e. , hester j. j. , fabian a. c. , winkler p. f. , 1995 , _ apj _ , * 440 * , 706 .
rees n. , 1990 , _ mnras _ , * 243 * , 637 .
reich w. , frst e. , altenhoff w. j. , reich p. , junkes n. , 1985 , _
a&a _ , * 151 * , l10 reich w. , frst e. , reich p. , junkes n. , 1988 , in _ supernova remnants and the interstellar medium _ ,
iau colloquium 101 , eds r. s. roger & t. l. landecker , cambridge university press , p. 293 . reich w. , frst e. , reich p. , reif k. , 1990 , _
a&as _ , * 85 * , 633 .
reynoso e. m. , dubner g. m. , goss w. m. , arnal e. m. , 1995 , _ aj _ , * 110 * , 318 .
rho j. , petre r. , 1998 , _ apj _ , * 503 * , l167 .
roberts d. a. , goss w. m. , kalberla p. m. w. , herbstmeier u. , schwarz u. j. , 1993 , _ a&a _ , * 274 * , 427 . routledge d. , landecker t. l. , vaneldik j. f. , 1986 , _ mnras _ , * 221 * , 809 . routledge d. , dewdney p. e. , landecker t. l. , vaneldik j. f. , 1991 , _ a&a _
, * 247 * , 529 .
safi - harb s. , harrus i. m. , petre r. , pavlov g. g. , koptsevich a. b. , sanwal d. , 2001 , _ apj _ , * 561 * , 308
. saikia d. j. , thomasson p. , roy s. , pedlar a. , muxlow t. w. b. , 2004 , _ mnras _ , in press .
saken j. m. , fesen r. a. , shull j. m. , 1992 , _ apjs _ , * 81 * , 715 . sanbonmatsu k. y. , helfand d. j. , 1992 , _ aj _ , * 104 * , 2189
. sarma a. p. , goss w. m. , green a. j. , frail d. a. , 1997 , _ apj _ , * 483 * , 335 .
schaudel d. , becker w. , aschenbach b. , trmper j. , reich w. , weisskopf m. , 2002 , in _ neutron stars , pulsars and supernova remnants _ , eds w. becker , h. lesch & j. trmper , mpe report 278 , p. 26 .
seward f. d. , 1990 , _ apjs _ , * 73 * , 781 . sollerman j. , ghavamian p. , lundqvist p. , smith r. c. , 2003 , _ a&a _ ,
* 407 * , 249 .
sramek r. a. , cowan j. j. , roberts d. a. , goss w. m. , ekers r. d. , 1992 , _ aj _ , * 104 * , 704 .
stephenson f. r. , green d. a. , 2002 , _ historical supernovae and their remnants _ , oxford university press .
stil j. m. , irwin j. a. , 2001 , _ apj _ , * 563 * , 816 .
stothers r. , frogel j. a. , 1974 , _ aj _ , * 79 * , 456 .
strom r. g. , 1994 , _
a&a _ , * 288 * , l1 .
tatematsu k. , fukui y. , landecker t. l. , roger r. s. , 1990 , _
a&a _ , * 237 * , 189 .
taylor a. r. , gibson s. j. , peracaula m. , martin p. g. ; landecker t. l. , brunt c. m. , dewdney p. e. , dougherty s. m. , gray a. d. , higgs l. a. , kerton c. r. , knee l. b. g. , kothes r. , purton c. r. , uyaniker b. , wallace b. j. , willis a. g. , durand d. , 2003 , _ aj _ , * 125 * , 3145 .
trimble v. , 1973 , _ pasp _ , * 85 * , 579 .
trushkin s. a. , 1998 , _ bull .
special astrophys .
_ , * 46 * , 62 . uyaniker b. , kothes r. , 2002 , _ apj _ , * 574 * , 805 . uyaniker b. , kothes r. , brunt c. m. , 2002 , _ apj _ , * 565 * ,
van den bergh s. , 1988a , _ pasp _ , * 100 * , 205 . van den bergh s. , 1988b , in _ supernova shells and their birth events _ , ed .
w. kundt , springer - verlag , berlin , p. 44 .
velzquez p. f. , dubner g. m. , goss w. m. , green a. j. , 2002 , _ aj _ , * 124 * , 2145 . welsh b. y. , sallmen s. , 2003 , _ a&a _ , * 408 * , 545 . whiteoak j. b. z. , green a. j. , 1996 , _ a&as _ , * 118 * , 329 .
winkler p. f. , gupta g. , long k. s. , 2003 , _ apj _ , * 585 * , 324 .
woermann b. , gaylard m. j. , otrupcek r. , 2001 , _ mnras _ , * 325 * , 1213 .
yar - uyaniker a. , uyaniker b. , kothes r. , 2004 , _ apj _ , in press ( astro - ph/0408386 ) .
this appendix presents a catalogue of galactic supernova remnants .
this catalogue is an updated version of those presented in detail in green ( 1984 , 1988 ) and in summary form in green ( 1991 , 1996b ) and stephenson & green ( 2002 ) .
detailed versions of this catalogue have been made available on the world - wide - web since 1993 november ( with subsequent versions of 1995 july , 1996 august , 1998 september , 2000 august , 2001 december and 2004 january ) .
this , the 2004 january version of the catalogue , contains 231 snrs .
the detailed version of this catalogue is available at + * * galactic coordinates * of the source centroid , quoted to the nearest tenth of a degree as is conventional .
( note : in this catalogue additional leading zeros are not used . ) * * right ascension * and * declination * of the source centroid .
the accuracy of the quoted values depends on the size of the remnant , for small remnants they are to the nearest few seconds of time and the nearest minute of arc respectively , whereas for larger remnants they are rounded to coarser values , but are in every case sufficient to specify a point within the boundary of the remnant .
these coordinates are almost always deduced from radio maps rather than from x - ray or optical observations , and are for j2000.0 . *
* angular size * of the remnant , in arcminutes , usually taken from the highest resolution radio map available .
the boundary of most remnants approximates reasonably well to a circle or an ellipse , a single value is quoted for the angular size of the more nearly circular remnants , which is the diameter of a circle with an area equal to that of the remnant , but for elongated remnants the product of two values is quoted , and these are the major and minor axes of the remnant boundary modelled as an ellipse . in a few cases an ellipse is not a satisfactory description of the boundary of the object ( refer to the description of the individual object given in its catalogue entry ) , although an angular size is still quoted for information . for ` filled - centre ' remnants
the size quoted is for the largest extent of the observed radio emission , not , as at times has been used by others , the half - width of the centrally brightened peak . * * type * of the snr : ` s ' or ` f ' if the remnant shows a ` shell ' or ` filled - centre ' structure , or ` c ' if it shows ` composite ' ( or ` combination ' ) radio structure with a combination of shell and filled - centre characteristics ; or ` s ? ' , ` f ? ' or ` c ? ' , respectively , if there is some uncertainty , or ` ? ' in several cases where an object is conventionally regarded as an snr even though its nature is poorly known or not well understood .
( note : the term ` composite ' has been used in a different sense by some authors , to describe snrs with shell radio and centrally - brightened x - ray morphologies .
an alternative term used to describe such remnants is ` mixed morphology ' , see rho & petre 1998 . ) * * flux density * of the remnant at 1 ghz in jansky .
this is _ not _ a measured value , but is deduced from the observed radio frequency spectrum of the source . the frequency of 1 ghz is chosen because flux density measurements at frequencies both above and below this value are usually available . * * spectral index * of the integrated radio emission from the remnant , @xmath89 ( here defined in the sense , @xmath90 , where @xmath67 is the flux density at a frequency @xmath91 ) , either a value that is quoted in the literature , or one deduced from the available integrated flux densities of the remnant . for several snrs a simple power law is not adequate to describe their radio spectra , either because there is evidence that the integrated spectrum is curved or the spectral index varies across the face of the remnant . in these cases
the spectral index is given as ` varies ' ( refer to the description of the remnant and appropriate references in the detailed catalogue entry for more information ) .
in some cases , for example where the remnant is highly confused with thermal emission , the spectral index is given as ` ? ' since no value can be deduced with any confidence .
* * other names * that are commonly used for the remnant .
these are given in parentheses if the remnant is only a part of the source .
for some remnants , notably the crab nebula , not all common names are given . in the detailed listings , available on the world - wide - web ,
notes on a variety of topics are given for each remnant .
first , it is noted if other galactic coordinates have at times been used to label it ( usually before good observations have revealed the full extent of the object ) , if the snr is thought to be the remnant of a historical sn , or if the nature of the source as an snr has been questioned ( in which case an appropriate reference is usually given later in the entry ) .
brief descriptions of the remnant from the available radio , optical and x - ray observations as applicable are then given , together with notes on available distance determinations , and any point sources or pulsars in or near the object ( although they may not necessarily be related to the remnant ) .
finally , appropriate references to observations are given for each remnant , complete with journal , volume , page , and a short description of what information each paper contains ( for radio observations these include the telescopes used , the observing frequencies and resolutions , together with any flux density determinations ) .
these references are _ not _ complete , but cover representative and recent observations of the remnant up to the end of 2003 and they should themselves include references to earlier work . the references do not generally include large observational surveys of particular interest in this respect are : the effelsberg 100-m survey at 2.7 ghz of the galactic plane @xmath92 , @xmath2 by reich et al .
( 1990 ) and frst et al .
( 1990 ) , reviews of the radio spectra of some snrs by kassim ( 1989 ) , kovalenko , pynzar & udaltsov ( 1994 ) and trushkin ( 1998 ) , the parkes 64-m survey at 2.4 ghz of the galactic plane @xmath93 , @xmath2 by duncan et al .
( 1995 ) and duncan et al.(1997 ) , the molonglo galactic plane survey at 843 mhz of @xmath94 , @xmath4 by green et al .
( 1999 ) , the survey of @xmath95 , @xmath96 at 8.35 and 14.35 ghz by langston et al .
( 2000 ) , reviews of _ einstein _ x - ray imaging and spectroscopic observations of galactic snrs by seward ( 1990 ) and lum et al .
( 1992 ) respectively , surveys of _ iras _ observations of snrs and their immediate surroundings by arendt ( 1989 ) and by saken , fesen & shull ( 1992 ) , the survey of emission towards snrs by koo & heiles ( 1991 ) , and the catalogue by fesen & hurford ( 1996 ) of uv / optical / infra - red lines identified in snrs .
the detailed version of the catalogue also including notes on the objects no longer thought to be snrs , and on many possible and probable remnants in the literature .
it should also be noted that : ( i ) some radio continuum and loops in the galactic plane ( e.g. berkhuijsen 1973 ) may be parts of very large , old snrs , but they have not been included in the catalogue ( see also combi et al .
1995 ; maciejewski et al . 1996
; kim & koo 2000 ; normandeau et al . 2000 ; woermann , gaylard & otrupcek 2001 ; stil & irwin 2001 ; uyaniker & kothes 2002 ) , ( ii ) the distinction between filled - centre remnants and pulsar wind nebula is not clear , and isolated , generally faint , pulsar wind nebulae are also not included in the catalogue ( e.g. gaensler et al .
1998b ; giacani et al .
2001 ; jones , stappers & gaensler 2002 ; braje et al . 2002 ;
gaensler et al . 2003 ) .
|
a catalogue of 231 galactic supernova remnants ( snrs ) is presented , and the selection effects applicable to the identification of remnants at radio wavelengths are discussed .
in addition to missing low surface brightness remnants , small angular size
i.e. young but distant remnants are also missing from the current catalogue of galactic snrs .
several statistical properties of galactic snrs are discussed , including the surface - brightness / diameter ( @xmath0 ) relation .
it is concluded that the wide range of intrinsic properties of galactic remnants with known distances , together with the observational selection effects , means that use of the @xmath0 relation to derive diameters and hence distances for individual snrs , or for statistical studies , is highly uncertain . the observed distribution of bright snrs , which are thought to be largely free from selection effects ,
is also used to derive a simple model for the distribution of snrs with galactocentric radius .
[ firstpage ] supernova remnants catalogues radio continuum : ism galaxy : structure ism : general
|
in the last few decades it has become apparent that many problems in information theory have analogies to certain problems in the area of statistical physics of disordered systems .
such analogies are useful because physical insights , as well as statistical mechanical tools and analysis techniques can be harnessed in order to advance the knowledge and the understanding with regard to the information
theoretic problem under discussion .
one important example of such an analogy is between the statistical physics of disordered magnetic materials , a.k.a .
spin glasses , and the behavior of certain ensembles of random codes for source coding ( see , e.g. , @xcite , @xcite , @xcite , @xcite ) and for channel coding ( see , e.g. , @xcite and references therein , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite ) . among the various models of interaction disorder in spin glasses , one of the most fascinating models is the _ random energy model _ ( rem ) , invented by derrida in the early eighties @xcite , @xcite , @xcite ( see also , e.g. , @xcite , @xcite , @xcite , for later developments ) .
the rem is on the one hand , extremely simple and easy to analyze , and on the other hand , rich enough to exhibit phase transitions .
according to the rem , the different spin configurations are distributed according to the boltzmann distribution , namely , their probabilities are proportional to an exponential function of their negative energies , but the configuration energies themselves are i.i.d . random variables , hence the name random energy model .. ] in ( * ? ? ? * chap .
6 ) , mzard and montanari draw an interesting analogy between the rem and the statistical physics pertaining to _ finite temperature decoding
_ @xcite of ensembles of random block codes .
the relevance of the rem here is due to the fact that in this context , the partition function that naturally arises has the log likelihood function ( of the channel output given the input codeword ) as its energy function ( hamiltonian ) , and since the codewords are selected at random , then the induced energy levels are random variables .
consequently , the phase transitions of the rem are ` inherited ' by ensembles of random block codes , as is shown in @xcite . in @xcite , this subject
was further studied and the free energies corresponding to the various phases were related to random coding exponents of the probability of error at rates below capacity and to the probability of correct decoding at rates above capacity .
while the rem is a very simple and interesting model for capturing disorder , as described above , it is not quite faithful for the description of a real physical system .
the reason is that according to the rem , any two distinct spin configurations , no matter how similar and close to each other , have independent , and hence unrelated , energies .
a more realistic model must take into account the geometry and the structure of the physical system and thus allow dependencies between energies associated with closely related configurations .
this observation has motivated derrida to develop the _ generalized random energy model _ ( grem )
@xcite ( see also , e.g. , @xcite , @xcite , @xcite , @xcite , @xcite , @xcite , for later related work ) .
the grem extends the rem in that it introduces an hierarchical structure in the form of a tree , by grouping subsets of ( neighboring ) spin configurations in several levels , where the leaves of this tree correspond to the various configurations . according to the grem , for every branch in this tree
, there is an associated independent randomly chosen energy component .
the total energy of each configuration is then the sum of these energy components along the branches that form the path from the root of the tree to the leaf corresponding to this configuration .
this way , the degree of dependency between the energies of two different configurations depends on the ` distance ' between them on the tree : more precisely , it depends on the number of common branches shared by their paths from the root up to the node at which their paths split .
the grem is somewhat more complicated to analyze than the rem , but not substantially so .
it turns out that the number of phase transitions in the grem depends on the parameters of the model .
if the tree has @xmath0 levels , there can be up to @xmath0 phase transitions , but there can also be a smaller number .
for example , in the case @xmath1 , under a certain condition , there is only one phase transition and the behavior of the free energy in both phases is just like in the ordinary rem . in analogy to the above described relationship between the rem and the statistical physics of random block codes , the natural question that now arises
is whether the grem and its phase transitions can give us some insights about the behavior of code ensembles with some hierarchical structure ( e.g. , tree
structured codes , successive refinement codes , etc . ) .
in particular , in what way do these phase transitions guide us in the choice of the design parameters of these codes ?
it is the purpose of this paper to explore these questions and to give at least some partial answers .
we demonstrate that there is indeed an intimate relationship between the grem and certain ensembles of hierarchical codes .
consider , for example , a two stage rate
distortion code of block length @xmath2 , where the first @xmath3 components of the reproduction vector , at rate @xmath4 , depend only on the first @xmath5 bits of the compressed bitstream , and the last @xmath6 symbols of the reproduction codeword , at rate @xmath7 , depend on the entire bitstream of length @xmath8 .
the overall rate of this code is , of course , the weighted average of @xmath4 and @xmath7 with weights proportional to @xmath3 and @xmath6 , respectively .
an ensemble of codes with this structure is defined as follows : first , we randomly draw a rate @xmath4 codebook of block length @xmath3 according to some distribution .
then , for each resulting codeword of length @xmath3 , we randomly draw a rate @xmath7 codebook of block length @xmath6 . and
@xmath6 , operating independently . ]
thus , the code has a tree structure with two levels , like a two level grem .
the overall distortion of the code along the entire @xmath9 symbols is the sum of partial distortions along the two segments , in analogy to the above described additivity of the partial energies along the branches of the tree pertaining to the grem , and since the codewords are random , then so are the distortions they induce .
the motivation for this class of codes , especially when the idea is generalized from two parts to a larger number of @xmath0 parts , say , of equal length ( @xmath10 ) , is that the delay , at least at the decoder , is reduced from @xmath9 to @xmath11 , because the decoder is causal in the level of segments of length @xmath11 .
the following questions now arise : is there any inherent penalty , in terms of performance , for this ensemble of reduced delay decoding codes ?
if so , how can we minimize this penalty ? if not , how should we choose the design parameters ( i.e. , @xmath12 and @xmath13 , @xmath14 , for a given overall average rate @xmath15 ) such that this code will ` behave ' like a full block code of length @xmath9 ? for simplicity ,
let us return to the case @xmath1 .
for a given @xmath15 and @xmath9 , we have two degrees of freedom : the choices of @xmath4 and @xmath3 ( which will then dictate @xmath7 and @xmath6 ) .
is it better to choose @xmath16 or @xmath17 , if at all it makes any difference ?
a similar question can be asked concerning @xmath3 and @xmath6 .
the answer depends , of course , on our figure of merit .
obviously , if one is interested only in the asymptotic distortion , the question becomes uninteresting , because then by choosing two independent codes for the two parts , both at rate @xmath15 , the overall distortion will be given by the distortion rate function , @xmath18 , just like that of the full unstructured code . for a given @xmath9 , of course
, the redundancies will correspond to the shorter blocks @xmath3 and @xmath6 , but this is a second order effect . here , we choose to examine performance in terms of the characteristic function of the overall distortion , @xmath19 $ ] .
this is , of course , a much more informative figure of merit than the average distortion , because in principle , it gives information on the entire probability distribution of the distortion .
in particular , it generates all the moments of the distortion by taking derivatives , and it is useful in deriving chernoff bounds on probabilities of large deviations events concerning the distortion . in the context of the analogy with statistical physics and the grem , this characteristic function can easily be related to the partition function whose hamiltonian is given by the distortion .
it turns out that the characteristic function of the distortion behaves in a rather surprisingly interesting manner and with a direct relation to the grem . for @xmath20 ,
when the corresponding grem has @xmath1 phase transitions , the characteristic function of the distortion behaves like that of two independent block codes of lengths @xmath3 and @xmath6 and rates @xmath4 and @xmath7 , thus the dependency between the two parts of the code is not exploited in terms of performance . for @xmath16 , which is the case where the analogous grem has
only one phase transition ( and behaves exactly like the ordinary rem , which is parallel to an ordinary random block code with no structure ) , the characteristic function behaves like that of a full unstructured optimum block code at rate @xmath15 across a certain interval of small @xmath21 , but beyond a certain point , it becomes inferior to that of a full code . for @xmath22 , it behaves like the unstructured code for the _ entire _ range of @xmath23 , but then one might as well use two independent block codes ( and reduce the search complexity at the encoder from @xmath24 to @xmath25 ) .
the choices of @xmath3 and @xmath6 are immaterial in that sense , as long as they both grow linearly with @xmath9 .
thus , the conclusion is that it is best to use @xmath26 , but if communication protocol constraints dictate different rates at different segments , then performance is better when @xmath16 than when @xmath20 .
these results can be extended to the case of @xmath0 stages .
a parallel analysis can be applied to analogous ensembles of ( reduced delay ) channel encoders of block length @xmath2 ( for the case @xmath1 ) , which have a similar tree structure : here , the first @xmath3 channel letters of each block depend only on the first @xmath5 information bits , whereas the other @xmath6 channel symbols depend on the entire information vector of length @xmath8 .
the random codebook is again drawn hierarchically in the same manner as before . if the code performance is judged in terms of the error exponent , then once again , the choice @xmath27 is always better than the choice @xmath28 . here , unlike the source coding problem
, there is an additional consideration : there are two types of incorrect codewords that are competing with the correct one in the decoding process : those for which the first @xmath3 channel inputs agree with those of the correct codeword ( the first segment is the same ) and those for which this is not the case . in this case
, @xmath7 has to be chosen sufficiently small so that the error term contributed by erroneous codewords of the first kind would not dominate the probability of error . considering the case
@xmath29 , if the overall average rate is not too small , it is possible to choose @xmath4 and @xmath7 so that the error exponent of this ensemble of codes is not worse than that of an ordinary random code with no structure .
this idea can be extended to @xmath0 stages in a straightforward manner .
in fact , we propose a systematic procedure to allocate rates to the different stages in a way that guarantees that the error exponent would be at least as good as that of the classical random coding error exponent pertaining to an ordinary random code at rate @xmath15 .
the outline of this paper is as follows . in section 2
, a few notation conventions are described . in section 3
, we provide some more detailed background in statistical physics , with emphasis on the rem and the grem .
finally , in section 4 , we present our main results on hierarchical code ensembles of the type described above , along with their relationship to the grem .
readers who are not interested in the relationship with statistical physics ( although this is one of the main points in the paper ) may skip section 3 and ignore , in section 4 , the comments on the statistical mechanical aspects , all this without essential loss of continuity .
throughout this paper , scalar random variables ( rv s ) will be denoted by capital letters , like @xmath30 and @xmath31 , their sample values will be denoted by the respective lower case letters , and their alphabets will be denoted by the respective calligraphic letters .
a similar convention will apply to random vectors and their sample values , which will be denoted with the same symbols in the boldface font .
thus , for example , @xmath32 will denote a random @xmath9-vector @xmath33 , and @xmath34 is a specific vector value in @xmath35 , the @xmath9-th cartesian power of @xmath36 .
sources and channels will be denoted generically by the letters @xmath37 and @xmath38 .
specific letter probabilities corresponding to a source @xmath38 will be denoted by the corresponding lower case letters , e.g. , @xmath39 is the probability of a letter @xmath40 .
a similar convention will be applied to the channel @xmath37 and the corresponding transition probabilities , @xmath41 , @xmath40 , @xmath42 .
the expectation operator will be denoted by @xmath43 .
the cardinality of a finite set @xmath44 will be denoted by @xmath45 . for two positive sequences @xmath46 and @xmath47
, the notation @xmath48 means that @xmath49 and @xmath50 are asymptotically of the same exponential order , that is , @xmath51 .
similarly , @xmath52 means that @xmath53 , etc .
information theoretic quantities like entropies and mutual informations will be denoted following the usual conventions of the information theory literature .
in this section , we provide some basic background in statistical physics , focusing primarily on the rem , along with its relevance to ordinary ensembles of source and channel block codes , and then we extend the scope to the grem . consider a physical system with a large number @xmath9 of particles , which can be in a variety of ` microstates ' pertaining to the various combinations of the microscopic physical states ( characterized by position , momentum , spin , etc . )
that these particles may have .
for each such microstate of the system , which we shall designate by a vector @xmath54 , there is an associated energy , given by an energy function ( hamiltonian ) @xmath55 .
one of the most fundamental results in statistical physics ( based on the law of energy conservation and the basic postulate that all microstates of the same energy level are equiprobable ) is that when the system is in equilibrium , the probability of a microstate @xmath54 is given by the boltzmann distribution @xmath56 where @xmath57 is the inverse temperature , that is , @xmath58 , @xmath59 being temperature , , where @xmath0 is boltzmann s constant , but following the common abuse of the notation , we redefine @xmath60 as temperature ( in units of energy ) . ] and @xmath61 is the normalization constant , called the _ partition function _ , which is given by @xmath62 or @xmath63 depending on whether @xmath54 is discrete or continuous .
the role of the partition function is by far deeper than just being a normalization factor , as it is actually the key quantity from which many macroscopic physical quantities can be derived , for example , the free energy is @xmath64 , the average internal energy ( i.e. , the expectation of @xmath55 where @xmath54 drawn is according ( [ bd ] ) ) is given by the negative derivative of @xmath65 , the heat capacity is obtained from the second derivative , etc .
one of the important examples of such a multi particle physical system is that of a magnetic material , in which each molecule has a magnetic moment , a three dimensional vector which tends to align with the magnetic field felt by that molecule .
in addition to the influence of a possible external magnetic field , there is also an effect of mutual interactions between the magnetic moments of various ( neighboring ) molecules .
quantum mechanical considerations dictate that the set of possible configurations of each magnetic moment ( spin ) is discrete : in the simplest case , it has only two possible values , which we shall designate by @xmath66 ( spin up ) and @xmath67 ( spin down ) .
thus , a spin configuration , i.e. , the vector of spins of @xmath9 molecules , is designated by a binary vector @xmath68 , where each component @xmath69 takes values in @xmath70 according to the spin of the @xmath71th molecule , @xmath72 .
when the spins of a certain magnetic material tend to align in the same direction , the material is called _ ferromagnetic _ , and a customary model of the hamiltonian , the _ ising model _ ,
is given by @xmath73 where the in first term , pertaining to the interaction , @xmath74 describes the intensity of the interaction with the summation being defined over pairs of neighboring spins ( depending on the geometry of the problem ) , and the second term is associated with an external magnetic field ( proportional to ) @xmath75 . when @xmath76 , the material is _
antiferromagnetic _ ,
namely , neighboring spins ` prefer ' to be antiparallel .
more general models allow interactions not only with immediate neighbors , but also more distant ones , and then there are different strengths of interaction , depending on the distance between the two spins . in this case , the first term is replaced , by the more general form @xmath77 , where now the sum can be defined over all possible pairs @xmath78 . here , in addition to the ferromagnetic case , where all @xmath79 , and the antiferromagnetic case , where all @xmath80 , there is also a situation where some @xmath81 are positive and others are negative , which is the case if a _
spin glass_. here , not all spin pairs can be in their preferred mutual position ( parallel / antiparallel ) , thus the system may be _ frustrated . _ to model situations of disorder , it is common to model @xmath81 as random variables ( rv s ) with , say , equal probabilities of being positive or negative . for example , in the edwards
anderson ( ea ) model @xcite , @xmath81 are taken to be i.i.d .
zero mean gaussian rv s when @xmath71 and @xmath82 are neighbors and zero otherwise .
in the sherrington kirkpatrick ( sk ) model @xcite , all @xmath83 are i.i.d .
zero mean gaussian rv s .
thus , the system has two levels of randomness : the randomness of the interaction coefficients and the randomness of the spin configuration given the interaction coefficients , according to the boltzmann distribution .
however , the two sets of rv s are normally treated differently .
the random coefficients are considered _ quenched _
rv s in the terminology of physicists , namely , they are considered fixed in the time scale at which the spin configuration may vary .
this is analogous to the situation of coded communication in a random coding paradigm : a randomly drawn code should normally be thought of as a quenched entity , as opposed to the randomness of the source and/or the channel . in @xcite,@xcite,@xcite , derrida took the above described idea of randomizing the ( parameters of the ) hamiltonian to an extreme , and suggested a model of spin glass with disorder under which the energy levels @xmath84 are simply i.i.d .
rv s , without any structure in the form of ( [ ham ] ) or its above described extensions .
in particular , in the absence of a magnetic field , the @xmath85 rv s @xmath84 are taken to be zero
mean gaussian rv s , all with variance @xmath86 , where @xmath87 is a parameter . to match the behavior of the hamiltonian ( [ ham ] ) with a limited number of interacting neighbors and random interaction parameters , which has a number of independent terms that is linear in @xmath9 . ]
the beauty of the rem is in that on the one hand , it is very easy to analyze , and on the other hand , it consists of sufficient richness to exhibit phase transitions .
the basic observation about the rem is that for a typical realization of the configurational energies @xmath84 , the number of configurations with energy about @xmath88 ( i.e. , between @xmath88 and @xmath89 ) , @xmath90 , is proportional ( up to sub exponential terms in @xmath9 ) to @xmath91 , as long as @xmath92 , whereas energy levels outside this range are typically not populated by spin configurations ( @xmath93 ) , as the probability of having at least one configuration with such an energy decays exponentially with @xmath9 .
thus , the asymptotic ( thermodynamical ) entropy per spin , which is defined by @xmath94 is given by @xmath95 the partition function of a typical realization of a rem spin glass is then @xmath96 whose exponential growth rate , @xmath97 behaves according to @xmath98\nonumber\\ & = & \max_{|e|\le e_0}\left[\ln 2- \left(\frac{e}{nj}\right)^2-\beta j \cdot\left(\frac{e}{nj}\right)\right].\end{aligned}\ ] ] solving this simple optimization problem , we find that @xmath99 is given by @xmath100 which means that the asymptotic free energy per spin , a.k.a . the _ free energy density _ , which is obtained by @xmath101 is given by ( cf .
* proposition 5.2 ) ) : @xmath102 thus , the free energy density is subjected to a phase transition at the inverse temperature @xmath103 . at high temperatures ( @xmath104 ) , which is referred to as the _ paramagnetic phase _ , the partition function is dominated by an exponential number of configurations with energy @xmath105 and the entropy grows linearly with @xmath9 .
when the system is cooled to @xmath106 and beyond , which is the _ glassy phase _ , the system freezes but it is still in disorder the partition function is dominated by a subexponential number of configurations of minimum energy @xmath107 .
the entropy , in this case , grows sublinearly with @xmath9 , namely the entropy per spin vanishes , and the free energy density no longer depends on @xmath57 .
further details about the rem can be found in @xcite and the references mentioned in the introduction .
as described in @xcite , there is an interesting analogy between the rem and the partition function pertaining to _ finite temperature decoding _
@xcite of ensembles of channel block codes ( see also @xcite ) . in particular , consider a codebook @xmath108 of @xmath109 binary codewords of length @xmath9 , @xmath110 , to be used across a binary symmetric channel ( bsc ) with crossover probability @xmath111 . given a binary vector @xmath112 at the channel output ,
consider the generalized posterior parametrized by @xmath57 : @xmath113 where @xmath114 , @xmath115 is the hamming distance between @xmath54 and @xmath112 , and where the real posterior is obtained , of course , for @xmath116 .
this is identified as a boltzmann distribution whose energy function ( which depends on the given @xmath112 ) is @xmath117 .
as described in @xcite and @xcite , there are a few motivations for introducing the temperature parameter @xmath57 here .
first , it allows a degree of freedom in case there is some uncertainty regarding the channel noise level ( small @xmath57 corresponds to high noise level ) .
second , it is inspired by the ideas behind simulated annealing techniques : by sampling from @xmath118 while gradually increasing @xmath57 ( cooling the system ) , the minima of the energy function ( ground states ) can be found .
third , by applying symbolwise map decoding , i.e. , decoding the @xmath119th symbol of @xmath54 as @xmath120 , where @xmath121 we obtain a family of _ finite temperature decoders _ parametrized by @xmath57 , where @xmath116 corresponds to minimum symbol error probability ( with respect to the true channel ) and @xmath122 corresponds to minimum block error probability . as in @xcite , we will distinguish between two contributions of @xmath123 : one is @xmath124 , where @xmath125 is the actual codeword transmitted , and the other is @xmath126 , pertaining to all incorrect codewords .
the former is typically about @xmath127 since @xmath128 concentrates about @xmath129 .
we next focus on the behavior of @xmath130 . to this end , consider a random selection of the code @xmath108 , where every bit of every codeword is drawn by an independent fair coin tossing . for a given @xmath112 ,
the energy levels @xmath131 pertaining to all incorrect codewords are rv s ( exactly like in the rem ) because of the random selection of these codewords .
now , the total number of correct codewords is about @xmath24 , and the probability that a randomly chosen @xmath54 would fall at distance @xmath132 from @xmath112 is exponentially @xmath133}$ ] , where @xmath134 then the typical number of codewords at normalized distance @xmath135 is about @xmath136}\ ] ] as long as @xmath137 and @xmath138 when @xmath139 .
thus , letting @xmath140 denote the small solution to the equation @xmath141 ( the gilbert
varshamov distance ) , we find that , with a clear analogy to the rem , the corresponding thermodynamical entropy is given by @xmath142 accordingly , the partition function @xmath130 of a typical code is given by @xmath143}\cdot e^{-\beta bn\delta}{\stackrel{\cdot } { = } } \exp\{n[r-\ln 2+\max_{\delta(r)\le\delta\le 1-\delta(r)}(h(\delta)- \beta b\delta)]\},\ ] ] and the free energy density pertaining to @xmath144 behaves according to @xmath145 where @xmath146 and @xmath147}{b},\ ] ] and where , again , the first line of @xmath148 corresponds to the paramagnetic phase with exponentially many codewords at distance ( energy ) @xmath149 from @xmath112 , and the second line is the glassy phase with subexponentially many codewords at distance @xmath150 . in @xcite ,
these free energies are related to random coding exponents as mentioned in the introduction . by the same token , in rate
distortion source coding , if one defines the partition function as @xmath151 with @xmath54 being the source vector , @xmath152 being the reproduction codevectors , and @xmath115 being the hamming distortion measure , then the same analysis takes place . in the sequel
, we will motivate this definition of the partition function of rate
distortion coding and use it .
as we have seen , the rem is an extremely simple model to analyze , but its simplicity is also recognized as a drawback from the aspect of faithfully modeling a spin glass .
the reason for this is the lack of structure which is needed to allow dependencies between energy levels of spin configurations that are closely related : for example , if @xmath54 and @xmath153 differ only in a single component , it is conceivable that the respective energies would be close , as suggested by ( [ ham ] ) . to this end , as described in the introduction , derrida proposed a generalized version of the rem the grem , which introduces dependencies between configurational energies in an hierarchical fashion .
we next briefly review the grem .
a grem with @xmath0 levels can best be thought of as a tree with @xmath85 leaves and depth @xmath0 , where each leaf represents one spin configuration .
this tree is defined by @xmath0 positive parameters , @xmath154 , which are all in the interval @xmath155 , and whose product , @xmath156 , equals @xmath157 .
the construction of this tree is as follows : the root of the tree is connected to @xmath158 distinct nodes , by integers . ] which will be referred to as first
level nodes .
each first level node is in turn connected to @xmath159 distinct second level nodes , thus a total of @xmath160 second level nodes . in the case
@xmath1 , these second level nodes are the leaves of the tree and @xmath161 . if @xmath162 , the process continues , and each second level node is connected to @xmath163 third level nodes , and so on . at the last step , each one of the @xmath164 nodes at level @xmath165
is connected to @xmath166 distinct leaves , thus a total of @xmath167 leaves .
the rem corresponds to the degenerate special case where @xmath168 .
the random selection of energy levels for the grem is defined by another set of @xmath0 parameters , @xmath169 , which are all positive reals that sum to unity .
the random selection is carried out in the following manner : for each one of the @xmath170 branches emanating from @xmath171th level nodes and connecting them to @xmath71th level nodes ( @xmath172 ) in the tree , we randomly choose an independent rv , henceforth referred to as a _
branch energy _ , which is a zero mean , gaussian rv with variance @xmath173 , where @xmath87 is like in the rem and where @xmath174 are as described above .
finally , the energy level of a given configuration is given by the sum of branch energies along the path from the root to the leaf that represents this configuration .
thus , the total energy , is the sum of @xmath0 independent zero mean gaussian rv s with variances @xmath173 , and so , it is zero mean gaussian rv with variance @xmath86 , exactly like in the rem .
however , now the energy levels of different configurations may be clearly correlated if the paths from the root to their corresponding leaves share some common branches before they split .
the degree of statistical dependence is according to their distance along the tree .
for example , if two configurations are first degree siblings , i.e. , they share the same parent node at level @xmath165 , then all their energy components are the same except their last branch energies , which are independent . on the other extreme , if their paths are completely distinct , then their energies are independent .
the grem for @xmath1 is analyzed in @xcite .
we next present the derivation for this case ( with a few more details than in @xcite ) .
let @xmath175 and @xmath176 be positive numbers whose product equals @xmath157 , and let @xmath177 and @xmath178 be positive numbers whose sum equals @xmath179 .
now , every configuration with energy @xmath88 has some first level branch energy @xmath180 and second level branch energy @xmath181 . for a typical realization of this grem ,
the number of first level branches with energy about @xmath180 is exponentially @xmath182\right\},\ ] ] provided that the expression in the square brackets is non negative , i.e. , @xmath183 , and @xmath184 otherwise .
therefore , the number of configurations with total energy about @xmath88 is exponentially @xmath185\right\},\ ] ] whose exponential rate ( the entropy per spin ) is given by @xmath186.\ ] ] note that @xmath187 is an even function , non increasing in @xmath188 , and it should be kept in mind that beyond the value of @xmath188 at which @xmath187 vanishes , denote it by @xmath189 , we have @xmath190 since @xmath191 is typically zero ( as was the case with the rem ) .
we shall get back to this point shortly , but for a moment , let us ignore it and solve the maximization problem pertaining to the above expression of @xmath187 , as is . denoting the resulting maximum by @xmath192 ( to distinguish from @xmath187 , where @xmath189 and the jump to @xmath193 are taken into account ) , we get : @xmath194 where @xmath195 .
taking now into account the above mentioned observation concerning the criticality of the point @xmath196 , we have to distinguish between two cases .
the first is the case where @xmath197 , namely , the first line of the above expression of @xmath192 vanishes for @xmath188 smaller than @xmath198 .
the first line vanishes for @xmath199 , so the condition for this case to hold is @xmath200 , or equivalently , @xmath201 . in this case
, we then have : @xmath202 which is exactly the same behavior as in the ordinary rem ( @xmath168 ) .
consequently , the exponential rate of the partition function , which is given by @xmath203,\ ] ] is also the same as in the rem , namely , @xmath204 where @xmath205 is the above defined critical inverse temperature of the rem ( see subsection [ rem ] ) .
we next consider the complementary case where @xmath206 . in this case
, the expression of @xmath187 should take into account the fact that it vanishes ( and then becomes @xmath207 ) according to the second line of ( [ psibasic ] ) .
this amounts to : @xmath208 where @xmath209 . before we compute the corresponding partition function ,
we make the following observation : @xmath210 where the inequality follows from the well known inequality @xmath211/[\sum_{i=1}^m b_i]\le\max_{1\le i \le m}a_i / b_i$ ] for positive @xmath212 and @xmath213 ( * ? ? ?
* lemma 1 ) . in the same manner , using the similar inequality @xmath211/[\sum_{i=1}^m
b_i]\ge\min_{1\le i \le m}a_i / b_i$ ] , we get @xmath214 it follows then that the condition @xmath206 is equivalent to the condition @xmath215 . defining @xmath216 we then have @xmath217 .
let us examine how @xmath99 behaves as @xmath57 grows from zero to infinity . for small enough @xmath57 , the achiever of @xmath99 , call it @xmath218 , is still smaller in absolute value than @xmath219 , and then it is obtained from equating to zero the derivative of @xmath220 $ ] , with @xmath187 being according to first line of ( [ 2phasetransitions ] ) , thus @xmath221 .
this remains true as long as @xmath222 , which means @xmath223 . in this case , the partition function is dominated by @xmath224\}$ ] first
level branches with energy @xmath225 , each followed by @xmath224\}$ ] second level branches with energy @xmath226 , and this is a pure paramagnetic phase . as @xmath57 continues to grow beyond @xmath227 , but is still below @xmath228 , the partition function is dominated by a subexponential number of first
level branches of energy @xmath229 followed by @xmath224\}$ ] second level branches with energy @xmath230 .
this is a `` semi glassy '' phase , where the first
level branches are already glassy but the second level ones are still paramagnetic .
as @xmath57 exceeds @xmath228 , this becomes a pure glassy phase where the partition function is dominated by a subexponential number of first level branches with energy @xmath229 and a subexponential number of second level branches with energy @xmath231 .
accordingly , the function @xmath99 exhibits two phase transitions at inverse temperatures @xmath227 and @xmath228 : @xmath232 again , the free energy density is obtained by @xmath233 .
this different behavior of the grem for the two different cases will be pivotal to our later discussion on the parallel behavior of ensembles of codes . when there is a general number @xmath0 of levels , the above analysis of the grem becomes , of course , more complicated and there are more cases to consider , but the concepts remain the same
. there can be up to @xmath0 phase transitions , but there can be less , depending on the parameters of the model @xmath234 . for details ,
the reader is referred to @xcite,@xcite .
in analogy to the relationship between the rem and ordinary ensembles of block codes , as was described in subsection [ remc ] , it is natural to wonder about the possibility of similar relationships between the grem and more general ensembles of block codes , and to ask whether the fact that the grem exhibits different types of behavior ( as we have seen in subsection [ grem ] ) , has implications on the behavior of these ensembles of codes . since the grem is defined by an hierarchical ( tree ) structure , it is plausible to expect that if a relationship to coding exists , it will be in the context of ensembles of codes which have hierarchical structures as well .
hierarchically structured ensembles of codes are encountered in numerous applications in information theory , including block causal tree
structured source codes and channel codes of the type described informally in the introduction , successive refinement source codes @xcite,@xcite,@xcite , codes for the broadcast channel ( * ? ? ?
* chap . 15.6 ) and codes based on binning techniques ( see , e.g. , @xcite,@xcite,@xcite ) , just to name a few . in this paper
, we confine our attention to the first above mentioned class of codes .
the fact that the grem behaves , in some situations , like the rem , and the rem is analogous to an ordinary block code without any hierarchical structure ( cf .
[ remc ] ) , may hint that in the parallel situations in the realm of our coding problem , a typical code from the hierarchical ensemble will perform essentially as well as a typical ( good ) code without the hierarchical structure . in these situations then ( which can be imposed by a clever choice of certain design parameters ) , it would be interesting to explore the question whether we may enjoy the benefit that the hierarchical structure buys us ( in our case , reduced delay ) without essentially paying in terms of performance . as we show in this section
, the answer to this question turns out to be affirmative to a large extent , both in the source coding setting and in the channel coding setting .
finally , in closing this introductory part of section [ main ] , a more technical comment is in order : as in subsection [ remc ] , throughout the sequel , we confine ourselves to the memoryless binary symmetric source ( bss ) with the hamming distortion measure , in the context of source coding , and to the binary symmetric channel ( bsc ) in the context of channel coding
. the random coding distribution in both problems will be i.i.d . and
uniform , i.e. , each bit of each codeword will be drawn by independent fair coin tossing .
also , we will focus mostly on the case @xmath1 .
the reason for this is that our purpose is this paper is more to demonstrate certain concepts , and so , we prefer to slightly sacrifice generality at the benefit of simplicity , and so , better readability , and a smaller amount of space .
having said that , all the derivations can be extended to apply to more general memoryless sources , channels , and random coding distributions ( as was done in @xcite ) , as well as to a general number @xmath0 of stages .
consider the bss @xmath235 , @xmath236 ( @xmath71 positive integer ) and the hamming distortion measure between two binary @xmath9vectors @xmath54 and @xmath237 : @xmath238 where @xmath239 if @xmath240 and @xmath241 if @xmath242 , @xmath243 . before discussing ensembles of codes with hierarchical structures , let us first confine attention to an ordinary ensemble with no structure .
consider a random selection of a codebook of size @xmath109 ( @xmath15 being the coding rate in nats per source bit ) , @xmath244 , @xmath245 , @xmath246 , where each component of each codeword is drawn randomly by an independent fair coin tossing .
for a given source vector @xmath54 and for a given such randomly drawn codebook @xmath108 , let @xmath247 denote the distortion associated with encoding @xmath54 . instead of examining the expected distortion , @xmath248 , w.r.t .
both the source and the random codebook selection , as is traditionally done , we will concern ourselves with a more refined and more informative objective function , which is the characteristic function of @xmath249 , namely , @xmath250\},\ ] ] or in particular , its exponential rate @xmath251 focusing on the range @xmath23 .
as is well known , the characteristic function provides information not only on the expected distortion , @xmath248 , but also on every moment of @xmath249 ( by taking derivatives of @xmath252 at @xmath253 ) .
it is also intimately related to the tail behavior ( i.e. , large deviations probabilities ) of the distribution of @xmath249 via chernoff bounds . in order to analyze @xmath252 and then @xmath254 , first , for an ordinary ensemble , and later for an hierarchical structured ensemble , it is convenient to define , for given @xmath54 and @xmath108 , the partition function , the partition function @xmath255 induced by a typical codebook is exactly the same as in ( [ ze ] ) , with the minor modification that here @xmath57 is not scaled by @xmath75 as in ( [ ze ] ) . ]
@xmath256 the function @xmath252 is obtained from the partition function by @xmath257 in the definition of the ensemble behavior of @xmath254 , there are now two options .
the first is to think of the above defined expectation of @xmath258 as being taken w.r.t . both the source @xmath32 and the code ensemble @xmath259 , and then to define @xmath254 as above .
the second option is to define the above expectation of @xmath258 w.r.t .
the source only , while keeping @xmath108 fixed , and then to define @xmath254 as @xmath260 , where the latter expectation is across the ensemble of codebooks @xmath259 . the difference between meanings of the two approaches is in the point of view : in the former approach the randomness of both @xmath32 and @xmath108 are treated on equal grounds , and this makes sense if @xmath32 and @xmath108 vary on the same time scale ( e.g. , when the codebook varies frequently according to some secret key ) . in the parallel discussion on spin glasses ( cf .
section 3.1 ) , this is analogous to the double randomness of both the spin configuration and the interaction parameters , and in the language of statistical physicists , this is called _ annealed _ averaging .
the second approach , which physicists refer to as _ quenched _ averaging , fits better the paradigm where the code @xmath108 is held fixed over many realizations of the source @xmath32 . in the information theory literature , it is more customary to adopt an approach analogous to annealed averaging and so , we shall do the same here .
let us begin the with the calculation of the annealed version of @xmath254 , first , for a an ordinary non hierarchical code : @xmath261^{1/\theta}\right\}\nonumber\\ & = & { \mbox{\boldmath $ e$}}\left\{\left[\sum_{d=0}^nn(d)\cdot e^{-s\theta d}\right]^{1/\theta}\right\}\nonumber\\ & { \stackrel{\cdot } { = } } & { \mbox{\boldmath $ e$}}\left\{\sum_{d=0}^nn^{1/\theta}(d)\cdot e^{-sd}\right\}\nonumber\\ & = & \sum_{d=0}^n{\mbox{\boldmath $ e$}}\{n^{1/\theta}(d)\}\cdot e^{-sd}\end{aligned}\ ] ] where @xmath262 is the number of codewords whose normalized hamming distance from @xmath32 is exactly @xmath135 , and where the third ( exponential ) equality holds , even before taking the expectation , because the summation over @xmath263 consists of a _ subexponential _ number of terms , and so , both @xmath264^{1/\theta}$ ] and @xmath265 are of the same exponential order as @xmath266^{1/\theta}$ ] .
this is different from the original summation over @xmath108 which contains an _ exponential _ number of terms .
now , as is shown in subsection a.1 of the appendix ( see also @xcite ) , @xmath267 } & \delta < \delta(r)~~\mbox{or}~~ \delta > 1-\delta(r)\\ e^{n[r+h(\delta)-\ln 2]/\theta } & \delta(r)\le\delta\le 1-\delta(r ) \end{array}\right.\ ] ] where @xmath140 is defined ( cf .
subsection [ remc ] ) as the small solution to the equation @xmath141 , which is also the distortion
rate function of the bss .
this gives @xmath268}\cdot e^{-s\delta n}+ \sum_{\delta\ge\delta(r ) } e^{n[r+h(\delta)-\ln 2]/\theta}\cdot e^{-s\delta n}\nonumber\\ & { \stackrel{\delta } { = } } & a+b\end{aligned}\ ] ] now , as @xmath269 , the term @xmath75 tends to @xmath270 , which is of the exponential order of @xmath271 .
the term @xmath272 , which is independent of @xmath273 , is of the exponential order of @xmath274 , where @xmath275= \left\{\begin{array}{ll } s\delta(r ) & s\le s_r\\ v(s , r ) & s > s_r \end{array}\right.\ ] ] where @xmath276.\ ] ] and @xmath277 since @xmath278 never exceeds @xmath279 for @xmath280 , the dominant term is @xmath272 , and therefore , for the ordinary block code ensemble , we have : @xmath281 it is not difficult to show also , using sphere covering considerations , that @xmath282 is the best achievable performance in terms of the exponential rate of the characteristic function of the distortion .
the function @xmath282 is depicted qualitatively in fig .
[ gen ] .
we proceed to define the ensemble of hierarchical codes and to analyze its performance with relation to the grem .
let @xmath2 , where @xmath9 , @xmath3 and @xmath6 are positive integers . for a given @xmath4 ,
consider a random selection of a codebook of size @xmath283 , @xmath284 , @xmath285 , @xmath286 , where each component of each codeword is drawn randomly by an independent fair coin tossing .
next , given @xmath7 , for each @xmath286 , consider a similar random selection of a codebook of size @xmath287 , @xmath288 , @xmath289 , @xmath290 .
the encoder works as follows : given a source vector @xmath291 , it finds a pair of indices @xmath292 , @xmath286 , @xmath290 , such that the distortion between @xmath54 and the concatenation of the codewords @xmath293 is minimum .
the index @xmath71 is encoded by @xmath5 nats and the index @xmath82 ( given @xmath71 ) is encoded by @xmath294 nats , thus a total of @xmath295 nats , where @xmath15 is the overall rate , given by @xmath296 the decoder can , of course , generate the first stage reproduction @xmath297 based on the first @xmath5 nats received , without having to wait for the @xmath294 following ones .
the extension of this hierarchical structure to a larger number of stages @xmath0 should be obvious .
in particular , as mentioned in the introduction , if @xmath0 divides @xmath9 and the @xmath9block is divided to @xmath0 sub blocks of length @xmath11 each , then the decoder can generate chunks of the reproduction at a reduced delay of @xmath11 instead of @xmath9
. the analogy of this structure with the grem should also be obvious .
the code has a tree structure and the configurational energies of the grem play the same role as the distortion here , as the overall distortion is the cumulative sum of the per stage distortions .
also , the coding rate @xmath13 here plays the same role as @xmath298 of the grem ( @xmath299 ) .
thus , it is natural to expect that the partition function @xmath255 of this code ensemble would behave analogously to that of the grem , as we shall see next . for the sake of simplicity ,
we return to the case @xmath1 , with the understanding that our derivations can be extended without any essential difficulties to a general @xmath0 . before analyzing the characteristic function of the distortion along with its exponential rate , it is instructive to examine the partition function @xmath255 for a given @xmath54 and address the analogy with that of the grem . for a given @xmath54 and a typical code in the ensemble ,
there are @xmath300}$ ] first - stage codewords @xmath301 at distance @xmath302 from the vector formed by the first @xmath3 components of @xmath54 , provided that @xmath303 and @xmath304 otherwise .
for each one of these first stage codewords , there are @xmath305}$ ] second stage codewords @xmath306 at distance @xmath307 from the vector formed by the last @xmath6 components of @xmath54 , provided that @xmath308 .
thus , the total number of concatenated codewords @xmath309 at distance @xmath310 ( that is , @xmath311 ) from @xmath54 is given by @xmath312}\cdot e^{n_2[r_2+h((\delta-\lambda\delta_1)/(1-\lambda))-\ln 2]}\nonumber\\ & { \stackrel{\cdot } { = } } & \exp\left\{n\max_{\delta(r_1)\le\delta_1\le 1-\delta(r_1)}\left[r+\lambda h(\delta_1 ) + ( 1-\lambda)h\left(\frac{\delta-\lambda\delta_1}{1-\lambda}\right)-\ln 2\right]\right\}.\end{aligned}\ ] ] consequently , the exponential growth rate of @xmath313 is given by @xmath314.\ ] ] for large @xmath135 , the constraint @xmath315 is inactive and the achiever of @xmath316 is @xmath317 , and then @xmath318 if we now gradually reduce @xmath135 , the behavior depends on whether we first encounter the value @xmath319 , below which @xmath317 no longer satisfies the constraint , or the the value @xmath320 , below which @xmath321 vanishes .
this in turn depends on whether @xmath322 is larger or smaller than @xmath140 , or equivalently , if @xmath323 or @xmath324 .
consider the case @xmath324 first . in this case ,
@xmath325 , and we have : @xmath326 exactly like in the ordinary , non hierarchical ensemble ( cf .
( [ srem ] ) ) , and then the corresponding exponential rate of the partition function is as in subsection [ remc ] , except that here @xmath57 is not scaled by @xmath75 , i.e. , @xmath327 .
the other case is @xmath323 , which is equivalent to @xmath328 . here , in analogy to the grem with two phase transitions
, we have : @xmath329 & \beta > \beta(r_2 ) \end{array}\right.\ ] ] we now identify the first line as the purely paramagnetic phase , the second line as the `` semi glassy '' phase ( where @xmath301 are glassy but @xmath306 are paramagnetic ) , and the third line as the purely glassy phase .
note that the glassy phase here behaves as if the two parts of the code , at rates @xmath4 and @xmath7 , were operating independently , namely , as if @xmath330 were all identical , in which case , the distortion would have been minimized separately over the two segments .
we will get back to this point in the sequel .
we have seen then that the ensemble behaves substantially differently depending on whether @xmath27 or @xmath20 . in the former case
, the above calculation may indicate that the ensemble performance is similar to that of an ordinary block code of length @xmath9 without any structure .
we next carry out a detailed analysis of the characteristic function and its exponential rate , which we shall denote by @xmath331 . similarly as before
, we first compute @xmath332 : @xmath333^{1/\theta}\right\}\nonumber\\ & { \stackrel{\cdot } { = } } & \sum_{d_1=0}^{n_1}\sum_{d_2=0}^{n_2}{\mbox{\boldmath $ e$}}\{n^{1/\theta}(d_1,d_2)\}\cdot e^{-s(d_1+d_2)},\end{aligned}\ ] ] where @xmath334 is the number concatenated codewords @xmath309 for which the first stage contributes distance @xmath335 and the second stage contributes distance @xmath336 . for the moments @xmath337 , or equivalently , @xmath338 ,
the following is proven in section a.2 of the appendix : @xmath339\ } & \delta_1 \in{{\cal i}}^c(r_1),~\delta_2\in{{\cal i}}^c(r_2)\\ \exp\{n[\lambda w_1 + ( 1-\lambda)w_2/\theta]\ } & \delta_1\in{{\cal i}}^c(r_1),~\delta_2\in{{\cal i}}(r_2)\\ \exp\{n[\lambda w_1+(1-\lambda)w_2]/\theta\ } & \delta_1 \in{{\cal i}}(r_1),~\delta_2\in{{\cal i}}(r_2)\\ \exp\{n\eta[\lambda w_1+(1-\lambda)w_2]\ } & \delta_1 \in{{\cal i}}(r_1),~\delta_2\in{{\cal i}}^c(r_2 ) \end{array}\right.\ ] ] where @xmath340 , @xmath341\setminus{{\cal i}}(r)$ ] , @xmath342 , @xmath299 , with @xmath343 being defined as @xmath344 and @xmath345 therefore , @xmath346}\times\nonumber\\ & & e^{-sn[\lambda\delta_1+(1-\lambda)\delta_2]}+\nonumber\\ & & \sum_{\delta_1\in{{\cal i}}^c(r_1)}\sum_{\delta_2\in{{\cal i}}(r_2 ) } e^{n[\lambda(r_1+h(\delta_1)-\ln 2)+(1-\lambda)(r_2+h(\delta_2)-\ln 2)/\theta]}\times\nonumber\\ & & e^{-sn[\lambda\delta_1+(1-\lambda)\delta_2]}+\nonumber\\ & & \sum_{\delta_1\in{{\cal i}}(r_1)}\sum_{\delta_2\in{{\cal i}}(r_2 ) } e^{n[\lambda(r_1+h(\delta_1)-\ln 2)+(1-\lambda)(r_2+h(\delta_2)-\ln 2)]/\theta}\times\nonumber\\ & & e^{-sn[\lambda\delta_1+(1-\lambda)\delta_2]}+\nonumber\\ & & \sum_{\delta_1\in{{\cal i}}(r_1)}\sum_{\delta_2\in{{\cal i}}^c(r_2 ) } e^{n\eta[\lambda(r_1+h(\delta_1)-\ln 2)+(1-\lambda)(r_2+h(\delta_2)-\ln 2)]}\times\nonumber\\ & & e^{-sn[\lambda\delta_1+(1-\lambda)\delta_2]}\nonumber\\ & { \stackrel{\delta } { = } } & a+b+c+d\end{aligned}\ ] ] let us now handle each one of these four terms and take the limit @xmath269 .
this results in : @xmath347}\right]\cdot \left[\sum_{\delta_2\in{{\cal i}}^c(r_2)}e^{n_2[r_2+h(\delta_2)-\ln 2-s\delta_2]}\right]\nonumber\\ & { \stackrel{\cdot } { = } } & e^{-n_1u(s , r_1)}\cdot e^{-n_2u(s , r_2)}\nonumber\\ & = & e^{-n[\lambda u(s , r_1)+(1-\lambda)u(s , r_2)]},\end{aligned}\ ] ] @xmath348}\right]\cdot \left[\sum_{\delta_2\in{{\cal i}}(r_2)}e^{-n_2s\delta_2}\right]\nonumber\\ & { \stackrel{\cdot } { = } } & e^{-n_1u(s , r_1)}\cdot e^{-n_2\delta(r_2)}\nonumber\\ & = & e^{-n[\lambda u(s , r_1)+(1-\lambda)\delta(r_2)]},\end{aligned}\ ] ] @xmath349},\ ] ] and @xmath350 where @xmath351- \mu(\delta_1,\delta_2)[r+\lambda h(\delta_1)+(1-\lambda)h(\delta_2)-\ln 2]\}\ ] ] and where @xmath352 among the terms @xmath272 , @xmath75 , and @xmath353 , the term @xmath272 is exponentially the dominant one . to check whether or not @xmath272 dominates also @xmath354
, we will have to investigate the function @xmath355 .
this is done in subsection a.3 of the appendix , where it is shown that this function is as follows : for @xmath16 : @xmath356 and for @xmath20 : @xmath357 & 0\le s\le s_{r_2}\\ \lambda s\delta(r_1)+(1-\lambda)v(s , r_2 ) & s > s_{r_2}\end{array}\right.\ ] ] finally , the overall exponential rate of the characteristic function , @xmath358 , we have to take into account the contribution of @xmath272 , as mentioned above .
this gives : @xmath359 where @xmath360 .
now , in the case @xmath361 , for small @xmath21 , the function @xmath362 is linear with slope @xmath140 , whereas the function @xmath363 is linear with a slope of @xmath364 which is larger .
thus , @xmath362 is smaller in some interval of small @xmath21 .
however , for larger @xmath21 , @xmath362 continues to have a linear term with slope @xmath365 whereas @xmath363 never exceeds the level of @xmath366 .
thus , there must be a ( unique ) point of intersection @xmath367 .
consequently , for @xmath361 , we have @xmath368 where @xmath355 is as in ( [ fr1gr2 ] ) . concerning the case @xmath20 , both @xmath362 ( of eq .
( [ fr1lr2 ] ) ) and @xmath363 start as linear functions of the same slope of @xmath364 .
however , while the latter begins its curvy part at @xmath369 , the former continues to be linear until the point @xmath370 . in this case , then it is easy to see that @xmath331 is dominated by @xmath363 across the entire range @xmath23 , i.e. , @xmath371 we see then that the ensemble performance is substantially different in the two cases : for @xmath20 , @xmath331 is exactly the same as if we used two _ independent _ block codes of lengths @xmath3 and @xmath6 at rates @xmath4 and @xmath7 , respectively . in particular , the corresponding average distortion is @xmath364 which is , of course , larger than @xmath140 . in other words , we are gaining nothing from the tree structure and the dependence between the two parts of the code . for @xmath16 , on the other hand , there is at least a considerable range of small @xmath21 for which @xmath372 , namely , the ensemble performance is exactly like that of the ordinary ensemble of full block code of length @xmath9 and rate @xmath15 , without any structure ( which is also the best achievable exponential rate ) .
however , beyond a certain value of @xmath21 , there is some loss in comparison to the ordinary ensemble .
the case @xmath22 can be obtained as the limiting behavior of both @xmath373 and @xmath361 , by taking both rates to be arbitrarily close to each other . in this case
, we obtain @xmath372 throughout the _ entire _ range @xmath23 ( cf . the discussion on this in the introduction ) .
the conclusion then is that if we use an hierarchical structure of the kind we consider in this paper , it is best to assign equal rates at the two stages , but then we might as well abandon the tree structure of the code altogether , and just encode the two parts independently , both at rate @xmath15 ( this will moreover save complexity at the encoder ) . if , however , certain considerations dictate different rates at different segments , then it is better to encode at a larger rate in the first segment and at a smaller rate in the second .
this derivation can be extended , in principle , to any finite number @xmath0 of stages .
the analysis is , of course , more complicated but conceptually , the ideas are the same
. we will not carry out this extension in this paper . in complete duality to the source coding problem
, one may consider a channel code ( for the bsc ) with a similar hierarchical structure : given a binary information vector of length @xmath374 nats , we encode it in two parts : the first segment , of length @xmath5 nats , is encoded to a binary channel input vector of length @xmath3 , independently of the forthcoming @xmath294 nats ( thus , the channel encoder is of reduced delay ) .
then , the remaining @xmath294 nats are mapped to another binary channel input vector of length @xmath6 and it depends on the entire information vector of length @xmath375 . the ensemble of codebooks is drawn similarly as before : first , a randomly drawn first stage codebook of size @xmath376 , and then , for each one of its codewords , another codebook of size @xmath377 is drawn independently . once again , each bit of each codeword is drawn by independent fair coin tossing . the decoder applies maximum likelihood ( ml ) decoding based on the entire channel output vector @xmath112 of length @xmath2 , pertaining to the input @xmath54 of length @xmath9 .
the analogy with the grem is that here , the energy function is the log
likelihood , which is additive over the two stages by the memorylessness of the channel .
in full analogy to the grem and the source coding problem of subsection [ lossysourcecoding ] , and as an extension to the derivation in subsection [ remc ] , here too , the partition function @xmath130 has exactly the same two different types of behavior , depending on whether @xmath27 or @xmath28 .
therefore , we will not repeat this here . concerning the aspect of performance evaluation of this ensemble of codes , and a comparison to the ordinary ensemble , here the natural figure of merit is gallager s random coding error exponent , which can be analyzed using methods similar to those that we used in subsection [ lossysourcecoding ] .
we will not carry out a very refined analysis as we did before , but we will make a few observations in this context , although not quite directly related to the grem . in the hierarchical case with @xmath1 stages , the probability of error consists of two contributions .
the first pertains to all incorrect codewords @xmath385 whose first segment @xmath153 agrees with that of the correct codeword , and the second one is associated with all other incorrect codewords . as for the former type of codewords ,
the ml decoder actually compares the likelihood scores of the second segment only ( as those of the first segment are the same and hence cancel out ) , and so , these incorrect codewords contribute a term of the order of @xmath386 to the average error probability , where @xmath387 is the gallager s random coding error exponent function ( * ? ? ?
* , eq . ( 5.6.16 ) ) . concerning the second set of incorrect codewords
, we can apply an upper bound as above , except that the expectations have to be taken w.r.t . the hierarchical ensemble .
however , it is easy to see that the expectation of @xmath388 is exactly the same as in the ordinary ensemble , and thus , so is the upper bound for this set of codewords , which is then @xmath389 .
the total average error probability is then upper bounded by @xmath390 this gives further motivation why @xmath7 should be chosen smaller than @xmath4 : if @xmath391 , the second term definitely dominates the exponent , because both @xmath392 and @xmath393 and so @xmath394 . for a given @xmath15 and @xmath395 , can we , and if so how , assign the segmental rates @xmath4 and @xmath7 such that the second term would not be dominant , i.e. , @xmath396 ? if @xmath15 is large enough this is possible .
for example , one way to do this is to select @xmath397 , where @xmath353 is the channel capacity . in this case , we have , by the convexity of @xmath398 : @xmath399 for this strategy to be applicable , @xmath15 must be at least as large as @xmath400 .
how does this discussion extend to a general number of stages @xmath0 and is there a more systematic approach to allocate the segmental rates @xmath401 for a given overall rate @xmath15 ?
for simplicity , let us suppose that the segment lengths are all the same , i.e. , @xmath402 .
the extension turns out to be quite straightforward : in the case of @xmath0 stages there are @xmath0 types of incorrect codewords : those that agree with the correct codeword in all stages except the last stage , those that agree in all stages except the last two stages , etc . accordingly , using the same considerations as above , it is easy to see then that the upper bound on the average error probability consists of @xmath0 contributions whose exponents are @xmath403 for convenience , let us denote @xmath404 under what conditions and how can we assign the segmental rates such that @xmath405 for all @xmath406 ? first , we must select @xmath407 sufficiently small such that @xmath408 .
as @xmath15 is given , this will dictate the choice of @xmath4 according to the identity @xmath409 next , we choose @xmath410 small enough such that @xmath411 as @xmath407 has already been chosen , this will dictate the choice of @xmath7 according to the identity @xmath412 and so on .
this procedure continues until in the last step we choose @xmath413 such that @xmath414 , which dictates the choice of @xmath415 via @xmath416 , where @xmath417 was selected in preceeding step .
an obvious condition for this procedure to be applicable is that @xmath15 would be large enough such that @xmath418 .
note that if some of the segmental rates exceed capacity ( or even the log alphabet size ) , this is not a problem , as long as the averages @xmath419 are all small enough .
we begin with a simple large deviations bound regarding the distance enumerator , which appears also in @xcite , but we present here too for the sake of completeness . for @xmath420 $ ] ,
consider the binary divergence @xmath421.\end{aligned}\ ] ] to derive a lower bound to @xmath422 , let us use the inequality @xmath423 and then @xmath424 for every given @xmath112 , @xmath425 is the sum of the @xmath426 independent binary random variables , @xmath427 , where the probability that @xmath428 is exponentially @xmath429}$ ] .
the event @xmath430 , for @xmath431 , means that the relative frequency of the event @xmath432 is at least @xmath433 .
thus , by the chernoff bound : @xmath434})\right\}\nonumber\\ & { \stackrel{\cdot } { \le } } & \exp\left\{-e^{nr}\cdot e^{-n(r - a)}(n[(\ln 2-r- h(\delta)+a]-1)\right\}\nonumber\\ & \le & \exp\left\{-e^{na}(n[\ln 2-r - h(\delta)+a]-1)\right\}.\end{aligned}\ ] ] denoting by @xmath435 the interval @xmath436 and by @xmath437 , the complementary range @xmath438\setminus{{\cal i}}(r)$ ] , we have , for @xmath439 : @xmath440}+ e^{nrs}\cdot e^{-(n\epsilon-1)e^{n\epsilon}}.\end{aligned}\ ] ] one can let @xmath180 vanish with @xmath9 sufficiently slowly that the second term is still superexponentially small , e.g. , @xmath441 .
thus , for @xmath439 , @xmath442 is exponentially bounded by @xmath443}$ ] independently of @xmath21 . for @xmath444 , we have : @xmath445}\cdot \mbox{pr}\{n(n\delta)\le e^{n[r+h(\delta)-\ln 2+\epsilon]}\}+\nonumber\\ & & e^{nrs}\cdot \mbox{pr}\{n(n\delta)\ge e^{n[r+h(\delta)-\ln 2+\epsilon]}\}\nonumber\\ & \le&e^{ns[r+h(\delta)-\ln 2+\epsilon ] } + e^{nrs}\cdot e^{-(n\epsilon-1)e^{n\epsilon}}\end{aligned}\ ] ] where again , the second term is exponentially negligible . to see that both bounds are exponentially tight ,
consider the following lower bounds .
for @xmath439 , @xmath446^{e^{nr}-1}\nonumber\\ & { \stackrel{\cdot } { = } } & e^{nr}e^{-n[\ln 2-h(\delta)]}\cdot \left[1-e^{-n[\ln 2-h(\delta)]}\right]^{e^{nr}}\nonumber\\ & = & e^{n[r+h(\delta)-\ln 2]}\cdot\exp\{e^{nr}\ln[1-e^{-n[\ln 2-h(\delta)]}]\}.\end{aligned}\ ] ] using again the inequality in ( [ lnineq ] ) , the second factor is lower bounded by @xmath447}/(1-e^{-n[\ln 2-h(\delta)]})\ } = \exp\{-e^{-n[\ln 2-r - h(\delta)]}/(1-e^{-n[\ln 2-h(\delta)]})\}\ ] ] which clearly tends to unity as @xmath448 for @xmath439 .
thus , @xmath442 is exponentially lower bounded by @xmath443}$ ] . for @xmath444 , and an arbitrarily small @xmath449
, we have : @xmath450}\cdot \mbox{pr}\{n(n\delta)\ge e^{n[r+h(\delta)-\ln 2-\epsilon]}\}\nonumber\\ & = & e^{ns[r+h(\delta)-\ln 2-\epsilon]}\cdot\left(1- \mbox{pr}\{n(n\delta ) < e^{n[r+h(\delta)-\ln 2-\epsilon]}\}\right)\end{aligned}\ ] ] where @xmath451}\}$ ] is again upper bounded , for an internal point in @xmath435 , by a double exponentially small quantity as above . for @xmath135 near the boundary of @xmath435 , namely , when @xmath452 , we can lower bound @xmath442 by slightly reducing @xmath15 to @xmath453 ( where @xmath449 is very small ) .
this will make @xmath135 an internal point of @xmath454 for which the previous bound applies , and this bound is of the exponential order of @xmath455}$ ] . since @xmath456 is still very close to zero
, then @xmath455}$ ] is of the same exponential order as @xmath457}$ ] since both are about @xmath458 .
it should be noted that a similar double exponential bound can be obtained for the probability of the event @xmath459 , where @xmath460 and @xmath461 .
here we can proceed as above except that the in the lower bound on divergence @xmath422 we should take the second line of ( a.3 ) ( rather than the third ) , which is of the exponential order of @xmath462}$ ] ( observe that here @xmath463 is exponentially larger than @xmath363 , as opposed to the earlier case ) .
thus , we obtain @xmath464 at the second level exponent , and so the decay is double exponential as before .
first , let us write @xmath465 as follows : @xmath466 where @xmath153 and @xmath467 designate @xmath468 and @xmath469 , respectively , and where @xmath470 denotes the indicator function of an event .
we now treat each one of the four cases pertaining to the combinations of both @xmath471 and @xmath472 being or not being members of @xmath473 and @xmath474 , respectively .
+ for a given , arbitrarily small @xmath449 , consider the event @xmath477 .
if both the number of indices @xmath71 for which @xmath478 is less than @xmath479 and for each @xmath71 , @xmath480 , then clearly , the event @xmath481 does not occur .
thus , for @xmath481 to occur , at least one of these events must occur .
in other words , either the number of indices @xmath71 for which @xmath478 is larger than @xmath479 or there exist @xmath71 for which @xmath482 .
the probability of the former event is upper bounded by @xmath483 ( cf .
subsection a.1 ) .
similarly , the probability of the latter , for a given @xmath71 , is bounded by @xmath484 .
thus , the probability of the union of events @xmath485 is upper bounded by @xmath486 , which is still double exponential in @xmath9 .
thus , @xmath487 therefore , @xmath488 which is exponentially upper bounded by @xmath489}$ ] since @xmath180 is arbitrarily small , @xmath490}$ ] , and the last term is double exponential . to obtain the compatible lower bound , we use @xmath491 now , the event @xmath492 is the event that there is exactly one value of @xmath71 such that @xmath493 , and that for this @xmath71 , there is exactly one @xmath82 such that @xmath494 . as shown in subsection a.1 , the probability of the former is exponentially @xmath495}$ ] and the probability of the latter is exponentially @xmath305}$ ] .
thus , by independence , @xmath496 is the product , which is exponentially @xmath489}$ ] .
define now the event @xmath44 as @xmath498\}\right\}.\ ] ] as we have argued before , the probability of @xmath44 is doubly exponentially close to unity ( since the probability of @xmath499 is upper bounded by the sum of exponentially many doubly - exponentially small probabilities ) .
now , clearly , if @xmath44 occurs , @xmath500\}\cdot\sum_{i=1}^{m_1 } 1\{d_h({\mbox{\boldmath $ x$}}',{\hat{\mbox{\boldmath $ x$}}}_i)=n_1\delta_1\}.\ ] ] thus , @xmath501\}\times\right.\right.\nonumber\\ & & \left.\left.\sum_{i=1}^{m_1 } 1\{d_h({\mbox{\boldmath $ x$}}',{\hat{\mbox{\boldmath $ x$}}}_i)=n_1\delta_1\}\right]^{1/\theta}\right\}\nonumber\\ & & + e^{nr/\theta}\cdot\mbox{pr}\{{{\cal a}}^c\},\end{aligned}\ ] ] where the second term is again doubly exponentially small .
as for the first term , we bound @xmath502 by unity and @xmath503\}\sum_{i=1}^{m_1 } 1\{d_h({\mbox{\boldmath $ x$}}',{\hat{\mbox{\boldmath $ x$}}}_i)=n_1\delta_1\}\right]^{1/\theta}\right\}\nonumber\\ & = & \exp\{n_2[r_2+h(\delta_2)-\ln 2+\epsilon]/\theta\}\cdot{\mbox{\boldmath $ e$}}\left\{\left[\sum_{i=1}^{m_1 } 1\{d_h({\mbox{\boldmath $ x$}}',{\hat{\mbox{\boldmath $ x$}}}_i)=n_1\delta_1\}\right]^{1/\theta}\right\}\end{aligned}\ ] ] where the latter expectation ( cf .
subsection a.1 ) is of the exponential order of @xmath495}$ ] if @xmath475 ( case 2 ) and @xmath495/\theta}$ ] if @xmath504 ( case 3 ) .
thus , in both cases , we obtain the desired exponential order as an upper bound . for the lower bound
, we argue similarly that the probability of the event @xmath505\}\right\}\ ] ] is doubly exponentially close to unity , and so , @xmath506\}\sum_{i=1}^{m_1 } 1\{d_h({\mbox{\boldmath $ x$}}',{\hat{\mbox{\boldmath $ x$}}}_i)=n_1\delta_1\}\right]^{1/\theta}\right\},\ ] ] and we again use the above result on the moments of @xmath507 in both cases of @xmath471 .
since @xmath504 , then the event @xmath508}\le \sum_{i=1}^{m_1}1\{d_h({\mbox{\boldmath $ x$}}',{\hat{\mbox{\boldmath $ x$}}}_i)=n_1\delta_1\ } \le e^{n_1[r_1+h(\delta_1)-\ln 2+\epsilon]}\right\},\ ] ] has a probability which is doubly exponentially close to unity .
thus , given that @xmath44 occurs , there are @xmath509}\le l\le e^{n_1[r_1+h(\delta_1)-\ln 2+\epsilon]}\ ] ] indices @xmath510 for which @xmath478 . given @xmath511 and given these indices , @xmath465 is the sum of @xmath512+n_2r_2}$ ] i.i.d .
bernoulli trials , @xmath513 , whose probability of success is exponentially @xmath514}$ ] . thus , similarly as in the derivation in subsection a.1 , @xmath515 or , equivalently , in the notation of eq .
( [ nd1d2 ] ) : @xmath516\ } & \lambda w_1+(1-\lambda)w_2 < 0\\ \exp\{n[\lambda w_1+(1-\lambda)w_2]/\theta\ } & \lambda w_1+(1-\lambda)w_2 \ge 0\end{array}\right.\ ] ] the total expectation should , of course , account for @xmath499 as well , but since the probability of this event is doubly exponentially small , then the contribution of this term is negligible .
first , we observe that the constraints @xmath504 and @xmath476 can be replaced by their one - sided versions @xmath517 and @xmath518 , respectively , since values of @xmath471 and @xmath472 beyond @xmath519 can not be better than their corresponding reflections @xmath520 and @xmath521 . next observe that @xmath355 can be rewritten as follows : @xmath522 where @xmath523\ ] ] subject to the constraints @xmath517 , @xmath518 , and @xmath524 , and @xmath525 + ( 1-\lambda)[s\delta_2-r_2-h(\delta_2)+\ln 2]\}\ ] ] subject to the constraints @xmath517 , @xmath518 , and @xmath526 .
note that the optimization problem associated with @xmath527 is a convex problem , but the one pertaining to @xmath528 is not , because of its last constraint which is not convex . when @xmath361 , we have @xmath532 . as for @xmath533 ,
it is easy to see that @xmath534 is a solution that satisfies the necessary and sufficient kuhn tucker conditions for optimality of a convex problem , and so , @xmath535 .
consider next the function @xmath528 .
let us ignore , for a moment , the non convex constraint @xmath536 , and refer only to the constraints @xmath303 and @xmath518 .
denote by @xmath537 the corresponding maximum without the non
convex constraint .
the maximization problem associated with @xmath538 is now convex and it is to see that @xmath539 and @xmath540 satisfy the necessary and sufficient conditions for optimality , where @xmath541 .
this is also a solution for @xmath542 if it satisfies the non convex constraint , namely , if @xmath543 whether or not this condition is satisfied depends on @xmath21 .
since we are assuming @xmath361 , we then have @xmath544 , where we remind that @xmath545 . consequently , there are three different ranges of @xmath21 : @xmath546 , @xmath547 , and @xmath548 .
when @xmath549 , this is equivalent to @xmath550 in which case the above necessary condition ( [ cond ] ) becomes @xmath551 to check whether this condition is satisfied , observe that @xmath552 , and so this is equivalent to the condition @xmath553 , which is @xmath554 , in agreement with the assumption on the range of @xmath21 . therefore , the above solution is acceptable for @xmath542 and by substituting it back into the objective function , we get : @xmath555 + ( 1-\lambda)[s\nu_s - r_2-h(\nu_s)+\ln 2]\nonumber\\ & = & \lambda s\delta(r_1)+(1-\lambda)v(s , r_2)\end{aligned}\ ] ] when @xmath556 , this is equivalent to @xmath557 , in which case the condition ( [ cond ] ) becomes @xmath558 , or equivalently , @xmath559 , which is @xmath280 .
however , @xmath560 is between @xmath561 and @xmath562 , and so , the conclusion is that the non convex constraint is satisfied only in upper part of the interval @xmath563 $ ] , i.e. , @xmath564 $ ] . in this range ,
@xmath565 , and this yields @xmath566 . for @xmath567 , the condition ( [ cond ] ) no longer holds . in this case
, the optimum solution should be sought on the boundary of the non convex constraint , namely , under the equality constraint @xmath568 , but this coincides then with the solution to @xmath533 which was found on this boundary as well .
thus , for @xmath569 $ ] , we have @xmath570 . summarizing our results for @xmath542 over the entire range of @xmath23
, we have @xmath571 or , equivalently , @xmath572 finally , @xmath362 should be taken as the minimum between @xmath533 and @xmath542 .
now , @xmath533 is linear and @xmath542 is concave ( as it is the minimum of a linear function in @xmath21 ) , coinciding with @xmath533 along @xmath573 $ ] .
thus @xmath542 can not exceed @xmath533 for any @xmath21 , and so , @xmath574 .
thus , @xmath575 in this case , @xmath576 . once again , @xmath533 is associated with a convex program whose conditions for optimality are easily seen to be satisfied by the solution @xmath577 and @xmath578 .
thus , @xmath579.\ ] ] as for @xmath542 , let us examine again the various ranges of @xmath21 , where this time , @xmath580 . for @xmath581 , we have @xmath582 and then the condition ( [ cond ] ) is equivalent to @xmath583 , which is @xmath554 , in agreement with the assumption .
this corresponds to @xmath577 and @xmath584 , which yields @xmath585 for @xmath586 , which means @xmath587 , condition ( [ cond ] ) is satisfied with equality , and the corresponding solution is @xmath577 and @xmath578 , which yields @xmath588.\ ] ] for @xmath589 , eq .
( [ cond ] ) is not satisfied , and we resort again to the boundary solution , which , as mentioned earlier , is the same as @xmath533 . summarizing our findings for the case @xmath20 , and applying similar concavity considerations as before ( telling us that @xmath574 ) , we have : @xmath590 & 0\le s\le s_{r_2}\\ \lambda s\delta(r_1)+(1-\lambda)v(s , r_2 ) & s > s_{r_2}\end{array}\right.\ ] ] t. c. dorlas and j. r. wedagedera , `` phase diagram of the random energy model with higher order ferromagnetic term and error correcting codes due to sourlas , '' _ phys .
_ , vol .
83 , no . 21 , pp .
44414444 , november 1999 .
n. merhav , `` relations between random coding exponents and the statistical physics of random codes , '' submitted to _
ieee trans . inform .
theory _ , august 2007 .
available on line at : [ http://www.ee.technion.ac.il/people/merhav/papers/p117.pdf ] .
n. merhav , `` error exponents of erasure / list decoding revisited via moments of distance enumerators , '' submitted to _
ieee trans . inform . theory _ ,
november 2007 . also , available on line at [ http://www.ee.technion.ac.il/people/merhav/papers/p119.pdf ] .
|
in an earlier work , the statistical physics associated with finite temperature decoding of code ensembles , along with the relation to their random coding error exponents , were explored in a framework that is analogous to derrida s random energy model ( rem ) of spin glasses , according to which the energy levels of the various spin configurations are independent random variables .
the generalized rem ( grem ) extends the rem in that it introduces correlations between energy levels in an hierarchical structure . in this paper
, we explore some analogies between the behavior of the grem and that of code ensembles which have parallel hierarchical structures .
in particular , in analogy to the fact that the grem may have different types of phase transition effects , depending on the parameters of the model , then the above mentioned hierarchical code ensembles behave substantially differently in the various domains of the design parameters of these codes .
we make an attempt to explore the insights that can be imported from the statistical mechanics of the grem and be harnessed to serve for code design considerations and guidelines .
+ + * index terms : * spin glasses , grem , phase transitions , random coding , error exponents .
department of electrical engineering + technion - israel institute of technology + haifa 32000 , israel +
|
there is a growing interest in extended charge transport models for semiconductor devices .
our paper emerges from the the fact that today s submicron semiconductor devices like e.g. mesfets and hemts are operated under strong electric fields and at high frequencies .
information transmission using an electromagnetic wave at very high frequencies will have a direct impact on how we design active and passive components in different fields of micro- and optoelectronics .
in such cases , quasi - static semiconductor device models like the energy balance model ( ebm ) are no longer adequate . especially in gaas and related materials used for high - speed device design ,
inertia effects play an important role since the momentum and energy relaxation times of the electron gas are close to the picosecond range
. the most elaborate and practicable approach for the description of charge transport in semiconductors used for device simulation would be the monte carlo ( mc ) method @xcite .
the advantage of this technique is a complete picture of carrier dynamics with reference to microscopic material parameters , e.g. effective masses and scattering parameters . but
the method must be still considered as very time consuming and hence too uneconomical to be used by device designers .
besides the simplest concept which is the traditional drift - diffusion model ( ddm ) , there is a much more rigorous approach to the problem , namely the so - called hydrodynamic model ( hdm ) . a simplified version of the hdm is the ebm , which is implemented in most of today s commercial semiconductor device simulators .
the hdm makes use of electron temperature for the description of charge transport and takes inertia effects into account as well .
starting from the boltzmann equation , blotekjaer @xcite and many others presented a derivation of such equations under the assumption of parabolic band structures . especially for silicon , satisfactory results were obtained this way @xcite .
but the results have often been unsatisfactory when mc models based on a nonparabolic structure were compared to hdm results based on an empirical choice of model parameters .
therefore , it is quite natural to improve the hdm by incorporating energy - dependent relaxation times and effective masses obtained from mc bulk simulations @xcite .
this is a strategy we pursue in this paper . in the first part of this work , we give a short definition of the hydrodynamic model for gaas .
we emphasize that a thorough analysis of the physical features of the charge carrier transport models is the basis for a clear understanding of their limits of applicability .
then we give a simple discretization scheme for the full hydrodynamic model in two dimensions , which will be applied to a gaas mesfet structure in the last part .
there , we compare the hdm results also to results obtained from the simpler ebm , and we investigate the influence of the grid resolution on the results .
our active device model is based on the _ single - gas _ hydrodynamic equations .
this is a simplification of the two - valley hydrodynamic equations , since strictly speaking , in gaas and other semiconductors with similar band structure like inp , there exists an electron gas with different thermal distribution function for each conduction band valley ( i.e. the @xmath0- and @xmath1-valleys ) .
the equations for each valley are , however , coupled through collision terms since electrons can scatter between two different valleys .
the corresponding relaxation rates may be of the order of a picosecond and are therefore relatively large .
this is the main drawback of the single - gas approximation , and it would be desirable to implement at least a two - valley hydrodynamic model .
reliable extensive two - valley simulations have been performed only for the one - dimensional case so far due to the large amount of equations and parameters involved in such a model .
a hydrodynamic two - valley simulation of gaas mesfets is the subject of one of our forthcoming papers .
the hdm equations consist of the continuity equation @xmath2 for negligible charge carrier generation and recombination , the momentum balance equation given by @xmath3 or alternatively ( only for the x - component ) @xmath4 and the energy balance equation @xmath5 @xmath6 where @xmath7 , @xmath8 ( @xmath9 ) , and @xmath10 are the electron density , the electron energy density ( average electron energy ) and the electron drift velocity , respectively .
@xmath11 is the x - component of the electron drift velocity and @xmath12 the momentum density .
corresponding equations are valid for the y- ( and z- ) components .
@xmath13 is the electron temperature , @xmath14 the elemental charge and @xmath15 boltzmann s constant .
@xmath16 is the average thermal equilibrium energy of electrons , where @xmath17 is the lattice temperature .
the ebm uses only a simplified energy balance equation ( see below ) .
the electronic current density @xmath18 inside the active device is given by @xmath19 the total current density is @xmath20 the momentum relaxation time @xmath21 is related to the mobility of the electrons via @xmath22 , and the energy relaxation time @xmath23 describes the exchange of energy between the heated electron gas and the lattice . @xmath24 and @xmath25 and the effective electron mass @xmath26 are assumed to be functions of the mean electron energy .
we performed steady - state monte carlo simulations in order to get the correct values for these parameters .
the hydrodynamic equations , together with poisson s equation ( @xmath27 @xmath28 is the number of ( ionized ) donors ) @xmath29 form a complete set of equations that can be used to solve for the electron density , velocity , energy and electric field for given boundary conditions , if we use a closing relation for the mean electron energy @xmath30 , the electron temperature @xmath13 and velocity @xmath31 : @xmath32 the reason for the double index of the electron mass will soon become clear . the last term in eq .
( [ closrel ] ) accounts for the fact that a minimum energy @xmath33 ev is necessary to excite an electron to an upper conduction band .
@xmath34 is the relative fraction of electrons in the l - band for the stationary homogeneous case . the term @xmath35 is often neglected in the literature , but this may lead to an overestimation of the electron temperature of _ more than 1000 k _ at high energies .
we point out again the important fact that we are using a single - gas approximation for the hydrodynamic model .
this means that the closing relation eq .
( [ closrel ] ) is a crude approximation which allows the calculation of the electron temperature from the total electron energy and electron drift velocity .
some authors neglect also the influence of the electron velocity on the temperature @xcite by directly relating the electron temperature @xmath13 to the average electron energy @xmath30 from stationary mc simulations .
the transition from the two - gas model to the single - gas approximation has to be done carefully ; therefore , we present here a short discussion of the problem which is usually not mentioned in the literature .
we assume parabolic @xmath36 and @xmath37valleys for the sake of brevity , but in our simulations , we took the effect of the non - parabolicity of the energy bands into account by using the kane model @xcite , which generalizes the parabolic relation for the electron energy @xmath38 and electron crystal momentum @xmath39 @xmath40 to @xmath41 with the non - parabolicity coefficient @xmath42 which has different values for the different energy bands .
@xmath43 is the effective electron mass at the bottom of the energy band under consideration .
very often , an energy - dependent electron mass is defined via @xmath44 i.e. the electron mass increases with growing energy .
the electron mass @xmath45 must not be confused with the energy - dependent electron mass used in the hydrodynamic model . in that case , the masses are a kind of _ average _ masses depending on the _ average _ electron energy .
there is even a further aspect related to the notion of the electron mass .
the crystal velocity of an electron is given for spherical bands by @xmath46 a short calculation for the kane model shows that @xmath47 which implies that crystal velocity @xmath31 and crystal momentum @xmath48 are related by @xmath49 i.e. a different definition of the energy - dependent electron mass applies if the electron velocity is calculated from the crystal momentum . in the single particle two - band mc simulations , the random walk of an electron inside the semiconductor material is monitored over a sufficiently long time . as a result , the probability @xmath50 that an electron resides in the @xmath36valley is obtained as a function of the applied constant homogeneous electric field @xmath51 or as a function of the mean electron energy @xmath30 , and the probability of finding the electron in an upper @xmath37valley is then @xmath52 .
also the values for for the average electron velocities @xmath53 and @xmath54 in the different valleys are obtained as well .
then it is reasonable to define the average electron velocity by @xmath55 the average electron momentum @xmath56 is given by @xmath57 hence the ( energy - dependent ) electron mass which must be used in the hydrodynamic model in order to relate average electron velocity and electron momentum is calculated from @xmath58 but we suggest that a _ different mass _
@xmath59 should be used for the calculation of the average kinetic electron energy in eq .
( [ closrel ] ) .
we identify @xmath60 and therefore @xmath61 @xmath26 and @xmath62 are _ not distinguished _ in the literature . it is tempting to use a naive definition for the electron mass @xmath63 it is interesting that data in the literature for the energy - dependent electron mass are usually in better agreement with this definition .
it is clear that the hdm is still an approximative description of charge carrier dynamics inside a semiconductor , and the different assumptions which are inherent in the derivation of the model may already cause larger errors in the simulation results than using only one mass .
therefore we do not claim that our discussion leads to improved simulation results , but it rather shows the limits of the frequently used model and points out that it is mandatory to maintain always a highest possible degree of consistency .
figure 1 shows @xmath26 , @xmath62 and @xmath64 normalized to the free electron mass @xmath65 as functions of the average electron energy for a gaas lattice temperature @xmath66 k and a low doping density @xmath67 .
the results from mc simulations were smoothed by a polynomial fit and transferred afterwards into the hydrodynamic simulation program .
we used @xmath68 and @xmath69 . for high energies when nearly all electrons are in the upper bands , the electron mass even exceeds the value @xmath70 due to the non - parabolicity of the energy bands .
figure 2 shows the average energy @xmath30 of an electron in a constant homogeneous electric field e for gaas . for each data point , the electron was scattered one million times ( including so - called self - scattering ) , therefore the resulting curve is already quite smooth .
it is clear that mc simulations deliver no data for average electron energies below @xmath71 mev , since the mean electron energy @xmath30 has this value if no electric field is applied to the crystal , and @xmath30 grows for increasing electric field .
this is no major problem , since the low energy region is of minor importance for the hydrodynamic simulation and the necessary data can be obtained from theoretical considerations @xcite .
-relation for gaas with @xmath66 k and @xmath72.,width=10 ] in order to complete the set of data which is necessary for hydrodynamic simulations , the electron velocity and energy relaxation times are depicted in figs . 3 and 4 for doping densities @xmath67 and @xmath73 .
the characteristic shape of the velocity curve can be explained by the fact that at high energies the electrons jump into the @xmath37bands where the electrons have a lower mobility than in the @xmath36band .
finally , we need an expression for the thermal conductivity of the electron gas , which is given by theoretical considerations @xmath74 several different choices for @xmath75 can be found in the literature , and many authors @xcite even neglect heat conduction in their models . as a matter of fact , heat conduction does not influence the simulation results very much if @xmath75 remains within a reasonable range , but baccarani and wordeman point out in @xcite that neglecting thermal conductivity completely can lead to nonphysical results and mathematical instability .
although their work is directed to si , their remarks should be equally valid for gaas since the equations have a similar form in both cases . in our simulations ,
we have chosen @xmath76 .
the ebm is obtained as a simplification of the full hdm by neglecting the convective terms of the momentum balance equation ( [ mb ] ) . additionally , the energy balance equation ( [ eb ] ) is simplified by the assumption that the time derivative of the mean electron energy @xmath77 is small compared to the other terms and that the kinetic part in @xmath8 can also be neglected , i.e. @xmath78 or , if we take the two - valley structure of gaas into account , @xmath79 the energy balance equation then becomes @xmath80 @xmath81 and the momentum balance equation simplifies to the well - known current equation @xmath82 the continuity equation and the poisson equation remain of course unchanged in the ebm .
neglecting the time derivative of the current density is equivalent to the assumption that the electron momentum is able to adjust itself to a change in the electric field within a very short time .
while this assumption is justified for relatively long - gated field effect transistors , it needs to be investigated for short - gate cases . setting the electron temperature
@xmath13 equal to the ( constant ) temperature of the semiconductor material @xmath17 leads to the drift - diffusion model .
such a simplification is clearly not justified for the case studied in this paper ( see also @xcite ) .
today , many elaborate discretization methods are available for the ddm equations or ebm equations .
the well - known scharfetter - gummel method @xcite for the ddm makes use of the fact that the current density is a slowly varying quantity .
the current equation is then solved exactly under the assumption of a constant current density over a discretization cell , which leads to an improved expression for the current density than it is given by simple central differences .
it is therefore possible to implement physical arguments into the discretization method .
similar techniques have been worked out for the ebm @xcite .
but due to the complexity of the hdm equations , no satisfactory discretization methods which include physical input are available for this case .
therefore , we developed a shock - capturing upwind discretization method , which has the advantage of being simple and reliable .
for our purposes , it was sufficient to use a homogeneous mesh and a constant time step .
but the method can be generalized to the non - homogeneous case .
time discretization is done for all equations by forward euler differencing , i.e. the discretization scheme is fully explicit . the discretization is always written down only for the x - component of vectorial quantities in the sequel , since the corresponding expressions for y - components are then easy to derive .
the constant timestep @xmath83 used in our simulations was typically of the order of a few tenths of a femtosecond , and quantities at time @xmath84 carry an upper integer index @xmath85 .
the rectangular simulated region of the mesfet gets discretized into @xmath86 rectangular cells @xmath87 of equal size @xmath88 .
scalar quantities at timestep @xmath85 like @xmath89 and @xmath90 , where @xmath91 and @xmath92 are located at the center of the cells , whereas vectorial quantities like e.g. the electric field components @xmath93 , @xmath94 or the velocity components @xmath95 , @xmath96 are always calculated first at midpoints between the scalar quantities . if necessary , we can define intermediate values , e.g. @xmath97 by @xmath98 but a different definition applies to e.g. @xmath99 , as we shall see .
the fundamental variables that we will have to compute at each timestep are @xmath100 , @xmath101 , @xmath102 ( or @xmath103 ) , @xmath104 and @xmath105 , always respecting the imposed boundary conditions .
all other variables used in the sequel should be considered as derived quantities .
the momentum balance equation is discretized in the following way : @xmath106 where @xmath107 and @xmath108 and the same discretization strategy is applied to the y - component of the electron velocity . from the momentum density
we obtain the new particle current density by @xmath109 and the momentum density at @xmath110 is extrapolated from neighbouring points in the direction of the electron flow x - component @xmath111 and finally we get @xmath112 @xmath113 we found that the purely heuristic choice @xmath114 in the equations above improves the stability of our code .
the electron temperature is related to the energy density by the relation @xmath115 and can therefore be regarded as a dependent variable .
the upwind discretization is of the energy balance equation is given by @xmath116 where we have defined the energy currents @xmath117 @xmath118 and @xmath119 having obtained the new values for the mean electron energy , the transport parameters and energy - dependent masses are then also updated .
the current continuity equation is discretized in a conservative way , using the particle current density @xmath120 @xmath121 @xmath122 i.e. particles that leave cell @xmath110 in x - direction enter cell @xmath123 and analogously for the y - direction ; therefore , the total number of electrons inside the mesfet can only be changed at the boundary of the simulation region ( mainly at the contacts ) .
the poisson equation , which is coupled to the hydrodynamic equations only through the particle density @xmath7 , can be solved by any convenient method which is fast enough , since the computational effort should be kept as small as possible .
therefore , we used a multigrid method to perform this task .
fortunately , the poisson equation has not to be solved at each timestep . since the relaxation times of gaas are of the order of some tenths of a picosecond , a timestep of about ten femtoseconds is fully sufficient for the update of the electric field .
fortunately , the stability of the discretization scheme is not affected that way , and allows an enormous reduction of the computational costs . for the basic quantities @xmath124 and @xmath8 we imposed dirichlet boundary conditions at each timestep at the contacts , e.g.
the potential @xmath125 at the source and the drain was fixed by the applied voltage @xmath126 analogously , we assumed charge neutral contacts at the source and the drain , such that the charge carrier density was given there by the fixed doping density .
the gate contact was modelled by assuming a schottky barrier height of 0.8 v. for further details concerning the metal - semiconductor contact modelling we refer to standard textbooks @xcite . as far as the energy density is concerned , we imposed the boundary conditions directly on the electron temperature by assuming that the electron gas is in thermal equilibrium with the drain / source contacts .
the artificial boundaries were modelled using von neumann type boundary conditions .
we present an explicit example in order to illustrate this point .
we assume that the discrete values of the electron density at the mesfet boundary between the source and the gate ( see fig .
5 ) are given by @xmath127 : the first index denotes the horizontal direction , whereas the second index starts with @xmath128 at the top of the mesfet .
then , after each update of the density according to eq .
( [ update ] ) , we enforce @xmath129 corresponding to the von neumann condition that the normal component of the electron density vanishes at the specified boundaries . as mentioned above , we used a multigrid algorithm to calculate the electric potential . also there , the mixed dirichlet / von neumann boundary conditions were imposed on the subgrids at each intermediate step of the calculations . a fortran90 program which calculates the potential by a multigrid algorithm can be obtained from the first author s address
the gaas mesfet structure used in our simulation is shown in fig .
the structure consists of a 0.1 @xmath130m - thick active layer with a doping concentration of @xmath72 on a 0.3 @xmath130 m buffer layer ( @xmath67 ) .
the doping profile is abrupt between the two layers , the lattice temperature is @xmath66 k. for steady - state results , we used long simulation times of 30 ps such that the steady state was _ de facto _ reached .
the length of the drain and source contacts is 0.5 @xmath130 m , the gate - source separation 0.5 @xmath130 m , the gate - drain separation is 1.0 @xmath130 m and the gate length is 0.8 @xmath130 m .
the schottky barrier height is assumed to be 0.8 v. in order to obtain stable and physically meaningful results , values like @xmath131 nm for a grid size of @xmath132 were typical values used in the simulations .
the rather large mesh was manageable due to the effective multigrid algorithm used for the solution of the poisson equation .
we found that the mesh size used in @xcite was too coarse for accurate results , although the authors improve accuracy and convergence speed of their calculations by using a non - homogeneous grid .
in fact , a non - homogeneous grid necessitates additional calculational costs which reduce the speed of the simulation , and the timesteps also depend on the size of the smallest cells .
furthermore , the poisson equation was solved by a conventional successive overrelaxation method .
we present results for a gate - source bias @xmath133 v and a drain - source bias @xmath134 v. fig .
6 shows the electron temperature inside the mesfet for the stationary case .
fig . 7 and 8
show the electron velocity and the electron temperature along the channel of the mesfet ( 0.077 @xmath130 m below the contacts ) . due to the high temperature which is reached under the gate ,
the electron mobility is strongly lowered in this region .
the electrons , which overshoot under the gate , are therefore deaccelerated abruptly to a lower velocity in the high - temperature region . the results obtained from an energy balance calculation are in good quantitative agreement ,
the differences are mostly pronounced in the region where the electron velocity is high , as it is expected from the different treatment of the energy density in the hdm and ebm . in the very close vicinity of the gate
, it was necessary to reduce artificially the electric field or the velocity of the electrons in order to stabilize our code .
we checked that this does not strongly affect the simulations results outside this region due to the very low density of the electron gas near the contact ; a similar procedure was also necessary in @xcite .
the velocity and temperature curves are in fact very similar to those of one - dimensional simulations of ballistic diodes ( @xmath135-structures ) , which were used as simplified models for fet channels .
it is interesting to observe in fig . 7 that our simulation results for a coarse grid ( grid size @xmath136 ) are very close to those presented in fig . 5 in @xcite , where a non - uniform mesh of typical size @xmath137 was used for a similar mesfet geometry .
this may be considered as a confirmation of the results given in @xcite for a relatively coars grid , but demonstrates the fact that stability of the code does not automatically imply accuracy of the results , and an investigation of the dependence of the results on the grid resolution is indispensable .
the channel below the gate ( see also fig .
10 ) , which is the most interesting region in the mesfet , is relatively small and must be resolved sufficiently . due to the strong heating of the electron gas in the channel region ,
most electrons are excited to the l - band .
this leads to a cooling effect of the electron gas , since the excitation energy is missing in the thermal energy balance . the dash - dotted curve in fig .
9 shows the electron temperature that would be obtained from the hdm if the energy term @xmath138 in eq .
( [ closrel ] ) were neglected .
the energy @xmath33 ev which is necessary to excite an electron to the upper conduction band corresponds to a temperature difference @xmath139 ; the observed error is of the same size . finally , fig .
10 shows a surface plot of the electron density inside the device . clearly visible
is the mesfet channel under the gate . ) .
the plot extends over the whole length of the device ( 537 grid points ) , but shows only a relevant layer of about 0.2 @xmath130 m below the contacts.,width=10 ]
the feasibility of two - dimensional hydrodynamic simulations is demonstrated for the case of a gaas mesfet structure .
although the single - gas hydrodynamic model is superior to the drift - diffusion or energy balance model , it is desirable to direct the efforts of future research in the direction of multi - valley hydrodynamic models .
models like the ebm will no longer be adequate for the physical description of high - speed submicron devices in the near future . it is obvious that future attempts to model submicron devices will face many more problems which have not been touched in this paper .
one difficulty is the fact that the components of semiconductor devices are often of very different size and material composition .
this necessitates the use of adaptive discretization grids or the hybridization of different numerical methods , but both strategies are hampered by severe problems like numerical instabilities or huge computational efforts for realistic simulations .
another class of problems is due to the fact that the physics of semiconductor materials is very complex , and therefore hard to implement such that the physical behavior of the device is satisfactorily described .
attempts to describe quantum effects in device modelling should be considered as tentative in the case of heterostructure devices .
there is no optimal solution for these problems , and the numerical and physical models have to be adapted to the problem under study .
we hope that our detailed description of a hydrodynamic simulation may serve also as a help for researchers entering the field of hydrodynamic semiconductor device modelling
. 1 k. tomizawa , _ numerical simulation of submicron semiconductor devices_. artech house : london , boston , 1993 . k. blotekjaer,``transport equations for electrons in two - valley semiconductors , '' _ ieee trans .
electron dev .
38 - 47 , jan .
gardner , `` numerical simulation of a steady - state electron shock wave in a submicrometer semiconductor device , '' _ ieee trans .
electron dev .
392 - 398 , feb . 1991 .
feng , a. hintz , `` simulation of submicrometer gaas mesfets using a full dynamic transport model , '' _ ieee trans .
electron dev .
9 , pp . 1419 - 1431 , sept .
alsunaidi , s.m .
hammadi , s.m .
el - ghazaly , `` a parallel implementation of a two - dimensional hydrodynamic model for microwave semiconductor device including inertia effects in momentum relaxation , '' _ int .
fields _ , vol .
107 - 119 , march - april 1997 .
el - ghazaly , _ private communication_. e.o .
kane , `` band structure of indium antimonide , '' _
chem . solids _ , vol .
249 - 261 , 1957 .
g. baccarani , m.r .
wordemann , `` an investigation of steady - state velocity overshoot in silicon , '' _ solid - state electronics _ ,
407 - 416 , 1985 .
a. aste , r. vahldieck , `` time - domain simulation of the full hydrodynamic model'',_int .
fields _ , vol .
16 , no.2 , pp . 161 - 174 , 2003 . d.l .
scharfetter , h.k .
gummel , `` large - signal analysis of a silicon read diode oscillator , '' _ ieee trans .
electron dev .
16 , no.1 , pp .
64 - 77 , jan . 1969 .
t. tang , `` extension of the scharfetter - gummel algorithm to the energy balance equation , '' _ ieee trans .
electron dev .
1984 _ , vol .
1912 - 1914 , dec . 1984 . *
andreas aste * received the diploma degree in theoretical physics from the university of basel , basel , switzerland , in 1993 , and the ph.d .
degree from the university of zrich , zrich , switzerland , in 1997 . from 1997 to 1998 he was a post doctoral assistant at the institute for theoretical physics in zrich . from 1998 to 2001 he was a research assistant and project leader in the laboratory for electromagnetic fields and microwave electronics of the swiss federal institute of technology eth . since 2001 , he is working as a researcher at the institute for theoretical physics at the university of basel .
aste is a member of the american physical society aps .
* rdiger vahldieck * received the dipl .- ing . and
dr .- ing .
degrees in electrical engineering from the university of bremen , germany , in 1980 and 1983 , respectively . from 1984
to 1986 , he was a research associate at the university of ottawa , ottawa , canada . in 1986
, he joined the department of electrical and computer engineering , university of victoria , british columbia , canada , where he became a full professor in 1991 . during fall and spring 1992 - 1993
, he was visiting scientist at the ferdinand - braun - institute fr hochfrequenztechnik in berlin , germany . since 1997
, he is professor of field theory at the swiss federal institute of technology , zrich , switzerland .
his research interests include numerical methods to model electromagnetic fields in the general area of electromagnetic compatibility ( emc ) and in particular for computer - aided design of microwave , millimeter wave and opto - electronic integrated circuits .
vahldieck , together with three co - authors , received the 1983 outstanding publication award presented by the institution of electronic and radio engineers . in 1996
, he received the 1995 j. k. mitra award of the institution of electronics and telecommunication engineers ( iete ) for the best research paper . since 1981
he has published more than 230 technical papers in books , journals and conferences , mainly in the field of microwave cad .
he is the past - president of the ieee 2000 international zurich seminar on broadband communications ( izs2000 ) and since 2003 president and general chairman of the international zurich symposium on electromagnetic compatibility .
he is a member of the editorial board of the ieee transaction on microwave theory and techniques . from 2000 until 2003
he served as an associate editor for the ieee microwave and wireless components letters and from january 2004 on as the editor - in - chief . since 1992
he serves on the technical program committee of the ieee international microwave symposium , the mtt - s technical committee on microwave field theory , and in 1999 on the tpc of the european microwave conference . from 1998 until 2003 professor vahldieck was the chapter chairman of the ieee swiss joint chapter on mtt , ap and emc .
0.4 cm * marcel rohner * received the dipl .- ing . and
dr .- ing .
degrees in electrical engineering from the swiss federal institute of technology , switzerland , in 1993 and 2002 , respectively . since 1993
he was a research assistant in the electronics laboratory of the swiss federal institute of technology eth , where he has been working in the area of semiconductor device modeling , monte - carlo device simulations , electro - optical sampling , digital filtering , and switched - current circuits . in 2002 , he joined the corporate technology center of leica geosystems ag in the field of electronic distance measurements .
|
a finite difference upwind discretization scheme in two dimensions is presented in detail for the transient simulation of the highly coupled non - linear partial differential equations of the full hydrodynamic model , providing thereby a practical engineering tool for improved charge carrier transport simulations at high electric fields and frequencies .
the discretization scheme preserves the conservation and transportive properties of the equations .
the hydrodynamic model is able to describe inertia effects which play an increasing role in different fields of micro- and optoelectronics , where simplified charge transport models like the drift - diffusion model and the energy balance model are no longer applicable .
results of extensive numerical simulations are shown for a two - dimensional mesfet device .
a comparison of the hydrodynamic model to the commonly used energy balance model is given and the accuracy of the results is discussed .
|
the geometric structure of a random coordinate projection of a class of functions plays an important role in empirical processes theory , where it is used to determine whether the uniform law of large numbers or the uniform central limit theorem is satisfied by the given class .
indeed , if @xmath2 is a class of functions on a probability space @xmath4 , and if @xmath5 is an independent sample distributed according to @xmath6 , then the `` complexity '' of the random set @xmath7 is the key parameter in addressing both these questions . in this context ,
if @xmath8 are independent , symmetric , @xmath9-valued random variables , then the complexity is governed by the expectation of the supremum of the bernoulli process indexed by @xmath10 , defined by @xmath11 and in particular , on the way this expectation grows as a function of @xmath1 for a typical sample of cardinality @xmath1 ( see , e.g. , @xcite and references therein ) .
the structure of such coordinate projections is central to questions in asymptotic geometric analysis as well .
for example , let @xmath12 be a convex , symmetric set ( i.e. , if @xmath13 then @xmath14 ) and put @xmath15 to be the class of linear functionals indexed by @xmath16 . if @xmath17 is a measure on @xmath18 , then @xmath10 is the set @xmath19 , where @xmath20 is the random operator @xmath21 .
fundamental questions on the geometry of convex , symmetric sets , such as dvoretzky s theorem and low-@xmath22 estimates have been answered by obtaining accurate , quantitative information on the structure of such coordinate projections , and by using very similar complexity parameters to ( [ eq : intro - bernoulli ] ) ( e.g. , @xcite ) . for both these reasons ,
a lot of effort has been invested in understanding various notions of complexity for a typical coordinate projection of a class of functions .
a well studied direction is to obtain quantitative estimates on the way in which ( [ eq : intro - bernoulli ] ) is related to two other complexity parameters , the combinatorial dimension and covering numbers .
roughly speaking , the combinatorial dimension of @xmath23 at scale @xmath24 , denoted by @xmath25 , is the largest dimension of a coordinate projection of @xmath26 that contains a `` cube '' of scale @xmath24 ( see definition [ def : comb - dim ] for an exact formulation ) .
if @xmath27 is a metric space then the covering number at scale @xmath24 , which we denote by @xmath28 , is the smallest cardinality of a subset @xmath29 such that for every @xmath30 , there is some @xmath31 for which @xmath32 . connections between ( [ eq : intro - bernoulli ] ) and the combinatorial dimension or the covering numbers of @xmath10 are rather well understood .
for example , a straightforward chaining argument ( see , e.g. , @xcite ) shows that for every @xmath23 , @xmath33 where @xmath34 is the euclidean metric on @xmath3 , @xmath35 is the diameter with respect to the same metric and @xmath36 is an absolute constant , independent of the dimension @xmath1 and of the set @xmath26 .
some of the other relations between these parameters are far more involved .
first , controlling the @xmath37 covering numbers using the combinatorial dimension was resolved in @xcite , where it was shown that if @xmath26 is a subset of the unit cube @xmath38 and @xmath17 is any probability measure on @xmath39 , then for every @xmath40 , @xmath41 where @xmath16 and @xmath36 are absolute constants .
also , the solution of the sign embedding of @xmath42 problem ( see @xcite ) implies that if @xmath43 and @xmath44 , then @xmath45 . in other words , under a normalization condition ( @xmath43 ) , the only reason that @xmath46 is almost extremal is that @xmath26 contains a high - dimensional cubic structure . in this article , we study a related geometric parameter the discrepancy of a typical coordinate projection .
discrepancy was introduced in a combinatorial context ( see the book @xcite for an extensive survey on this topic ) and is defined as follows .
[ def : disc ] if @xmath23 , then the discrepancy of @xmath26 is @xmath47 and the infimum is taken with respect to all signs @xmath48 .
we denote by @xmath49 the hereditary discrepancy of @xmath26 , given by @xmath50 where @xmath51 is the coordinate projection of @xmath26 onto @xmath52 .
observe that if @xmath53 is the convex hull of @xmath54 , then @xmath55 .
hence , from the geometric viewpoint , the discrepancy of @xmath26 is proportional with a constant @xmath56 to the minimal width of @xmath53 in a direction of a vertex of the combinatorial cube @xmath57 .
the hereditary discrepancy is governed by a similar minimal width , but of the `` worst '' coordinate projection of @xmath53 .
our goal here is to study the discrepancy using the covering numbers and the combinatorial dimension of @xmath26 , but we will focus on sets @xmath26 that are random coordinate projections of a class of function @xmath2 , which gives them more structure .
a natural question in this context is to identify conditions on @xmath2 under which there is a gap between @xmath58 and @xmath59 for a typical coordinate projection of @xmath2 , as a function of the sample size @xmath1 . to that end
, we will develop dimension dependent bounds on the discrepancy of @xmath60 ( and in particular , bounds that are not asymptotic ) .
note that the metric structure of @xmath23 is not enough to determine its discrepancy . indeed ,
if @xmath61 and @xmath62 then @xmath63 . on the other hand , if @xmath64 , which is linearly isometric to @xmath65 , then @xmath66 .
the reason for the large gap in the discrepancy between the two isometric sets is that @xmath67 consists of a vector that is `` well spread '' while @xmath65 consists of a `` peaky '' vector with respect to the underlying coordinate structure . in that sense
, @xmath67 is in a much better position than @xmath65 .
note that in this example , @xmath68and for the set @xmath67 , which is in a `` good position '' there is gap between the expectation for signs and the discrepancy .
we will show that this is a general phenomenon : it is well known that @xmath69 is determined by the euclidean metric structure of @xmath26 ( up to a logarithmic factor in the dimension ) , and therefore , it is almost invariant under a linear isometry ( i.e. , a change in the coordinate structure ) .
thus , the expectation almost does not change when applying an isometry or a good isomorphism of @xmath34 .
as we will explain here , the situation with the discrepancy is rather different and the position of the set matters a great deal . since the sets @xmath26 that we will be interested in are not arbitrary but have some structure as random coordinate projections of well behaved function classes
, they will be much closer in nature to @xmath67 than to @xmath65 .
our main result is that if the canonical gaussian process indexed by @xmath70 is continuous and if the class satisfies a subgaussian condition [ i.e. , if the @xmath71 norm is equivalent to the @xmath72 norm on @xmath2 , see definition [ def : sub - gauss - set ] ] , then a typical coordinate projection of @xmath2 behaves as a set of vectors in a `` general position . '' as such , and just like the set @xmath67 , a typical coordinate projection exhibits certain shrinking properties that will be explained in section [ sec : shrink ] , and which causes the discrepancy of such a set to be much smaller than the average over signs .
[ theoa ] let @xmath70 be a class of mean zero functions .
assume further that the canonical gaussian process indexed by @xmath2 is continuous and that the @xmath72 and @xmath71 norms are equivalent on @xmath2 .
then @xmath73 in probability . to put theorem [ theoa ] in the right perspective , observe that if the @xmath74 and @xmath37 norms are equivalent on a class of mean zero functions @xmath2 , then for every integer @xmath1 there is a subset of @xmath75 of probability at least @xmath36 on which @xmath76 where @xmath36 depends only on the equivalence constant between the @xmath77 and @xmath37 norms on @xmath2 , @xmath78 is an absolute constant and @xmath79
. hence , there is a true gap between the discrepancy and the mean of a typical coordinate projection .
although the formulation of theorem [ theoa ] is asymptotic , the result itself is quantitative in nature , as a function of the dimension of the coordinate projection .
the proof of theorem [ theoa ] is , in fact , a dimension dependent estimate on the sequences @xmath80 , for which , with probability at least @xmath81 , @xmath82 .
we will show that the sequences @xmath80 are given using metric parameters that measure the continuity of the gaussian process indexed by @xmath2talagrand s @xmath83 functionals @xcite .
the @xmath83 functionals will be defined in section [ sec : pre ] , but for now let us mention that under mild measurability assumptions on the class , the canonical gaussian process indexed by @xmath2 is continuous if and only if @xmath84 .
we will prove that for every @xmath85 and @xmath86 there are constants @xmath36 and @xmath87 that depend on @xmath88 , @xmath89 and on the equivalence constant between the @xmath71 and @xmath72 norms on @xmath2 , such that for every @xmath1 , @xmath90 where @xmath91 is the diameter of @xmath2 with respect to the @xmath72 norm . and , in particular , under the assumptions of theorem [ theoa ] , for every @xmath86 , @xmath92 .
moreover , the proof of theorem [ theoa ] actually shows that for every @xmath1 , with @xmath6-probability of at least @xmath81 , @xmath93 where @xmath87 and @xmath36 depend on @xmath88 , @xmath89 and the equivalence constant between the @xmath72 and @xmath71 norms on @xmath2 .
the proof of theorem [ theoa ] is based on two ingredients .
the first is a new estimate on the discrepancy of an arbitrary set @xmath94 .
it is a combination of the entropy method , which is often used to control the combinatorial discrepancy ( see , e.g. , @xcite ) , and talagrand s generic chaining mechanism @xcite , which was introduced to establish the connection between the @xmath83 functionals and the continuity of gaussian processes .
the combination of these two methods will be explained in section [ sec : disc ] .
it allows one to find a good choice of signs for roughly @xmath95 coordinates , while the error incurred by considering the sum taken only on these coordinates is determined by the @xmath83 functional for @xmath96 . repeating this argument , one obtains a bound on the discrepancy of @xmath26 in terms of a sum of @xmath83 functionals of coordinate projections of @xmath26 and for values @xmath97 that depend on the dimension of each projection , and those dimensions decrease quickly .
the second component required for the proof of theorem [ theoa ] is that the sets we are interested in are not general .
we will obtain a structural result on a typical @xmath10 that allows us to bound the @xmath83 functionals of its coordinate projections using the @xmath72 structure of @xmath2 .
indeed , we will show that if the @xmath72 and @xmath71 norms are equivalent on @xmath2 then a typical coordinate projection @xmath10 has a rather regular structure it is a subset of a minkowski sum of two sets .
the first one is small , with a bounded diameter in the weak @xmath98 space @xmath99 .
the other set is a subset of @xmath10 itself and can be viewed as a set of vectors in a `` general position . '' in particular , further coordinate projections of the latter set shrink distances between any two of its elements .
the structural result we obtain is of independent interest and can be used to derive information on the geometry of convex sets .
for example , consider a symmetric probability measure @xmath17 on @xmath100 .
we say that @xmath17 isotropic and @xmath101-subgaussian if a random vector @xmath102 distributed according to @xmath17 satisfies that for every @xmath103 , @xmath104 simple examples of isotropic , @xmath101-subgaussian measures on @xmath100 are the gaussian measure and the uniform measure on the vertices of the cube @xmath105 , where in both cases @xmath101 can be taken to be an absolute constant , independent of the dimension .
let @xmath106 be independent random vectors , distributed according to @xmath17 and consider the random operator @xmath107 defined by @xmath108 .
[ corolb ] for any @xmath109 there are constants @xmath110 and @xmath111 that depend only on @xmath101 , for which the following holds .
let @xmath112 and set @xmath113 . then
, for every @xmath114 , with probability at least @xmath115 , for every @xmath116 , @xmath117 where @xmath118 are independent , standard gaussian random variables , and both expectations are taken with respect to those variables .
corollary [ corolb ] shows that the random operator @xmath20 maps an arbitrary @xmath26 to a set of vectors in a `` general position '' in a strong sense , since it implies that for most vectors in @xmath119 , mutual distances are shrunk by any further coordinate projection .
let us note that we will prove a stronger result than corollary [ corolb ] , namely that the @xmath83 functionals associated with @xmath119 display this type of shrinking phenomenon .
the final result we present has to do with the reverse direction of theorem [ theoa ] .
assume that @xmath120 is a convex , symmetric set , which satisfies that the canonical gaussian process @xmath121 is bounded and that the @xmath72 and @xmath71 norms are equivalent on @xmath122 .
we will show that if the logarithm of the @xmath72 covering numbers of @xmath122 grows like @xmath123 then for a typical sample @xmath5 selected according to @xmath6 , @xmath124 it is standard to verify ( see lemma [ lemma : disc - vs - vc ] ) that if @xmath23 , then @xmath125 therefore , if @xmath2 is a class of mean - zero functions and @xmath126 satisfies the above , then @xmath127 , complementing the upper bound established in theorem [ theoa ] . although this is not exactly the reverse direction of theorem [ theoa ] , it is very close to it .
indeed , if @xmath70 indexes a bounded gaussian process then so does @xmath126 , and the logarithm of the covering numbers of @xmath122 can not grow faster than @xmath128 . on the other hand , if the log - covering numbers grow a little slower , even by a suitable logarithmic factor , then @xmath129 .
in fact , this is as close as one can get to a covering numbers characterization of the fact that @xmath130 ( see , e.g. , @xcite ) .
this result not only shows that @xmath131 is large if the gaussian process @xmath2 indexes is bounded but not continuous , it also shows why . under a boundedness assumption on the gaussian process [ which implies that @xmath132 is bounded ] , the reason the hereditary discrepancy of @xmath10 is extremal is because a typical coordinate projection of @xmath133 contains a high dimensional , large cubic structure , and that forces the hereditary discrepancy to be large .
the proof of this result , which is presented in section [ sec : upp - low ] , is based on the observation that if @xmath2 is convex and symmetric then the richness of @xmath2 at scale @xmath134 is exhibited by the existence of cubes of scale @xmath134 and of dimension @xmath135 in a typical coordinate projection of @xmath2 of dimension @xmath1 .
it thus should be viewed as a `` small scale '' version of the sign embedding theorem which was mentioned above .
unfortunately , the optimal estimate in the sign embedding theorem can not be used directly in our case , firstly because @xmath10 is unlikely to be a subset of @xmath38 , and secondly , because a typical coordinate projection of @xmath2 satisfies that @xmath136 hence , the optimal estimate in the sign embedding theorem has to be used for @xmath137 , and that only ensures that @xmath10 contains a cube of scale @xmath134 and of constant dimension , which is far from what we need .
the proof of the existence of a cube in @xmath10 is based on two localization arguments , one with respect to the @xmath37 norm and the other with respect to the @xmath138 norm .
the first localization shows that if the @xmath72 covering number of @xmath2 at scale @xmath134 is of the order of @xmath139 then the richness of a typical coordinate projection of @xmath2 of dimension @xmath135 originates from the set @xmath140 that is , functions in @xmath2 of @xmath72 norm at most @xmath141 . in the second localization ,
one shows that the complexity of a typical coordinate projection actually comes from a further pointwise truncation of the functions in @xmath2 , and @xmath142 in ( [ eq : intersect ] ) can essentially be replaced by @xmath143the unit ball in @xmath144 .
this article is organized as follows . in section [ sec : pre ] , we present further preliminaries , most of them concerning subgaussian variables and the @xmath83 functionals . in section [ sec : disc ] , we develop bounds on the discrepancy of an arbitrary subset of @xmath100 .
section [ sec : shrink ] is devoted to the proof of the structural results on coordinate projections of subgaussian processes and its corollaries , including corollary [ corolb ] .
theorem [ theoa ] is proved in section [ sec : main ] and its converse and the resulting lower bound on the hereditary discrepancy of a typical coordinate projection is proved in section [ sec : upp - low ] .
throughout , absolute constants ( i.e. , fixed , positive numbers ) will be denoted by @xmath145 etc .
their values may change from line to line .
we denote by @xmath146 constants that depend only on the parameter @xmath147 and we set @xmath148 to be constants that will remain fixed throughout the article . by @xmath149 , we mean that there are constants @xmath36 and @xmath87 such that @xmath150 , and we write @xmath151 if @xmath152 .
we will consider a single , fixed euclidean structure on all finite - dimensional spaces @xmath100 and denote the corresponding euclidean norms by without specifying the dimension . with a minor abuse of notation ,
the cardinality of a set and the absolute value are denoted in the same way .
if @xmath153 is a normed space , let @xmath154 be its unit ball , and for @xmath155 we denote the unit ball by @xmath156 .
if @xmath157 let @xmath158 be the empirical measure supported on @xmath159 , set @xmath160 to be the corresponding @xmath37 space , and for @xmath161 let @xmath162 be the coordinate subspace of @xmath34 spanned by @xmath163 .
the situation we will study here is as follows .
let @xmath2 be a class of real valued functions on a probability space @xmath4 , let @xmath164 be independent random variables distributed according to @xmath17 and set @xmath5 .
let @xmath165 be the coordinate projection of @xmath2 defined by @xmath159 and for every @xmath116 let @xmath166 be the coordinate projection of @xmath2 on the subset of coordinates @xmath167 .
sometimes , for the sake of simplicity , we shall omit the superscript @xmath159 . here
, we will describe properties of sums of independent random variables that have quickly decaying tails .
[ def : psi ] let @xmath168 be a functions defined on a probability space @xmath4 . for @xmath169 , define the @xmath170-orlicz norm by @xmath171 for basic facts regarding orlicz norms , we refer the reader to @xcite .
it is well known that a random variable has a bounded @xmath172 norm for @xmath169 if and only if it has a well behaved tail ; that is , there is an absolute constant @xmath36 such that for every @xmath173 and every @xmath174 , @xmath175 conversely , there is an absolute constant @xmath78 such that if @xmath168 displays a tail behavior dominated by @xmath176 for @xmath169 then @xmath177 .
there are several basic properties of sums of independent random variables we require .
the proofs of these facts can be found , for example , in @xcite .
note that if @xmath168 has a subexponential tail then its empirical means concentrate around its true mean , with a tail behavior that is a mixture of subgaussian and subexponential .
indeed , the following result is a version of bernstein s inequality and shows just that .
[ thm : bern ] there exists an absolute constant @xmath36 for which the following holds .
let @xmath4 be a probability space and set @xmath178 to be a function with a bounded @xmath179 norm .
if @xmath164 are independent and distributed according to @xmath17 then for every @xmath180 , @xmath181 if a function has mean zero and a bounded @xmath74 norm , one can obtain a purely subgaussian tail .
[ lemma : psi_2 ] there exists an absolute constant @xmath36 for which the following holds .
let @xmath182 be independent random variables of mean zero .
then , for every @xmath183 , @xmath184 in particular , if @xmath106 are independent random variables distributed according to @xmath17 and @xmath168 has zero mean , then for every @xmath185 , @xmath186 where @xmath78 is an absolute constant . in what follows
, we will assume that the @xmath74 and @xmath37 norms are equivalent on @xmath2 in the following sense .
[ def : sub - gauss - set ] a set @xmath70 is @xmath101-subgaussian if @xmath187 and @xmath188 for every @xmath189 . next , let us turn to the definition of the @xmath83 functionals @xcite .
let @xmath27 be a metric space .
an _ admissible sequence _ of @xmath26 is a sequence of subsets of @xmath26 , @xmath190 , such that @xmath191 and for every @xmath192 , @xmath193 .
[ def : gamma-2 ] for a metric space @xmath27 and an integer @xmath194 , let @xmath195 where the infimum is taken with respect to all admissible sequences of @xmath26 .
set @xmath196 .
let @xmath197 be a metric projection function onto @xmath198 , that is , @xmath199 is a nearest point to @xmath200 in @xmath198 with respect to the metric @xmath201 .
it is easy to verify that for every admissible sequence , every @xmath30 , and any @xmath202 , @xmath203 and that the diameter of @xmath26 satisfies @xmath204 .
moreover , it is clear that the @xmath83 functionals are decreasing in @xmath97 and are subadditive in @xmath26 in the following sense .
let @xmath102 be a normed space and consider two sets @xmath205 .
if @xmath206 is the minkowski sum of @xmath207 and @xmath208 , then for every integer @xmath97 , @xmath209 there is a close connection between the @xmath83 functionals with respect to @xmath37 norms and properties of gaussian processes ( see @xcite for expositions on these connections ) .
indeed , let @xmath210 be a centered gaussian process indexed by a set @xmath26 and for every @xmath211 define a metric on @xmath26 by @xmath212 .
one can show that under mild measurability assumptions on @xmath26 , @xmath213 where @xmath78 and @xmath214 are absolute constants .
the upper bound is due to fernique @xcite and the lower bound is talagrand s majorizing measures theorem @xcite .
the proof of both parts can be found in @xcite .
thus , the @xmath215 functional is finite if and only if the gaussian process indexed by @xmath26 is bounded .
note that if @xmath112 and @xmath216 then @xmath217 and therefore @xmath218 just like @xmath219 determines the supremum of the canonical gaussian process indexed by @xmath220 ( which we will always assume to satisfy the necessary measurability assumptions ) , the continuity of that process is determined by properties of the sequence @xmath83 .
[ def : pregaussian ] let @xmath70 be a class of mean zero functions . set @xmath221 to be the centered gaussian process indexed by @xmath2 with a covariance structure endowed by @xmath72 ; that is , for every @xmath189 , @xmath222 .
we say that @xmath2 is @xmath17-pregaussian if it has a version with all sample functions bounded and uniformly continuous with respect to the @xmath72 metric .
[ thm : cont - gauss ] let @xmath223 be a centered gaussian process and endow @xmath26 with the @xmath37 metric given by the covariance structure of the process , as above . under measurability assumptions ,
the following are equivalent : 1 .
the map @xmath224 is uniformly continuous on @xmath26 with probability @xmath225 .
2 . @xmath226 .
3 . there exists an admissible sequence of @xmath26 such that @xmath227 in other words , @xmath26 is pregaussian if and only if @xmath228 .
theorem [ thm : cont - gauss ] is not proved in @xcite but only stated there , and its formulation in @xcite was done using the notion of majorizing measures rather than with the @xmath83 functionals . since the proof of the continuity theorem follows from an application of the majorizing measures theorem and since the latter is proved in @xcite using the language of the @xmath215-functional , it is not difficult to convert the proof of the continuity theorem from @xcite and obtain theorem [ thm : cont - gauss ] . moreover , as shown in @xcite , there is a quantitative connection between the modulus of continuity of @xmath223 and the sequence @xmath229 .
since we will not use this quantitative estimate here , we will not formulate it . finally , let us define the covering and packing numbers of a metric space .
[ def : cov - and - ent ] let @xmath27 be a metric space .
the covering number of @xmath26 at scale @xmath230 with respect to the metric @xmath201 is the smallest number of open balls of radius @xmath24 needed to cover @xmath26 , and is denoted by @xmath28 .
we set @xmath231 .
@xmath232 are called the entropy numbers of @xmath26 .
a set @xmath233 is called @xmath24-separated if the distance between any two of its elements is at least @xmath24 .
we denote by @xmath234 the cardinality of a maximal @xmath24-separated subset of @xmath26 .
it is standard to verify that for every @xmath230 , @xmath235 , @xmath236 , and thus one can use either one of the two notions freely .
we begin this section with a technical lemma which is at the heart of the proof of theorem [ theoa ] .
the lemma allows one to find a good choice of signs on roughly half of the coordinates , while the error incurred by the choice of coordinates and signs can be controlled using the geometric structure of @xmath26 .
a preliminary result we need has to do with bernoulli processes the well - known hffding inequality ( see , e.g. , @xcite ) .
[ thm : hoeffding ] let @xmath237 be independent , symmetric , @xmath9-valued random variables . then , for every @xmath238 and every @xmath180 , @xmath239 in particular , @xmath240 let us formulate the main lemma .
[ lemma : gen - disc ] let @xmath241 there exist absolute constants @xmath242 and @xmath243 for which the following holds .
assume that @xmath244 is an increasing positive sequence tending to infinity , @xmath245 is a positive sequence and @xmath246 is an integer such that @xmath247 let @xmath112 for which @xmath248 , set @xmath249 to be a sequence of subsets of @xmath26 and let @xmath250 .
consider maps @xmath197 that satisfy that : 1 .
for every @xmath251 , @xmath252 , 2 .
for every @xmath30 , @xmath253
. then , there exists @xmath254 such that @xmath255 , and for every @xmath30 , @xmath256 the proof is a combination of a chaining argument and the entropy method , which is frequently used in discrepancy theory ( see , e.g. , @xcite ) . in the chaining mechanism ,
one takes the sets @xmath198 to be finer and finer approximations of the set @xmath26 and @xmath199 is a nearest element to @xmath200 in @xmath198 with respect to the underlying metric ( which is , in our case , the @xmath257 metric ) .
recall that the entropy of a discrete random variable @xmath102 taking values in a countable set @xmath258 is @xmath259 where @xmath260 .
the entropy function @xmath261 indicates how close @xmath102 is to being equally distributed ; the more equally distributed @xmath102 is , the larger @xmath261 is .
the three facts we will need regarding the entropy are well known and we omit their proofs . first , if @xmath262 then there is a value of @xmath102 that is attained with probability at least @xmath263 .
second , if @xmath102 attains at most @xmath1 values then @xmath264 , and finally , if @xmath265 is a random vector then @xmath266 . in the entropic argument we will use , each `` link '' in each chain in @xmath26 is assigned a random variable @xmath267 that depends on the link and on the chain .
the idea is to show that with probability at least @xmath268 , for every @xmath170 , each random variable @xmath269 falls in an interval @xmath270 whose length is at most @xmath271 .
one would like to make these scaling factors @xmath272 as small as possible while still ensuring that conditions @xmath225 and @xmath273 hold , since those conditions imply that the intersection of the level sets of all the random variables @xmath269 has the desired measure .
more details on the way entropic arguments have been used in the context of discrepancy theory may be found in @xcite . before presenting the proof
, one should mention that a chaining argument was implicit in matouek s result on the discrepancy of a subset of @xmath274 with a bounded vc dimension @xcite .
the first step in the proof of lemma [ lemma : gen - disc ] is the following entropy estimate .
we denote by @xmath275 $ ] the integer value of @xmath276 .
[ lemma : entropic - estimate ] there exists an absolute constant @xmath36 for which the following holds .
let @xmath238 , set @xmath277 and put @xmath278.\ ] ] then @xmath279 by hffding s inequality ( theorem [ thm : hoeffding ] ) , for every @xmath280 , @xmath281 also , @xmath282 implying that @xmath283 consider @xmath284 for which @xmath285 .
since @xmath286 is increasing in @xmath287 $ ] , it follows that for such values of @xmath288 , @xmath289 fix an integer @xmath1 which satisfies that @xmath290 and which will be named later , and observe that if we set @xmath291 then @xmath292 clearly , @xmath293 , and thus , if @xmath294 ( i.e. , if @xmath295 ) , then @xmath296 otherwise , @xmath297 , implying that @xmath298 moreover , @xmath299 therefore , @xmath300 now , consider the following three cases . first , if @xmath301 , take @xmath302 , and thus @xmath303 if @xmath304 set @xmath1 to be a suitable absolute constant and if @xmath305 , put @xmath306 .
therefore , in both these cases @xmath307 and our claim follows .
proof of lemma [ lemma : gen - disc ] without loss of generality , assume that @xmath26 is finite .
recall that @xmath248 and that @xmath250 , consider the sets @xmath198 and the maps @xmath197 , let @xmath308 and put @xmath309 .
let @xmath244 and @xmath245 be as in the assumptions of the lemma and set @xmath237 to be independent , symmetric , @xmath9-valued random variables .
consider the bernoulli process @xmath310 . since @xmath311 is linear in @xmath200 and @xmath312 , then for every @xmath313 , @xmath314 for every @xmath251 and @xmath315 define @xmath316.\ ] ] observe that @xmath317 is a vector that takes a finite number of values .
since the entropy is subadditive then @xmath318 suppose that one can find @xmath245 for which @xmath319 . by the properties of the entropy
, this implies that there are numbers @xmath320 such that @xmath321 since @xmath322 , there will be at least two vectors @xmath237 and @xmath323 in @xmath207 that differ on at most @xmath324 coordinates and on at least @xmath325 of them .
the desired sequence will then be @xmath326 .
indeed , for @xmath315 , @xmath327 implying that every @xmath30 satisfies @xmath328 hence , to complete the proof , it remains to show that for a sequence @xmath245 that satisfies the assumptions of the lemma , @xmath329 .
applying lemma [ lemma : entropic - estimate ] for @xmath330 , and since @xmath331 , it is evident that @xmath332 .
thus , @xmath333 proving our claim
. we will apply lemma [ lemma : gen - disc ] in two typical situations .
the first case will lead to a bound on the discrepancy of a set using the @xmath83 functionals of the set and of its coordinate projections .
the second will result is an entropy integral type bound , presented in section [ sec : entropy - integral ] , which will then be used to re - prove spencer s result on the discrepancy of a finite set system @xcite and matouek s vc theorem @xcite .
corollary [ cor : choice - of - q - for - gamma-2 ] below will play a central part in the proof of theorem [ theoa ] .
since it follows from a simple computation , we omit its proof .
[ cor : choice - of - q - for - gamma-2 ] there exist absolute constants @xmath334 , @xmath335 , @xmath336 for which the following holds .
let @xmath337 , assume that @xmath248 , set @xmath338 and put @xmath198 to be a collection of subsets of @xmath26 with @xmath193 . then , if @xmath339 there exists @xmath254 such that latexmath:[$n/4 \leq latexmath:[\[\biggl|\sum_{i=1}^n \eta_i t_i \biggr| \leq\sum_{s=1}^\infty q_s point to @xmath200 in @xmath198 . in this section , we will prove an analog of dudley s entropy integral bound ( see , e.g. , @xcite ) in the context of discrepancy .
the entropy integral is often used to upper bound latexmath:[${\sup_{t \in t } } @xmath237 , but here we will present a modified version that allows one to control @xmath342 , where the infimum is taken with respect to all @xmath343 for which roughly half the coordinates are nonzero . let @xmath337 and recall that for every @xmath230 , @xmath344 is the cardinality of a maximal @xmath24-separated subset of @xmath26 .
also , set @xmath345 [ thm : entropy - integral - disc ] there exist an absolute constant @xmath36 for which the following holds . if @xmath337 and @xmath248 , then there exist @xmath254 , such that @xmath255 and for every @xmath30 , @xmath346 recall that dudley s entropy integral bound shows that @xmath347 for a suitable absolute constant @xmath78 .
clearly , this entropy integral may be considerably larger than the quantity we have in theorem [ thm : entropy - integral - disc ] .
it is also evident that if one could iterate theorem [ thm : entropy - integral - disc ] for the set @xmath348 , where @xmath349 , and continue in the same manner , then one would likely improve upon the bound resulting from the standard entropy integral bound that holds for a typical choice of signs , if indeed distances in @xmath350 shrink relative to distances in @xmath26 .
the proof of theorem [ thm : entropy - integral - disc ] is based on lemma [ lemma : gen - disc ] .
it requires two additional simple results . since their proofs
are standard , we shall not present them here .
[ lemma : q - est-1 ] there exist absolute constants @xmath78 , @xmath214 , @xmath111 and @xmath351 for which the following holds .
let @xmath337 , set @xmath352 to be the largest integer @xmath97 satisfying @xmath353 and define @xmath354 then conditions and of lemma [ lemma : gen - disc ] hold if one selects @xmath355 [ lemma : integral - vs - sum ] let @xmath356 and @xmath168 be nonincreasing , nonnegative functions and let @xmath357 be a decreasing sequence . if for every @xmath192 , @xmath358 , and if there is @xmath359 such that for every @xmath251 , @xmath360 then @xmath361 proof of theorem [ thm : entropy - integral - disc ] let @xmath244 and @xmath245 be as in lemma [ lemma : q - est-1 ]
. without loss of generality assume that @xmath26 is a finite set and define the sets @xmath198 iteratively , as follows .
set @xmath362 to be the first integer such that @xmath363 , let @xmath364 for @xmath365 and set @xmath366 . for @xmath367 , let @xmath368 and put @xmath369 to be a maximal @xmath370-separated subset of @xmath371 whose cardinality is at most @xmath372 . continue in this way to construct the sets @xmath198 for @xmath373 . for every @xmath97 ,
let @xmath199 be a nearest point to @xmath374 in @xmath198 .
let @xmath375 , and since the sets @xmath198 are nested , then @xmath376 and @xmath377 for every @xmath378 .
therefore , applying lemmas [ lemma : gen - disc ] and [ lemma : q - est-1 ] , there is a choice @xmath254 with @xmath379 such that for every @xmath30 , @xmath380 it remains to bound the sums in ( [ eq : upper.sums ] ) by the appropriate integrals , using lemma [ lemma : integral - vs - sum ] .
first , for @xmath381 let @xmath382},\qquad g({\varepsilon})= \sum_{s=\nu_n+1}^{m } 2^{(s-\nu_n)/2 } { \mathbh{1}}_{({\varepsilon}_{s},{\varepsilon}_{s-1}]}.\ ] ] clearly , in @xmath383=[0,{\varepsilon}_{\nu_n}]$ ] , @xmath168 and @xmath356 are nonincreasing and nonnegative , for every @xmath24 in that range @xmath384 and the conditions of lemma [ lemma : integral - vs - sum ] hold . since @xmath385 , then @xmath386 for the other term in ( [ eq : upper.sums ] ) , if @xmath387 then @xmath388 and the sum is estimated in a similar way .
let us show how theorem [ thm : entropy - integral - disc ] can be used to prove a version of spencer s celebrated result from @xcite ( see also @xcite ) .
[ thm : spencer ] there exists an absolute constant @xmath36 such that if @xmath389 is of cardinality @xmath390 , then @xmath391 without loss of generality , assume that @xmath248 . using the notation of theorem [ thm : entropy - integral - disc ] , for every @xmath230 , @xmath392 , and since @xmath389 then @xmath393 .
hence , there are @xmath254 for which @xmath394 now the result follows by repeating this argument for @xmath395 , where @xmath396 , an so on .
a well - known measure of complexity for subsets of @xmath274 is the @xmath397 dimension of the set ( its real value counterpart will be used in section [ sec : upp - low ] ) .
let @xmath398 .
we say that @xmath399 is shattered by @xmath26 if @xmath400that is , if the coordinate projection @xmath401 is the entire combinatorial cube on these coordinates .
define @xmath402 to be the maximal cardinality of a subset of @xmath403 that is shattered by @xmath26 . in @xcite
, matouek proved that the discrepancy of a vc class is polynomially better than could be expected from a random choice of signs .
he obtained the best possible estimate for the discrepancy of vc - subsets of @xmath274 as a function of the dimension @xmath246 .
[ thm : mat ] for every integer @xmath201 , there is a constant @xmath404 for which the following holds . if @xmath398 and @xmath405 , then @xmath406 . to prove matouek s theorem ,
recall the following fundamental property of a vc class , due to haussler @xcite .
[ lemma : vc - cover ] if @xmath398 and @xmath405 , then for every @xmath407 and every @xmath408 , @xmath409 where @xmath404 is a constant that depends only on @xmath201 .
proof of theorem [ thm : mat ] again , we may assume that @xmath248 and view @xmath26 as a subset of @xmath100 .
let @xmath410 .
therefore , @xmath411 .
a change of variables shows that @xmath412 hence , there is a choice of @xmath413 such that for every @xmath30 @xmath414 and if we set @xmath415 then @xmath416 .
since @xmath417 then repeating the same argument for the set @xmath395 , there are @xmath418 such that for every @xmath30 , latexmath:[$|{\sum_{i \in i_1 } \eta_i^1 t_i } of signs @xmath237 such that for every @xmath30 , latexmath:[\[\biggl|\sum_{i=1}^n { \varepsilon}_i t_i \biggr| \leq c_3(d ) \sum_j fact that for every @xmath288 , @xmath421 . the proof of theorem [ thm : mat ] illustrates once again the main property we used to bound the discrepancy of a subset of @xmath100 .
it is not enough for the set to be small in the sense of its metric entropy ; what is needed is additional control on the `` size '' of all of the set s coordinate projections .
one way of controlling those coordinate projections is by taking into account information about the position of vectors in the set , since coordinate projections of vectors in a good position shrink norms and mutual distances .
it is clear from our estimate on the discrepancy of a set @xmath422 that it would be useful to control the @xmath162 distances between points in @xmath26 for every @xmath407that is , distances between coordinate projections of elements of @xmath26 .
one would be able to obtain a good bound on @xmath423 if @xmath26 is not too rich and if for every @xmath407 and every @xmath424 , @xmath425 is significantly smaller than @xmath426 .
unfortunately , usually this is not true even for a single vector @xmath427 . indeed ,
if @xmath428 is supported in @xmath52 then @xmath429 does not `` shrink '' at all . on the other hand ,
if the coordinates of @xmath428 are roughly equal , then the coordinate projection onto any @xmath52 shrinks the norm of @xmath428 by a factor of @xmath430 .
it is well known that a strong shrinking phenomenon is exhibited by vectors in a general position . in this section
, we will show that if a class of functions @xmath2 is @xmath101-subgaussian , then a shrinking phenomenon happens for a typical set @xmath431 uniformly for all coordinate projections of @xmath26 . as a starting point ,
let us describe the so - called `` standard shrinking '' phenomenon for a single function @xmath168 .
let @xmath168 be a function for which @xmath432 .
then , concentration implies that with high probability , @xmath433 however , as we mentioned above , the shrinking phenomenon one needs here is more general that for every subset @xmath434 , the @xmath435 norm of @xmath168 is upper bounded ( possibly up to a logarithmic factor ) by @xmath436 ( which translates in the @xmath98 normalization to the shrinking of the norm ) .
the following lemma shows that this stronger claim is true as well whenever @xmath168 is @xmath101-subgaussian .
[ lemma : shrinking - single ] for every @xmath86 and @xmath109 , there is a constant @xmath437 for which the following holds . if @xmath438 then for every integer @xmath1 , with probability at least @xmath81 , for every @xmath116 , @xmath439 fix @xmath1 and @xmath440 .
since @xmath441 , then by bernstein s inequality , for every @xmath180 , @xmath442 let @xmath443 and recall that there are at most @xmath444 subsets of @xmath39 of cardinality @xmath362 .
hence , it suffices to take @xmath445 and obtain that with probability of at least @xmath446 , for every subset @xmath52 of @xmath39 of cardinality @xmath362 , @xmath447 therefore , summing the probabilities with respect to @xmath362 , it follows that for the correct choice of @xmath448 , with probability at least @xmath81 , ( [ eq : single - function ] ) is true for all subsets of @xmath39 of cardinality at most @xmath449 .
the claim now easily follows .
when one attempts to generalize this simple shrinking argument to a class of functions , one faces a problem : the probabilistic estimate obtained in the proof of lemma [ lemma : shrinking - single ] does not allow one to control many functions simultaneously .
thus , a naive extension of that result is simply too weak to lead to a function class analog of the shrinking phenomenon . to formulate the shrinking phenomenon for an @xmath101-subgaussian class of functions ,
let us recall some notation .
for any two sets @xmath207 and @xmath208 in a vector space , @xmath450 , and for a class of functions @xmath2 , a random sample @xmath5 and @xmath161 , @xmath451 for every integer @xmath362 , let @xmath452 , where @xmath453 is a monotone nonincreasing rearrangement of @xmath454 .
thus , @xmath455 is the unit ball of the weak @xmath98 space @xmath456 .
denote by @xmath457 the collection of all subsets of @xmath39 of cardinality at most @xmath362 and set @xmath458 to be the smallest integer @xmath97 such that @xmath459 .
[ thm : decomposition - of - f ] for every @xmath86 and @xmath109 there exist constants @xmath78 , @xmath214 , @xmath111 and @xmath460 depending only on @xmath101 and @xmath89 for which the following holds
. let @xmath2 be an @xmath101-subgaussian class of functions and assume that for each @xmath461 , @xmath462 for some @xmath463 .
then , for every integer @xmath1 and every @xmath464 , there are sets @xmath465 and @xmath466 with the following properties .
first , @xmath467 ; second , with @xmath6-probability of at least @xmath81 , if @xmath5 then : 1 . for every integer @xmath464 and every @xmath116 of cardinality @xmath362 , @xmath468 2 . for every @xmath469 and every @xmath434 of cardinality @xmath362 , @xmath470 3 .
if @xmath471 then for every @xmath472 and every @xmath473 , @xmath474 the way theorem [ thm : decomposition - of - f ] should be understood is as follows .
consider a typical @xmath5 and let @xmath475 .
for every @xmath434 the further coordinate projection satisfies @xmath476 where @xmath477 depend only on the cardinality of @xmath52 and not on @xmath52 itself , and @xmath478 .
the set @xmath479 captures the `` peaky '' part of @xmath350 and is contained in a relatively small set : a ball in @xmath480 whose radius depends on the `` complexity '' of the class @xmath2 .
the set @xmath67 consists of vectors that satisfy the desired shrinking property . indeed , for every @xmath481 and every @xmath116 of cardinality @xmath362 one has @xmath482 where the last inequality holds if @xmath483 .
proof of theorem [ thm : decomposition - of - f ] fix an integer @xmath1 . for every integer @xmath464 ,
let @xmath484 be an almost optimal admissible sequence of @xmath2 with respect to @xmath485 , and set @xmath486 to be the metric projection onto @xmath487 with respect to the @xmath74 norm .
for every such @xmath362 we will construct two sets of functions , @xmath465 and @xmath488 such that @xmath489 as follows : let @xmath490 and set @xmath491 [ and from here on we will omit the superscript @xmath362 and write @xmath492 instead of @xmath493 . note that this choice of decomposition depends only on @xmath362 and does not depend on @xmath1 . for every @xmath494 set @xmath495 and observe that @xmath496 since the expectation of all the functions in @xmath2 is the same .
thus , for every @xmath461 , @xmath497 has mean zero , and for every @xmath498 and every @xmath185 , @xmath499 let @xmath500 for @xmath501 , where @xmath78 is a constant to be named later .
because of our choice of @xmath97 , @xmath502 and @xmath503 , and thus @xmath504 hence , summing over @xmath505 , it follows that with probability at least @xmath506 for every @xmath461 and every @xmath494 @xmath507 summing the probabilities for all possible integers @xmath508 and noting that for every @xmath509 , @xmath510 , it is evident that for @xmath501 there is a set @xmath511 with probability at least @xmath512 for which the following holds . for every @xmath513 , every @xmath509 , every @xmath514 and every @xmath494 @xmath515 where we have used the fact that @xmath484 is an almost optimal admissible sequence with respect to @xmath485 .
fix @xmath516 , @xmath509 , @xmath434 of cardinality @xmath362 and @xmath517 .
consider the sets @xmath518 and @xmath519 and note that both are in @xmath457 .
since @xmath520 is increasing , then on the set @xmath521 @xmath522 in particular , if @xmath523 is a nonincreasing rearrangement of @xmath524 then by ( [ eq : f-1 ] ) applied to the set @xmath525 consisting of the @xmath526 largest elements of @xmath524 , @xmath527 thus , @xmath528 .
turning our attention to the sets @xmath529 , we will show that with high probability , for every @xmath509 and every @xmath434 of cardinality @xmath362 , the coordinate projection @xmath530 has a well behaved lipschitz constant . to that end
, fix @xmath509 , set @xmath531 and recall that for every function @xmath356 , @xmath532 .
hence , by bernstein s inequality , for every @xmath533 and every @xmath534 , @xmath535 let @xmath536 be the collection of subsets of @xmath39 of cardinality @xmath362 .
since @xmath537 and @xmath538 , then by taking @xmath539 and @xmath540 , @xmath541 summing over all possible @xmath509 , there is a subset @xmath542 of probability at least @xmath543 on which the following holds . for every @xmath509 , every @xmath544 and every @xmath545 , @xmath546\\[-8pt ] & \leq & 2l^2 u \log({ek}/{m } ) \cdot\|f_1-f_2\|_{l_2}^2.\nonumber\end{aligned}\ ] ] thus , fix a `` legal '' choice of @xmath547 for which @xmath548 .
since both ( [ eq : f-1 ] ) and ( [ eq : f-2 ] ) hold on that event , the proof of the first and second claims is evident . for the third part ,
fix @xmath549 to be named later .
again , by bernstein s inequality and since @xmath2 is @xmath101-subgaussian , then with probability at least @xmath550 , for every @xmath544 @xmath551 thus , taking @xmath552 , for @xmath553 and @xmath554 , it is evident that with probability at least @xmath555 , for every @xmath556 , @xmath557 as claimed . the first corollary of theorem [ thm : decomposition - of - f ] we shall present here is a shrinking property of @xmath558 .
[ thm : herd ] for every @xmath86 there exists a constant @xmath559 for which the following holds .
let @xmath2 be an @xmath101-subgaussian class of functions on a probability space @xmath4 and assume that for every @xmath461 , @xmath560 for some @xmath561 . then , with probability at least @xmath81 , for every @xmath116 and every integer @xmath97 that satisfies @xmath562 , @xmath563 before proving theorem [ thm : herd ] , recall the following well - known result on the expectation of a monotone rearrangement of independent standard gaussian variables ( see , e.g. , @xcite ) .
[ lemma : gauss - arr ] let @xmath564 be independent standard gaussian variables and denote by @xmath565 the nonincreasing rearrangement of @xmath566 .
then , @xmath567 moreover , @xmath568 proof of theorem [ thm : herd ] fix @xmath86 and let the sets @xmath569 be as in the proof of theorem [ thm : decomposition - of - f ] . take any @xmath570 , let @xmath571 and set @xmath572 . since @xmath573 , then by the sub - additivity of @xmath83 , it is evident that for every integer @xmath97 , @xmath574 by ( [ eq : f-2 ] ) , the mapping @xmath575 is a lipschitz function with a constant @xmath576 .
therefore , recalling that @xmath577 , @xmath578\\[-8pt ] & \leq & c_1\sqrt{\log(ek / m ) } \gamma_{2,s}(f , l_2),\nonumber\end{aligned}\ ] ] where @xmath579 . to conclude the proof ,
observe that by theorem [ thm : decomposition - of - f ] , @xmath580 , where @xmath581 and @xmath582 .
since the @xmath83 functionals are monotone with respect to inclusion and are decreasing in @xmath97 , and since @xmath583 applying the majorizing measures theorem and lemma [ lemma : gauss - arr ] @xmath584 hence , for every @xmath97 , @xmath585 , implying that for every @xmath586 , @xmath587 . combining this with ( [ eq : f_2 ] ) , it follows that for every @xmath116 , @xmath588 as claimed .
the proof of theorem [ thm : herd ] yields a stronger result than the one formulated .
it shows that with probability @xmath81 , for every @xmath116 and every @xmath589 , @xmath590 observe that in some sense , the range @xmath591 [ i.e. , @xmath592 is the interesting range of @xmath97 , since @xmath593 which decreases exponentially in @xmath97 for @xmath594 .
another outcome of theorem [ thm : decomposition - of - f ] was formulated as corollary [ corolb ] in the .
[ cor : rn ] for every @xmath86 and @xmath109 , there exist a constant @xmath437 such that the following holds .
let @xmath17 be an isotropic , @xmath101-subgaussian measure on @xmath100 , set @xmath106 to be independent , distributed according to @xmath17 and consider the random operator @xmath595 .
if @xmath596 and @xmath597 , then with @xmath6-probability at least @xmath81 , for every @xmath434 , @xmath598 where the expectation on both sides is with respect to the gaussian variables .
the proof of corollary [ cor : rn ] follows from theorem [ thm : decomposition - of - f ] and the majorizing measures theorem .
proof of corollary [ cor : rn ] since @xmath17 is an @xmath101-subgaussian measure on @xmath100 , each @xmath599 corresponds to a function @xmath600 , @xmath601 , for which @xmath602 .
let @xmath603 , set @xmath604 and put @xmath605 for which the assertion of theorem [ thm : decomposition - of - f ] holds .
fix @xmath116 of cardinality @xmath362 .
since @xmath2 is a class of linear functionals , the decomposition of @xmath2 given in theorem [ thm : decomposition - of - f ] actually implies a decomposition of @xmath26 which we denote by @xmath606 and @xmath607 .
thus , for every @xmath30 , @xmath608 , where @xmath609 for @xmath610 . since @xmath611 then @xmath612 clearly , for any @xmath613 , @xmath614 and by the shrinking property of @xmath607 , for every @xmath615 , @xmath616 therefore , by slepian s lemma ( see , e.g. , @xcite ) and since @xmath617 , @xmath618 also , recall that @xmath619 , and , just as in the proof of theorem [ thm : herd ] and by the isotropicity of @xmath17 , @xmath620 applying the majorizing measures theorem , @xmath621 and thus @xmath622 as claimed . to put corollary [ cor : rn ] in the right context , even if one considers the case where @xmath17 is the canonical gaussian measure on @xmath100 , the standard concentration estimate for the norm of a gaussian vector around its mean ( used in @xcite to prove the result for @xmath623 ) is not strong enough to allow a uniform control over all subsets of @xmath39 .
what allows one to bypass this obstacle and obtain a result even in a subgaussian setup ( in which case such a concentration result does not exist , and thus , even the result for @xmath623 is not obvious ) is the application of a cardinality - sensitive deviation argument rather than a concentration based method .
note that the logarithmic term in corollary [ cor : rn ] can not be removed .
for example , if @xmath624 and @xmath17 is the canonical gaussian measure on @xmath100 then the vector @xmath625 has the same distribution as @xmath626 , where @xmath627 are independent standard gaussian variables [ that are also independent of @xmath564 ] .
recall that @xmath536 is the collection of subsets of @xmath39 of cardinality @xmath362 and observe that @xmath629 where the last assertion is the second part of lemma [ lemma : gauss - arr ] .
therefore , with probability at least @xmath78 , there will be some @xmath545 for which @xmath630 showing that indeed , one can not remove the logarithmic term .
as we explained in previous sections , our method of selecting signs in a way that is better than choosing typical signs depends on two properties .
one is that the complexity of the set ( as captured , e.g. , by @xmath83 or the metric entropy of the set ) is small , and the other is that the set is in a good position ( e.g. , if coordinate projections shrink the set s complexity ) .
our results thus far indicate that for a subgaussian class and a typical @xmath631 , @xmath10 is essentially a set in a good position .
thus , it seems likely that the ability to choose signs that outperform the typical behavior of signs will be governed solely by the complexity of @xmath2 .
as theorem [ theoa ] , which we reformulate below , shows , this is indeed the case .
although the proof of theorem [ theoa ] is rather technical , the basic idea behind it is simple .
it follows from a combination of the two main results of the previous sections .
first of all , that a typical coordinate projection of a subgaussian class is contained in the minkowski sum of a small set and a set that satisfies a strong shrinking property .
second , that the discrepancy of sets that satisfy a shrinking property may be bounded in a nontrivial manner using their metric complexity .
[ thm : main ] for any @xmath86 , @xmath85 and @xmath109 there are constants @xmath78 and @xmath214 that depend on @xmath89 , @xmath88 and @xmath101 and for which the following holds .
let @xmath70 be an @xmath101-subgaussian class , consisting of mean zero functions .
then , for every @xmath1 there is a set @xmath632 with @xmath633 such that for every @xmath634 and every @xmath116 , @xmath635 where for every @xmath636 @xmath637 before proving theorem [ thm : main ] , let us recall the following notation . for every integer @xmath362
, @xmath638 is the largest integer @xmath97 such that @xmath639 .
if @xmath464 , then @xmath458 is the first integer for which @xmath640 . in particular , for every @xmath509 , @xmath641 ( but of course , @xmath458 could be much larger ) .
we will also say that for @xmath5 , a function class @xmath2 satisfies the shrinking property on @xmath116 with a constant @xmath36 if for every @xmath642 , @xmath643 proof of theorem [ thm : main ] fix @xmath86 and consider @xmath5 for which the assertions of theorem [ thm : decomposition - of - f ] hold . fix any integer @xmath636 and let @xmath644 be of cardinality @xmath246 . using the notation of theorem [ thm : decomposition - of - f ] , we may decompose @xmath645 , where @xmath646 , @xmath647 , and @xmath648 satisfies the shrinking property on every @xmath434 of cardinality @xmath246 with a constant @xmath649and in particular
, it does so on @xmath650 . for every @xmath461
, choose @xmath651 and @xmath652 such that @xmath653 .
hence , for every @xmath654 and any @xmath461 , @xmath655 let @xmath656 be an admissible sequence of @xmath648 which will be specified later and set @xmath657 to be a nearest point to @xmath658 in @xmath659
. as in corollary [ cor : choice - of - q - for - gamma-2 ] , if @xmath660 then there exist @xmath661 such that @xmath662 and for every @xmath663 , @xmath664 since functions in @xmath648 satisfy the shrinking property with a constant @xmath36 , then for every @xmath665 @xmath666 implying that @xmath667 let @xmath668 and continue in the same manner : first decompose @xmath669 , then apply the fact that @xmath670 is contained in an appropriate weak @xmath98 ball , and finally , since @xmath671 satisfies the shrinking property on @xmath672 , use corollary [ cor : choice - of - q - for - gamma-2 ] again , and so on . as a result of iterating this argument ,
there are nested subsets of @xmath39 , @xmath673 , with @xmath674 and of cardinalities @xmath675 and vectors @xmath676 , such that @xmath677 with the following property . for every @xmath678 ,
let @xmath679 and for every @xmath461 , @xmath680 , @xmath681 , @xmath682 one has @xmath683 therefore , there are signs @xmath684 such that , @xmath685 where the last term comes from a trivial estimate on the discrepancy of a projection of @xmath2 onto the set of coordinates @xmath686 and the shrinking phenomenon . to complete the proof
, one has to bound ( [ eq : disc - error - subgauss - proj ] ) from above .
to that end , set @xmath687 and recall that @xmath688 . to estimate the second term in ( [ eq : disc - error - subgauss - proj ] ) , since @xmath689 then @xmath690 .
therefore , @xmath691 turning our attention to the first term in ( [ eq : disc - error - subgauss - proj ] ) , for every @xmath692 let @xmath693 , set @xmath694 and @xmath695 . in other words , @xmath696 consists of all the integers @xmath288 for which @xmath697 ; @xmath698 is the largest cardinality of such a set and @xmath699 is the smallest one . since @xmath700
then for every @xmath701 , the sequence @xmath702 satisfies @xmath703 and we denote this sequence @xmath704 . since @xmath705 decays exponentially , then @xmath706 set @xmath707 , fix @xmath708 and let @xmath709 be the largest integer such that @xmath710 .
then , for every @xmath711 , @xmath712 and for @xmath713 , @xmath714 . observe that for every @xmath715 , @xmath716 and for every @xmath717 , @xmath718 .
therefore , @xmath719 for an almost optimal choice of @xmath720 .
now , for every @xmath721 and using that @xmath722 for @xmath723 and @xmath714 for @xmath724 , it is evident that @xmath725 note that @xmath726 and thus @xmath727 .
therefore , there is an absolute constant @xmath728 such that if @xmath721 then @xmath729 and @xmath730 hence , there is a constant @xmath731 such that for every @xmath732 , @xmath733 and thus , @xmath734 since @xmath735 , the claim follows .
let @xmath85 . under the assumptions of theorem [ thm : main ] and using its notation , for every @xmath736 @xmath737 and @xmath738 in particular , if @xmath739 ( i.e. , if @xmath2 is @xmath17-pregaussian ) , then @xmath740 converges in probability to 0 .
let us mention once again that the reason that theorem [ thm : main ] is meaningful is because for a typical @xmath106 , a class of mean zero functions that is @xmath101-subgaussian satisfies that @xmath741 thus , there is a true gap between the discrepancy ( or even the hereditary discrepancy ) of a typical coordinate projection and the average over signs of a coordinate projection of a pregaussian , subgaussian class @xmath2 .
in this section , our aim is to show that if @xmath2 is a subgaussian class that indexes a bounded gaussian process , then the reason for the gap between the expectation over signs of a random coordinate projection and the infimum over signs is indeed that @xmath742 . to be more precise , we show the following .
[ thm : optimal ] for every @xmath86 and @xmath743 there is a constant @xmath744 for which the following holds .
let @xmath745 be a class of mean zero functions such that @xmath133 is @xmath101-subgaussian . if @xmath746 and if the entropy numbers satisfy that @xmath747 then there is a sequence of integers @xmath748 tending to infinity , such that for every @xmath749 , with probability at least @xmath81 in @xmath750 , @xmath751 where @xmath752 is selected according to @xmath753 . in particular , @xmath132 does not converge to @xmath754 in probability . observe that this is almost the reverse direction of theorem [ theoa ] .
indeed , it is well known ( see , e.g. , @xcite , chapter 9 ) that there is no entropic characterization of classes that index a bounded gaussian process which is not continuous ; such a characterization is given by a majorizing measures argument @xcite .
however , because @xmath221 is a bounded process with a covariance structure endowed by @xmath72 , then by sudakov s inequality ( see , e.g. , @xcite , chapter 3 ) , @xmath755 on the other hand , since @xmath756 is not @xmath17-pregaussian , one can show that @xmath757 thus , up to a logarithmic factor , the entropy numbers of @xmath2 are as in theorem [ thm : optimal ] . whether theorem [ thm : optimal ] remains true using only the assumption that @xmath758 is not clear .
the idea behind the proof of theorem [ thm : optimal ] is to find a cube in a typical coordinate projection of @xmath133 .
we will first show that if @xmath133 has a `` large '' separated set with respect to the @xmath72 metric at scale @xmath759 , then its typical coordinate projection of dimension @xmath1 contains a cubic structure of dimension @xmath1 and scale @xmath759 .
the cubic structure we will be interested in is captured by the combinatorial dimension .
[ def : comb - dim ] let @xmath2 be a class of functions on @xmath258 .
for every @xmath230 , a set @xmath760 is said to be @xmath24-shattered by @xmath2 if there is some function @xmath761 , such that for every @xmath762 there is some @xmath763 for which @xmath764 if @xmath765 , and @xmath766 if @xmath767 . define the combinatorial dimension at scale @xmath24 by @xmath768 note that if @xmath2 is a @xmath769-class of functions then @xmath770 .
also , in a similar way one may define the combinatorial dimension of a subset of @xmath100 , when each vector is viewed as a function defined on @xmath403 .
it is standard to verify that if @xmath771 , then the coordinate projection @xmath772 , defined by the shattered set @xmath773 , contains a subset of cardinality @xmath774 which is @xmath775-separated with respect to the @xmath776 norm ( recall that @xmath777 is the uniform probability measure supported on @xmath773 ) , and that @xmath778 ( see lemma [ lemma : disc - vs - vc ] ) . as we mentioned in the
, the reverse direction is also true , and if @xmath779 contains a large well - separated set in @xmath72 that it must have a large combinatorial dimension at a scale that is proportional to the scale of the separation ( see @xcite for an exact statement and proof ) . a fact that will be used here and which is based on this reverse direction is the following .
[ thm : mv ] there exist absolute constants @xmath78 and @xmath214 for which the following holds .
let @xmath780 and assume that @xmath781 .
then , @xmath782 hence , the only reason that @xmath783 is almost extremal is that @xmath119 contains a large cube in a high - dimensional coordinate projection .
the key observation of this section is the following theorem .
[ thm : cube ] for every @xmath743 and @xmath86 there exist constants @xmath78 and @xmath214 that depend on @xmath784 and @xmath89 for which the following holds .
let @xmath756 be a convex , symmetric , @xmath101-subgaussian set of mean zero functions .
suppose that @xmath785 and that there is some @xmath1 for which @xmath786 . then , there is a set @xmath787 such that @xmath788 and for every @xmath789 , @xmath790 theorem [ thm : cube ] implies theorem [ thm : optimal ] because of the next lemma .
[ lemma : disc - vs - vc ] if @xmath112 , then @xmath125 first , note that @xmath791 , and thus we may assume that @xmath26 is convex and symmetric . now , let @xmath407 be @xmath89-shattered by @xmath26 with the level function @xmath97 .
fix @xmath792 and without loss of generality assume that @xmath793 .
since @xmath52 is @xmath89-shattered by @xmath26 , there is some @xmath794 for which @xmath795 when @xmath796 and @xmath797 when @xmath798 .
thus , @xmath799 as claimed . hence , from here on we may assume without loss of generality that the class @xmath2 is convex and symmetric , and that it is @xmath101-subgaussian .
the proof of theorem [ thm : cube ] requires several additional facts . to formulate them , denote for @xmath800 @xmath801 and if @xmath802 , set @xmath803 to be the minimal number of translates of @xmath208 needed to cover @xmath207 .
the first lemma we need is taken from @xcite .
[ lemma : lmpt ] let @xmath804 be a convex , symmetric set
. for @xmath805 , set @xmath806 and @xmath807 .
then , @xmath808 the second result was proved in @xcite ( theorem 2.3 ) .
although it was formulated there for subsets of @xmath100 , its proof shows that the claim is true for any subgaussian class of functions .
it implies that a random coordinate projection of @xmath2 , viewed as a mapping between @xmath72 and @xmath160 , is almost norm preserving for functions with a sufficiently large @xmath72 norm .
[ thm : mpt ] there exist absolute constants @xmath78 and @xmath214 for which the following holds .
let @xmath70 be a convex , symmetric , @xmath101-subgaussian class of functions .
for every @xmath809 and any positive integer @xmath1 , set @xmath810 then , with probability at least @xmath811 , for every @xmath461 such that @xmath812 , @xmath813 [ cor : iso ] for every @xmath109 , there are constants @xmath814 and @xmath815 that depend only on @xmath101 , for which the following holds .
let @xmath2 be an @xmath101-subgaussian , convex and symmetric class of functions for which @xmath816 . then , with probability at least @xmath817 ,
if @xmath461 and @xmath818 then @xmath819 in particular , if @xmath820 is an @xmath24-separated set in @xmath72 for @xmath821 then with probability at least @xmath817 , @xmath822 is @xmath823-separated in @xmath160 .
let @xmath78 and @xmath214 be as in theorem [ thm : mpt ] .
observe that @xmath824 and apply theorem [ thm : mpt ] for @xmath825 .
thus , @xmath826 , implying that if @xmath827 then with probability at least @xmath828 , if @xmath829 then @xmath830 turning to the second part , note that if @xmath820 is @xmath24-separated in @xmath72 for @xmath831 , then for every @xmath832 , @xmath833 and @xmath834 .
thus , the second part follows from the first one .
now we can formulate the first localization result , showing that the richness of a typical coordinate projection comes from the intersection of @xmath2 with a ball of radius @xmath134 .
[ thm : intersection ] for every positive @xmath207 , @xmath208 , @xmath101 and @xmath86 , there are constants @xmath835 , @xmath78 @xmath214 and @xmath111 depending on @xmath207 , @xmath208 , @xmath101 and @xmath89 for which the following holds .
let @xmath745 be a convex , symmetric , @xmath101-subgaussian class of mean zero functions such that @xmath836 .
fix an integer @xmath1 and assume that @xmath786 . then , with probability at least @xmath837 , @xmath838 since @xmath2 is @xmath101-subgaussian and by applying sudakov s inequality , we may assume without loss of generality that @xmath839 .
let @xmath122 be a maximal @xmath840 separated set in @xmath2 with @xmath841 .
let @xmath842 for a constant @xmath843 to be named later .
since @xmath122 is @xmath844 separated in @xmath72 , then by corollary [ cor : iso ] , with probability at least @xmath845 , if @xmath846 then @xmath822 is @xmath823-separated in @xmath847
. moreover , if @xmath461 satisfies @xmath848 then @xmath849 clearly , the condition on @xmath24 holds if @xmath850 , and since @xmath835 it follows that @xmath851 . consider the set @xmath852 . by the majorizing measures theorem and a simple application of theorem [ thm : decomposition - of - f ] , with probability at least @xmath81 for @xmath853 , @xmath854 let @xmath855 be in the intersection of the two events given by ( [ eq : iso ] ) and ( [ eq:4.2 ] ) , set @xmath856 and note that @xmath857 .
therefore , @xmath858\\[-8pt ] & \equiv&\log n(v , 8\rho b_2^{k^\prime}),\nonumber\end{aligned}\ ] ] where @xmath859 ( and thus @xmath860 as well ) .
if @xmath861 , then by lemma [ lemma : lmpt ] , ( [ eq:4.2 ] ) and ( [ eq:4.3 ] ) , @xmath862 solving this inequality for @xmath863 , it is evident that there exists a constant @xmath864 ( where @xmath865 depends on @xmath101 and @xmath89 ) for which @xmath866 . since @xmath2 is convex and symmetric and @xmath820 , then @xmath867 and by ( [ eq : iso ] ) , @xmath868 hence , there is a constant @xmath869 for which with probability at least @xmath870 , @xmath871 implying that @xmath872 the next step in the proof of theorem [ thm : cube ] is a second localization argument .
theorem [ thm : intersection ] shows that under our assumptions , there is a small ball ( of radius @xmath759 ) in @xmath2 that causes coordinate projections of @xmath2 of dimension @xmath1 to be `` rich . ''
now , one has to localize even further by truncating the functions in @xmath873 .
[ def : beta ] for every @xmath874 and every @xmath461 , let @xmath875 and @xmath876 .
for every @xmath5 let @xmath877 proof of theorem [ thm : cube ] first , by theorem [ thm : intersection ] , with probability at least @xmath837 , @xmath878 where @xmath835 .
set @xmath879 and note that by the proof of theorem [ thm : decomposition - of - f ] for the class @xmath122 and @xmath880 , each @xmath881 can be written as @xmath882 , where @xmath883 ( by the convexity and symmetry of @xmath2 ) , and @xmath884 .
moreover , if we write @xmath885 then with @xmath886 probability @xmath81 , @xmath887 , where @xmath888 . by a standard concentration argument
similar to the one used in theorem [ thm : decomposition - of - f ] , since @xmath889 then for every @xmath890 , @xmath891 .
thus , @xmath892 , and since @xmath893 , then @xmath894 where @xmath895 .
let @xmath896 for which the above estimates hold , fix @xmath448 to be named later and let @xmath897 and @xmath898 be as in definition [ def : beta ] for the set @xmath122 . consider the set @xmath899 and observe that @xmath900 .
therefore , if we set @xmath901 and select @xmath448 to satisfy that @xmath902 then , @xmath903 for an appropriate choice of @xmath904 . since @xmath905 , then @xmath906 therefore , @xmath907 note that @xmath908 therefore , by the optimal estimate in the sign - embedding theorem @xcite , there are constants @xmath909 and @xmath910 such that @xmath911 in other words , there is a set @xmath912 , @xmath913 and a vector @xmath914 such that for every @xmath915 , there is @xmath916 for which @xmath917 and it is standard to verify that @xmath918 . it remains to show that @xmath167 is @xmath919-shattered by @xmath2 itself . to that end , fix any @xmath915 , and let @xmath920 be the function for which @xmath921 observe that @xmath922 .
indeed , if there were some @xmath923 for which @xmath924 and @xmath925 , then on one hand , @xmath926 , but on the other , @xmath927 , which is impossible . in a similar fashion , @xmath928 . finally , fix @xmath929 .
if @xmath930 and @xmath931 then @xmath932 . otherwise , @xmath933 , implying that @xmath934
hence , for every @xmath929 , @xmath935 and by the same argument , for every @xmath936 , @xmath937 therefore , @xmath938 , as claimed .
the author would like to thank a. libman , m. kozdoda and the anonymous referees for their careful reading of the manuscript and for many valuable suggestions and comments .
the research leading to the results presented here has received funding from the european research council under the european community s seventh framework programme ( fp7/2007 - 2013)/erc grant agreement [ 203134 ] , from the israel science foundation grant 666/06 and from the australian research council grant dp0986563 .
|
we show that for a typical coordinate projection of a subgaussian class of functions , the infimum over signs @xmath0 is asymptotically smaller than the expectation over signs as a function of the dimension @xmath1 , if the canonical gaussian process indexed by @xmath2 is continuous . to that end
, we establish a bound on the discrepancy of an arbitrary subset of @xmath3 using properties of the canonical gaussian process the set indexes , and then obtain quantitative structural information on a typical coordinate projection of a subgaussian class . .
|
topological landau - ginzburg models are an important class of 2d topological field theories ( 2d tfts ) . when coupled to topological gravity , they give soluble examples of noncritical string backgrounds .
they are also closely related to topological sigma - models describing superstring propagation in calabi - yau manifolds ( the so - called cy / lg correspondence ) . c. vafa showed how to compute topological correlators in landau - ginzburg ( lg ) models on riemann surfaces of arbitrary genus @xcite .
remarkably , the answer is given by a simple closed formula .
the goal of this paper is to generalize the results of ref .
@xcite to riemann surfaces with boundaries .
topological boundary conditions for a 2d tft are also called topological d - branes ( of type b , in this case ) . in the case of lg models
, they have been studied in refs .
@xcite . in order to get a simple formula for topological correlators ,
one first needs a simple description of topological d - branes .
recently such a description was proposed by m. kontsevich ( unpublished ) . according to this proposal , topological d - branes are cdg modules over a certain cdg algebra which depends on the superpotential @xmath0 of the landau - ginzburg model .
spaces of open strings stretched between d - branes are identified with the space of morphisms in the derived category of cdg modules .
the ope algebra of boundary operators is encoded in the composition law of morphisms .
it should be noted that the language of derived categories is very suited to the discussion of topological d - branes and has been used in the context of open string field theory ( see e.g. refs .
@xcite ) . in our previous paper
@xcite , we have given a physical derivation of kontsevich s conjecture .
we also performed some explicit computations in the case when the lg model is massive .
the derivation is reviewed in section [ cdg ] below , with some improvements in exposition .
the main physical idea is to represent topological d - branes as bound states of branes and anti - branes of maximal dimension .
kontsevich s conjecture concerning the boundary ope follows from the analysis of the world - sheet action for the lg model with boundaries in the presence of both branes and anti - branes .
> from the physical viewpoint , computing the boundary ope is only part of the problem .
one would also like to know the topological metric on the space of open strings . together with the boundary ope ,
this provides complete information about topological boundary correlators on a disk .
more generally , one may consider disk correlators involving both bulk and boundary insertions .
even more generally , one may consider correlators on a riemann surface with arbitrary number of boundaries and handles , and arbitrary insertions of bulk and boundary operators . in this paper
we give explicit formulas for all these correlators . in some special cases ,
topological disk correlators have been computed in ref .
@xcite .
> from the mathematical viewpoint , an interesting result of this paper is a simple explicit formula for the topological open - string metric , i.e. for a non - degenerate pairing between vector spaces @xmath1 and @xmath2 , where @xmath3 and @xmath4 are arbitrary b - branes ( i.e. cdg modules ) , and @xmath1 is the space of morphisms from @xmath3 to @xmath4 in the derived category of cdg modules .
the non - degeneracy of the pairing follows from the properties of the path integral defining topological lg models , or equivalently , from the axioms of 2d tft with boundaries @xcite .
the pairing is natural , in an obvious sense . in particular , the algebra of endomorphisms of any b - brane is naturally a graded frobenius algebra .
unlike in the closed - string case , this algebra is not supercommutative , in general . to appreciate better the significance of the open - string metric , recall that in the case of b - branes on a calabi - yau manifold @xmath5 this metric is given by the usual pairing between @xmath6 and @xmath7 , where @xmath3 and @xmath4 are objects of the bounded derived category of @xmath5 .
the pairing is non - degenerate as a consequence of the serre duality theorem . on a general compact complex manifold serre duality
tells us that there is a non - degenerate pairing between @xmath6 and @xmath8 , where @xmath9 is the canonical line bundle .
guided by this example , the authors of ref .
@xcite introduced the notion of a serre functor on a triangulated category .
basically , a serre functor @xmath10 on a triangulated category @xmath11 is an autoequivalence of @xmath11 such that there exists a natural non - degenerate pairing between @xmath1 and @xmath12 .
a serre functor on @xmath11 , if it exists , is unique . using this terminology
, one can say that all categories of topological branes have a trivial serre functor .
this is interesting , because such categories are hard to come by .
for example , the category of b - branes in a landau - ginzburg model must have a trivial serre functor , and our result for the open - string metric gives an explicit formula for the corresponding pairing .
the non - degeneracy of this pairing is far from obvious , and it would be interesting to find a direct mathematical proof of this fact .
in this section we review the physical derivation of the boundary ope algebra in the topologically twisted lg model .
let @xmath13 , and @xmath0 be a holomorphic function on @xmath5 ( the superpotential ) .
the worldsheet bulk lagrangian density is given by ( cf .
@xcite ) @xmath14 for definiteness , we have chosen the standard flat kahler metric on @xmath5 .
if the worldsheet @xmath15 has no boundary , the action is invariant up to a total derivative under the @xmath16 worldsheet supersymmetry .
what will be important to us is the type - b @xmath17 susy transformation @xmath18 our conventions are the same as in ref .
the presence of worldsheet boundaries breaks @xmath17 susy in general .
for instance , a brane wrapped on the whole target space @xmath5 is well known to be non - supersymmetric .
however it is not unreasonable to expect that all supersymmetric d - branes ( in our case , b - branes ) can be obtained by the process of tachyon condensation on top - dimensional brane - anti - brane systems .
in fact this is the basic assumption that we will make in this paper . as we shall see , the requirement of type - b @xmath17 susy invariance after tachyon condensation naturally gives rise to the mathematical structure of boundary ope algebra mentioned above .
more specifically , we want to add a boundary term to the action for the tachyon and gauge fields associated with a brane - anti - brane system which wrap the whole space @xmath5 , such that the type - b supersymmetry can be restored .
to this end , we first modify the bulk lagrangian density by a total derivative @xmath19 this does not change the physics in the bulk , although it makes easier to write down the boundary action . taking the standard neumann boundary conditions for all bulk fields , the modified bulk action , @xmath20 , has the following variation under type - b susy : @xmath21 on the other hand
, the modified bulk action is invariant under an @xmath22 subalgebra : @xmath23 .
this means that the boundary term we should add to recover @xmath17 invariance must be @xmath22-supersymmetric by itself .
let the brane - anti - brane system be described by a pair of hermitian vector bundles @xmath24 with unitary connections .
we may regard @xmath25 as a @xmath26-graded vector bundle . since @xmath13 , these bundles are necessarily trivial . for simplicity ,
let us assume that @xmath27 and @xmath28 have rank one .
the generalization to higher rank is discussed below .
the boundary degrees of freedom are described by a boundary fermion @xmath29 , which is taken to be a section of @xmath30 .
this field couples to a boundary tachyon @xmath31 , which is a section of the bundle @xmath32 , and the gauge field @xmath3 which is a connection on the line bundle @xmath30 .
the gauge field @xmath3 is the difference of gauge fields on the brane and the anti - brane .
the boundary action also depends on the sum of the two gauge fields , which we denote @xmath33 .
the field @xmath33 is a connection on the line bundle @xmath34 .
the @xmath22-supersymmetric boundary action has the standard form @xcite @xmath35\ ] ] where @xmath36 , and @xmath37 , @xmath38 are superfield completions : @xmath39 expressing in terms of components after integrating out the massive field @xmath40 , one gets @xmath41 \label{eq : sb}\ ] ] where @xmath42 is the covariant derivative associated with the connection @xmath3 , @xmath43 is given by @xmath44 and @xmath45 is defined similarly , but with @xmath46 replaced with @xmath47 .
now we require that the total action @xmath48 preserves the type - b @xmath17 supersymmetry .
the variations of @xmath31 and @xmath3 are determined by variations of the bulk fields @xmath49 .
the only field whose @xmath17 variation is not known is the boundary fermion @xmath29 .
however we know that its variation can not involve @xmath50 since such a term destroys @xmath17 invariance .
gauge invariance then dictates the following general form @xmath51 where @xmath52 is an unknown function , and @xmath53 are a pair of sections on @xmath30 which do not depend on @xmath29 .
it is not difficult to see that @xmath17 invariance implies @xmath54 , which in turn is satisfied if and only if @xmath55 in other words , @xmath17 invariance requires that the connection @xmath3 defines a holomorphic structure on the line bundle @xmath30 .
similarly , @xmath17 invariance requires that the curvature of @xmath33 be of type @xmath56 .
thus both @xmath27 and @xmath28 must be holomorphic line bundles . to have @xmath17 invariance one must further demand that all @xmath29-dependent ( resp .
@xmath57-dependent ) terms in @xmath58 cancel each other .
it can be shown that this amounts to two conditions on @xmath59 and @xmath60 .
first , they must be purely bosonic , i.e. they depend only on @xmath61 and not on @xmath62 .
second , they must satisfy the following equations @xmath63 these two equations have covariant solutions if and only if @xmath31 takes the following form @xmath64 together with the identification @xmath65 one can also show that these conditions are sufficient for cancellation of @xmath29-dependent and @xmath50-dependent terms in @xmath58 .
the final requirement of @xmath66 invariance comes from demanding that the remaining terms in @xmath58 , which now do not involve @xmath29 at all , exactly cancel the bulk variation @xmath67 .
it is straightforward to show that the variation of @xmath68 is given by @xmath69 which cancels the bulk variation if and only if @xmath70 if the bundles @xmath27 and @xmath28 have rank higher than one , there is an additional complication .
namely , the integrand in the string path integral is no longer an exponential of a local action : there is also a non - local factor @xmath71 where dots denote terms depending on the fermions and the boundary tachyon . in order to figure out the conditions imposed by @xmath17 susy
, it is desirable to have a local action .
this can be achieved by representing the path - ordered exponential by a path - integral over auxiliary degrees of freedom living on the boundary of the worldsheet .
in fact , there are several way to do this ; for example , the boundary degrees of freedom can be bosonic or fermionic .
the action for all these fields , and their transformation properties under @xmath17 susy will be described elsewhere .
the results are the following .
first of all , both @xmath27 and @xmath28 must be holomorphic vector bundles .
second , the boundary tachyon , which is still a section of @xmath32 , must have the form @xmath72 where @xmath73 is a holomorphic section of @xmath32 , and @xmath74 is a holomorphic section of @xmath30 .
finally , @xmath73 and @xmath74 must satisfy @xmath75 this implies , among other things , that @xmath27 and @xmath28 must have the same rank . in the b - twisted model , the anti - holomorphic fermions @xmath76 and @xmath77 become 0-forms on the world - sheet , while the holomorphic components @xmath78 and
@xmath79 combine into a 1-form @xmath80 .
the brst transformation is obtained from type - b @xmath66 variation by setting @xmath81 .
the brst charge @xmath82 is the sum of the bulk and boundary contributions .
the bulk contribution is standard .
the boundary part @xmath83 must generate the correct transformation @xmath84 because the connection defines a holomorphic structure on @xmath32 , one can locally trivialize the connection to @xmath85 . since @xmath5 is affine , one can actually take @xmath86 globally .
the boundary brst charge in this gauge is given by @xmath87 since @xmath86 , @xmath73 and @xmath74 are simply holomorphic matrix - valued functions of @xmath49 .
one may check explicitly that the brst invariance is restored after adding the boundary contribution .
the bulk brst charge is known to be given by @xmath88 where @xmath89 .
the bulk charge is not nilpotent but satisfies @xmath90 the total brst charge squares to zero by virtue of @xmath91 , as well as @xmath92 , which comes from the fact that @xmath73 and @xmath74 are holomorphic .
physical states in the topological theory must be annihilated by the total brst charge .
since we are working with a topological theory , the zero - mode approximation is adequate .
the zero - mode hilbert space consists of sections of @xmath93 .
the operators @xmath29 and @xmath57 are realized by the block matrices @xmath94 the boundary part of the brst operator is therefore identified with @xmath95 while the bulk part reduces to the standard dolbeault operator @xmath96 a section @xmath61 of @xmath93 is brst - invariant with the bulk fields @xmath49 . ] if and only if @xmath61 is holomorphic and super - commutes with @xmath83 .
if @xmath61 is a super - commutator of @xmath83 and a holomorphic section @xmath97 , then it is brst trivial .
let us denote by @xmath98 the space of holomorphic sections of @xmath93 .
since we are working on @xmath99 , this is simply the space of @xmath100 block matrices with holomorphic entries .
it has a natural decomposition into even and odd ( diagonal and off - diagonal ) components .
let @xmath101 denote the following operator on @xmath98 : @xmath102\ ] ] where the brackets denote the super - commutator .
it squares to zero by virtue of @xmath103 in @xmath0 in what follows .
] in view of the above , the space of physical states can be identified with the cohomology of @xmath101 .
obviously , the ope algebra corresponds simply to matrix multiplication of elements of @xmath98 .
now let us reformulate the structure which appeared above in mathematical terms .
a cdg algebra is a triple @xmath104 , where @xmath3 is a @xmath26-graded associative algebra , @xmath105 is an odd derivation of @xmath3 , and @xmath4 is an even element of @xmath3 satisfying @xmath106 $ ] for any @xmath107 . here
the brackets denote the super - commutator .
a cdg - module over a cdg algebra is a pair @xmath108 , where @xmath109 is a @xmath26-graded module over @xmath3 , and @xmath110 is an odd endomorphism of the @xmath26-graded vector space @xmath109 which satisfies the super - leibniz identity @xmath111and the `` twisted differential '' condition @xmath112 in the special case @xmath113 cdg algebras and modules are called dg algebras and modules . in the above definition one can replace @xmath26-grading with @xmath114-grading .
then one requires @xmath105 and @xmath110 to be of degree one , and @xmath4 must be of degree two . in
what follows , we will be working with @xmath26 gradings , unless specified otherwise .
a well - known example of a ( @xmath114-graded ) dg algebra is the algebra of differential forms on a smooth manifold , equipped with the de rham differential .
more generally , one may consider the algebra of sections of the endomorphism bundle of a flat vector bundle with a connection , equipped with the flat covariant differential ( the so - called twisted de rham complex ) .
cdg algebras appear when we replace flat connections with arbitrary connections . in this example
@xmath4 is simply the curvature 2-form of the connection .
this is the origin of the term `` cdg algebra '' ( cdg means `` curved differential graded '' ) .
let us make three trivial remarks about the definitions of cdg algebras and modules .
first , if @xmath115 , then a cdg algebra can not be naturally regarded as a cdg - module over itself , since the equation @xmath106 $ ] is in conflict with the desired @xmath116 .
second , if the algebra @xmath3 is super - commutative , then the pair @xmath117 is a dg - algebra , but dg modules over this dg - algebra are not the same as cdg - modules over the cdg - algebra @xmath104 , unless @xmath113 . indeed , in the former case we have the condition @xmath118 , while in the latter case we have @xmath119 .
third , if @xmath120 , but @xmath115 , then @xmath110 is simply an odd endomorphism of the module @xmath109 ( i.e. @xmath121 for all @xmath122 and @xmath123 ) satisfying @xmath124 .
the results of the previous section can be rephrased as follows .
consider a cdg algebra @xmath125 , where @xmath126 is the algebra of polynomials on @xmath99 , and @xmath0 is the superpotential .
the graded bundle @xmath127 which describes the brane - anti - brane system is a graded module over @xmath126 , and the tachyon field @xmath128 gives it the structure of a cdg module over @xmath129 . since we are working on @xmath99 , the module @xmath109 is simply a sum of several copies of @xmath126 ( i.e. a free module ) , and @xmath110 can be viewed as a matrix with elements in @xmath126 . in order to rephrase in this language the boundary ope algebra ,
we need to discuss endomorphisms of cdg modules .
we are interested in the situation when @xmath3 is purely even and commutative , and @xmath120 .
consider first endomorphisms of @xmath109 regarded simply as an @xmath3-module .
these are linear maps from @xmath109 to @xmath109 which commute with the action of @xmath3 . if @xmath109 is a free module of rank @xmath130 ( i.e. a sum of @xmath130 copies of @xmath3 regarded as a module over itself ) , @xmath131 is the algebra of @xmath132 matrices with entries in @xmath3 .
this algebra has natural @xmath26-grading : endomorphisms which do not change the grading on @xmath109 are declared even , while those which change the grading are declared odd . in other words , if @xmath133 where @xmath134 and @xmath135 are free modules of rank @xmath136 and @xmath137 , @xmath138 , then endomorphisms which map @xmath134 to @xmath134 and @xmath135 to @xmath135 are even , while those which map @xmath134 to @xmath135 and vice versa are odd . in the matrix notation , we can write an endomorphism @xmath61 as a block matrix @xmath139 where @xmath140 are matrices with entries in @xmath3 of sizes @xmath141 , @xmath142 , @xmath143 , and @xmath144 , respectively . with the above grading ,
@xmath3 and @xmath110 are even , and @xmath4 and @xmath145 are odd .
the endomorphism algebra has a natural odd derivation which squares to zero .
it is given by @xmath146,\ ] ] where the brackets denote the supercommutator .
note that @xmath147 because of the super - jacobi identity and the fact that @xmath148 to summarize , if @xmath109 is a cdg module , then @xmath131 is a dg algebra .
it is easy to see that in the case @xmath149 , @xmath150 , this dg algebra is precisely the algebra of boundary operators in the zero - mode approximation .
the differential @xmath101 is identified with the boundary part of the brst operator .
thus the space of physical states is simply the cohomology of @xmath101 , and the boundary ope structure is simply the algebra structure on the cohomology of @xmath101 .
one can easily generalize this to the case when strings begin on one brane and end on another one . in this case
we have a pair of cdg modules @xmath109 and @xmath151 , and the space of morphisms between them is a graded vector space @xmath152 equipped with a differential @xmath101 defined as follows : @xmath153 this vector space is identified with the space of operators which change the boundary conditions ( in the zero - mode approximation ) .
the cohomology corresponding to the differential @xmath101 is identified with the brst cohomology ( i.e. the space of physical boundary changing operators modulo brst trivial ones ) .
if we have three different branes , say , @xmath154 , and @xmath9 , we can compose morphisms from @xmath109 to @xmath151 and from @xmath151 to @xmath9 to obtain morphisms from @xmath109 to @xmath9 .
this composition descends to the cohomology of @xmath101 , as one can easily see , and corresponds to composition of boundary - changing operators .
consider now a category whose objects are b - branes , whose spaces of morphisms are spaces of physical states for strings stretched between branes , and where composition of morphisms is defined by means of product of boundary - changing operators .
consider also a category whose objects are free cdg modules over @xmath149 , whose morphisms are cohomology classes of the differential @xmath101 , and where composition is the obvious one .
the above discussion can be restated by saying that these two categories are equivalent .
the second category coincides with the derived category of cdg modules over the cdg algebra @xmath155 .
thus we have derived kontsevich s conjecture .
the main physical assumption in the derivation is that any b - brane can be obtained as a bound state of branes and anti - branes of maximal dimension ( d - branes wrapping the whole target space ) .
> from the mathematical point of view , there is the following subtlety : the derived category is not simply a category , but has some extra structure ( the shift functor and the so - called distinguished triangles ) . this structure makes the derived category into a triangulated category .
ideally , one would like to show that triangulated structures also agree .
one can easily see that the shift functor on the derived category corresponds to the functor on the b - brane category which exchanges branes and anti - branes , i.e. exchanges @xmath73 and @xmath74 in eq .
( [ eq : dfg ] ) . as for distinguished triangles
, they should show up in physics if we consider stability conditions for b - branes @xcite .
we hope to return to this issue in the future . as discussed in ref .
@xcite , the category of b - branes is non - zero only if the set @xmath156 is the critical level set of @xmath0 , i.e. it contains some critical points of @xmath0 .
furthermore , it is sensitive only to the infinitesimal neighborhood of the critical points .
it turns out one can make a very precise mathematical statement about how this localization occurs , i.e. one can describe the category of b - branes entirely in terms of the geometry of the critical level set @xcite .
unfortunately , the mathematical statement is not so easy to understand for a non - mathematician .
nevertheless , we will try to give some idea about it , because in some cases it can be used very effectively to determine the structure of the category of b - branes ( see below ) .
consider the critical level @xmath156 .
we stress that it should be regarded as an algebraic variety , contains nilpotent elements .
however we shall use the less accurate but hopefully more familiar term _ variety _ in this paper to avoid scaring away the readers . here
we make the standard ( but often unjustified ) assumption that someone is still reading the paper .
] rather than as a set . for example , if @xmath157 , the set defined by @xmath158 consists of a single point , and does not depend on @xmath159 . on the other hand ,
the algebra of functions on the algebraic variety @xmath160/w)$ ] is @xmath161/z^n$ ] , and depends on @xmath159 .
if @xmath162 is a critical value of @xmath0 , the variety @xmath163 given by @xmath156 is singular . in ref .
@xcite the following construction was proposed .
consider the bounded derived category of coherent sheaves on @xmath163 .
its objects are bounded complexes of coherent sheaves on @xmath163 .
a perfect complex is an object of @xmath164 which is quasi - isomorphic to a bounded complex of locally free sheaves .
one may consider a full subcategory @xmath165 of @xmath164 whose objects are perfect complexes , and also the quotient category @xmath166 .
the point of this construction is that for a smooth @xmath163 the category @xmath165 coincides with the whole @xmath164 , and @xmath167 is trivial .
thus @xmath167 is only sensitive to singularities of @xmath163 . the main result of ref .
@xcite is that the category of b - branes is equivalent to @xmath167 where @xmath163 is the critical variety corresponding to @xmath0 .
this result gives a precise meaning to the physical statement that strings are confined to the critical points of @xmath0 .
in particular , in the case @xmath157 , corresponding to the a - series of @xmath168 minimal models , the variety @xmath163 is zero - dimensional , and one can completely describe the category @xmath167 . as shown in ref .
@xcite , the category of b - branes is semi - simple , i.e. every object is a finite direct sum of indecomposable objects .
furthermore , there are @xmath169 indecomposable b - branes labeled by @xmath170 , and the endomorphism algebra of the @xmath171-th brane ( i.e. the algebra of boundary operators which do not change the boundary condition ) is @xmath172/(z^p,\theta^2+z^{n-2p})$ ] , where @xmath173 is an even ( bosonic ) variable , @xmath174 is an odd ( fermionic ) variable , and @xmath175 .
we discuss this and other examples in detail elsewhere . here
we just remark that although it is easy to construct all irreducible b - branes as cdg modules , it is far from obvious that _ any _ b - brane is isomorphic to a sum of these . on the other hand , the results of ref .
@xcite quickly reduce the problem of classification of branes to the problem of classification of matrix solutions to the equation @xmath176 .
the latter is solved by the jordan normal form theorem .
in this section we derive a closed - form formula for disk correlators of arbitrary bulk and boundary topological operators .
the crucial element of the derivation is to represent the b - brane on which the boundary of the disk ends as a brane - anti - brane system with tachyon fields turned on , as discussed in section 2 . for simplicity
, we assume the target space @xmath5 to be affine .
the brane - anti - brane pair is denoted by @xmath24 with @xmath177 being equal rank holomorphic vector bundles .
the formula we seek to prove is @xmath178}{\part_1w\part_2w\ldots\part_nw } \label{eq : form}\ ] ] a few comments are in order .
the bosonic bulk chiral fields are renamed here by @xmath179 , and @xmath97 is an arbitrary bulk operator represented by an element of @xmath180/i$ ] , with @xmath181 being the ideal generated by all the differentials @xmath182 .
the twisted differential @xmath110 is defined as in ( [ eq : dfg ] ) , and @xmath183 is an arbitrary boundary operator , i.e. a representative of a class in the @xmath101-cohomology .
the supertrace `` @xmath184 '' is a twisted trace operator in the fundamental representation of @xmath185 : @xmath186 where @xmath187 is the grading operator defined by @xmath188 thus the integrand in eq .
( [ eq : form ] ) is a meromorphic @xmath159-form on @xmath99 .
finally , the contour integral is performed over a union of @xmath159-dimensional lagrangian tori which enclose all the critical points of @xmath189 @xcite .
this integral can be regarded as a generalized residue , and can also be written as an integral of a @xmath190-form over a large @xmath190-dimensional sphere in @xmath99 . the @xmath190-form is the product of the numerator in eq .
( [ eq : form ] ) , which is a holomorphic @xmath159-form , and a certain @xmath169-form depending on @xmath0 alone whose explicit form can be found in ref @xcite .
this way of writing the generalized residue emphasizes that it can be computed using only the asymptotic behavior of @xmath191 , and @xmath61 .
in particular , no explicit knowledge of the critical points of @xmath0 is required . in the case
@xmath192 , the two ways of writing the formula become almost identical : in both cases one has an integral of a meromorphic 1-form along a contour in the complex plane , and the only difference is whether the contour consists of several small circles enclosing all the singularities of the 1-form , or is a single large circle enclosing all the singularities . in the next section
, we shall generalize ( [ eq : form ] ) to riemann surfaces with arbitrary number of handles and boundaries . in order to evaluate the path integral on a disk
, we will use the fact that it is a topological invariant and so does not depend on the worldsheet metric @xmath193 .
therefore we may rescale the metric by @xmath194 and evaluate the integral in the limit @xmath195 .
it is also well - known that the path integral is localized around instantons which are fixed points of the brst operator @xcite . in our case
these instanton configurations are given by @xmath196 around each of these instantons , the contributions of bosonic and fermionic non - zero modes cancel each other , and the path integral is reduced to an ordinary integral over the zero modes . for the b - twisted lg theory on a riemann surface without boundaries , the zero modes come from constant scalars @xmath197 , and closed 1-forms @xmath198 . for the disk topology , there are no @xmath199 zero modes .
furthermore , although the boundary condition does not affect the bosonic zero modes , it leads to the following relation among the fermionic zero modes : @xmath200 the path integral now reduces to an ordinary integral with measure @xmath201 the integrand consists of two factors . the modified bulk action ( cf .
section 2 ) contributes , apart from an unimportant numerical factor , @xmath202 although we record explicit @xmath203-dependence here , it is understood that we take @xmath204 in the end .
if the theory were bosonic , the presence of a boundary would introduce an extra factor , the holonomy of the connection @xmath3 along @xmath205 : @xmath206 what we need is a supersymmetric extension which incorporates the the tachyon .
such an extension is well - known and is given by ( see for example @xcite ) @xmath207 with @xmath208 in general , @xmath109 also contains @xmath209 and @xmath210 in the diagonal entries , but for the zero - mode integration they do not appear .
the former term is not there since @xmath211 .
the latter term is not there because the curvature is of type @xmath56 and must couple to @xmath212 ; however , there is no @xmath213 zero mode .
note also that the supersymmetric analogue of the wilson loop involves a super - trace , rather than the ordinary trace .
this arises as a consequence of the gso projection generalized to the case of the brane - anti - brane system @xcite .
since we are dealing with zero modes , the path ordering has no effect . recalling that @xmath214 and taking the trivialization @xmath215 , the disk partition function in the limit @xmath195 becomes @xmath216\ ] ] to evaluate this finite - dimensional integral
, we will use the same trick as that employed by vafa in the closed string case @xcite .
namely , we will first consider the situation where all critical points of @xmath0 are isolated and non - degenerate .
as explained in ref .
@xcite , in this case topological correlators are sums over all critical points , and each term in the sum can be evaluated by expanding everything in taylor series to a first non - trivial order .
in particular , the superpotential can be approximated by a quadratic function , and we expect that b - branes can be `` approximated '' by clifford modules .
below we will show very explicitly how clifford modules arise .
once we have an explicit expression for the integral , we show that it can be written as a generalized residue of a meromorphic @xmath159-form on @xmath99 , namely , as eq .
( [ eq : form ] ) .
then we can infer eq .
( [ eq : form ] ) in the general case by appealing to the holomorphic dependence of topological correlators on the coefficients of @xmath0 .
so let us choose coordinates @xmath217 centered on the critical point , and let @xmath218 . here
@xmath219 is a non - degenerate symmetric bi - linear form , and dots denote terms of higher order in @xmath173 .
the twisted differential @xmath110 must vanish at @xmath220 for the contribution to the integral to be non - zero ( see above ) , so its taylor series expansion starts as follows : @xmath221 here @xmath222 are @xmath173-independent odd operators on the graded vector space @xmath109 , where @xmath109 is defined by the condition that the graded holomorphic vector bundle @xmath223 be @xmath224 .
the equation @xmath225 implies that @xmath226 satisfy the clifford algebra relations @xmath227 thus @xmath109 is a graded clifford module , as anticipated .
all graded clifford modules are isomorphic to modules of the form @xmath228 , where @xmath20 is an irreducible clifford module ( spinor module , see appendix of ref .
@xcite ) , and @xmath98 is a vector space .
actually , there is a small difference between even and odd @xmath159 : for @xmath159 even there are two inequivalent spinor modules which differ by a parity reversal , while for @xmath159 odd there is only one spinor module , since the parity reversal can be undone by an automorphism .
this means that for @xmath159 even @xmath98 should be a graded vector space , while for @xmath159 odd @xmath98 is not graded .
we replace @xmath61 by its value at @xmath220 which we denote @xmath229 . since @xmath61 supercommutes with @xmath110 , @xmath229 must supercommute with all @xmath226 . in the case
when @xmath159 is even , this implies that @xmath230 has the form @xmath231 , where @xmath232 . in the case
when @xmath159 is odd , we have two possibilities for @xmath229 : either @xmath233 , or @xmath234 here @xmath235 denotes the grading operator on @xmath20 , and @xmath122 is an arbitrary element of @xmath236 . to simplify the integral still further ,
note that although superficially it depends both on the ( flat ) khler metric on @xmath99 and a hermitian metric on the clifford module @xmath109 , in fact topological correlators can not depend on either .
this is a general property of topological correlators in the b - model , but we can make it explicit in the case at hand .
we can rewrite the integral over bosonic and fermionic variables as an integral of an inhomogeneous differential form : @xmath237}\alpha(0)\phi(0)\right ] dz_1\ldots dz_n.\ ] ] hermitian metric enters only through @xmath238 , and the latter appears only in the anti - commutator with @xmath239 .
now recall that the latter combination is the total ( bulk plus boundary ) brst charge in the zero - mode approximation .
a standard argument then implies that the integral does not depend on the choice of hermitian metrics .
we can use this freedom to simplify the integral further .
first , we can use a linear change of coordinates @xmath240 to bring the matrix @xmath82 to the standard form ( identity matrix ) .
the effect of this is to multiply the integral by @xmath241 , where @xmath242 .
second , we can choose the metric on the vector space @xmath109 so that @xmath243 .
the integral thus becomes @xmath244dz_1\ldots dz_n.\ ] ] up to an unessential numerical factor , this is equal to @xmath245 note that this is independent of @xmath246 , as expected on general grounds .
returning to the original coordinates , we get : @xmath247 this is a contribution of a single isolated nondegenerate critical point .
if there are several such points @xmath248 , each contributes independently : @xmath249 now recall that one of the properties of the generalized residue is the following identity , holding whenever the superpotential has only non - degenerate critical points : @xmath250 here @xmath251 is an arbitrary holomorphic function on @xmath99 .
since @xmath252 , it follows that @xmath163 is given by eq .
( [ eq : form ] ) .
suppose now that @xmath0 has degenerate critical point(s ) .
consider the following deformation @xmath253 where @xmath254 is a holomorphic function such that the deformed superpotential has only non - degenerate critical points and the same asymptotic behavior as @xmath189 ( i.e. @xmath254 does not bring in extra critical points from the infinity ) . in the limit @xmath255 one recovers the original theory as some or all non - degenerate critical points of the deformed theory merge .
it is quite reasonable to expect that all b - branes associated with a degenerate critical point derive from b - branes associated with the merging simple critical points for certain choices of @xmath254 . by holomorphy , the disk correlators
should still be given by ( [ eq : form ] ) .
this concludes the derivation of eq .
( [ eq : form ] ) .
it is instructive to check that eq .
( [ eq : form ] ) enjoys the properties expected from a topological disk correlator .
we first show that the formula ( [ eq : form ] ) leads to the correct selection rule for the disk correlators .
recall that a generic superpotential leaves unbroken a @xmath26 r - symmetry , which gives the @xmath26-grading of the boundary operators . for odd @xmath159
this symmetry is anomalous on a disk , thus the disk correlator vanishes unless @xmath256 it is easy to see that the formula ( [ eq : form ] ) obeys this selection rule . we proceed to show that ( [ eq : form ] ) vanishes whenever @xmath97 or @xmath61 is brst - exact . for the bulk insertion @xmath97
, this means that the residue must vanish if @xmath97 belongs to the ideal generated by @xmath257 .
this follows from the basic property of the residue : @xmath258 vanishes whenever @xmath259 belongs to the above - mentioned ideal .
now let us assume that @xmath61 is brst exact , i.e. @xmath260 $ ] , with @xmath261 satisfying @xmath262 .
note that the twisted differential satisfies @xmath263 since an insertion of @xmath264 makes the residue vanish , one can effectively treat @xmath265 and @xmath110 as anti - commuting .
it is then an easy computation to show that the disk correlator given by ( [ eq : form ] ) for @xmath266 $ ] indeed vanishes .
we make one more consistency check before closing this section .
in general there is an ordering issue if multiple boundary operators are inserted , and the correlator must preserve a super - cyclic symmetry .
this is already visible with two boundary operators @xmath267 .
suppose @xmath122 and @xmath268 are initially close to each other . by moving @xmath122 around @xmath205 to the other side of @xmath268
, the disk correlator must flip sign if both @xmath122 and @xmath268 have odd degrees , and must be unchanged otherwise . furthermore , if one interprets @xmath265 s appearing in ( [ eq : form ] ) as `` operators '' inserted on the boundary ,- cohomology ] they should super - commute with true boundary operators for consistency . in the following
we show that ( [ eq : form ] ) does satisfy these consistency requirements .
first , let @xmath122 be a boundary operator , then @xmath269=0 $ ] . acting by @xmath270 ,
one gets @xmath271 = - [ d , \part a]\ ] ] since @xmath272 are holomorphic sections of @xmath273 , the right hand side is a brst - exact operator .
therefore @xmath265 and @xmath122 can be regarded as super - commuting , just as anticipated .
now assume that @xmath122 and @xmath268 are two brst - invariant boundary operators .
we want to show that the disk correlator for @xmath274 $ ] as given by ( [ eq : form ] ) vanishes .
the proof is straightforward .
as always , it suffices to consider the case when @xmath275 we already showed that one can make the following substitution @xmath276 { \rightarrow}(-1)^{n\cdot\deg\,a}{{\rm str}}\big[a\,(\part d)^n b\big].\ ] ] using the identity @xmath277 , one immediately gets @xmath278\,\big\rangle_{\rm disk } \;=\ ; 0.\ ] ] this establishes the fact that ( [ eq : form ] ) posesses the super - cyclic symmetry , as expected on general grounds .
let us now consider the case of a general oriented riemann surface @xmath15 , which has @xmath279 handles and @xmath280 boundary components @xmath281 each of which is diffeomorphic to a circle .
the orientation of @xmath15 induces a natural orientation on the @xmath145 s . without loss of generality we may consider the case when there is a single bulk operator insertion @xmath97 , and a single boundary operator insertion @xmath282 at every boundary component @xmath283 .
we first focus on the case @xmath284 , i.e. a disk with several handles attached .
one could in principle evaluate the path integral explicitly , as is done above , but there is a far more efficient way to arrive at the answer .
we note that this correlator can be thought of as an inner product of two closed string states , using the usual closed - string topological metric . the first closed string state is obtained by performing a path integral over a two - dimensional world - sheet which looks like a genus-@xmath285 oriented surface with a circular hole , and an insertion of @xmath97 .
we will denote this state by @xmath286 from the results of ref .
@xcite it follows easily that the corresponding element in @xmath180/\partial w$ ] is represented by @xmath287 where @xmath288 is the hessian of @xmath189 .
the second closed string state is the boundary state of the b - brane with an insertion of @xmath61 on the boundary .. ] from our result for the @xmath289 correlator we infer that this boundary state is @xmath290 here it is understood that the overall factor @xmath291 is removed , so that the boundary state is a holomorphic function , rather than a holomorphic @xmath159-form .
hence the correlator of interest is given by @xmath292}{\part_1w\part_2w\ldots\part_nw}\ ] ] if we have more than one boundary component , we can follow a similar strategy .
we take a genus-@xmath285 closed string amplitude with @xmath293 operator insertions , and replace @xmath294 of them with the boundary states corresponding to @xmath294 boundary circles @xmath283 .
we will denote by @xmath295 the boundary part of the brst operator for the @xmath296 brane .
it follows that for general @xmath285 and @xmath294 the topological correlator is given by @xmath297.\ ] ] this generalizes vafa s formula to the open string case .
note that since we removed the factors @xmath298 from the boundary states , the integrand is a meromorphic function . to compute the correlator
, one has to compute its residue , i.e. to multiply it by @xmath298 and integrate over a union of lagrangian tori given by @xmath299 .
we have computed all correlators in the b - twisted landau - ginzburg model with boundaries .
perhaps the most interesting extension of these results would be a computation of the space - time superpotential in type ii compactifications on calabi - yau manifolds with d - branes . at special points in the moduli space
such compactification are described by landau - ginzburg orbifolds ( gepner models ) , and one could hope that the methods of this paper could be used to determine the _ exact _ superpotential for both open and closed - sector moduli . to achieve this goal , one must generalize the formalism to orbifolds . let us sketch how this could be done .
suppose that a finite group @xmath300 acts on @xmath99 while preserving both the complex structure and the superpotential @xmath0 .
then we can form a semi - direct product of the cdg algebra @xmath155 and the group algebra of @xmath300 .
it is natural to expect that cdg modules over this cdg algebra are topological d - branes in the corresponding landau - ginzburg orbifold .
one can also think of these cdg modules as equivariant cdg modules over the original cdg algebra @xmath155 .
let us explain what this means in concrete terms . given a cdg module @xmath133
, we need to specify on @xmath134 and @xmath135 representations @xmath301 and @xmath302 of @xmath300 such that the twisted differential @xmath110 intertwines them : @xmath303 for example , consider the case @xmath192 , @xmath304 .
this lg model is known as the @xmath305 minimal model . as mentioned above ,
any topological d - brane in this model is equivalent to a direct sum of several indecomposable d - branes , which correspond to the following factorization of @xmath0 : @xmath306 the modules @xmath134 and @xmath135 are free modules of rank one for all these branes .
consider now the group @xmath307 which acts by @xmath308 , where @xmath309 is the primitive @xmath171-th root of unity .
any representation of @xmath300 on @xmath310 is specified by a single integer @xmath311 ( weight ) which runs from @xmath312 to @xmath313 .
further , since @xmath73 has weight @xmath314 , @xmath300-equivariance requires that the weight of @xmath135 be equal to the weight of @xmath134 plus @xmath314 .
thus an indecomposable d - brane in an orbifolded theory is a pair @xmath315 , where @xmath109 is an indecomposable d - brane in the original theory , and @xmath311 is an integer modulo @xmath171 . since orbifolding creates @xmath171 copies of each d - brane , the total number of indecomposable d - branes in the orbifolded model is @xmath316 .
we can check this result using mirror symmetry .
the mirror of this lg orbifold is the unorbifolded lg model with the same superpotential @xmath304 , hence we expect that there are @xmath316 a - branes in the latter model .
this is indeed the case @xcite .
furthermore , one can check the category of b - branes in the orbifolded lg model is equivalent to the category of a - branes in the unorbifolded model @xcite .
this amounts to a proof of the homological mirror symmetry conjecture in a simple ( non - geometric ) case .
many string orbifolds ( including gepner models ) involve not only geometric action on the target space , but also some world - sheet symmetries .
one such symmetry is @xmath317 , where @xmath318 is the left - moving fermion number .
the topological version of @xmath317 is the symmetry which reverses the grading of every cdg module @xmath109 and exchanges @xmath73 and @xmath74 .
orbifolding with respect to this symmetry does the following .
first of all , we need to keep only cdg modules which are invariant with respect to this symmetry . in
the above example this means that we keep only sums of d - branes correponding to @xmath319 and @xmath320 , and , for even @xmath171 , also a brane with @xmath321 .
second , an object of the new category should really be a pair @xmath322 , where @xmath109 is a cdg module , and @xmath323 is an isomorphism between @xmath109 and its grade - reversal . in the above example
, this has the effect that the self - dual brane with @xmath321 actually splits into two inequivalent branes .
it is an interesting problem to compute general bulk - boundary correlators in lg orbifolds , including bulk twisted - sector states .
we hope to address this problem elsewhere .
we are grateful to vladimir baranovsky and dmitri orlov for useful conversations .
this work was supported in part by the doe grant de - fg03 - 92-er40701 .
|
we compute topological correlators in landau - ginzburg models on a riemann surface with arbitrary number of handles and boundaries .
the boundaries may correspond to arbitrary topological d - branes of type b. we also allow arbitrary operator insertions on the boundary and in the bulk .
the answer is given by an explicit formula which can be regarded as an open - string generalization of c. vafa s formula for closed - string topological correlators .
we discuss how to extend our results to the case of landau - ginzburg orbifolds .
hep - th/0305136
|
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.