sample_number int64 8 203k | abstract stringlengths 21 17.4k | article stringlengths 0 18.9k | compressed_article stringlengths 0 10.9k | compression_time float64 0 0.78 | original_tokens int64 0 3.79k | compressed_tokens int64 0 2.29k | token_saving stringclasses 2 values |
|---|---|---|---|---|---|---|---|
67,240 | the evolution of the multipolar structure of the magnetic field of isolated neutron stars is studied assuming the currents to be confined to the crust .
lower orders ( @xmath0 ) of multipole are seen to evolve in a manner similar to the dipole suggesting little or no evolution of the expected pulse shape .
we also study the multifrequency polarization position angle traverse of psr b0329 + 54 and find a significant frequency dependence above 2.7 ghz .
we interpret this as an evidence of strong multipolar magnetic field present in the radio emission region .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in | strong multipolar components of the magnetic field have long been thought to play an important role in the radio emission from pulsars .
multipolar components present even in a small degree can significantly change the radius of curvature of the field lines near the polar cap and determine the illumination of the pulsar beam .
this in turn might be responsible for the observed complexity in pulse profiles .
significant evolution in the structure of the magnetic field may therefore lead to simplification of pulse profile with age .
we explore the ohmic evolution of multipolar magnetic fields in isolated neutron stars assuming the flux to be confined within the crust .
multipolar magnetic fields would also cause the polarisation position angle ( ppa ) traverse to depart from the well known rotating vector model of radhakrishnan and cooke ( 1969 ) ( rc ) . as a result of radius - to - frequency mapping ( rfm ) radio emission at higher frequencies originate closer to the stellar surface where multipole fields are stronger .
this would lead to a frequency dependence of the ppa traverse , departures from the rc model being stronger at higher frequencies .
we monitor the ppa of psr b0329 + 54 at several frequencies to investigate this effect .
in order to calculate the evolution of the multipolar magnetic field we solve the ohmic diffusion equation numerically . we assume the currents responsible for the entire field structure to be confined to the crust of the neutron star .
we adopt a well motivated choice for the electrical conductivity as a function of depth in the neutron star interior and use a standard cooling curve to represent the evolution of temperature ( for details see mitra et al .
1999 ) . fig .
[ 1 ] ( left panel ) shows the evolution of different multipole orders assuming the same initial strength for all orders .
it is evident from the figure that except for very high orders the reduction in the field strength is similar to that of a dipole .
hence no significant evolution of the pulse shape is expected due to the evolution of the multipolar structure of the field .
we investigate the multifrequency behaviour of the polarization position angle of psr b0329 + 54 at a number of frequencies over a wide range ( from 408 mhz to 4.8 ghz ) .
the ppa curves at 408 mhz and 1.4 ghz are in agreement with the dipolar rc model but at 2.7 ghz and 4.8 ghz the ppa appears to have developed a kinky feature ( fig . [ 1 ] right panel ) which is absent at lower frequencies .
assuming rfm , the observed kinkiness might be due to the presence of nondipolar magnetic fields in the emission region .
the continuous line as shown in fig .
[ 1 ] is a model for the nondipolar magnetic field structure ( a combination of a dipole and a quadrupole located at the center of the star ) which provides a good fit to the observed ppa ( mitra 1999 ) .
mitra , d. , 1999 , phd thesis submitted to jawaharlal nehru university , new delhi .
+ mitra , d. , konar , s. & bhattacharya , d. , 1999 , , 307 , 459 .
+ radhakrishnan , v. & cooke d. j. , 1969 , astrophysics letters , 3 , 225 + | multipolar components magnetic field in radio emission from pulsars components change radius curvature field lines near polar cap determine illumination pulsar beam responsible for complexity pulse profiles significant evolution structure magnetic field may simplification pulse profile with age explore ohmic evolution of multipolar magnetic fields in neutron stars flux confined within crust multipolar magnetic fields cause polarisation position angle ppa traverse depart from rotating vector model of radhakrishnan and cooke radio emission higher frequencies closer to stellar surface fields stronger to frequency dependence ppa traverse departures from rc model stronger at higher frequencies monitor ppa of psr b0329 + 54 at frequencies calculate evolution multipolar magnetic field solve ohmic diffusion equation numerically assume currents for field structure confined to crust neutron star adopt choice for electrical conductivity function depth neutron star interior use standard cooling curve evolution temperature mitra 1999 fig . [ 1 ] shows evolution of multipole orders same initial strength reduction in field strength similar to dipole no significant evolution of pulse shape expected due to multipolar structureinvestigate multifrequency behaviour polarization position angle psr b0329 + 54 frequencies 408 mhz to 4.8 ghz ppa curves 408 mhz 1.4 ghz dipolar rc model 2.7 4.8 ghz ppa kinky feature fig . 1 ] absent lower frequencies kinkiness due nondipolar magnetic fields emission region continuous line . 1 ] model nondipolar magnetic field structure dipole quadrupole center star good fit observed ppa mitra 1999 ) mitra 1999 phd thesis jawaharlal nehru university new delhi mitra konar bhattacharya 1999 307 , 459 radhakrishnan cooke d. 1969 astrophysics letters , 3 , 225 | 0.136067 | 718 | 366 | , Saving $0.0 in GPT-4. |
67,255 | we study thermodynamic properties and speed of sound in a free energy evolution of quark - gluon plasma ( qgp ) with one loop correction factor in the mean - field potential .
the values of the thermodynamic properties like pressure , entropy and specific heat are calculated for a range of temperatures .
the results agree with the recent lattice results .
the speed of sound is found to be @xmath0 independent of parameters used in the loop correction which matches almost with lattice calculations .
+ + pacs number(s ) : 25.75 ld , 12.38 mh + keywords : quark - gluon plasma , sound speed ; department of physics and astrophysics , university of delhi , delhi - 110007 , india + | the matter formation during the early stage of universe expansion is considered to be a complicated phenomena . to solve this problem
there are a number of experimental laboratories set up in the last two decades .
it is believed that there is a phase transition during the early universe expansion to a phase of confining matter of hadrons from deconfined matter of free quarks and gluons @xcite .
the process of transition is investigated in the experiments like relativistic heavy - ion collider ( rhic ) at bnl and large hadron collider ( lhc ) at cern .
the study makes it attractive for the experimentalists and theorists to collaborate in this exciting field in the present day of heavy ion collider physics @xcite . in this short paper
, we focus on the qgp evolution through the free energy expansion of the system with one loop correction . the free energy is constructed through a density of state which is obtained with a one loop corrected mean - field potential .
so the free energy contribution from every constituent particles of the system is considered in this calculation . moreover ,
the correction factor in the potential is obtained through coupling value , @xcite and due to this correction value , there are changes in the free energy expansion of qgp fireball , and it also impacts in the stability of droplet with the variation of dynamical quark and gluon flow parameters @xmath1 . in this short article we calculate thermodynamic parameters like pressure , entropy and specific heat through the free energy . then we calculate the speed of sound in a qgp with one loop correction in mean field potential which is extensively dealt with in our earlier paper @xcite .
the construction of density of states and the free energy with the loop correction has been explained in detail with these papers . in conclusion ,
we give the details of thermodynamic properties of qgp fireball like pressure , entropy , specific heat and speed of sound with the flow parameter @xmath2 .
the density of states in phase space with loop correction in the interacting potential is obtained through a generalized thomas- fermi model as @xcite : @xmath3^{2 } \frac{dv_{conf}}{dq}~,\ ] ] where , @xmath4- \frac{m_{0}^{2}}{2 q}.\ ] ] @xmath5 and @xmath6 are quark and gluon flow parameters chosen in fitting the evolution of qgp droplets @xcite .
then we finally get the density of state as : @xmath7^{3 } g^{6}(q)a,\ ] ] where @xmath8\ ] ] and @xmath9 is the volume occupied by the qgp .
@xmath10 is the relativistic four - momentum in natural units .
@xmath11 where , @xmath12 is the coupling value of quark and gluon with degree of freedom @xmath13 defined as @xmath14 in which @xmath15 is qcd parameter measured in the scale of lattice qcd and it is taken as @xmath16gev .
the coefficient @xmath17 in the confining potential of the above expression is one loop correction in their interactions and it is given as @xcite : @xmath18 where @xmath19 is the number of light quark elements @xcite .
now we calculate free energy for the system using the expression as in @xcite : @xmath20 with minimum energy cut off as : @xmath21 where @xmath22 $ ] .
the minimum cut off in the model makes the integral to have a finite value by avoiding the infra - red divergence while taking the magnitude of @xmath15 and @xmath23 as of the same order as in lattice qcd .
@xmath24 is the corresponding degeneracy factors for the subsequent free energies .
the inter - facial energy obtained through a scalar weyl - surface in ramanathan et al .
@xcite with a suitable modification to take care of the hydrodynamic effects is given as : @xmath25 the energy is used in replacing the role of bag energy of mit model and it minimizes the drawback produced in numerical techniques by mit model . the hadronic free energy is @xcite @xmath26 for quark free energy we consider quark masses @xmath27 and @xmath28 [ 15 ] .
the hadronic particles are taken having the masses below @xmath29gev and their contribution to the interaction is considered zero in nature .
we consider these values as it has a dominant component in the contribution of the hadronized phase .
we can thus compute the total modified free energy @xmath30 as , @xmath31 where @xmath32 stands for @xmath33 , @xmath34 and @xmath35 quark and gluon .
the thermodynamic properties like pressure , entropy and specific heat of the system can be calculated from the total free energy .
the standard thermodynamics give the following relations @xcite : @xmath36 @xmath37 @xmath38 the behavior of entropy and specific heat with temperature indicate the nature of phase transition of the system .
so , we calculate the entropy and specific heat for a flow parameter @xmath39 and gluon parameter @xmath40 . at this particular value
we can get the stability of the droplet with maximum radius of @xmath41 . the stability is found more with increasing gluon flow parameter up to @xmath42 . with the increasing gluon parameter the pressure of the system is increased and very much effective in the free energy expansion , yet the entropy and specific heat is not effective with the increasing gluon parameter . now using these entropy and specific heat
we calculate the speed of sound of qgp .
the speed of sound is given as the ratio of these two thermodynamic parameters @xcite : @xmath43 the speed of sound is found to be almost constancy over the range of temperature and the model parameter .
vs. t at @xmath44 , @xmath45 at the critical radius.,width=264 ] vs. t at @xmath44 , @xmath45 at the critical radius.,width=264 ] vs. t at @xmath44 , @xmath45 at the critical radius.,width=264 ]
the thermodynamic properties like pressure , entropy and specific heat are calculated for quark and gluon flow parameter with the introduction of one loop correction factor in the interacting mean - field potential .
they are calculated for a particular chosen flow parameter which fit well in the evolution of free energies with the stable largest droplet size .
free energy expansion can be produced with different stable droplet size with increasing gluon flow parameter up to @xmath42 .
there is change in the pressure but the parameter does not affect in the entropy and specific heat .
the outputs from the calculation are shown in the figures .
the figure @xmath46 shows the variation of @xmath47 with the temperature @xmath23 for three flavor and two flavor of quarks .
the variation in the figure determines the equation of state ( eos ) of the system .
the equation of state is an essential component to understand the nature of qgp and to model the behavior of qgp in early universe formation .
the variation agrees with the recent outcomes of lattice results with insignificant difference . in fig.@xmath48
we show the plot of @xmath49 with the temperature .
the plot indicates the variation at lower temperature .
it again shows the constancy over higher temperature .
the result almost follows the lattice predictions for three and two flavors of quarks and it produced similar results with our earlier works without the loop correction @xcite . again
@xmath50 shows specific heat changes with the temperature .
the specific heat is almost independent with higher temperature and the plot matches with the recent calculation of specific heat .
moreover , it is indicated by lattice qcd that formation of qgp droplets take place under the condition of roll - over phase rather than a sharp jump as temperature varies @xcite . from the two thermodynamic parameters we use to calculate the speed of sound of the system .
the result is shown in the figure @xmath51 . from the figure
, the speed of sound is well behaved in nature in terms of the lattice results of sound speed and almost the same to our earlier calculation of speed of sound without the loop correction .
the result is approximately close to @xmath52 which is obtained with this range of temperature .
it is quite similar to the latest value of speed of sound of lattice qcd @xcite .
we conclude that due to the introduction of one loop correction in the mean field potential , we can describe the eos through pressure and behavior of entropy and specific heat .
so , we further obtained the speed of sound from these two thermodynamic parameters as @xmath0 which is independent from the model parameter and temperature .
it is a good result in comparison with recent lattice results .
e. karsch , a. peikert and e. laermann , nucl .
b * 605 * , 579 ( 2001 ) .
l. p. csernai , j. i. kapusta , e. osnes , phys .
d * 67 * , 045003 ( 2003 ) ; a. tawfik , phys .
d * 71 * , 054502 ( 2005 ) .
h. satz , _ cern - th-2590 _ , 18pp ( 1978 ) ; f. karsch , e. laermann , a. peikert , ch . schmidt and s. stickan ,
b * 94 * , 411 ( 2001 ) .
m. g. mustafa , d. k. srivastava and b. sinha , euro .
j. c * 5 * , 711 ( 1998 ) ; f. karsch and h. satz , nucl .
a * 702 * , 373 ( 2002 ) .
a. peshier , b. k@xmath53mpfer , o. p. pavlenko and g. soff , phys .
b * 337 * , 235 ( 1994 ) ; v. goloviznin and h. satz , z. phys .
c * 57 * , 671 ( 1993 ) .
r. ramanathan , k. k. gupta , a. k. jha and s. s. singh , pram . j. phys .
* 68 * , 757 ( 2007 ) .
s. s. singh , d. s. gosain , y. kumar and a. k. jha , pram .
* 74 * , 27 ( 2010 ) .
s. s. singh , k. k. gupta and a. k. jha , int .
a * 29 * , 1450094 ( 2014 ) ; s. s. singh and r. ramanathan , prog .
103do2 * , 2014 .
r. ramanathan , y. mathur , k. k. gupta and a. k. jha , phys .
c * 70 * , 027903 ( 2004 ) .
e. fermi , zeit f. physik * 48 * , 73 ( 1928 ) ; l. h. thomas , proc .
* 23 * , 542 ( 1927 ) ; h. a. bethe , rev . mod
* 9 * , 69 ( 1937 ) .
n. brambilla , a. pineda , j. soto and a. vairo , phys .
rev d * 63 * , 014023 ( 2001 ) .
k. melnikov and a. yelkhovsky , nucl .
b * 528 * , 59 ( 1998 ) ; a. h. hoang , phys .
rev d * 59 * , 014039 ( 1999 ) .
w. fischler , nucl .
b * 129 * , 157 ( 1977 ) ; a. billoire , phys .
lett b * 92 * , 343 ( 1980 ) .
a. v. smirnov , v. a. smirnov , m. steinhauser , phys .
b * 668 * , 293 ( 2008 ) ; a. v. smirnov , v. a. smirnov and m. steinhauser , phys .
* 104 * , 112002 ( 2010 ) .
g. neergaad and j. madsen , phys . rev .
d * 60 * , 05404 ( 1999 ) ; m. b. christiansen and j. madsen , j. phys .
g * 23 * , 2039 ( 1997 ) .
i. mardor and b. svetitsky , phys .
d * 44 * , 878 ( 1991 ) ; h. t. elze and w. greiner , phys .
b * 179 * , 385 ( 1986 ) .
h. weyl , _ nachr .
wiss gottingen _ 110 ( 1911 ) .
r. balian and c. block , ann .
( ny ) * 60 * , 401 ( 1970 ) .
r. v. gavai , s. gupta and s. mukherjee , pram .
* 71 * , 487 ( 2008 ) ; s. borsnyi , nucl .
a * 904 - 05 * , 270c ( 2013 ) . a. a. khan , @xmath54 , phys .
d * 64 * , 074510 ( 2001 ) ; y. aoki , z fodor , s. d. katz and k. k. szab , j. high energy phys .
* 01 * , 089 ( 2006 ) .
d. s. gosain , s. s. singh , int .
* 53 * , 2688 ( 2014 ) .
y. aoki , g. endrdi , z fodor , s. d. katz and k. k. szab , nature * 443 * , 675 ( 2006 ) .
f. karsch , nucl .
a * 783 * , 13 ( 2007 ) ; h. b. meyer , euro .
j. a * 47 * , 86 ( 2011 ) . | matter formation early universe expansion complicated experimental laboratories set up last two decades believed phase transition early universe expansion confining matter hadrons from deconfined free quarks gluons transition investigated in experiments relativistic heavy - ion collider bnl large hadron collider at cern study attractive for experimentalists theorists collaborate in heavy ion collider physics paper on qgp evolution through free energy expansion with one loop correction free energy constructed through density of state obtained with one loop corrected mean - field potential free energy contribution from particles considered correction factor potential obtained through coupling value changes in free energy expansion qgp fireball impacts stability of droplet dynamical quark gluon flow parameters calculate thermodynamic parameters pressure entropy specific heat through free energy speed of sound in qgp with one loop correction mean field potential construction of density of states free energy with loop correction explained detail papers conclusion details of thermodynamic properties of qgp fireball like pressure entropy heat speed of sound with flow parameterdensity of states phase space loop correction interacting potential obtained thomas- fermi model @xmath3^{2{dq} @xmath4-^ @xmath5 @xmath6 quark gluon flow parameters evolution qgp droplets density of state @xmath7^{3 } g^{6}(q)a @xmath8 @xmath9 volume qgp @xmath10 relativistic four - momentum natural units @xmath11 @xmath12 coupling value quark gluon degree freedom @xmath13 @xmath14 @xmath15 qcd parameter lattice qcd @xmath16gev coefficient @xmath17 confining potential loop correction interactions @xcite @xmath18 @xmath19 number light quark elements calculate free energy system expression @xmath20 minimum energy cut off @xmath21 @xmath22 minimum cut off integral finite value infra - red divergence magnitude @xmath15 @xmath23 same order lattice qcd @xmath24 degeneracy factors for free energies inter - facial energy scalar weyl - surface ramanathan et al @xcite modification hydrodynamic effects @xmath25 energy bag energy mit model minimizes drawback numerical techniques hadronic free energy is @xcite @xmath26 quark free energy consider quark masses @xmath27 @xmath28 hadronic particles masses below @xmath29gev contribution interaction zero dominant component contribution hadronized phase compute total modified free energy @xmath30 as @xmath31 @xmath32 for @xmath33 @xmath34 @xmath35 quark gluon thermodynamic properties pressure entropy specific heat calculated from total free energy standard thermodynamics relations @xcite : @xmath36 @xmath37 @xmath38 behavior entropy specific heat with temperature phase transition calculate entropy specific heat for flow parameter @xmath39 gluon parameter @xmath40 stability droplet maximum radius @xmath41 stability increasing gluon flow parameter up to @xmath42 increasing gluon parameter pressure increased effective in free energy expansion entropy specific heat not effective increasing gluon parameter calculate speed of sound of qgp speed ratio of thermodynamic parameters @xcite : @xmath43 speed of sound constancy over temperature model parameter vs t @xmath44 @xmath45 critical radius@xmath44 @xmath45 critical radius thermodynamic properties pressure entropy specific heat calculated for quark gluon flow one loop correction factor interacting mean - field potential calculated for flow parameter evolution free energies stable droplet size free energy expansion different droplet size increasing gluon flow parameter up to @xmath42 change in pressure affect entropy specific heat outputs in figures figure @xmath46 shows variation @xmath47 with temperature @xmath23 for three two flavor quarks variation determines equation of state system essential understand nature qgp model behavior early universe formation variation agrees with recent outcomes lattice results insignificant difference fig.@xmath48 plot of @xmath49 with temperature variation at lower temperature constancy over higher temperature result follows lattice predictions for three two flavors quarks similar results without loop correction @xmath50 shows specific heat changes with temperature independent with higher temperature plot matches recent calculation specific heat formation qgp droplets roll - over phase sharp jump temperature varies thermodynamic parameters calculate speed of sound system result figure @xmath51figure speed sound behaved lattice results same earlier calculation without loop correction result close to @xmath52 obtained temperature similar latest value speed lattice qcd @xcite loop correction field potential describe eos pressure entropy heat obtained speed sound thermodynamic parameters @xmath0 independent from model parameter temperature good result recent lattice results karsch peikert e laermann b * 605 *, 579 ( 2001 ) csernai kapusta osnes phys . d * 67 *, 045003 ( 2003 ) tawfik d * 71 *, 054502 ( 2005 ) h. satz cern - th-2590 18pp ( 1978 ) karsch laermann peikert schmidt stickan b * 94 * , 411 ( 2001 ) mustafa srivastava sinha c * 5 * , 711 ( 1998 ) karsch h satz 702 *, 373 ( 2002 ). peshier pavlenko soff phys * 337 * , 235 ( 1994 ) goloviznin h. satz phys c * 57 * , 671 ( 1993 ) ramanathangupta a k jha singh 68 757 ( 2007 ) singh gosain y kumar a k jha pram 74 27 ( 2010 ) singh gupta jha int. 29 1450094 ( 2014 ) singh prog 103do2 2014 ramanathan mathur gupta a k jha phys . 70 027903 ( 2004 ) e fermi 48 73 ( 1928 ) l h thomas proc. 23 542 ( 1927 ) bethe rev 9 69 ( 1937 ) brambilla. pineda j. soto a. vairo . 63 014023 ( 2001 ) k melnikov a. yelkhovsky nucl 528 59 ( 1998 ) a h. hoang phys 59 014039 ( 1999 ) fischler nucl 129 157 ( 1977 ) billoire phys. 92 343 ( 1980 ) v. smirnov smirnov m steinhauser phys . 668 293 ( 2008 ) smirnov steinhauser phys . 104 112002 ( 2010 ). neergaad j.madsen phys 60 05404 1999 ) christiansen j madsen phys 23 2039 ( 1997 ) mardor svetitsky phys 44 878 ( 1991 ) elze w greiner 179 385 ( 1986 ) weyl wiss gottingen 110 ( 1911 ) balian c. block 60 401 ( 1970 ) gavai s gupta mukherjee pram 71 487 ( 2008 ) borsnyi nucl 904 - 05 270c ( 2013 ) khan phys d 64 * 074510 ( 2001 ) y aoki fodor. katz k k szab high energy phys 01 089 ( 2006 ) gosain s s singh int. 53 * 2688 ( 2014 ) y aoki endrdi fodor katz szab nature 675 ( 2006 ) karsch nucl 783 13 ( 2007 ) b meyer euro 47 86 ( 2011 ) | 0.445911 | 3,171 | 1,614 | , Saving $0.1 in GPT-4. |
67,259 | the glasgow reweighting method is evaluated for su(2 ) lattice gauge theory at nonzero @xmath0 and finite @xmath1 .
we establish that the overlap problem of su(3 ) measurements , in which the transition points determined from thermodynamic observables have an unphysical dependence on the value of @xmath0 used to generate ensembles for reweighting , persists for su(2 ) . by combining the information from different lattice ensembles we alleviate sampling bias in the fugacity expansion , and identify the lee yang zeros associated with the transition to a high density phase that can plausibly be associated with diquark condensation .
we also confirm the existence of a line of first order transitions above a critical point in the @xmath2 phase plane previously predicted by effective chiral lagrangian calculations . | recent speculation over bcs type - instabilities in the fermi surface at high density @xcite@xcite has led to resurgence of interest in two colour qcd .
lattice models with pseudoreal representations are attractive candidates for simulation at finite density since for such theories the dirac matrix is positive definite , which permits the use of existing monte carlo techniques at @xmath3 . for su(3 ) at finite density the lattice action is otherwise complex @xcite , and so the probabilistic importance sampling step of such methods is therefore undefined .
reweighting methods have proved a useful means of addressing this issue , where the @xmath0 dependence of the grand canonical partition function @xmath4 can be made semi - analytic in a fugacity expansion , as with the glasgow method @xcite .
the complex action issue is thus avoided , by generating a lattice ensemble in an accessible regime of the parameter space ( eg .
@xmath5 for su(3 ) ) .
naively , one might anticipate that the specific lattice ensemble used in the reweighting has little impact on numerical evaluations of the expansion .
in fact quite the reverse is true . for the su(3 )
model with dynamical quarks , even at intermediate coupling where one might expect thermal fluctuations to enhance the frequency of sampling physically relevant states , an ensemble generated at @xmath6 reproduces a similar phase structure to the quenched model @xcite@xcite .
this has severe consequences since quenched finite density qcd is understood to be the zero flavor limit ( @xmath7 ) of a theory with equal numbers of quarks and conjugate quarks .
the lowest lying baryonic state in the model is thus the unphysical `` baryonic pion '' , formed from quark - conjugate quark pairs rather than the lightest three quark state @xcite@xcite .
however , this is of no consequence in two colour qcd @xcite as the baryonic pion and baryon propagators are equivalent at @xmath3 .
pseudoreal models are not entirely free from the effects of the reweighting overlap pathologies , however .
models with quarks in the adjoint representation ( with @xmath8 odd ) suffer from the related reweighting pathology : the sign problem . since @xmath9 is always real for two colour qcd , although reweighting is no longer mandatory it provides a useful opportunity to investigate the overlap issue ( where the correct physics may be more easily extracted from lattice measurements by conventional means ) , and to quantify the signatures of the sampling numerical discrepancies therein .
the @xmath10flavor symmetry of su(2 ) qcd given by quarks in their fundamental representation @xmath11 , is @xmath12 rather than @xmath13 as might be anticipated , as is demonstrated by a change of basis in the free field langrangian .
@xmath14 @xmath15 where @xmath16 is a @xmath11 doublet .
the inclusion of explicit symmetry breaking terms in @xmath17 and @xmath0 can be shown similarly to lead to the symmetry breaking patterns tabulated above @xcite@xcite . @xmath18 .global @xmath10flavor symmetries of the two colour qcd action , in the continuum and with staggered quarks for @xmath19 . [ cols="^,^,^ " , ]
despite expectations , generating an ensemble for the reweighting method with a value of @xmath20 arbitrarily close to @xmath21 still leads to an overlap problem in a model with a pseudoreal representation evaluated at @xmath22 .
in fact , for an exploratory study ( in which @xmath21 is unknown ) , the fugacity expansion coefficients are more effectively sampled through the combination of terms from a covering series of ensembles .
even with the real monte carlo measure of two colour qcd the overlap problem is still pathological for the glasgow reweighting method . with new multi - parameter ( @xmath23 , @xmath0 ) reweighting approaches to su(3 )
@xcite we would therefore expect there to be similar sampling bias in the coefficients across the full fugacity expansion . in order to successfully investigate the possibility of there being similar tricritical behavior in the @xmath24 phase plane of su(3 ) @xcite with the glasgow method ,
this issue of effective sampling across the full range of the polynomial this overlap problem should then be addressed . as we have , seen substantially different transition points
are determined from reweighting measurements at intermediate coupling from ensembles generated with @xmath20 arbitrarily close to @xmath21 .
lee yang zero analysis allows the simple identification of a first order transitions in @xmath0 with lattice measurements on comparatively small volumes . to extend the rigor of this approach and determine the critical exponents of the transition at @xmath21 , however , it will be necessary to increase the lattice size as the zeros scaling at @xmath25 is not without finite volume effects
it will also be interesting to repeat this volume scaling analysis for @xmath26 and to compare the measured critical exponents for the diquark phase transition with those predicted with the chiral langragian approach @xcite , believed to be second order .
+ thanks to m. alford for useful discussions . | speculation over bcs type instabilities in fermi surface at high density led to interest in two colour qcd lattice models with pseudoreal attractive for simulation at finite density dirac matrix positive permits monte carlo techniques at @xmath3 for su(3 ) at finite density lattice action complex probabilistic importance sampling step undefined reweighting methods issue dependence of canonical partition function semi - analytic in fugacity expansion complex action issue avoided by generating lattice ensemble in accessible regime parameter space @xmath5 for su(3 ) specific lattice ensemble reweighting on numerical evaluations expansion reverse for su(3 ) model with dynamical quarks ensemble generated at @xmath6 reproduces similar phase structure to quenched model consequences quenched finite density qcd zero flavor limit of theory with equal numbers quarks lowest lying baryonic state is unphysical baryonic pion '' formed from conjugate pairs lightest three quark state no consequence in two colour qcd baryonic pion and baryon propagators equivalent at @xmath3pseudoreal models not free reweighting overlap pathologies models with quarks representation @xmath8 odd suffer from reweighting pathology sign problem @xmath9 real for two colour qcd reweighting mandatory investigate overlap quantify sampling numerical discrepancies @xmath10flavor symmetry of su(2 ) qcd quarks @xmath11 is @xmath12 @xmath13 demonstrated by change of basis in free field langrangian @xmath14 @xmath15 @xmath16 is @xmath11 doublet symmetry breaking terms in @xmath17 @xmath0 to symmetry breaking patterns @xmath18 @xmath10flavor symmetries of two colour qcd action continuum staggered quarks for @xmath19 generating ensemble reweighting with value @xmath20 close to @xmath21 leads to overlap problem in model pseudoreal representation at @xmath22 @xmath21 unknown fugacity expansion coefficients sampled through combination of terms from series ensembles real monte carlo measure two colour qcd overlap problem pathological for glasgow reweighting methodnew multi - parameter @xmath23 @xmath0 ) reweighting approaches su(3 ) @xcite similar sampling bias coefficients across full fugacity expansion investigate tricritical behavior @xmath24 phase plane su(3 ) @xcite glasgow method effective sampling range polynomial overlap problem addressed different transition points determined reweighting measurements intermediate coupling ensembles @xmath20 close to @xmath21 lee yang zero analysis first order transitions @xmath0 small volumes extend rigor determine critical exponents transition @xmath21 increase lattice size zeros scaling @xmath25 finite volume effects repeat volume scaling analysis @xmath26 compare critical exponents diquark phase transition chiral langragian approach @xcite second order thanks m. alford discussions | 0.186979 | 1,126 | 590 | , Saving $0.0 in GPT-4. |
67,265 | we propose a 2-temperature radial dynamical model of plasma flow near sgr a * and fit the bremsstrahlung emission to extensive quiescent x - ray chandra data .
the model extends from several arcseconds to black hole ( bh ) gravitational radius , describing the outer accretion flow together with the infalling region .
the model incorporates electron heat conduction , relativistic heat capacity of particles and feeding by stellar winds .
stellar winds from each star are considered separately as sources of mass , momentum and energy .
self - consistent search for the stagnation and sonic points is performed .
most of gas is found to outflow from the region .
the accretion rate is limited to below @xmath0 of bondi rate due to the effect of thermal conduction enhanced by entropy production in a turbulent flow .
the x - ray brightness profile proves too steep near the bh , thus a synchrotron self - compton point source is inferred with luminosity @xmath1 erg / s .
we fit the sub - mm emission from the inner flow , thus aiming at a single model of sgr a * accretion suitable at any radius . | our galaxy is known to host a supermassive black hole ( bh ) with mass @xmath2 at a distance @xmath3 kpc @xcite .
the bh exhibits very low luminosity state probably due to inefficient feeding and cooling .
the bh is fed by stellar winds within several arcseconds from the compact object roughly around the radius of bh gravitational influence @xcite .
the stellar winds are expelled at large speeds .
they collide , heat up to @xmath4 k and emit bremsstrahlung x - rays , observed by chandra @xcite .
a small fraction of mass accretes onto the black hole is thus producing the emission in sub - mm and other wavebands .
however , the inferred accretion rate within several schwarzschild radii is 2 orders of magnitude lower @xcite than the inferred bondi accretion rate @xcite at several arcseconds .
this disparity is resolved in a present work with a point source revealed coincident with sgr a*. a brief account of observations is made in [ s_obs ] .
the dynamical model is outlined in [ s_model ] .
the results are discussed in [ s_res ] .
we analyze @xmath5 ms of chandra exposure of sgr a * and central arcseconds @xcite significantly improving over the previously analyzed @xmath6 ks exposure @xcite . as we are interested in the quiescent emission , we bin the observations in @xmath7 s bins and exclude the flaring states with counts triple the mean .
we also subtract the point sources and extended bright emission zones , thus extracting the quiescent count surface brightness profile within @xmath8 having the extensive data we are able to perform the subpixel spatial binning in rings @xmath9 thickness owing to dithering of spacecraft .
the counts from four @xmath10 ring segments centered at sgr a * are compared in order to test the viability of the radial model .
it appears that within @xmath11 the counts do not differ significantly between ring segments , but the variation was found at @xmath12 the point spread function ( psf ) is extracted by observing the nearby binary j174540.9 - 290014 .
feeding of the black hole should be a starting point of any accretion model .
this approach helps to eliminate a number of arbitrary boundary conditions .
a set of @xmath13 wind emitters is believed to supply almost all the matter into the feeding region of sgr a*. following @xcite , we identify the important wind emitters , find the wind speeds and ejection rates .
we obtain the orbital data from @xcite , assuming the stars either belong to the disk or taken to have the minimum eccentricities . as we are constructing the radial model , the radial feeding function @xmath14 is produced by smoothing wind inputs over radius between the apocenter and the pericenter for each star ( see fig . [ fig_feed ] ) .
the averaged wind velocity is found as a root - mean - square average over stars weighed with the ejection rate .
we do not account for orbital velocities of stars in energy input as feeding is dominated by only a few stars close to the bh .
s2 star is included into the calculation as it may eject more matter @xcite , than falls onto the bh .
the dynamical model has sources of mass , radial momentum and energy due to winds @xcite .
the main feature of the model is the electron thermal conduction proportional to the temperature gradient with conductivity @xmath15 @xcite .
here we have included the factor of @xmath16 inhibition of conductivity in turbulent magnetic field @xcite .
so computed heat flux appears to be below the saturated flux @xcite . in reality
the inner accretion flow is collisionless and the outer accretion is weakly collisional , however , the single prescription for conductivity gives the reasonable approximation . the realistic behavior of electrons in the inner flow is achieved by using their relativistic heat capacity in the equations of motion , which naturally leads to the ratios @xmath17 up to 10 even in adiabatic flows . for completeness of the theory
we add the direct energy transfer into electrons and protons equivalent to the entropy production , which happens due to viscosity in the rotating flow or due to dissipation of turbulence @xcite .
we assume that the fractions @xmath18 and @xmath19 of available gravitational energy goes to electrons and protons .
the coulomb collisions are included for numerical stability to balance the electron and proton temperature in the outer flow , though they do not have the significant dynamical effect .
the effect of gas composition is accounted for by introducing the effective mass @xmath20 per electron and correspondent reduction in ion gas pressure .
these numbers are taken for the solar abundance of elements , which seems to be reasonable for stellar winds @xcite .
paczhynski - wiita gravitational potential is employed .
the proposed system of equations has no artificial boundary conditions , but it appears to have an unmatched complexity .
we self - consistently solve for the positions of the stagnation point , where gas velocity is zero , and the inner isothermal sonic point @xcite .
the heat flux is set to zero at the point , where @xmath21 in the bondi solution near the bh .
the outer boundary is either taken to be the isothermal sonic point in the outflow or the point with slightly higher density ( for numerical stability ) .
the relaxation technique is used for the 2-temperature system between the inner boundary and the stagnation point , whereas only shooting works outside the stagnation point .
having produced a bunch of dynamical models , we convert the temperature and density radial profiles into the surface brightness profile .
we take the up - to - date bremsstrahlung emissivities ( see @xcite and errata ) and account for emission by heavy elements , excluding iron .
we calculate the spectrum along each ray and convert it to counts by applying the solar metallicity interstellar absorption @xcite with hydrogen column @xmath22@xmath23 and convolving with the response of chandra .
then we apply the psf blurring and compare with the observed surface brightness profile .
the outer surface brightness profile can be reasonably fitted by a model with @xmath24 @xmath25 which also leads to the ratio @xmath26 near the bh .
however , we find that the inner part of surface brightness curve is too steep for any extended emission .
there should be a point source in the center accounting for @xmath27 to @xmath28 of central surface brightness and having the unabsorbed luminosity about @xmath29 erg / s of monoenergetic @xmath30 kev photons .
about the same luminosity is expected from the synchrotron self - compton ( ssc ) process near the bh .
the search for the best model with the point source continues .
we want the best model to reproduce the observed faraday rotation measure @xmath31@xcite and optically thick flux @xmath32 jy at @xmath33 ghz @xcite .
an order of magnitude consistency is achieved on the way .
the accretion rate appears to be self - consistently limited to @xmath34 of bondi value , thus the connection between the inner accretion flow and the outer accretion flow is established .
the future versions will include the angular momentum and use x - ray spectral information .
baganoff , f. k. , et al .
2003 , , 591 , 891 bondi , h. 1952 , mnras , 112 , 195 cuadra , j. , nayakshin , s. , & martins , f. 2008 , mnras , 383 , 458 cowie , l. l. , mckee , c. f. 1977 , , 211 , 135 ghez , a. m. , et al .
2008 , , 689 , 1044 gould , r. j. 1980 , , 238 , 1026 johnson , b. m. , quataert , e. 2007 , , 660 , 1273 krichbaum , t. p. , graham , d. a. , bremer , m. , alef , w. , witzel , a. , zensus , j. a. , eckart , a. 2006 , jphcs , 54 , 328 lu , j. r. , ghez , a. m. , hornstein , s. d. , morris , m. r. , becklin , e. e. , & matthews k. 2009 , , 690 , 1463 marrone , d. p. , moran , j. m. , zhao , j. , & rao r. , 2007 , , 654l , 57 martins , f. , genzel , r. , hillier , d. j. , eisenhauer , f. , paumard , t. , gillessen , s. , ott , t. , & trippe s. 2007 , a&a , 468 , 233 martins , f. , gillessen , s. , eisenhauer , f. , genzel , r. , ott , t. , & trippe s. 2008 , , 672 , l119 muno , m. p. , et al .
2008 , , 673 , 251 narayan , r. , medvedev , m. v. , , 562 , l129 morrison , r. , mccammon , d. 1983 , , 270 , 119 najarro , f. , figer , d. f. , hillier , d. j. , kudritzki , r. p. 2004
, , 611 , l105 paumard t. , et al .
2006 , , 643 , 1011 quataert , e. 2004 , , 613 , 322 shcherbakov , r. v. 2008 , , 177 , 493 lamers , h. j. g. l. m. , cassinelli , j. p. 1999
, `` introduction to stellar winds '' , ( new york : cambridge university press ) | galaxy supermassive black hole bh ) mass @xmath2 distance @xmath3 kpc @xcite low luminosity inefficient feeding cooling fed by stellar winds arcseconds winds expelled large speeds collide heat @xmath4 k emit x - rays observed small fraction mass accretes black hole emission sub - mm wavebands inferred accretion rate schwarzschild radii 2 orders lower @xcite bondi accretion rate @xcite arcseconds disparity resolved point source coincident with sgr a* account observations s_obs ] dynamical model outlined s_model ] results discussed s_res ] analyze @xmath5 ms chandra exposure sgr a * central arcseconds @xcite improving @xmath6 ks exposure @xcite interested quiescent emission bin observations @xmath7 s bins exclude flaring states triple mean subtract point sources extended bright emission zones quiescent count surface brightness profile @xmath8 subpixel spatial binning rings @xmath9 thickness counts from four @xmath10 ring segments sgr a * compared test radial model@xmath11 counts differ between ring segments variation at @xmath12 point spread function extracted observing nearby binary j174540.9 - 290014 feeding black hole starting point accretion model arbitrary boundary conditions @xmath13 wind emitters supply matter into feeding region sgr a* identify wind emitters find wind speeds ejection rates obtain orbital data from @xcite stars to disk or minimum eccentricities constructing radial model feeding function @xmath14 smoothing wind inputs over radius apocenter pericenter each star fig averaged wind velocity root - mean - average over stars ejection rate for orbital velocities stars few stars close to bh s2 star eject more matter dynamical model sources mass radial momentum energy due to winds @xcite electron thermal conduction proportional to temperature gradient with conductivity @xmath15 @xcite factor @xmath16 inhibition conductivity in turbulent magnetic field @xcite computed heat flux below saturated flux @xcite inner accretion flow collisionless outer accretion weakly collisional single prescription for conductivity reasonable approximationrealistic behavior of electrons in inner flow achieved using relativistic heat capacity in equations motion leads to ratios @xmath17 up to 10 in adiabatic flows direct energy transfer into electrons protons equivalent to entropy production due to viscosity rotating flow or dissipation turbulence fractions @xmath18 and @xmath19 of gravitational energy goes to electrons protons coulomb collisions included for stability balance electron proton temperature significant dynamical effect effect gas composition effective mass @xmath20 per electron reduction in ion gas pressure numbers taken for solar abundance of elements reasonable for stellar winds paczhynski - wiita gravitational potential employed proposed system equations no artificial boundary conditions unmatched complexity solve for positions stagnation point inner isothermal sonic point heat flux set to zero at point outer boundary isothermal sonic point or higher density for relaxation technique used for 2-temperature system between inner boundary stagnation point works outside stagnation convert temperature density radial profiles into surface brightness profile bremsstrahlung emissivities account for emission by heavy elements excluding ironcalculate spectrum ray convert to counts solar metallicity interstellar absorption @xcite hydrogen column @xmath22 response chandra apply psf blurring compare surface brightness profile outer surface brightness profile fitted model @xmath24 @xmath25 leads ratio @xmath26 near bh inner surface brightness curve steep for extended emission point source center @xmath27 to @xmath28 central surface brightness unabsorbed luminosity @xmath29 erg / s monoenergetic @xmath30 kev photons same luminosity synchrotron self - compton process near bh search best model point source continues reproduce faraday rotation measure @xmath31@xcite optically thick flux @xmath32 @xmath33 ghz @xcite order magnitude consistency achieved accretion rate limited to @xmath34 bondi value connection inner outer accretion flow established future versions include angular momentum use x - ray spectral information baganoff 2003 591 , 891 bondi 1952 mnras 112 , 195 martins 2008 383 , 458 1977 211 , 1352008 689 , 1044 gould 1980 238 , 1026 johnson quataert. 2007 660 , 1273 krichbaum graham bremer alef witzel zensus eckart 2006 jphcs 54 , 328 lu j ghez hornstein morris becklin matthews k. 2009 690 , 1463 marrone d. moran zhao j rao r. 2007 654l , 57 martins genzel hillier eisenhauer paumard gillessen ott trippe s. 2007&a 468 , 233 martins gillessen eisenhauer genzel ott trippe s. 2008 672 , l119 muno m. p.. 2008 673 , 251 narayan medvedev m.. 562 , l129 morrison mccammon d. 1983 270 , 119 najarro figer hillier. kudritzki r..2004 611 l105 paumard 2006 643 1011 quataert 2004 613 , 322 shcherbakov 2008 177 , 493 lamers cassinelli j.. 1999 introduction stellar winds new cambridge university press | 0.356034 | 2,228 | 1,163 | , Saving $0.1 in GPT-4. |
67,273 | an accelerated particle sees the minkowski vacuum as thermally excited , which is called the unruh effect .
due to an interaction with the thermal bath , the particle moves stochastically like the brownian motion in a heat bath .
it has been discussed that the accelerated charged particle may emit extra radiation ( the unruh radiation @xcite ) besides the larmor radiation , and experiments are under planning to detect such radiation by using ultrahigh intensity lasers @xcite .
there are , however , counterarguments that the radiation is canceled by an interference effect between the vacuum fluctuation and the radiation from the fluctuating motion . in this and another reports @xcite , we review our recent analysis on the issue of the unruh radiation . in this report , we particularly consider the thermalization of an accelerated particle in the scalar qed , and derive the relaxation time of the thermalization .
the interference effect is discussed separately in @xcite . | quantum field theories in the space - time with horizons exhibit interesting thermodynamic behavior .
the most prominent phenomenon is the hawking radiation and the fundamental laws of thermodynamics hold in the black hole background .
a similar phenomenon occurs for a uniformly accelerated observer in the ordinary minkowski vacuum @xcite .
this is called the unruh effect .
if a particle is uniformly accelerated in the minkowski space with an acceleration @xmath0 , there is a causal horizon ( the rindler horizon ) and no information can be transmitted from the other side of the horizon .
because of the existence of the rindler horizon , the accelerated observer sees the minkowski vacuum as thermally excited with the unruh temperature @xmath1.\end{aligned}\ ] ] since the unruh temperature is very small for ordinary acceleration , it was very difficult to detect the unruh effect . but with the ultra - high intensity lasers , the unruh effect can be experimentally accessible . in the electro - magnetic field of a laser with intensity @xmath2 ,
the unruh temperature is given by @xmath3.\end{aligned}\ ] ] the extreme light infrastructure project @xcite is planning to construct peta watt lasers with an intensity as high as @xmath4 $ ] . the expected unruh temperature becomes more than @xmath5 .
so it is time to ask ourselves how we can experimentally observe such high unruh temperature of an accelerated electron in the laser field .
chen and tajima proposed that one may be able to detect the unruh effect by observing quantum radiation @xcite from the electron .
it is called the unruh radiation .
since a uniformly accelerated electron feels the vacuum ( with quantum virtual pair creations of particles and anti - particles ) as thermally excited with the unruh temperature , the motion of the electron fluctuates and is expected to become thermalized(fig . [ fluctuation ] ) .
this fluctuating motion of the electron changes the acceleration of the electron and may produce additional radiation ( the unruh radiation ) to the ordinary larmor radiation .
the rough estimation @xcite suggested that the strength of the unruh radiation is much smaller than the classical one by @xmath6 , but the angular dependence becomes quite different . especially in the direction along the acceleration
there is a blind spot for the larmor radiation while the unruh radiation is expected to be radiated more spherically .
hence they proposed to detect the unruh radiation by setting a photon detector in this direction .
the above argument seems intuitively correct , but there are two problems that should be clarified .
the first problem is the _ thermalization time _ of the fluctuation .
the electromagnetic field of laser are not constant but oscillating .
one may approximate the electron s motion around the turning points by a uniform acceleration .
this approximation is valid only when the period of the laser is large enough compared to the relaxation time ( or thermalization time ) of the particle s fluctuation . using a stochastic approach
, we obtained the relaxation time of the fluctuation and showed that the relaxation time is longer than the period of the laser .
in such a case , we must fully analyze the transient dynamics of the fluctuation to calculate the radiation in the laser field .
the second problem is the _ interference effect_. since the unruh radiation originates in the interaction with the particle with the quantum fluctuations of the vacuum , we can not neglect the interference of the unruh radiation and the vacuum quantum fluctuations . in a simpler model , it has been known that the unruh radiation is completely canceled by the interference effect .
the cancellation was shown for the unruh detector in both 1 + 1 and 3 + 1 dimensions@xcite .
there was no calculation of the interference effect in the case of the uniformly accelerated charged particle since the calculation needs some technicalities . in the paper
@xcite we calculated the interference effect for the charged particle in the scalar qed and found that some of the unruh radiation is canceled by the interference effect , but the cancellation occurs only partially .
so we still have a possibility to detect additional radiation from the uniformly accelerated charged particle , but the complete understanding needs more detailed analysis .
in the rest of this report we first briefly review the stochastic model of a uniformly accelerated charged particle and then show how the thermalization of the fluctuation occurs by solving the stochastic equation .
finally we briefly sketch the calculation of the radiation , particularly put emphasis on the interference effect .
more details of the calculation of the interference effect and the unruh radiation are reviewed in another report of the same authors in the proceedings @xcite .
we consider the scalar qed .
the model is analyzed in @xcite and here we briefly review the settings and the derivation of the stochastic abraham - lorentz - dirac ( ald ) equation .
the system composes of a relativistic particle @xmath7 and the scalar field @xmath8 .
the action is given by @xmath9 = & -m \int d\tau \
, \sqrt{\dot{z}^{\mu } \dot{z}_\mu } + \int d^4x \ , \frac{1}{2}(\partial_\mu \phi)^2 \nonumber \\ & + \int d^4x \ , j(x;z ) \phi(x ) .\end{aligned}\ ] ] the scalar current @xmath10 is defined as @xmath11 we choose the parametrization @xmath12 to satisfy @xmath13 .
the equation of motion of the particle is given by @xmath14 where we have added the external force @xmath15 so as to accelerate the particle uniformly ; @xmath16 .
then a classical solution of the particle ( in the absence of the coupling to @xmath17 ) is given by @xmath18 the equation of motion of the radiation field @xmath19 is solved by using the retarded green function @xmath20 as @xmath21 where @xmath22 is the homogeneous solution of the equation of motion and represents the vacuum fluctuation .
it is responsible for the particle s fluctuating motion under a uniform acceleration . inserting the solution ( [ sol - phi ] ) into ( [ particle - eom ] )
, we have the following stochastic equation for the particle @xmath23 where @xmath24 } - \ddot{z}_\mu$ ] , which comes from the deviation of the current @xmath25 the homogeneous part @xmath26 of the scalar field describes the gaussian fluctuation of the vacuum , hence , the first term in the parenthesis of ( [ stochastic ] ) can be interpreted as random noise to the particle s motion @xmath27 it is essentially quantum mechanical , but if it is evaluated on a world line of a uniformly accelerated particle @xmath28 , it behaves as the ordinary finite temperature noise .
the second term in the parenthesis of ( [ stochastic ] ) is a functional of the total history of the particle s motion @xmath29 for @xmath30 , but it can be reduced to the so called radiation damping term of a charged particle coupled with radiation field .
it is generally nonlocal , but since the green function damps rapidly as a function of the distance @xmath31 , the term is approximated by local derivative terms .
after the mass renormalization , we get the following generalized langevin equation for the charged particle , @xmath32 where @xmath33 .
this equation is an analog of the ald equation for a charged particle interacting with the electromagnetic field .
the dissipation term is induced by the effect of the backreaction of the particle s radiation to the particle s motion .
note that , if the noise term is absent , the classical solution ( [ classical - trajectory ] ) with a constant acceleration is still a solution to the equation ( [ stochastic - particle ] ) .
the stochastic equation ( [ stochastic - particle ] ) is nonlinear and difficult to solve .
here we consider small fluctuations around the classical trajectory induced by the vacuum fluctuation @xmath22 . especially we consider fluctuations in the transverse directions .
first we expand the particle s motion around the classical trajectory @xmath34 as @xmath35 the particle is accelerated along the @xmath36 direction . in the following we consider small fluctuation in transverse directions . by expanding the stochastic equation ( [ stochastic - particle ] ) , we can obtain a linearized stochastic equation for the transverse velocity fluctuation @xmath37 as , @xmath38 performing the fourier transformation with respect to the trajectory s parameter @xmath12 @xmath39 the stochastic equation can be solved as @xmath40 where @xmath41 the vacuum 2-point function along the classical trajectory can be evaluated from ( [ phi - correlation ] ) as @xmath42 it has originated from the quantum fluctuations of the vacuum , but it can be interpreted as finite temperature noise if it is evaluated on the accelerated particle s trajectory @xcite .
the fourier transformation of the symmetrized two point function is evaluated as @xmath43 where @xmath44 which is an even function of @xmath45 .
the correlator @xmath46 should be regularized at the uv , which is large @xmath45 or short proper time difference , where quantum field theoretic effects of electron become important .
full qed treatment is necessary there . for small @xmath45 ,
it is expanded as @xmath47 the expansion corresponds to the derivative expansion @xmath48 with this expansion , the expectation value of the square of the transverse velocity fluctuation can be evaluated as @xmath49 here we consider the acceleration of the electron to be at the order 0.1 ev , which is much smaller than the electron mass 0.5 mev . with the assumption @xmath50 , one can evaluate the integral and get the following result , @xmath51 here we have recovered @xmath52 and @xmath53 .
this gives the equipartition relation for the transverse momentum fluctuations in the unruh temperature @xmath54 .
the thermalization process of the stochastic equation ( [ stochastic - transverse ] ) can be also discussed . for simplicity , we approximate the stochastic equation by dropping the second derivative term
. then it is solved as @xmath55 where @xmath56 is given by @xmath57 the relaxation time is @xmath58 the velocity square can be also calculated as @xmath59 for @xmath60 , it approaches the thermalized average ( [ equipartition ] ) .
the relaxation time in the proper time can be estimated , for the parameter @xmath61 ev and @xmath62 mev , @xmath63 let s compare this relaxation time with the laser frequency . the planned wavelength of the laser at eli is around @xmath64 nm and the oscillation period of the laser field is very short ; @xmath65 seconds .
so the relaxation time is much longer and the charged particle can not become thermalized during each oscillation .
hence the assumption of the uniform acceleration in the laser field is not good .
even in such a situation , if the electron is accelerated in the laser field for a long time , an electron may feel an averaged temperature .
the position of the particle in the transverse directions also fluctuates like the ordinary brownian motion in a heat bath .
the mean square of the transverse coordinate @xmath66 is calculated as @xmath67 with the diffusion constant @xmath68 in the ballistic region where @xmath69 , the mean square becomes @xmath70 while in the diffusive region ( @xmath71 ) , it is proportional to the proper time as @xmath72 as the ordinary brownian motion , the mean square of the particle s transverse position grows linearly with time .
if it becomes possible to accelerate the particle for a sufficiently long period , it may be possible to detect such a brownian motion in future laser experiments .
now we are ready to calculate the radiation emanated from the uniformly accelerated charged particle .
an important point is the interference effect between the quantum fluctuations of the vacuum @xmath22 and the radiation induced by the fluctuating motion in the transverse directions @xmath73 .
first let s consider the two point function @xmath74 the unruh radiation estimated in @xcite corresponds to calculating the 2-point correlation function of the inhomogeneous terms @xmath75 .
( the same term also contains the larmor radiation . ) however , this is not the end of the story . as it has been discussed in @xcite , the interference terms @xmath76 may possibly cancel the unruh radiation in @xmath75 after the thermalization occurs .
this is shown for an internal detector , but it is not obvious whether the same cancellation occurs for the case of a charged particle we are considering .
the energy - momentum tensor of the radiation field can be obtained from the 2-point function @xmath77 it is written as a sum of the classical part and the fluctuating part @xmath78 .
the classical part corresponds to the larmor radiation while the fluctuating part contains both of the unruh radiation and the interference terms . in @xcite
we calculated the 2-point function including the interference term , and obtained the energy - momentum tensor .
the result we have obtained is summarized in @xcite in this proceedings .
some terms are partially canceled but not all .
hence , it seems that the uniformly accelerated charged particle emits additional radiation besides the larmor radiation .
the remaining terms after the partial cancellation are proportional to @xmath79 and suppressed compared to the larmor radiation .
it has a different angular distribution , but the additional radiation also vanishes in the forward direction .
we have systematically studied the thermalization of a uniformly accelerated charged particle in the scalar qed using the stochastic method , and calculated the radiation by the particle .
two main messages of are the fluctuation of the particle does nt become thermalized during the period of the laser , and we need to study transient dynamics to obtain the radiation from an electron accelerated in the oscillating laser field .
the issue of the interference terms are more subtle . since the particle fluctuation originates in the quantum vacuum fluctuation of the radiation field , it can be by no means neglected .
our result shows that the interference term partially cancels the unruh radiation , but some of them survives .
the remaining unruh radiation is smaller compared to the larmor radiation by a factor @xmath0 ( acceleration ) and has a different angular distribution . in this sense , it is qualitatively consistent with the proposal @xcite .
but as we briefly review in @xcite , the additional radiation also vanishes in the forward direction . and
it seems difficult to detect such additional radiation experimentally so far as the transverse fluctuation is concerned .
please beware that the longitudinal fluctuations which we have not calculated yet ( because of technical difficulties related to a choice of gauge ) may change the situation . | quantum field theories horizons exhibit thermodynamic behavior prominent phenomenon hawking radiation fundamental laws thermodynamics in black hole background similar phenomenon occurs for uniformly accelerated observer in minkowski vacuum @xcite unruh effect particle uniformly accelerated in minkowski space @xmath0 causal horizon rindler horizon no information transmitted from other side accelerated observer sees vacuum thermally excited with unruh temperature unruh temperature small for ordinary acceleration difficult to detect ultra - high intensity lasers unruh effect accessible in - field laser with intensity @xmath2 unruh temperature given by @xmath3. extreme light infrastructure project peta watt lasers intensity high @xmath4 expected unruh temperature more than @xmath5 observe high unruh temperature of accelerated electron in laser field chen tajima detect unruh effect by observing quantum radiation @xcite from electron called unruh radiation uniformly accelerated electron feels vacuum thermally excited with unruh temperature motion fluctuates expected to become thermalized fluctuating motion changes acceleration produce additional radiation unruh radiationestimation @xcite suggested strength unruh radiation smaller than classical by @xmath6 angular dependence different direction along acceleration blind spot for larmor radiation unruh radiation radiated spherically proposed to detect unruh radiation by setting photon detector in this direction argument correct two problems first problem thermalization time of fluctuation electromagnetic field laser not constant oscillating approximate electron s motion by uniform acceleration valid when period laser large compared to relaxation particle s fluctuation stochastic approach obtained relaxation time fluctuation showed longer than period laser must analyze transient dynamics fluctuation to calculate radiation in laser field second problem interference effect_ unruh radiation originates in interaction with particle quantum fluctuations vacuum can not neglect interference simpler model unruh radiation completely canceled by interference effect cancellation shown for unruh detector in 1 + 1 and 3 + 1 dimensions@xcite no calculation of interference effect uniformly accelerated charged particle calculated interference effect for charged particle in scalar qed some unruh radiation canceled by interference effect cancellation occurs partiallypossibility to detect additional radiation from uniformly accelerated charged particle complete understanding needs detailed analysis report review stochastic model of uniformly accelerated charged particle show thermalization of fluctuation by solving stochastic equation sketch calculation radiation emphasis on interference effect details calculation interference effect unruh radiation reviewed in another report same proceedings @xcite consider scalar qed model analyzed in @xcite review settings derivation of stochastic abraham - lorentz - dirac ( ) equation system relativistic particle @xmath7 scalar field @xmath8 action given by @xmath9 = -m\tau^4x scalar current @xmath10 defined as @xmath11 choose parametrization @xmath12 satisfy @xmath13 equation of motion particle given by @xmath14 added external force @xmath15 to accelerate particle uniformly @xmath16classical solution particle coupling @xmath17 by @xmath18 equation motion radiation field @xmath19 solved retarded green function @xmath20 @xmath21 @xmath22 homogeneous solution represents vacuum fluctuation responsible for particle s fluctuating motion under uniform acceleration inserting solution [ sol - phi into particle - eom ] stochastic equation for particle @xmath23 where @xmath24 from deviation current @xmath25 homogeneous part @xmath26 scalar field describes gaussian fluctuation vacuum first term stochastic as random noise to particle s motion quantum mechanical accelerated ordinary finite temperature noise second term stochastic ] ) functional of history particle s motion @xmath29 @xmath30 reduced to radiation damping term of charged particle radiation field nonlocal green function damps rapidly distance @xmath31 term approximated by local derivative terms after mass renormalization generalized langevin equation for charged particle @xmath32 where @xmath33 analog equation charged particle interacting with electromagnetic field dissipation term induced by backreaction particle s radiation to motionif noise term absent classical solution constant acceleration solution to equation stochastic - particle stochastic equation nonlinear difficult to solve consider small fluctuations around classical trajectory by vacuum fluctuation fluctuations transverse directions expand particle s motion around trajectory accelerated along @xmath36 direction consider small fluctuation in transverse directions expanding stochastic equation linearized stochastic equation for transverse velocity fluctuation @xmath37 @xmath38 fourier transformation trajectory parameter stochastic equation solved as @xmath40 vacuum 2-point function along classical trajectory evaluated phi - correlation @xmath42 originated from quantum fluctuations interpreted as finite temperature noise on accelerated particle s trajectory fourier transformation of symmetrized two point function evaluated as @xmath43 even function of @xmath45 correlator @xmath46 regularized at uv large time difference quantum field theoretic effects of electron important full qed treatment necessary@xmath45 expanded as @xmath47 corresponds to derivative expansion @xmath48 expectation value square transverse velocity fluctuation evaluated as @xmath49 acceleration electron 0.1 ev smaller than electron mass 0.5 mev assumption @xmath50 evaluate integral result @xmath51 recovered @xmath52 and @xmath53 equipartition relation for transverse momentum fluctuations in unruh temperature @xmath54 thermalization process stochastic equation discussed approximate stochastic equation dropping second derivative term solved as @xmath55 @xmath56 given by @xmath57 relaxation time @xmath58 velocity square calculated as @xmath59 @xmath60 approaches thermalized average relaxation time estimated parameter @xmath61 ev @xmath62 mev @xmath63 compare relaxation time with laser frequency wavelength laser @xmath64 nm oscillation period short @xmath65 seconds relaxation time longer charged particle not thermalized during each oscillation assumption uniform acceleration in laser field not good electron accelerated laser field feel averaged temperature position particle in transverse directions fluctuates like motion heat bathmean square transverse coordinate @xmath66 calculated as @xmath67 diffusion constant @xmath68 ballistic region @xmath69 square becomes @xmath70 diffusive region @xmath71 proportional to proper time as @xmath72 ordinary brownian motion mean square particle s transverse position grows linearly with time accelerate particle long detect brownian motion in future laser experiments ready calculate radiation from uniformly accelerated charged particle interference effect between quantum fluctuations vacuum @xmath22 radiation fluctuating motion transverse directions @xmath73 consider two point function @xmath74 unruh radiation estimated @xcite corresponds to 2-point correlation function inhomogeneous terms @xmath75 larmor radiation not interference terms @xmath76 cancel unruh radiation @xmath75 after thermalization shown internal detector charged particle energy - momentum tensor radiation field from 2-point function @xmath77 sum of classical part fluctuating part @xmath78 classical part corresponds larmor radiation fluctuating part contains unruh radiation interference terms calculated 2-point function interference term obtained energy - momentum tensor result summarized in @xciteterms partially canceled not all uniformly accelerated charged particle emits additional radiation besides larmor remaining terms after cancellation proportional to @xmath79 suppressed compared to larmor radiation different angular distribution additional radiation vanishes forward direction studied thermalization of uniformly accelerated charged particle in scalar qed stochastic method calculated radiation by particle fluctuation does nt thermalized during laser need study transient dynamics radiation from electron accelerated in oscillating laser field issue interference terms subtle particle fluctuation originates in quantum vacuum fluctuation radiation neglected result shows interference term partially cancels unruh radiation some survives remaining unruh radiation smaller larmor by factor @xmath0 ( different angular distribution qualitatively consistent with proposal @xcite additional radiation vanishes in forward direction difficult to detect additional radiation transverse fluctuation longitudinal fluctuations not calculated technical difficulties may change situation | 0.511219 | 3,160 | 1,560 | , Saving $0.1 in GPT-4. |
67,311 | based on simple physics arguments it is shown that the concept of quark - gluon plasma , a state of matter consisting of uncorrelated quarks , antiquarks , and gluons , has a fundamental problem . | the existence of quark - gluon plasma ( qgp ) , a state of matter in which quarks are free to move in space , was postulated by cabibbo and parisi in @xcite in response to the concept of limiting temperature of hagedorn @xcite .
a transition from hadronic matter to qgp is supposed to occur at an energy density of about 1 gev/@xmath0 which can be reached either by heating or by compressing or both .
intensive search for qgp in collisions of lead and gold nuclei at energies of up to @xmath1 17.2 gev at the cern sps yielded `` compelling evidence for the existence of a new state of quark - gluon matter ( ... ) in which quarks are liberated to roam freely '' @xcite .
the subsequent au+au collision experiments at rhic , albeit with much higher energies of @xmath2 200 gev , resulted in a somewhat weaker statement reporting only a `` new form of nuclear matter '' with an `` energy density and temperature clearly exceeding the critical values predicted by qcd calculations '' @xcite . in the same report
it is stressed that the observed medium behaves like a strongly coupled fluid rather than the expected gas of free uncorrelated quarks .
the common expectation is that the latter can still exist at energies several times higher than the critical energy density @xcite .
this is supported by lattice calculations which show that with increasing temperature the system ( slowly ) approaches the ideal gas limit @xcite albeit deviations from ideal gas still occur at temperatures as high as ( 2 - 3 ) @xmath3 @xcite . the evidence and the possible reasons for these deviations
were extensively discussed by a pioneer of the field in @xcite . in this letter
i use simple physics arguments to show that the concept of qgp , a state of matter with liberated quarks , at any temperature has a fundamental problem . the problem , which does not manifest itself during creation of qgp but only during the transition back to hadrons , consists in the fact that simultaneous hadronization in regions separated by space - like intervals must in some cases lead to single quarks left at the borders between hadronization domains because there is no way to synchronize this process without violating causality .
the problem is exposed in detail in section 2 by means of a gedanken experiment . in section 3
i will discuss possible solutions of the paradox .
i start from the assumption that the qgp , large ( comparing to a nucleon ) volume filled with uncorrelated quarks , antiquarks , and gluons , can exist .
i will then use a certain amount of qgp to perform a gedanken experiment during which i only do things which are not forbidden by physics laws .
the final state after the experiment , nevertheless , will be one with isolated quarks separated by a macroscopic distance which is not allowed by qcd . here
are the steps of the procedure .
\i ) i create one cubic mm of qgp with a temperature well above the critical temperature and a total net baryon number @xmath4=0 .
i stretch it to dimensions of 10 fm x 10 fm x 1000 light years , keeping the density constant .
i connect both ends such as to form a ring .
\ii ) i break the qgp ring at one point by allowing the qgp to expand and cool such that the hadronization starts there .
the phase boundary propagates along the ring in both directions with the velocity of , say , 0.05 c. for the problem under discussion it does not matter whether the propagation of the phase boundary is spontaneous ( rarefaction wave , moving with the speed of sound ) or imposed from outside ( removing the bonds ) .
the hadronization proceeds until the last chunk of qgp , on the opposite side of the ring , turns into hadrons .
+ i can repeat this gedanken experiment , pictorially represented in fig . [
fig : ring1 ] , many times .
the hadronization is always successful in the sense that all quarks in the system are turned into hadrons .
now , however , i introduce a little modification in the second step : ii ) as before , i break the qgp ring at one point by allowing the qgp to expand and cool such that the hadronization starts there . at the same time
, however , my assistant does the same at the opposite end of the ring .
this has no immediate influence at what is happening at my end of the ring because the two points are separated by light years .
now i have two separate blobs of qgp .
the four phase boundaries propagate until two small chunks of qgp remain ( fig .
[ fig : ring2 ] ) . + obviously , there is a 33% chance that these two chunks have integer net baryon numbers . with the remaining probability of 67% they have fractional ones .
so , if i repeat our second experiment many times , sooner or later i will end up with two objects with fractional baryon numbers , separated by light years . this state is not allowed by qcd .
we started from an allowed state , we never did anything forbidden by physics laws , and we ended up in a state which is forbidden .
in the next section i will discuss possible resolutions of the paradox . before doing this ,
several remarks are in order regarding the technical aspects of the presented gedanken experiment .
first , while i was considering here the baryon numbers of the outcoming particles and requiring it to be integer , alternatively , one could monitor the color of the final particles , and require them to be white . in both cases
the reasoning is equivalent and leads to the same conclusions . for technical reasons i decided to base the argument on baryon number and not on color except in figs .
[ fig : zoo ] and [ fig : zoo2 ] where color is better to explain the point .
second , the amount of qgp used in the described gedanken experiment is much higher than the one we are familiar with , i.e. the one expected in a relativistic heavy ion collision , and the ring - like shape is something one would not expect to be very frequent in nature . on the other hand , a simple calculation using
the mass of the observable universe indicates that the amount of qgp during big bang was much higher than the one considered here .
what concerns the ring shape , while a ring is best to illustrate the problem , the problem remains the same even if one squeezes the ring such that the two hadronization starting points get close to each other . in this case
the qgp blob resembles in shape the elongated fireball created in a heavy ion collision , with the hadronization starting in the middle ( also quite possible in a heavy ion collision ) .
the information does not need to propagate from the other side of the ring but only across some 5 - 10 fm , so the situation quantitatively is much less dramatic .
qualitatively , however , the problem is the same - whether or not given quarks are allowed to form a hadron depends not only on themselves and on their direct neighborhood but also on remote parts of the qgp volume .
one could argue that this is not a problem on the scale of 5 - 10 fm .
however , the qgp blob created in the early universe had the same problem if the hadronization , caused by the expansion , was happening in the whole volume . converting quarks into hadrons in the whole volume at the same time can be compared to trying to reach a homogeneous magnetization in a bulk ferromagnetic by lowering the temperature .
third , since the plasma temperature does not enter explicitely the problem persists for all temperatures above the critical temperature .
in particular , one can not argue that the paradox is restricted to the cases with temperatures close to the critical temperature .
fourth , one could try to dismiss the depicted experiment by arguing that splitting qgp in two parts is like splitting conducting material in electrodynamics .
there , any residual charge imbalance is removed by the current flowing through the last point of contact between the two halves . however , this is true only if the detaching proceeds slowly .
if e.g. two metals with different work functions are detached quickly non - zero net charges remains on the two parts . for electric charge , unlike baryon number , this is not a problem .
it is worth considering whether the described problem could be just another case of the famous epr paradox .
the qgp ring is an entangled state of quarks and gluons .
when hadronization starts at one point the wave function collapses and from now on every point of the ring `` knows '' whether starting hadronization there is allowed or not .
however , in this case one could use the ring for superluminal transfer of information . indeed ,
if upon starting hadronization i observe a string then this means that my assistant did start hadronization , even if he did it only one second before me and even if he is many light years away from me .
it is commonly believed , however , that the entangled states and the epr paradox do not offer a way of propagating information at superluminal speed @xcite . the second way out would be to assume that the qgp properties are such that a plasma blob can not have a hole , and
the hadronization can only happen at the surface .
volume hadronization , e.g. caused by density dropping uniformly in the entire volume during hubble - like expansion , would be forbidden .
the qgp would be resistant against attempts of pulling apart pieces of it , i.e. it would behave like a liquid with infinite surface tension . for heavy ion collisions it would mean that the hadronization starts at both ends of the elongated fireball , in spite of the fact that the particles there have the largest lorentz gamma factor . for the early universe the consequences are much more dramatic .
since the phase boundary can not proceed faster than the speed of sound , and certainly not faster than the speed of light in vacuum , and since the observable universe in the qgp state had dimensions comparable to the size of the solar system , its hadronization must have taken minutes .
since the entire universe may be much larger than the observable universe ( maybe infinite ) the actual time needed might be even longer .
the third possibility is that local correlations between quarks make some cutting surfaces more probable than the others when it comes to cutting the ring and starting the hadronization . obviously , in absence of such correlations the qgp ring basically looks like in fig .
[ fig : zoo ] and no preferred breaking points can be recognized . if , however , some kind of interactions lead to clustering of quarks and gluons into ( white ) objects of integer baryon numbers like in fig . [
fig : zoo2 ] then starting hadronization from several points of the ring at the same time will not lead to any problem .
however , this kind of matter would be hadron resonance matter rather than the qgp .
the degrees of freedom would not be the ones of quarks and gluons , expected from a genuine quark - gluon plasma .
arguing that the plasma may look like in fig .
[ fig : zoo ] at high temperatures and like in fig .
[ fig : zoo2 ] close to the phase transition does not resolve the paradox because the transition from uncorrelated quarks to clusters again has to take a time comparable to the size of the qgp volume divided by the speed of light .
the much shorter time scales of `` whitening of the qgp '' , obtained in @xcite , were based on statistical considerations in which the problem discussed in this letter can not show up .
i demonstrated that the concept of qgp , state of matter with uncorrelated quarks , antiquarks , and gluons , leads to isolated objects with fractional baryon numbers , unless superluminal signalling is allowed , or , by some mechanism , the hadronization is restricted to the surface of the qgp volume , meaning that e.g. the hadronization in the early universe took at least minutes rather than a couple of microseconds .
the third , obvious , way of avoiding the paradox is to declare the uncorrelated qgp as non - existent , and to replace it by a state consisting of quark clusters with integer baryon numbers ( resonance matter ) .
both the surface - hadronization and the resonance matter options result in a liquid- rather than a gas - like structure of the matter .
this agrees with the hydrodynamical character of the matter created in nuclear collisions at rhic and , at the same time , indicates that this character will be preserved at higher temperatures .
i gratefully acknowledge useful discussions with peter braun - munzinger , pawe danielewicz , staszek mrwczyski , sergei voloshin , and , especially , uli heinz . at the same time
i would like to stress that they carry no responsibility for the views expressed in this letter .
i would also like to thank bengt friman and anton andronic for reading the manuscript and helping to spot its many deficiencies .
99 n. cabibbo and g. parisi , phys .
r. hagedorn , nuovo cimento suppl . 3 ( 1965 ) 147 .
r. hagedorn and j. ranft , nuovo cimento suppl . 6
( 1968 ) 169 .
http://newstate-matter.web.cern.ch/newstate-matter ; see also u. heinz and m , jacob , arxivnucl - th/0002042v1 .
report bnl-73847 - 2005 , brookhaven national laboratory , 2005 j.i .
kapusta , arxiv:0705.1277 [ nucl - th ] f. karsch , proceedings of qm2006 , shanghai , arxiv : hep - ph/0701210v1 .
f. karsch , nucl.phys.a783:13-22,2007 .
e. shuryak , arxiv : hep - ph/0703208v1 .
tiwari , superluminal phenomena in modern perspective , rinton press , 2003 .
c. manuel and s. mrwczyski , phys.rev .
d70 ( 2004 ) 094019 . | existence of quark - gluon plasma quarks free move postulated by cabibbo parisi limiting temperature transition from hadronic matter to qgp at energy density 1 gev/@xmath0 reached by heating or compressing search for qgp in collisions lead gold nuclei at to 17.2 gev at cern sps yielded evidence for new state quark - gluon matter quarks liberated roam subsequent au collision experiments at rhic higher energies 200 gev resulted in weaker statement new form of nuclear matter with energy density temperature exceeding critical values calculations observed medium behaves like strongly coupled fluid expected gas of free uncorrelated quarks latter can exist at energies higher than critical energy density supported by lattice calculations with increasing temperature system approaches ideal gas limit deviations from ideal gas occur at temperatures 2 - 3 ) @xmath3 evidence reasons for deviations discussed by pioneer in physics arguments concept of qgp state with liberated quarks at temperature has fundamental problemproblem not during creation qgp only during transition to hadrons in simultaneous hadronization in regions separated by space intervals lead to single quarks left at borders between hadronization domains no way to synchronize without violating causality problem exposed in section 2 gedanken experiment in section 3 discuss solutions paradox start from assumption qgp large volume filled with uncorrelated quarks antiquarks gluons can exist use amount of qgp to gedanken experiment not forbidden by physics laws final state experiment with isolated quarks separated by macroscopic distance not allowed by qcd steps procedure create one cubic mm of qgp with temperature above critical temperature total net baryon number @xmath4=0 stretch to dimensions 10 fm x 10 fm x 1000 light years density constant connect both ends form ring break qgp ring to expand cool hadronization starts phase boundary propagates along ring in both directions with velocity of 0.05 c matter propagation spontaneous or imposed hadronization proceeds until last chunk of qgp turns into hadrons repeat gedanken experiment in fig . many timeshadronization successful all quarks turned into hadrons now introduce modification in second step break qgp ring allowing qgp expand and cool hadronization starts assistant does same opposite end no immediate influence my end two points separated by light years now two separate blobs of qgp four phase boundaries propagate until two small chunks qgp remain ( fig . fig : ring2 ] ) 33% chance two chunks have integer net baryon numbers remaining probability 67% have fractional ones if repeat second experiment end up with two objects with fractional baryon numbers separated by light years state not allowed by qcd started from allowed state never did forbidden ended up in state forbidden next section discuss possible resolutions of paradox remarks regarding technical aspects of gedanken experiment considering baryon numbers integer alternatively monitor color final particles require them be white reasoning equivalent leads to same conclusions for decided base argument on baryon number not color except in figs . zoo ] and zoo2 ] where color better amount of qgp used in described gedanken experiment higher than familiar withexpected in relativistic heavy ion collision ring - like shape not frequent in nature simple calculation using mass universe indicates qgp during big bang higher than considered here ring shape ring problem problem remains same even if squeezes ring two hadronization starting points close qgp blob resembles elongated fireball in heavy ion collision hadronization starting in middle possible information propagate from other side of ring only across 5 - 10 fm situation quantitatively less dramatic qualitatively problem same - quarks form hadron depends on direct neighborhood remote parts of qgp volume not problem on scale of 5 - 10 fm qgp blob in early universe had same problem if hadronization in whole volume converting quarks into hadrons in whole volume compared to homogeneous magnetization in bulk ferromagnetic by lowering temperature plasma temperature enter problem persists for all temperatures above critical temperature paradox restricted to cases with temperatures close to critical temperature experiment splitting qgp in two parts like splitting conducting material in electrodynamics residual charge imbalance removed by current flowing through last point of contact between two halves true only if detaching proceeds slowlymetals different work functions detached quickly non - zero net charges remains on two parts electric charge unlike baryon number not a problem considering described problem case epr paradox qgp ring is entangled state of quarks gluons hadronization starts wave function collapses every point ring knows starting hadronization allowed could use ring for superluminal transfer information if starting hadronization observe string means assistant hadronization one second before light years away commonly believed entangled states epr paradox propagating information at superluminal speed second way assume qgp properties plasma blob have hole hadronization only at surface volume hadronization density dropping expansion forbidden qgp resistant against pulling apart pieces behave like liquid with infinite surface tension heavy ion collisions hadronization starts at both ends of elongated fireball particles have largest lorentz gamma factor early universe consequences more dramatic phase boundary not proceed faster than speed sound not speed light in vacuum observable universe in qgp state had dimensions comparable to size solar system hadronization must have taken minutes entire universe may be larger than observable universe actual time needed might be longerthird possibility local correlations between quarks make cutting surfaces probable ring starting hadronization correlations qgp ring looks like fig . zoo ] no preferred breaking points recognized if interactions lead to clustering quarks gluons into white objects of integer baryon numbers like fig zoo2 starting hadronization from several points problem matter hadron resonance matter qgp degrees of freedom not of quarks gluons from genuine quark - gluon plasma plasma look like fig . zoo at high temperatures close to phase transition resolve paradox transition from uncorrelated quarks to clusters time comparable to size qgp volume divided by speed light shorter time scales of whitening of qgp in @xcite based on statistical considerations problem show up concept of qgp uncorrelated quarks gluons leads to isolated objects with baryon numbers unless superluminal signalling allowed or hadronization restricted to surface qgp volume hadronization in early universe took minutes microseconds third way avoiding paradox declare uncorrelated qgp as non - existent replace by state quark clusters with integer baryon numbers resonance mattersurface - hadronization resonance matter liquid gas structure agrees hydrodynamical character nuclear collisions higher temperatures discussions with peter braun - munzinger pawe danielewicz staszek mrwczyski sergei voloshin uli heinz no responsibility views letter bengt friman anton andronic manuscript deficiencies cabibbo g. parisi hagedorn nuovo cimento suppl . 3 ( 1965 ) 147 ranft nuovo cimento 6 ( 1968 ) 169-matter.web.cern.ch u heinz jacob arxivnucl - th/0002042v1 report bnl-73847 - 2005 brookhaven national laboratory kapusta arxiv:0705.1277 f karsch proceedings qm2006 arxiv hep - ph/0701210v1 karsch nucl.phys.a783:13-22,2007 shuryak arxiv hep - ph/0703208v1 tiwari superluminal phenomena modern perspective rinton press 2003 manuel s mrwczyski phys.rev . d70 ( 2004 ) 094019 | 0.447368 | 3,042 | 1,513 | , Saving $0.1 in GPT-4. |
67,325 | in the town of saratov where he was prisonner , poncelet , continuing the work of euler and steiner on polygons simultaneously inscribed in a circle and circumscribed around an other circle , proved the following generalization .
* theorem * _ let @xmath0 and @xmath1 be two smooth conics in @xmath2 .
if @xmath1 passes through the @xmath3 vertices of a complete polygon with @xmath4 sides tangent to @xmath0 then @xmath1 passes through the vertices of infinitely many such polygons .
_ according to berger @xcite this theorem is the nicest result about the geometry of conics . even if it is , there are few proofs of it . to my knowledge
there are only three .
the first proof , published in 1822 and based on infinitesimal deformations , is due to poncelet ( @xcite ) .
later , jacobi proposed a new proof based on finite order points on elliptic curves ; his proof , certainly the most famous , is explained in a modern way and in detail by griffiths and harris ( @xcite ) . in 1870
weyr proved a poncelet theorem in space ( more precisely for two quadrics ) that implies the one above when one quadric is a cone ; this proof is explained by barth and bauer ( @xcite ) . our aim in this short note is to involve vector bundles techniques to propose a new proof of this celebrated result .
poncelet did not appreciate jacobi s for the reason that it was too far from the geometric intuition .
i guess that he would not appreciate our proof either for the same reason . | in all this text the ground field is @xmath5 .
a set of @xmath4 vertices connected by @xmath4 distinct lines form a @xmath4-gon , when a set consisting of @xmath4 distinct lines with their @xmath3 vertices form a complete @xmath4-gon .
-gon inscribed in a cubic , a complete @xmath6-gon circumscribed around a conic and a pentagon simultaneously inscribed in a conic and circumscribed around another , height=245 ] we say that a @xmath4-gon ( respectively a complete @xmath4-gon ) is inscribed in a given curve if this curve passes through the @xmath4-vertices ( respectively the @xmath3 vertices ) .
we say that a @xmath4-gon , or a complete @xmath4-gon , is circumscribed around a smooth conic @xmath0 if the sides of the polygon , i.e. the @xmath4 lines , are tangent to the @xmath0 .
first of all let us introduce a vector bundle @xmath7 naturally associated to any set of @xmath4 lines tangent to a fixed smooth conic @xmath8 ( they were defined by schwarzenberger in @xcite ) .
a set of @xmath4 lines tangent to @xmath0 corresponds by projective duality to a degree @xmath4 divisor on the dual conic @xmath9 . according to the isomorphism @xmath10 we can consider the subvariety @xmath11 defined by the equation @xmath12 ( @xmath13 are the homogeneous coordinates on @xmath14 and @xmath15 the homogeneous coordinates on @xmath16 ) and the projections @xmath17 and @xmath18 respectively on @xmath14 and @xmath16 . @xmath19 the variety @xmath20 is a double cover of @xmath14 ramified along @xmath0 . if @xmath21 then @xmath22 is a line in @xmath14 tangent to @xmath0 . if , instead of considering a point , we are considering a line bundle on @xmath16 we will find a vector bundle of rank two on @xmath14 by taking the direct image of its inverse image .
moreover , following scwharzenberger , we know a very explicit resolution of this bundle . indeed , tensorizing the following exact sequence @xmath23 by @xmath24 and taking its direct image by @xmath17 we have : @xmath25 the map @xmath26 can be represented by the matrix of linear forms : @xmath27 let us show that the zero locus @xmath28 of a non zero section @xmath29 is the set of @xmath3 vertices of the @xmath4 tangent lines to @xmath0 given by the corresponding @xmath4 points on @xmath30 .
we denote by @xmath31 this set of @xmath4 points on @xmath30 .
since @xmath32 , the section @xmath33 corresponds to an hyperplane @xmath34 .
this hyperplane meets the image @xmath35 of @xmath36 in @xmath37 ( by the veronese imbedding @xmath38 ) along @xmath4 points which correspond to the points of the divisor @xmath31 .
the section @xmath33 induces a rational map @xmath39 which is not defined over the zero - scheme @xmath28 .
more precisely let @xmath40 be a point in @xmath14 and @xmath41 its dual line .
this dual line corresponds by the veronese morphism to a two - secant line of @xmath35 ( call it @xmath42 again ) . if @xmath42 is not a two secant line to @xmath31 there is exactly one intersection point @xmath43 which is the image of @xmath40 by @xmath44 .
conversely the map @xmath44 is not well defined when @xmath45 , i.e. when @xmath42 is a two - secant line to @xmath31 , or equivalently when @xmath40 is a vertex of two tangent lines to @xmath0 along @xmath31 .
we can prove now the so - called darboux theorem ( @xcite , page 248 ) .
[ da ] let @xmath46 be a curve of degree @xmath47 .
if there is a complete @xmath4-gon ( polygon with @xmath4 sides ) tangent to a smooth conic @xmath0 and inscribed into @xmath48 , then there are infinitely many of them .
i recall here a proof already written in @xcite .
a complete @xmath4-gon circumscribed around @xmath0 and inscribed into @xmath48 corresponds to a non - zero global section @xmath49 vanishing along its vertices @xmath28 : @xmath50 by the remark of the previous section ( [ prelim ] ) , the curve @xmath48 corresponds to a global section of @xmath51 .
since the map @xmath52 is surjective , there exists a non - zero section @xmath53 ( i.e. another @xmath4-gon ) such that the determinant @xmath54 is the equation of @xmath48 .
this proves the theorem since any linear combination of @xmath33 and @xmath55 vanish along the vertices of a complete @xmath4-gon .
these curves described by darboux are called _
poncelet curves_. when @xmath56 they are the so - called lurth quartics ( see @xcite ) .
let us now consider @xmath4-gons that are simultaneously inscribed in a smooth conic and circumscribed around a other one . for these configurations poncelet proved ( @xcite , page 362 ) :
let us consider one such @xmath4-gon .
it is the union of @xmath4 lines @xmath58 with the @xmath3 vertices @xmath59 for @xmath60 and @xmath61 .
there is a non - zero section @xmath29 vanishing along the vertices of these lines .
we denote by @xmath28 the set of these vertices .
let us tensorize the following exact sequence @xmath62 by @xmath63 .
since @xmath64 consists in @xmath4-points , it induces the following decomposition of @xmath7 along @xmath1 : @xmath65 according to this decomposition , we consider the following exact sequence : @xmath66 where @xmath67 is a rank two vector bundle over @xmath14 . taking the cohomology long exact sequence we verify immediately that @xmath68 .
then , let us consider a pencil of sections of @xmath67 and also the pencil of sections of @xmath7 induced by it .
we obtain a commutative diagram : @xmath69 the sheaf @xmath70 is supported by a curve @xmath71 of degree @xmath47 that is the determinant of a pencil of sections of @xmath7 .
this curve @xmath71 is a poncelet curve .
then a general point on @xmath71 is a vertex of a complete @xmath4-gon inscribed in @xmath71 and circumscribed around @xmath0 .
moreover any intersection point of the @xmath4 lines forming the @xmath4-gon with @xmath71 is a vertex of this @xmath4-gon ( it is clear by bzout theorem since @xmath72 ) .
let @xmath73 be the curve supporting the sheaf @xmath74 .
we have of course @xmath75 .
then @xmath1 is an irreducible component of a poncelet curve and by the way any general point on @xmath1 is the vertex of complete @xmath4-gon inscribed in @xmath71 .
then this configuration meets the conic @xmath1 in at least ( because there are @xmath4 lines ) and at most ( because they are vertices and the decompostion of the bundle along @xmath1 is fixed ) @xmath4-points , so exactly @xmath4-points , each counting doubly .
bos , h.j.m . , kers , c. , oort , f. , raven , d.w .
: poncelet s closure theorem , its history , its modern formulation , a comparison of its modern proof with those by poncelet and jacobi , and some mathematical remarks inspired by these early proofs .
expo . math . 5 ( 1987 ) , 289364 . | ground field is @xmath5 @xmath4 vertices @xmath4 lines form @xmath4-gon @xmath4 lines @xmath3 vertices form complete @xmath4-gon inscribed in cubic @xmath6-gon circumscribed around conic pentagon height=245 @xmath4-gon inscribed in curve if passes through @xmath4-vertices @xmath4-gon circumscribed around smooth conic @xmath0 if sides polygon @xmath4 lines tangent to @xmath0 vector bundle @xmath7 associated to @xmath4 lines tangent to conic @xmath8 @xmath4 lines tangent to @xmath0 corresponds projective duality to @xmath4 divisor on dual conic @xmath9 isomorphism @xmath10 subvariety @xmath11 defined equation @xmath12 @xmath13 homogeneous coordinates on @xmath14 @xmath15 @xmath16 projections @xmath17 and @xmath18 on @xmath14 @xmath16 variety @xmath20 double cover of @xmath14 @xmath0 @xmath22 line in @xmath14 tangent to @xmath0considering point line bundle @xmath16 vector bundle two @xmath14 inverse image following scwharzenberger explicit resolution bundle tensorizing sequence @xmath23 @xmath24 direct image @xmath17 @xmath25 map @xmath26 represented matrix linear forms @xmath27 zero locus @xmath28 non zero section @xmath29 @xmath3 vertices @xmath4 tangent lines @xmath0 @xmath4 points @xmath30 denote @xmath31 @xmath4 points @xmath30 section @xmath33 corresponds hyperplane @xmath34 hyperplane meets image @xmath35 @xmath36 @xmath37 @xmath4 points divisor @xmath31 section @xmath33 induces rational map @xmath39 not defined zero - scheme @xmath28 @xmath40 point in @xmath14 @xmath41 dual line dual line corresponds two - secant line @xmath35 if @xmath42 not two secant line @xmath31 one intersection point @xmath43 image @xmath40 by @xmath44 map @xmath44 not defined @xmath45@xmath42 two - secant line to @xmath31 @xmath40 vertex two tangent lines @xmath0 prove darboux theorem @xcite page 248 @xmath46 curve of degree @xmath47 complete @xmath4-gon tangent to smooth conic @xmath0 inscribed into @xmath48 infinitely many complete @xmath4-gon circumscribed around @xmath0 inscribed @xmath48 corresponds non - zero global section @xmath49 vanishing vertices @xmath28 @xmath50 curve @xmath48 corresponds global section @xmath51 map @xmath52 surjective non - zero section @xmath53 @xmath4-gon determinant @xmath54 equation of @xmath48 proves theorem linear combination @xmath33 @xmath55 vanish along vertices complete @xmath4-gon curves poncelet curves @xmath56 lurth quartics consider @xmath4-gons inscribed in smooth conic circumscribed other one page 362 @xmath4-gon union of @xmath4 lines @xmath58 @xmath3 vertices @xmath59 for @xmath60 @xmath61 non - zero section @xmath29 vanishing along vertices linesdenote @xmath28 vertices tensorize sequence @xmath62 by @xmath63 @xmath64 @xmath4-points induces decomposition @xmath7 along @xmath1 : @xmath65 consider sequence @xmath66 @xmath67 rank two vector bundle over @xmath14 cohomology sequence verify @xmath68 consider pencil sections @xmath67 pencil sections @xmath7 obtain commutative diagram : @xmath69 sheaf @xmath70 supported curve @xmath71 degree @xmath47 determinant pencil sections @xmath7 curve @xmath71 poncelet curve general point on @xmath71 vertex of @xmath4-gon @xmath71 circumscribed around @xmath0 intersection point @xmath4 lines @xmath4-gon @xmath71 vertex @xmath4-gon @xmath72 @xmath73 curve supporting sheaf @xmath74 @xmath75 @xmath1 component poncelet curve general point on @xmath1 vertex of complete @xmath4-gon @xmath71 configuration meets conic @xmath1 in @xmath4 lines vertices decompostion bundle @xmath1 fixed @xmath4-points each counting doublykers c oort f raven poncelet s closure theorem history modern formulation comparison proof poncelet jacobi mathematical remarks early proofs expo math 5 ( 1987 ) 289364 | 0.309849 | 1,871 | 1,059 | , Saving $0.0 in GPT-4. |
67,345 | the dynamics of folding of proteins is studied by means of a phenomenological master equation .
the energy distribution is taken as a truncated exponential for the misfolded states plus a native state sitting below the continuum .
the influence of the gap on the folding dynamics is studied , for various models of the transition probabilities between the different states of the protein .
we show that for certain models , the relaxation to the native state is accelerated by increasing the gap , whereas for others it is slowed down . | proteins are known to fold into a unique native state that is biologically active @xcite . despite their complexity and frustrated character ,
their folding rate is fast and the folding times vary from milliseconds to seconds . however , these times are long compared to the usual collapse time of a homopolymer ( a few microseconds ) @xcite .
this means that the diversity of the aminoacids and the resulting frustration play a crucial role in the slowing down .
the main picture that has emerged to describe this type of dynamics is that of a `` funnel - like '' phase space @xcite : the folding pathway of a protein in phase space is along a unique but rough funnel towards its ground state .
however , so far , this picture is not based on microscopic models .
numerical simulations that have been dealing with lattice models of short disordered polymers tend to support this funnel picture .
a number of these simulations seem to show that `` good folders '' ( sequences that fold rapidly to their ground state , and are thus good candidates to represent real proteins ) indeed follow a `` funnel '' in the energy landscape , whereas `` bad folders '' possess a large number of energetic traps along the dynamical path that slow down the folding process @xcite .
several studies have tried to relate the foldability of model proteins with their energy gap .
it has been argued in ref .
@xcite that model proteins which have a large energy gap between the native state and the first conformationally different ( compact ) state fold rapidly .
on the other hand , other studies @xcite have shown that the parameter which governs the rapidity of folding is the distance of the folding temperature to the collapse temperature .
the aim of this paper is to show that the situation is not so simple , and that in particular , the folding rate depends very much on the dynamics used for the simulations .
in particular , we show that for certain transition rates , a large gap accelerates the folding , whereas for other models , it may slow it down .
many models have stressed the analogy between protein folding and the thermodynamics of heteropolymers @xcite . in a mean field approach ,
some models of quenched disordered polymers are similar to a random energy model ( rem ) @xcite ; the low lying states of the protein are identified with the low - lying states of the rem , which are responsible for the slow dynamics . although this analogy is questionable as far as real proteins are concerned @xcite , we will adopt this framework for the dynamical models studied in the following . for the rem as well as for the sherrington kirkpatrick ( sk ) model @xcite
, it has been shown that i ) all low lying states have the same extensive energy and ii ) the distribution of the corrections to extensivity of their free energies is exponential ( bound at high energy but not at low energy ) . for proteins ,
the energies are bound below by that of the native state and above by that of a swollen coil . in addition , the energy landscape is known to be very rugged and the number of misfolded states ( at fixed energy @xmath0 ) is known to grow very rapidly with @xmath0 .
it is thus natural , by analogy with many disordered systems , to assume an exponential distribution for the energies of the lowest lying misfolded states and to isolate the native state below the continuum of energies . in our model
, we describe the phase space of the protein as consisting of @xmath1 misfolded states @xmath2 with @xmath3 and one native state @xmath4 .
the distribution of energies of the misfolded states is continuous , given by : @xmath5 where @xmath6 is a parameter ( related to the glass transition temperature of the rem ) and @xmath7 is the energy of the highest state . more precisely , the energy levels are such that @xmath8 , where @xmath9 is the energy of the first misfolded state . therefore , boltzmann weights @xmath10 vary between @xmath11 and @xmath12 . to account for the native state
, we include an additional state with energy @xmath13 and boltzmann weight @xmath14 . the energy gap @xmath15 is defined by @xmath16 .
the main consequence of the truncation of the energy distribution is that the system will never spend a very large time in one of the energy traps , and that @xmath17 will always relax towards its equilibrium value @xmath18 with a finite relaxation time at large times .
slow dynamics features such as aging wo nt appear either ; this is in contrast with the case when the distribution is not truncated , which has been studied extensively recently @xcite .
the calculations are made by taking first the limit @xmath19 before taking the limit of large times .
this is justified if the number of metastable conformations is large enough .
the dynamics is based on the master equation : @xmath20 where @xmath17 is the probability of occupation of the energy level @xmath21 at time @xmath22 , and @xmath23 is the transition rate from the energy level @xmath24 to the energy level @xmath25 . in all the cases we have studied ,
detailed balance is satisfied : @xmath26 when solving the master equation , the quantities that are usually calculated are averages over the distribution of disorder @xmath27 .
this is justified in the case of a macroscopic system with short range interactions , but not in the case of a protein , which is too small an object for self - averageness to hold .
this is why , in contrast with other studies , we have calculated quantities which are not averaged over the distribution of disorder .
we considered three cases , depending on the choice of transition rates between energy levels .
we find that if the transitions rates depend only on the final state , the relaxation is accelerated by a large gap whereas if they only depend on the initial state , the dynamics is slowed down ; in the intermediate case , the situation is more complex .
in this section , we present the results of the calculations , which will be detailed elsewhere . studies as the ones presented below have already been made for a non truncated distribution of energies : disorder - averaged quantities show stretched exponential or power law behavior at large times @xcite .
the same models have been used in ref .
@xcite in the context of heteropolymer folding .
if @xmath13 and a gap are included in the analysis , the master equation can easily be solved , leading to an exponential behavior , @xmath29 , where @xmath30 . since @xmath31 ( the bar denotes the average over the distribution of energies ) , where @xmath32 , the relaxation time is @xmath33 .
let us compare the dynamics of two protein sequences that differ only by their native energies @xmath13 keeping @xmath9 ( or @xmath34 ) constant . as seen above , the relaxation time @xmath35 decreases when the gap @xmath36 increases .
indeed , in such a dynamical scheme , jumps towards the native state are enhanced and the lower its energy , the faster it is populated .
the dynamics is illustrated by fig.[1 ] where the probability of occupation of the native state @xmath37 is plotted for different values of the gap .
on fig.[2 ] , we show the relaxation time as a function of the gap .
0.5 cm as was noted in ref .
@xcite , this corresponds to the case of a unique barrier at energy @xmath39 , through which the system must pass in order to make a transition from state @xmath40 to @xmath41 .
calculations can be made in both cases where the initial conditions are uniform ( all states are equally populated ) or delta - like ( the initial probability of occupation of a specific state is one ) .
they lead to the same conclusions concerning the relaxation times , and we give here the results for the case of uniform initial conditions . at short times , the probability of occupation of states follows a power - law : @xmath42 at longer times , the relaxation is exponential . for a non - native state @xmath43 , @xmath44 and for the native state ,
@xmath45 these expressions involve two time constants : @xmath46 , which is the relaxation time of the energy level @xmath47 in the absence of a gap , and @xmath48 .
this last relaxation time is the one that governs the long time dynamics of the native state .
contrarily to the previous case , keeping @xmath9 ( or @xmath34 ) fixed , @xmath49 now increases as @xmath36 increases . in this scheme
, the energy landscape can be viewed as a collection of energy traps ; as the energy @xmath4 decreases , it becomes more difficult to escape from this state ; low energy states are populated quite rapidly , and the system remains trapped in the native state ; as a result , the relaxation to the boltzmann distribution is slowed down .
this effect is more pronounced as the energy gap gets bigger ( see fig.[3 ] and [ 4 ] ) .
0.5 cm in this case , the transition rates depend both on the initial and final state through the above formula , where @xmath51 is a parameter between 0 and 1 .
relaxation to the native state is described by a new relaxation time @xmath49 : @xmath52 where @xmath53 .
the dependence of @xmath54 on the gap is now slightly more complicated .
one can distinguish two regimes : 1 .
if @xmath55 and @xmath56 , @xmath54 decreases as @xmath57 increases ( see fig .
[ 5 ] and [ 6 ] ) , and the relaxation is accelerated . on the other end , if @xmath58 , @xmath54 increases with @xmath15 for @xmath59 and decreases with @xmath36 for @xmath60 , where @xmath61 .
if @xmath62 , @xmath54 is an increasing function of @xmath36 for @xmath59 and a decreasing function of @xmath57 for @xmath60 , where @xmath63 .
it seems difficult to compare our results with experiments in real proteins , since no systematic study of the gap of given sequences has been carried out experimentally .
however , there has been a number of lattice simulations that lead to a variety of results and interpretations . klimov and thirumalai @xcite
claim that there seems to be no direct correlation between the gap and the folding dynamics . on the other hand , sali , shakhnovich and karplus @xcite have shown that the `` best '' folders are those with the largest energy gap . in all cases ,
the simulations are performed on very short chains ( 27 monomers at best ) and the interactions between monomers are random with a gaussian distribution .
the energy of a configuration is given by @xmath64 where @xmath65 if monomers @xmath66 and @xmath67 are neighbors on the lattice and @xmath68 with a negative parameter @xmath14 in order to mimic the hydrophobic character of the solvent .
these previous studies are applicable only to very short proteins
. for longer chains , there are no results indicating how a protein will find its folding path through its complicated energy landscape .
some attempts have been made to understand analytically the dynamics of random heteropolymers @xcite @xcite ; and at present , the debate regarding the interpretation of the simulations @xcite
@xcite @xcite is still open . our phenomenological approach based on a rem energy landscape ,
although far from realistic , might capture the long time relaxation laws of random chains .
the finiteness of objects such as proteins is taken into account by truncating the energy spectrum .
moreover , we show that the dynamics depends on the whole energy landscape , rather that only on one parameter , such as the gap .
according to the type of transition rates used in the dynamics , one can obtain different behaviors for the relaxation as a function of the gap .
it seems difficult to decide which one is more realistic .
it would be interesting to have systematic results on the dynamics of synthesized sequences , generated for example through mutation experiments @xcite . | proteins fold into unique native state biologically active despite complexity frustrated character folding rate fast times vary milliseconds to seconds times long compared to usual collapse time of homopolymer ( few microseconds diversity of aminoacids frustration crucial in slowing down main picture dynamics is `` funnel - like '' phase space folding pathway of protein in along unique rough funnel towards ground state picture not based on microscopic models numerical simulations short disordered polymers support funnel picture simulations show good folders '' fold rapidly to ground state follow `` funnel '' in energy landscape bad folders '' possess large energetic traps slow down folding process studies tried to relate foldability of model proteins with energy gap argued model proteins large energy gap between native state first conformationally different state fold rapidly other studies shown parameter governs rapidity folding is distance of folding temperature to collapse temperature aim of paper to show situation not so simple folding rate depends on dynamics used for simulations for certain transition rates large gap accelerates folding for other may slow downmodels stressed analogy between protein folding thermodynamics heteropolymers @xcite models of quenched disordered polymers similar to random energy model rem ) @xcite low lying states protein identified with rem responsible for slow dynamics analogy questionable real proteins adopt framework for dynamical models studied for rem sherrington kirkpatrick model @xcite shown all low lying states have same extensive energy distribution of corrections to extensivity free energies exponential bound at high not low energy proteins energies bound below native state above by swollen coil energy landscape rugged misfolded states rapidly with @xmath0 natural to assume exponential distribution for energies lowest lying misfolded states isolate native state below continuum energies model phase space of protein @xmath1 misfolded states one native state @xmath4 distribution of energies misfolded states continuous given by @xmath5 @xmath6 parameter related to glass transition temperature rem @xmath7 energy of highest state energy levels @xmath8 @xmath9 is energy first misfolded state boltzmann weights @xmath10 vary between @xmath11 and @xmath12account for native state include additional state with energy @xmath13 boltzmann weight @xmath14 energy gap @xmath15 defined by @xmath16 consequence truncation energy distribution system spend large time in energy traps @xmath17 towards equilibrium value @xmath18 finite relaxation time at large times slow dynamics features aging appear contrast with distribution not truncated studied extensively recently calculations taking limit @xmath19 before limit large times justified if metastable conformations large enough dynamics based on master equation : @xmath20 @xmath17 probability of occupation energy level @xmath21 @xmath22 @xmath23 transition rate from @xmath24 to @xmath25 detailed balance satisfied @xmath26 quantities calculated averages over distribution of disorder @xmath27 justified macroscopic system with short range interactions not protein too small object for self - averageness calculated quantities not averaged over distribution disorder considered three cases transition rates between energy levels if transitions rates depend on final state relaxation accelerated by large gap depend on initial state dynamics slowed intermediate case situation more complex present results of calculations detailed elsewherestudies made for non truncated distribution of energies disorder - averaged quantities show exponential power law behavior at large times @xcite same models used in heteropolymer folding if @xmath13 gap included master equation solved to exponential behavior @xmath29 @xmath30 @xmath31 average relaxation time is @xmath33 compare dynamics of two protein sequences by native energies @xmath13 @xmath9 @xmath34 constant relaxation time @xmath35 decreases when gap @xmath36 increases jumps towards native state enhanced lower energy faster populated dynamics illustrated by fig.[1 ] probability of occupation of native state @xmath37 for different values gap fig.[2 ] relaxation time function of gap 0.5 cm corresponds unique barrier at energy @xmath39 transition from state @xmath40 to @xmath41 calculations in initial conditions uniform or delta - like probability of occupation lead to same conclusions concerning relaxation times results for uniform initial conditions at short times probability of occupation of states follows power - law @xmath42 at longer times relaxation exponentialnon - native state @xmath43 @xmath44 native state @xmath45 expressions involve two time constants @xmath46 relaxation time energy level @xmath47 gap @xmath48 last relaxation time governs long time dynamics native state @xmath9 fixed @xmath49 increases as @xmath36 increases energy landscape energy traps energy @xmath4 decreases difficult to escape low energy states populated rapidly system remains trapped in native state relaxation to boltzmann distribution slowed effect pronounced energy gap bigger fig. 4 transition rates depend on initial final state @xmath51 parameter between 0 and 1 relaxation to native state new relaxation time @xmath49 : @xmath52 @xmath53 dependence of @xmath54 on gap complicated two regimes if @xmath55 @xmath56 @xmath54 decreases as @xmath57 increases relaxation accelerated @xmath58 @xmath54 increases with @xmath15 @xmath59 decreases with @xmath36 @xmath60 @xmath62 @xmath54 increasing function of @xmath36 @xmath59 decreasing function @xmath57 for @xmath60difficult to compare results with real proteins no systematic study of gap of sequences carried lattice simulations lead to variety results interpretations klimov thirumalai claim no direct correlation between gap folding dynamics sali shakhnovich karplus shown best folders are with largest energy gap simulations performed on short chains ( 27 monomers interactions between random with gaussian distribution energy of configuration given by @xmath64 @xmath65 @xmath66 @xmath67 neighbors @xmath68 with negative parameter @xmath14 to mimic hydrophobic character solvent previous studies applicable only to short proteins for longer chains no results indicating protein folding path through complicated energy landscape attempts to understand dynamics of random heteropolymers debate interpretation simulations open phenomenological approach based on rem energy landscape might capture long time relaxation laws of random chains finiteness of proteins by truncating energy spectrum dynamics depends on whole energy landscape parameter gap according to transition rates different behaviors for relaxation gap difficult to decide which one realistic interesting to have systematic results on dynamics of synthesized sequences through mutation experiments | 0.363666 | 2,604 | 1,292 | , Saving $0.1 in GPT-4. |
67,357 | we discuss the production of @xmath0 bosons in deep inelastic processes @xmath1 ( @xmath2 : @xmath3 proton , @xmath4 neutron ) , in the context of an electroweak model , in which the vector boson self interactions may be different from those prescribed by the electroweak standard model .
we present results which show the strong dependence of the cross section on the anomalous magnetic dipole moment @xmath5 of the @xmath6 .
we show that even small deviations from the standard model value of @xmath5 @xmath7 could imply large deviations in the cross section rates of @xmath0 production through the collision of an ultrahigh energy ( anti)neutrino on a target nucleon
. however , the enhancement of the cross section rates is not large enough to be detectable . | one of the most important features of the standard model of the electroweak interactions @xcite is the non - abelian gauge nature of the vector bosons . in the present paper ,
we discuss this point , namely the structure of the self couplings of the electroweak vector bosons . in order to do it ,
we analyze @xmath0 production in deep inelastic ( anti)neutrino - nucleon scattering by using the parton model @xcite , in the context of an electroweak model with non - standard vector boson self interactions .
such a model was proposed by m. kuroda _ et al . _
( kmss model ) @xcite . in this model the trilinear vector boson coupling constants depend on only one free parameter , @xmath5 , the anomalous magnetic dipole moment of the @xmath0 @xcite . the diagrams which contribute to the @xmath0 boson production cross section , at the quark level in the lowest order in @xmath8 , contain only three vector boson self interactions ,
hence the boson production rates depend only on @xmath5 . at the present time the best limits on @xmath5 are @xmath9 from a recent analysis of the l3 collaboration at lep @xcite . in a previous work @xcite ,
we have discussed the heavy boson production via the deep inelastic @xmath10 in the context of the standard model .
we pointed out there that the process @xmath11
( @xmath2 : @xmath3 proton , @xmath4 neutron ) is the only one which gets contribution from photon - exchange diagrams and that its total cross section reaches significant values for ultrahigh - energy ( uhe ) neutrinos@xcite colliding on a target nucleon .
therefore , we restrict ourselves our study in this paper to the mentioned process .- scattering can be obtained from those of @xmath10-scattering by the replacement @xmath12 , @xmath13 and @xmath14-type quarks @xmath15 @xmath16-type quarks ] in this work , we calculate in the context of the standard model separately the contribution to the process @xmath11 from the two different mechanisms which contribute at the lowest order in @xmath8 , keeping only photon exchange diagrams : production at the leptonic vertex and through the boson self interaction .
we show explicitly the compensation via destructive interference inherent to the standard model as a non - abelian gauge theory .
we find that this compensation reaches up to four orders of magnitude , for ultrahigh - energy neutrinos with energy @xmath17 colliding on a target nucleon .
hence , one could expect that even small deviations from the coupling structure of the standard model , like an anomalous dipole magnetic moment different from 1 , would lead to observable effects in the predictions for charged boson production in @xmath10 collisions .
this paper is organized as follows . in section
ii , we take the formulae given in [ 6 ] for heavy boson production in @xmath10-scattering and extended them for the case of @xmath18 boson production via the process @xmath11 , in the frame of an electroweak model , in which the trilinear vector boson coupling constants may deviate through an anomalous magnetic dipole moment term from those given by the standard model . in section iii , we describe briefly the non - standard electroweak model , which we will take for our study . in section iv
, we present and discuss our results for the total cross section of deep inelastic process @xmath11 , which is the only one that gets contribution from @xmath19-exchange diagrams .
finally , in section v , we summarize our conclusions .
we have discussed in [ 6 ] the heavy boson production via the deep inelastic @xmath10 collisions in the context of the standard model .
we showed there that the process @xmath21 is the only one which gets contribution from photon - exchange diagrams and that its total cross section reaches significant values for uhe neutrinos colliding on a target nucleon .
therefore , we restrict ourselves our study in this paper to the mentioned reaction . in fig .
1 are depicted the diagrams which contribute at the lowest order in @xmath8 at the quark parton level to the cross section of process ( 1 ) .
we have presented in detail the kinematics of heavy boson production in deep inelastic @xmath22 scattering in [ 6 ] and we pointed out there how to take care for the different ways in which the bosons are arranged in the non - abelian couplings diagrams .
now we will extend the formulae given there for the process ( 1 ) in the case in which the @xmath18 boson may have an anomalous magnetic moment @xmath23 different from that predicted by the standard model .
we will assume for the trilinear coupling constants of the vector bosons the following general form : @xmath24 the standard model self interactions are obtained by taking @xmath25 , @xmath26 , @xmath27 and @xmath28 ( @xmath29 being the electroweak mixing angle ) . in order to calculate @xmath18 boson production through process ( 1 ) with the non - standard non - abelian couplings given in ( 2 )
, we may take the formulas given in section iii . of [ 6 ] ,
provided the following changes and extensions are performed \1 ) the index @xmath30 runs now from 1 to 4 .
\2 ) @xmath31 , @xmath32 , @xmath33 for @xmath34 , and @xmath35 , @xmath36 , @xmath37 , @xmath38 remain as they are defined in [ 6 ] .
\3 ) @xmath39 \4 ) @xmath40 .
\5 ) @xmath41 where @xmath42,\end{aligned}\ ] ] with @xmath43 . for other polarizations
@xmath44 has to be replaced by @xmath45 in an appropriate way .
the values for @xmath46 and @xmath47 are given in table 1 .
we present in the appendix the explicit expressions for the quantities @xmath48 and we perform also there the summation over the polarizations of the produced boson . [ cols="^,^,^,^",options="header " , ] _ table 5 .
@xmath49 as a function of @xmath5 and @xmath50 ( @xmath51 ) .
we have analyzed the effects of a non - standard anomalous magnetic dipole moment @xmath5 of the @xmath6 in the charged boson production in unpolarized deep inelastic @xmath10(@xmath52)-scattering for uhe neutrinos colliding on a target nucleon , using the electroweak model proposed by m. kuroda _ et al . _
( kmss model ) , in which @xmath5 may be different from 1 , the value predicted by the electroweak standard model .
we found , that even small deviations of @xmath5 from its standard model value ( at the present time the best limits on @xmath5 are @xmath9 from a recent analysis of the l3 collaboration at lep ) could lead to large enhancement in the predicted cross section rates for @xmath53 production via @xmath10(@xmath52 ) collisions .
however , such enhanced rates remain too small to be detectable experimentally at the uhe neutrino telescopes .
we give in this appendix , taking all fermion masseless , the expressions for the quantities @xmath48 , which are defined as follows : @xmath54 with @xmath55 and @xmath56 , @xmath57 being the polarization of the initial quark and incoming neutrino , respectively . in general , these quantities are functions of scalar products of the momenta @xmath58 , @xmath59 , @xmath60 , @xmath61 , @xmath23 and the polarization vector @xmath62 of the produced boson . using the polarization sum for the massive vector boson @xmath63 and the definitions given in ( a1 ) for the @xmath64 we get @xmath65 with help of these expressions we obtain ( @xmath66 ) @xmath67 we can express @xmath68 as follows @xmath69 & \vspace{3 mm } \\ & 8[\varepsilon p\cdot p'q ' \cdot qk + \varepsilon p'\cdot pq \cdot q'k & \hspace{0.5 in } lr , rl\\ & \hspace{0.5 cm } + \varepsilon q\cdot p'q'\cdot pk+ \varepsilon q'\cdot pq\cdot p'k & \\ & \hspace{0.5 cm } -qq'(\varepsilon p\cdot p'k + \varepsilon p'\cdot pk ) & \\ & \hspace{0.5 cm } -pp'(\varepsilon q\cdot q'k + \varepsilon q'\cdot qk ) ] & \end{array } \right.\end{aligned}\ ] ] using ( a2 ) and the definitions given above in ( a1 ) for the @xmath64 we get @xmath70 for @xmath71 . hence @xmath72 the expressions for the remaining @xmath48 can be found in the appendix of [ 6 ] .
* fig . 1 * : feynman diagrams which contribute to process ( 1 ) : ( a ) boson production from the incoming neutrino , ( b ) the outgoing lepton , ( c ) the initial ( d ) and final quark , and ( e ) through the non - abelian couplings ( @xmath14 stands for @xmath14 , @xmath73 , @xmath74 , @xmath75 , @xmath76 ; @xmath16 for @xmath16 , @xmath77 , @xmath78 , @xmath79 , @xmath80 ) .
v. s. berezinsky , nucl . phys . *
b28a * , 352 ( 1992 ) ; v. s. berezinsky and g. zatsepin , proceedings of the 1976 dumand summer workshop ( a. roberts and r. donaldson , eds . ) , p. 215
( 1977 ) ; t. stanev , nucl . phys . * b14a * , 17 ( 1990 ) ; k. greisen , phys . rev . lett . * 16 * , 748 ( 1966 ) ; c. t. hill and d. n. schramm , phys . lett . *
b131 * , 247 ( 1983 ) ; and phys .
* d31 * , 564 ( 1985 ) ; r. gandhi _ et al . _ , phys . rev . *
d58 * , 093009 ( 1998 ) and references therein . , e. andres _ et al .
_ , nature * 410 * , 441 ( 2001 ) ; _ antares collaboration _ , y. becherini _ et al .
_ , e - print archive : hep - ph/0211173 ; _ auger collaboration _ , d. zavrtanik _ et al .
. suppl . * 85 * , 324 ( 2002 ) ; _ nestor collaboration _ , p. k. f. grieder _
et al . _ , nuovo cim . *
24c * , 771 ( 2001 ) ; _ rice collaboration _ ,
i. kravchenko _
et al . _ ,
* 19 * , 15 ( 2003 ) . | important standard model electroweak interactions @xcite non - abelian gauge vector bosons present paper discuss structure self couplings vector bosons analyze @xmath0 production deep inelastic anti)neutrino - nucleon scattering parton model @xcite electroweak model non - standard vector boson self interactions model proposed by kuroda et kmss model @xcite trilinear vector boson coupling constants depend on one parameter @xmath5 anomalous magnetic dipole moment diagrams boson production contain three vector boson self interactions boson production rates depend on @xmath5 best limits on @xmath5 are @xmath9 from recent analysis l3 collaboration lep @xcite previous work discussed heavy boson production deep inelastic @xmath10 standard model process @xmath11 ( proton neutron only contribution from photon - exchange diagrams total cross section reaches significant values for ultrahigh - energy neutrinos@xcite colliding on target nucleon restrict study to processscattering from @xmath10-scattering @xmath12 quarks quarks calculate standard model contribution to process @xmath11 from two mechanisms lowest order in @xmath8 photon exchange diagrams production at leptonic vertex boson self interaction compensation via destructive interference standard model non - abelian gauge theory compensation reaches four orders of magnitude for ultrahigh - energy neutrinos with @xmath17 colliding on target nucleon small deviations from coupling structure anomalous dipole magnetic moment lead to effects in predictions charged boson production in @xmath10 collisions paper organized section ii formulae for heavy boson production in @xmath10-scattering for @xmath18 boson production via @xmath11 electroweak model vector boson coupling constants deviate anomalous magnetic dipole moment section iii describe non - standard electroweak model study section iv discuss results for total cross section of deep inelastic process @xmath11 contribution from @xmath19-exchange diagrams section v summarize conclusions discussed heavy boson production via deep inelastic @xmath10 collisions standard modelshowed process @xmath21 contribution from photon - exchange diagrams total cross section reaches significant values for uhe neutrinos colliding on target nucleon restrict study to reaction fig . 1 diagrams lowest order @xmath8 quark parton level to cross section process ( 1 ) presented kinematics heavy boson production in deep inelastic @xmath22 scattering in [ 6 ] bosons in non - abelian couplings diagrams extend formulae for process ( 1 ) @xmath18 boson anomalous magnetic moment different standard model assume trilinear coupling constants vector bosons @xmath24 standard model self interactions obtained by @xmath25 @xmath28 @xmath29 electroweak mixing angle calculate @xmath18 boson production through process ( 1 ) with non - standard non - abelian couplings formulas in section iii . of [ 6 ] changes extensions performed index @xmath30 runs from 1 to 4 @xmath31 remain defined in [ 6 ] @xmath39 @xmath40 @xmath41 @xmath42 with @xmath43@xmath44 replaced by @xmath45 values for @xmath46 @xmath47 in table 1 appendix expressions for quantities @xmath48 summation over polarizations produced boson table 5 @xmath49 function of @xmath5 @xmath50 analyzed effects non - standard anomalous magnetic dipole moment @xmath5 @xmath6 in charged boson production in unpolarized deep inelastic @xmath10(@xmath52)-scattering for uhe neutrinos colliding on target nucleon using electroweak model kuroda @xmath5 different from 1 standard model small deviations of @xmath5 from standard model value best limits @xmath5 @xmath9 lead enhancement predicted cross section rates for @xmath53 production via @xmath10(@xmath52 ) collisions enhanced rates small detectable at uhe neutrino telescopes appendix expressions for quantities @xmath48 defined @xmath54 @xmath55 @xmath56 @xmath57 polarization of initial quark incoming neutrinoquantities functions scalar products @xmath58 polarization vector @xmath62 produced boson using polarization sum vector boson @xmath63 definitions ( a1 ) @xmath64 get @xmath65 obtain @xmath66 ) @xmath67 @xmath68 @xmath69 \vspace{3 mm\varepsilon p\cdot p'q qk +'\cdot pq q'k\hspace{0.5 in }{0.5 cm\varepsilon q\cdot p'q' pk+ pq p'k\hspace{0.5 cm\varepsilon p\cdot p'k + p'\cdot pk{0.5 cm\varepsilon q\cdot q'k + q'\cdot qk using ( a2 ) definitions a1 ) @xmath64 get @xmath70 @xmath71 @xmath72 expressions for remaining @xmath48 in appendix [ 6 ] fig feynman diagrams contribute process boson production incoming neutrino outgoing lepton initial final quark non - abelian couplings @xmath14 berezinsky . b28a 352 ( 1992 ) berezinsky zatsepin proceedings 1976 summer workshop a. roberts r. donaldson eds p. 215 ( 1977 ) stanev nucl . . b14a 17 ( 1990 ) k greisen phys . rev . 16 748 ( 1966 ) t. hill d n schramm phys . lett b131 247 ( 1983 ) phys . d31 564 ( 1985 ) r. gandhi et al . phys . rev . d58 093009 ( 1998 ) e. andres et al . nature 410 441 ( 2001 ) antares collaboration y. becherini et al . e - print archive hep - ph/0211173 auger collaboration d. zavrtanik et al suppl 85 324 ( 2002 ) nestor collaboration f grieder et al 24c 771 2001 ) rice collaboration kravchenko al 19 15 ( 2003 ) | 0.425408 | 2,558 | 1,327 | , Saving $0.1 in GPT-4. |
67,362 | the non - keplerian galactic rotational curves and the gravitational lensing data @xcite strongly indicate a significant dark matter component in the universe .
moreover , these data can be combined to deduce the equation of state of dark matter @xcite . yet
, the existence of dark matter has been challenged following the tradition of critical scientific spirit . in the process , the theory of general relativity itself has been questioned and various modified theories of gravitation have been proposed @xcite . within the framework of the einsteinian general relativity , here
i make the observation that if the universe is described by a spatially flat friedmann - robertson - walker ( frw ) cosmology with einsteinian cosmological constant then the resulting cosmology predicts a significant dark matter component in the universe .
the phenomenologically motivated existence proof refrains from invoking the data on galactic rotational curves and gravitational lensing , but uses as input the age of the universe as deciphered from studies on globular clusters .
the abstracted result arises from the crucial observation that in general , without invoking physics of structure formation ( i.e. primordial birth of galaxies and stars ) , it is impossible to constrain the ratio of the fractional matter and dark energy densities .
however , an exception occurs , for the spatially flat frw cosmology with einsteinian cosmological constant . in the ` matter - dominated ' epoch with cosmological constant
, this exception completely determines the ratio of the matter and dark energy densities as a function of the age of the universe . once the latter is brought in as an input , the said ratio is unambiguously fixed for the present epoch .
it turns out that the resulting fractional density for matter @xmath0 is too large by a factor of 4 to 7 than the known @xmath1 associated with the standard model of particle physics .
this circumstance then unambiguously predicts a significant non - standard model component to @xmath0 .
it is by rigorously establishing this outline in detail that the said claim is arrived at .
no uniqueness of the mentioned exception is implied ; that is , there may exist other cosmologies where a similar circumstance occurs .
i display the newtonian constant @xmath2 explicitly but set the speed of massless particles in vacuum @xmath3 to be unity .
the setting of our argument is as follows .
i consider the matter - dominated epoch with einsteinian cosmological constant . by the latter one
means that the dark energy equation of state @xmath4 is confined to the choice @xmath5 .
then , a time independent @xmath6 corresponds to the einsteinian cosmological constant .
further , i consider the cosmology which is spatially flat . that is , spatial curvature constant @xmath7 . in the matter
dominated epoch , i set the associated pressure to zero , @xmath8 .
this is quite realistic .
thus , the frw cosmology we consider is defined by the set : @xmath9 once the energy densities , @xmath10 and @xmath6 , and the pressures , @xmath11 and @xmath12 , associated with matter and cosmological constant @xmath13 are specified the scale factor @xmath14 , parameterizing the cosmological expansion , is obtained by solving the einstein field equations for a spatially flat frw cosmology : @xmath15 \frac{\ddot{a}}{a } = -\frac{4\pi g}{3 } \rho_\mathrm{m } - \frac{4\pi g}{3}\left(\rho^\lambda + 3 p^\lambda\right ) \ , .
\label{eq:2}\end{aligned}\ ] ] in writing the above equation i set @xmath16 for the matter dominated epoch , and defined @xmath17 equivalently , one can solve eq .
( [ eq:1 ] ) , and the equation for the conservation of energy - momentum ( which follows from eqs .
( [ eq:1 ] ) and ( [ eq:2 ] ) ) @xmath18 in obtaining the above equation @xmath19 was set to zero as required by the setting , and where in addition the last term on the left hand side now vanishes due to the fact that @xmath20 ( @xmath21 ) , giving @xmath22 the system of eqs .
( [ eq:1 ] ) and ( [ eq:4 ] ) now forms a closed system .
it governs the time evolution of @xmath14 and @xmath23 . for the spatially flat frw cosmology defined by the set ( [ eq : set ] ) the division of eq .
( [ eq:1 ] ) by the square of the hubble parameter @xmath24 yields @xmath25 where @xmath26 our task now is to obtain an expression for the time evolution of the ratio , @xmath27 . towards this end eq .
( [ eq:4 ] ) gives the temporal dependence of the matter density on the cosmic scale factor @xmath28 here , @xmath29 marks the beginning of the matter dominated epoch .
it is to be given as a physical input ( @xmath30 , temperature @xmath31 ) .
the time dependence of the scale factor @xmath14 is now obtained by substituting for @xmath32 from eq .
( [ eq : rhom ] ) into eq .
( [ eq:1 ] ) , and solving the resulting differential equation for @xmath14 .
this exercise yields the time evolution of the scale factor @xmath33 \ , .
\label{eq : a}\ ] ] this result agrees with that arrived at , e.g. , by frieman in ref .
( * ? ? ?
* eq . 3.6 ) . is proportional to @xmath34 $ ] .
however , as we next note , for the purposes of establishing the proposed thesis it is precisely the ` proportionality constant ' that turns out to carry a pivotal importance . ] from a physical perspective , note that the nonlinearity of the gravitational field equations has the consequence that the `` amplitude '' of the scale factor carries a unique dependence on the indicated densities . furthermore , eq . ( [ eq : a ] ) places a severe constraint on the relative initial densities by requiring the square bracket in eq .
( [ eq : a ] ) to take the value unity at @xmath35 : [ fig : zeta ] .
the label @xmath36 marks the present epoch with @xmath37 years in units of @xmath38 ( @xmath39 years , see text).,title="fig : " ] @xmath40^{-2}\label{eq : constraint}\ ] ] where we introduced @xmath41 . substituting back @xmath14 from eq .
( [ eq : a ] ) in eq .
( [ eq : rhom ] ) provides the general time evolution of the relevant energy densities for @xmath42 @xmath43^{-2}\,.\label{eq:12}\ ] ] the result ( [ eq : constraint ] ) also follows from the above equation for @xmath35 .
it reflects the internal consistency of the calculations . on dividing each of the densities in eq .
( [ eq:12 ] ) by the critical density , @xmath44 , the time evolution of the ratio @xmath27 follows @xmath45 where @xmath46 once the age of the universe is specified by some independent observations , the considered cosmology uniquely determines the ratio @xmath47 and it in addition predicts the fractional matter density to be @xmath48 this is the combined result of eqs .
( [ eq : one ] ) and ( [ eq : zeta ] ) .
since 1998 , there exists observational evidence for a positive cosmological constant @xcite .
the latest data and analysis of the luminosity data on supernovae 1a , in addition , obtains @xcite ( for a @xmath7 cosmology ) [ fig : omega ] .
the label @xmath36 marks the present epoch with @xmath37 years in units of @xmath38 ( @xmath39 years , see text ) .
, title="fig : " ] @xmath49 these observations give us confidence that a spatially flat frw cosmology with einsteinian cosmological constant may indeed correspond to the physical reality .
taking @xmath50 yields @xmath51 years .
it sets the time scale for us .
however , for the moment , one need not take any input from these data , except for working in units of @xmath38 ( for convenience ) .
instead , i use the present age of the universe @xmath37 years @xcite as determined by the studies on globular clusters as the input .
thus the present epoch corresponds roughly to the range : @xmath52 . in this range ,
figures 1 and 2 show that @xmath53 the latter may be compared with @xmath54 inferred by astier et al .
@xcite .
however , @xmath1 for the present epoch can only account for a small fraction , i.e. , one fourth to one seventh of the predicted @xmath55 as given in eq .
( [ eq : prediction ] ) .
this state of affairs leaves the fractional density @xmath56 to point towards the existence of some form of non - standard model matter in the non - relativistic form .
this , by definition , is the astrophysical / cosmic dark matter , with fractional density given by [ fig : thubble ] corresponds to @xmath57 .
the @xmath58 is defined as inverse of the hubble parameter for spatially flat frw cosmology with einsteinian cosmological constant.,title="fig : " ] @xmath59 for the present epoch , it yields roughly : @xmath60 . in the defined setting , the @xmath61 as well as the @xmath62 carry the same temporal evolution .
it is immediately read off from eqs .
( [ eq : rhom ] ) and ( [ eq : a ] ) . in particular ,
the ratio @xmath63 is frozen in time . before concluding it
may be noted that the hubble parameter as implied by the obtained scale factor given in eq .
( [ eq : a ] ) may also be evaluated .
the result is , @xmath64 it defines a characteristic time scale which is closely related to the age of the universe @xmath65 the range @xmath52 corresponds to @xmath66 ( see fig .
3 ) . a more rigorous analysis along the lines of ref .
( * ? ? ?
* sec .
3.2 ) yields a very similar result . despite the on going debate on whether or not the general theory of relativity needs a modification while confronting the astrophysical and cosmological scene , the phenomenological existence proof given in the preceding discussion makes a strong case that such modifications are not required by the data on the galactic rotational curves or gravitational lensing .
the dark matter required by these latter observations is the same dark matter that is here predicted within the general relativistic framework for the observationally favored spatially flat frw cosmology with einsteinian cosmological constant .
the tantalizing possibility exists that the pioneer anomaly may require a modification of the general theory of relativity in a subtle way @xcite .
but that call comes not from the dark matter .
the focus on new physics possibilities that dark matter offers perhaps ought to shift dramatically .
one such direction is provided by the new and entirely unexpected theoretical discovery of a mass dimension one fermionic field of spin one half @xcite .
that fermionic field , by virtue of its mass dimension , explains the darkness of dark matter as a natural consequence of the spacetime symmetries underlying the freely falling frames of general relativity . | i wish to thank daniel grumiller for his reading of the final draft , and for his input .
i also thank daniel sudarsky for a remark on a related work which led to the considerations presented here .
d. m. wittman , j. a. tyson , d. kirkman , i. dellantonio and g. bernstein , _ detection of weak gravitational lensing distortions of distant galaxies by cosmic dark matter at large scales , _ nature * 405 * , 143 ( 2000 ) [ arxiv : astro - ph/0003014 ] .
h. hoekstra , h. k. c. yee and m. d. gladders , _ properties of galaxy dark matter halos from weak lensing , _ astrophys .
j. * 606 * , 67 ( 2004 ) [ arxiv : astro - ph/0306515 ] . t. faber and m. visser , _ combining rotation curves and gravitational lensing : how to measure the equation of state of dark matter in the galactic fluid , _ arxiv : astro - ph/0512213 . v. sahni and a. a. starobinsky , _ the case for a positive cosmological @xmath13-term , _ int .
j. mod .
d * 9 * ( 2000 ) 373 [ arxiv : astro - ph/9904398 ] .
s. perlmutter _ et al .
_ [ supernova cosmology project collaboration ] , _ discovery of a supernova explosion at half the age of the universe and its cosmological implications , _ nature * 391 * ( 1998 ) 51 [ arxiv : astro - ph/9712212 ] .
a. g. riess _ et al .
_ [ supernova search team collaboration ] , _ observational evidence from supernovae for an accelerating universe and a cosmological constant , _ astron .
j. * 116 * , 1009 ( 1998 ) [ arxiv : astro - ph/9805201 ] . s. perlmutter _ et al .
_ [ supernova cosmology project collaboration ] , _ measurements of @xmath67 and @xmath13 from 42 high - redshift supernovae , _ astrophys .
j. * 517 * , 565 ( 1999 ) [ arxiv : astro - ph/9812133 ] .
j. l. tonry _ et al . _
[ supernova search team collaboration ] , _ cosmological results from high - z supernovae , _ astrophys . j. * 594 * ( 2003 ) 1 [ arxiv : astro - ph/0305008 ] .
a. g. riess _ et al .
_ [ supernova search team collaboration ] , _ type ia supernova discoveries at @xmath68 from the hubble space telescope : evidence for past deceleration and constraints on dark energy evolution , _ astrophys .
j. * 607 * , 665 ( 2004 ) [ arxiv : astro - ph/0402512 ] .
p. astier _ et al .
_ , _ the supernova legacy survey : measurement of @xmath69 , @xmath70 and @xmath71 from the first year data set , _ arxiv : astro - ph/0510447 .
d. n. spergel _ et al . _
[ wmap collaboration ] , _ first year wilkinson microwave anisotropy probe ( wmap ) observations : determination of cosmological parameters , _ astrophys .
j. suppl .
* 148 * ( 2003 ) 175 [ arxiv : astro - ph/0302209 ] .
s. g. turyshev , v. t. toth , l. r. kellogg , e. l. lau and k. j. lee , _ the study of the pioneer anomaly : new data and objectives for new investigation , _ int .
j. mod .
d ( in press ) .
arxiv : gr - qc/0512121 .
d. v. ahluwalia - khalilova and d. grumiller , _ dark matter : a spin one half fermion field with mass dimension one ?
, _ phys .
d * 72 * ( 2005 ) 067701 [ arxiv : hep - th/0410192 ] .
d. v. ahluwalia - khalilova and d. grumiller , _ spin half fermions with mass dimension one : theory , phenomenology , and dark matter , _ jcap * 0507 * ( 2005 ) 012 [ arxiv : hep - th/0412080 ] . | daniel grumiller final draft input daniel sudarsky remark work considerations wittman tyson kirkman dellantonio bernstein detection weak gravitational lensing distortions distant galaxies cosmic dark matter large scales nature 405 143 ( 2000 ) astro - ph/0003014 h hoekstra yee gladders properties galaxy dark matter halos weak lensing astrophys . j. 606 67 ( 2004 ) ph/0306515 t. faber m. visser combining rotation curves gravitational lensing measure equation state dark matter galactic fluid astro - ph/0512213 v. sahni starobinsky case positive cosmological @xmath13-term int . j. mod . 9 2000 ) 373 astro ph/9904398 perlmutter supernova cosmology project collaboration discovery supernova explosion half age universe cosmological implications nature 391 1998 ) 51 astro - ph/9712212 riess supernova search team collaboration observational evidence supernovae accelerating universe cosmological constant astron . j. 116 1009 ( 1998 ) ph/9805201 perlmuttersupernova cosmology project collaboration measurements @xmath67 @xmath13 42 high - redshift supernovae astrophys . j 517 565 1999 ) - ph/9812133 tonry supernova search team collaboration cosmological results high - z supernovae astrophys j 594 2003 ) astro - ph/0305008 riess supernova search team collaboration type supernova discoveries @xmath68 hubble space telescope evidence past deceleration constraints dark energy evolution astrophys . j 665 2004 ) astro - ph/0402512 astier supernova legacy survey measurement @xmath69 @xmath70 @xmath71 first year data set astro - ph/0510447 spergel wmap collaboration first year wilkinson microwave anisotropy probe observations determination cosmological parameters astrophys . j. 148 2003 ) 175 astro - ph/0302209 turyshev toth kellogg lau lee study pioneer anomaly new data objectives investigation int j - qc/0512121 ahluwalia - khalilova grumiller dark matter spin half fermion field mass dimension one72 2005 067701 arxiv hep - th/0410192 ahluwalia khalilova grumiller half fermions mass dimension theory dark matter 0507 2005 ) 012 arxiv hep - th/0412080 | 0.182782 | 1,069 | 533 | , Saving $0.0 in GPT-4. |
67,371 | the dispersion relation of longitudinal electrostatic oscillations in a relativistic plasma is studied in the context of the nonextensive statistics formalism proposed by tsallis [ c. tsallis , j. stat .
phys . * 52 * , 479 ( 1988 ) ] , where nonextensivity is characterized by a parameter @xmath0 in tsallis s entropy .
@xmath1 corresponds to the usual boltzmann - gibbs , extensive statistics formalism . in the nonrelativistic regime ,
normalizability of the equilibrium distribution function implies that @xmath2 .
we show that in the relativistic regime much tighter constraints must be satisfied , namely @xmath3 , where @xmath4 is the boltzmann constant , @xmath5 is the temperature of the plasma , and @xmath6 is the particle mass .
then we study longitudinal oscillations in a proton - electron plasma , assuming immobile protons , and electrons whose distribution function maximizes tsallis s entropy .
the dispersion relation of these oscillations is written in integral form for the long wavelength limit .
explicit expressions in terms of generalized hypergeometric functions can be found for all possibles values of @xmath0 in the ultra - relativistic regime . | traditionally , the equilibrium of statistical systems has been studied based on the boltzmann - gibbs entropy , @xmath7 where @xmath4 is the boltzmann constant , and @xmath8 is the probability of the @xmath9-th microstate .
however , an increasing amount of experimental , computational and theoretical evidence , shows that this formalism is not adequate for systems where long range interactions in time or space are important , and , in general , systems where the relevant phase space has fractal or multifractal structure .
tsallis @xcite proposed the following form for the entropy of such systems : @xmath10 where @xmath0 is a real number . for @xmath1
, @xmath11 reduces to @xmath12 .
@xmath11 has the basic property that for independent systems @xmath13 and @xmath14 , the entropy of the composite system @xmath15 is given by : @xmath16 thus , @xmath0 is a measure of the nonextensivity . among the various physical systems where
connections with the tsallis entropy formalism have been found are gravitational systems @xcite , long range hamiltonian systems @xcite , nonlinear inverse bremsstrahlung absorption in plasmas @xcite , turbulence @xcite , and granular systems @xcite , and many applications in astrophysics , fluid dynamics , biology , economy , quantum mechanics , and others .
( see , e.g. @xcite . )
an important property of the nonextensive formalism is that the distribution function which maximizes @xmath11 is non - maxwellian @xcite .
specifically , for @xmath17 , high energy states are more probable than in the extensive case ; for @xmath18 high energy states are less probable than in the extensive case , and there is a cutoff beyond which no states exist . for a unidimensional gas where
@xmath19 , this cutoff is given by @xcite @xmath20 velocity distributions in plasmas are often found to be non - maxwellian .
high energy tails in space and laboratory plasmas @xcite are a common occurrence .
metaequilibrium non - maxwellian radial density profiles in cylindrical turbulent pure electron plasmas are observed @xcite . as stated above ,
such distribution functions emerge naturally if the system is described by a nonextensive statistics , and several authors have successfully applied this approach to plasma systems @xcite .
lima et al .
@xcite studied the longitudinal oscillations of a collisionless thermal plasma , in the context of the nonextensive formalism based on .
they derived the dispersion relation for electrostatic waves , and calculated it explicitly for an equilibrium distribution function which maximizes .
they found that it fits experimental data better than if the energy distribution is given by a maxwellian .
all the previous analysis have been made for non - relativistic plasmas . however , relativistic extensions of the tsallis formalism are possible @xcite .
our purpose is to study the consequences of a relativistic nonextensive statistics in plasmas . in particular
, we will examine some general properties of the equilibrium distribution function and calculate the dispersion relation for electrostatic waves in the long wavelength limit for ultrarelativistic plasmas .
in the nonextensive description , the equilibrium distribution function for a relativistic plasma can be written @xcite @xmath21^{1/(q-1 ) } \ , \ ] ] where @xmath22 is the particle energy and @xmath23 is a normalization constant .
equation generalizes the nonrelativistic expression ( see , e.g. , @xcite ) . for @xmath1 , yields the usual boltzmann distribution function @xmath24 . for simplicity ,
in the following we will consider a one dimensional plasma , so that @xmath25 .
since the relativistic energy is positive , @xmath26 will be real and normalizable if @xmath27 satisfies the following conditions : @xmath28 i.e. , @xmath27 can take any possible value if @xmath29 , but the distribution function has an energy cutoff if @xmath30 . notice that both and hold for the extensive limit @xmath1 , in which case they give the expected constraint @xmath31
. due to the cutoff , if @xmath30 @xmath26 is always normalizable .
however , normalizability of the distribution function for @xmath32 when @xmath33 is not guaranteed .
the integral @xmath34 is finite if @xmath35 when @xmath36 , with @xmath37 . using , @xmath38 { } c p^{\frac{1}{q-1 } } \ , \ ] ] where @xmath39^{\frac{1}{q-1}}>0 $ ] .
thus , @xmath40 is normalizable if @xmath41 .
let us now analyze the case @xmath42 .
the cutoff in and give @xmath43 if @xmath44 then yields the cutoff in momentum space corresponding to the energy cutoff . in the limit case @xmath45 , the only momentum allowed is @xmath46 . if @xmath47 , then no real value of @xmath48 satisfies condition , and no distribution function exists .
we may interpret this saying that for a given temperature , only @xmath0 values consistent with are possible : @xmath49 summarizing : 1 .
+ forbidden @xmath0 .
@xmath26 is not normalizable .
2 . @xmath51 .
+ @xmath26 is normalizable .
energy can take values @xmath52 3 .
+ @xmath26 is normalizable .
energy can only take values @xmath54
i.e. @xmath55 4 .
+ forbidden @xmath0 .
no values of @xmath48 consistent with energy cutoff .
two main differences arise with respect to the non - relativistic case @xcite , where allowed values are @xmath57 . in the relativistic case constraints
are much tighter , @xmath58 , and depend on the ratio of thermal to rest energy . considering these constraints , the relativistic distribution function can be plotted .
the result is qualitatively similar to the nonrelativistic case : ) .
full circles : @xmath59 .
solid line : @xmath1 .
open circles : @xmath60 . ]
we consider a proton - electron relativistic plasma . assuming protons remain at rest , and that particles only move in one dimension ( @xmath61 ) , the dispersion relation for longitudinal oscillations is @xmath62 where @xmath63 is the plasma frequency
, @xmath64 is the wavenumber , @xmath65 is the argument of the laplace transform ( @xmath66 , where @xmath67 is the wave frequency , if there is no damping ) , @xmath6 the electron mass , @xmath68 is the particle momentum , and @xmath69 is the equilibrium electron distribution function , normalized to unity .
particle energy is @xmath70 . following ref .
@xcite , we now consider the long wavelength limit @xmath71 , where @xmath72 , and @xmath73 is the debye length .
the integrand has a pole at @xmath74 .
if @xmath75 , integration can be performed on the real axis : @xmath76 where @xmath77 is the maximum momentum allowed by if @xmath78 , or @xmath79 if @xmath80 . since @xmath26 is an even function of @xmath81 , the first term in vanishes . the dispersion relation can then be written : @xmath82 where @xmath83 . integrating by parts , and using the fact that @xmath84 , finally yields @xmath85 equation depends on @xmath0 in general , and therefore , unlike the nonrelativistic case @xcite , nonextensive corrections to the dispersion relation of longitudinal waves may appear to the lowest order in @xmath86 .
closed expressions in terms of known functions can be obtained for eq . in the ultrarelativistic case @xmath87 . as shown in sec .
[ distribution_function ] , @xmath0 can only take values between 0 and @xmath88 .
we consider two situations : since @xmath89 , in this case , the dispersion relation can be written @xmath90 noting that @xmath91 and @xmath92 , normalization of the distribution function @xmath93 @xmath94 yields @xmath95 where @xmath96 the integral in can be done analytically , yielding @xmath97 where @xmath98 \ , \end{aligned}\ ] ] @xmath99 is the hypergeometric function , @xmath100 with @xmath101 , and @xmath102 is the generalized hypergeometric function , @xmath103 now @xmath105 is given by eq . , and dispersion relation eq
. is @xmath106 noting that @xmath107 , normalization of the distribution function yields @xmath108^{-1 } \ , \ ] ] where @xmath109 again , integration in can be performed analytically , yielding @xmath110 where @xmath111^{-1 } \frac 1{(\alpha^2\tau^2 + 1)^2 } \\
\times \left [ ( 1+i\alpha\tau)\sqrt{1+\hat p^2 } \cdot f_1\left(1+\alpha;\frac32,\frac32;2+\alpha;\frac{\alpha\tau}{\alpha\tau+i } , \frac{\alpha\tau}{\alpha\tau - i}\right ) \right .
- \left(1-\frac{\hat p}{\alpha\tau}\right)^{1+\alpha } [ ( 1+i\alpha\tau)+\hat p(i+\alpha\tau ) ] \cdot f_1\left(1+\alpha;\frac32,\frac32;2+\alpha ; \frac{\alpha\tau-\hat p}{\alpha\tau+i } , \frac{\alpha\tau-\hat p}{\alpha\tau - i}\right ) \right ] \ , \end{gathered}\ ] ] where @xmath112 is the appell hypergeometric function of two variables : @xmath113
some consequences of describing a plasma based on a nonextensive statistical treatment have been discussed . in particular , the one dimensional plasma distribution function which maximizes tsallis s entropy @xmath11 has been considered . in the nonrelativistic case ,
normalizability of the distribution function constrains @xmath0 to values greater than or equal to @xmath114 .
however , in the relativistic case a much shorter range is possible , and the maximum value of @xmath0 depends on the ratio of temperature to particle rest energy .
then we derive the dispersion relation for longitudinal oscillations in a plasma composed of electrons and rest protons . in the long wavelength limit , and unlike the nonrelativistic case , nonextensive corrections appear to the lowest order .
finally , in the ultra - relativistic regime , this dispersion relation is written explicitly in terms of generalized hypergeometric functions . | equilibrium of statistical systems studied based on boltzmann - gibbs entropy @xmath7 @xmath4 boltzmann constant @xmath8 probability of @xmath9-th microstate evidence shows formalism adequate for long range interactions phase space fractal or multifractal structure tsallis @xcite proposed form for entropy : @xmath10 @xmath0 real number @xmath11 reduces to @xmath12 @xmath11 independent systems @xmath13 entropy of composite system @xmath15 given by @xmath16 @xmath0 measure of nonextensivity physical systems connections with tsallis entropy formalism gravitational systems long range hamiltonian systems nonlinear inverse bremsstrahlung absorption turbulence granular systems applications in astrophysics fluid dynamics biology economy quantum mechanics nonextensive formalism distribution function maximizes @xmath11 non - maxwellian @xcite for @xmath17 high energy states more probable @xmath18 less probable cutoff beyond no states unidimensional gas @xmath19 cutoff given by @xcite @xmath20 velocity distributions in plasmas non - maxwellianhigh energy tails space laboratory plasmas @xcite common metaequilibrium non - maxwellian radial density profiles in cylindrical turbulent electron plasmas observed distribution functions emerge system described nonextensive statistics authors applied approach plasma systems @xcite lima et al studied longitudinal oscillations collisionless thermal plasma nonextensive formalism derived dispersion relation electrostatic waves calculated equilibrium distribution function fits experimental data better maxwellian previous analysis non - relativistic plasmas relativistic extensions tsallis formalism possible @xcite purpose study consequences relativistic nonextensive statistics in plasmas examine properties equilibrium distribution function calculate dispersion relation electrostatic waves long wavelength limit ultrarelativistic plasmas equilibrium distribution function relativistic plasma written @xcite @xmath21^ @xmath22 particle energy @xmath23 normalization constant equation generalizes nonrelativistic expression @xmath1 yields usual boltzmann distribution function @xmath24 one dimensional plasma relativistic energy positive @xmath26 real normalizable if @xmath27 satisfies conditions@xmath27 value if @xmath29 distribution function energy cutoff if @xmath30 hold limit @xmath1 give constraint @xmath31 cutoff if @xmath30 @xmath26 normalizable normalizability distribution function for @xmath32 when @xmath33 not guaranteed integral @xmath34 finite if @xmath35 @xmath36 @xmath37 @xmath38 @xmath39{q-1}}>0 @xmath40 normalizable if @xmath41 analyze case @xmath42 cutoff @xmath43 @xmath44 yields cutoff momentum space energy cutoff limit case @xmath45 only momentum allowed @xmath46 if @xmath47 no value of @xmath48 satisfies condition no distribution function exists temperature only @xmath0 values possible @xmath49 forbidden @xmath0 @xmath26 not normalizable @xmath51 @xmath26 normalizable energy values @xmath52 only values @xmath54 no values of @xmath48 consistent with energy cutoff differences non - relativistic case @xcite allowed values are @xmath57 relativistic case constraints tighter @xmath58 depend on thermal to rest energy relativistic distribution function plotted result similar to nonrelativistic case full circles @xmath59 solid line @xmath1 open circles @xmath60 proton - electron relativistic plasma protons rest particles move one dimension @xmath61 dispersion relation longitudinal oscillations @xmath62 @xmath63 plasma frequency wavenumber @xmath65 argument laplace transform @xmath66 wave frequency no @xmath6 electron mass @xmath68 particle momentum equilibrium electron distribution function particle energy @xmath70 long wavelength limit @xmath71 @xmath72 @xmath73 debye length integrand pole at @xmath74 @xmath75 integration performed real axis @xmath76 @xmath77 maximum momentum @xmath78 @xmath79 @xmath80 @xmath26 even function @xmath81 first term vanishes dispersion relation @xmath82 @xmath83integrating parts @xmath84 yields @xmath85 depends on @xmath0 nonrelativistic nonextensive corrections dispersion relation longitudinal waves appear lowest order @xmath86 closed expressions known functions obtained ultrarelativistic @xmath87 @xmath0 values between 0 and @xmath88 two situations @xmath89 dispersion relation written @xmath90 @xmath91 @xmath92 normalization distribution function @xmath93 @xmath94 yields @xmath95 @xmath96 integral analytically @xmath97 @xmath98 @xmath99 hypergeometric function @xmath100 @xmath101 @xmath102 generalized hypergeometric function @xmath103 @xmath105 given eq dispersion relation@xmath106 @xmath107 normalization distribution function yields @xmath108^{-1 @xmath109 integration yielding @xmath110 @xmath111^{-1 1{(\alpha^2\tau^2 + 1)^2 } 1+i\alpha\tau){1+\hat p^2 }+\alpha\frac32;2+\alpha\alpha\tau+i }\alpha - i}-\frac{\alpha\tau){1+\alpha } 1+i\alpha\tau)+\hat p(i+\alpha\tau )+\alpha\frac32;2+\alpha\alpha\tau-\alpha\tau+i }\tau - i} @xmath112 hypergeometric function two variables @xmath113 consequences describing plasma nonextensive statistical treatment discussed one dimensional plasma distribution function maximizes tsallis s entropy @xmath11 considered nonrelativistic case normalizability distribution function constrains @xmath0 values @xmath114 relativistic case shorter range possible maximum value @xmath0 depends ratio temperature to particle rest energyderive dispersion relation longitudinal oscillations plasma electrons protons long wavelength limit nonrelativistic nonextensive corrections lowest order ultra - relativistic regime dispersion relation written generalized hypergeometric functions | 0.423666 | 2,417 | 1,274 | , Saving $0.1 in GPT-4. |
67,377 | the storage capacity of an incremental learning algorithm for the parity machine , the tilinglike learning algorithm , is analytically determined in the limit of a large number of hidden perceptrons .
different learning rules for the simple perceptron are investigated .
the usual gardner - derrida one leads to a storage capacity close to the upper bound , which is independent of the learning algorithm considered . | the storage capacity is one of the most important characteristics of a neural network .
it is the maximal number of random input - output patterns per input entry that a network is able to correctly classify with probability one .
this quantity is independent of the algorithm used to learn the weights of the network ; it only depends on its architecture
. we will refer to it as the _ architecture storage capacity _ @xmath0 in the following , to distinguish it from the algorithm - dependent one , hereafter called _ algorithm storage capacity _ , @xmath1 .
the simplest neural network , the perceptron , has the inputs directly connected to the output .
geometrical arguments @xcite and a statistical mechanics calculation @xcite determined that @xmath2 .
several perceptron learning algorithms , like the adatron @xcite , are known to achieve such storage capacity .
the capacity can be increased using networks with more complicated architectures .
the next one on increasing complexity is the extensively studied @xcite monolayer perceptron ( mlp ) , which has @xmath3 `` hidden ''
perceptrons connected to the output unit . as a mlp can store any function of its inputs , provided that the number of hidden units is adequate , it is not worth to consider more complex architectures for the storage problem . given an input pattern , the hidden units states define a @xmath3-dimensional vector , the pattern s _ internal representation _ ( ir ) .
the network s output is a function of the ir . in the following ,
we consider binary neurons , of states @xmath4 , and focus on the parity machine , whose output is the product of the @xmath3 components of the ir .
the main problem when training mlps is that the irs of the input patterns are unknown .
it has been proposed @xcite to build the hidden layer using a constructive procedure , called tilinglike learning algorithm ( tla ) , in which the hidden perceptrons are included one after the other and trained to correct the learning errors of the preceding unit .
as each unit can at least correct one error , convergence is ensured @xcite .
it is straightforward to show that the tla generates a parity machine @xcite , but the number of included hidden units depends crucially on the performance of the perceptron learning algorithm used to train them .
geometric arguments @xcite and a statistical mechanics replica calculation @xcite , showed that the architecture storage capacity of the parity machine in the limit of a large number of hidden units @xmath3 is @xmath5 .
however , it was not clear whether this storage capacity could be actually achieved with a learning algorithm . in this paper , we show that the tilinglike learning algorithm ( tla ) can reach a storage capacity close to the architecture storage capacity , provided that the hidden perceptrons are trained with an appropriate learning algorithm . in section [ sec.tla ]
we describe more precisely the setting and the tla .
the analytical expression of the algorithm storage capacity in the limit of large @xmath3 is determined in section [ sec.alpha ] , where we show that the learning algorithm used to train the hidden units must satisfy stringent condition for the tla to converge with a finite number of hidden units .
the results presented in section [ sec.algo ] show that these conditions rule out some perceptron learning algorithms , like the adatron .
the conclusions are presented in section [ sec.concl ] .
let us assume a training set @xmath6 of @xmath7 input - output patterns .
the inputs @xmath8 are random gaussian @xmath9-dimensional vectors with zero mean and unit variance in each direction .
the corresponding outputs @xmath10 are the learning targets .
their values @xmath11 are randomly selected with probability : @xmath12 with @xmath13 .
the rle of the bias @xmath14 introduced in ( [ proba ] ) will become clear in the following .
the probability of the targets @xmath15 is unbiased .
the tla constructs the parity machine by including successive perceptrons in the hidden layer .
each unit is connected to the input @xmath16 through weights @xmath17 where @xmath18 is a threshold .
the inputs are classified through @xmath19 .
thus , a perceptron separates linearly the input space with a hyperplane orthogonal to @xmath20 ( we assume @xmath21 ) at a distance @xmath18 to the origin .
the weights and the threshold are learned through the minimization of a cost function : @xmath22 where the potential @xmath23 is the contribution of each pattern to the cost , and @xmath24 is the distance of the pattern to the hyperplane . within the tla heuristics , the first perceptron @xmath25 is trained to learn targets @xmath26 .
after learning , its weights are @xmath27 ; its training error is @xmath28 where @xmath29 is the heaviside function and @xmath30 the perceptron s output to pattern @xmath31 .
if @xmath32 , the training set is correctly classified ; the tla stops with only one simple perceptron .
otherwise , a new perceptron is introduced .
the successive perceptrons @xmath33 are trained to learn training sets @xmath34 with targets @xmath35 , that is , @xmath36 if the pattern @xmath37 is correctly classified by the previous perceptron and @xmath38 otherwise . if the perceptron learning algorithm is correctly chosen it can be shown that the successive training errors @xmath39 are strictly decreasing @xcite .
thus , the tla procedure necessarily converges to a mlp with @xmath3 units , where the @xmath40 perceptron is the first one to meet the condition @xmath41 .
then , the product @xmath42 gives the correct output to the patterns of the training set @xmath43 .
the algorithm storage capacity of the tla , @xmath44 , is simply the inverse function of @xmath45 , the average number of perceptrons typically included by the tla when the training set has a size @xmath46 . in order to determine @xmath45 , consider the @xmath47 hidden unit : the probability of its targets @xmath48 depends on the training error @xmath49 of the previous perceptron .
although there exist some correlations between the outputs @xmath50 , due to the correlations in the weights of the successive perceptrons , in the limit of a large training sets ( @xmath51 ) they may be neglected @xcite .
thus , we may assume that the targets @xmath50 are independently drawn with probability ( [ proba ] ) , with a bias @xmath49 .
then , the successive training errors satisfy a simple recursive relation @xcite : @xmath52 where @xmath53 is the training error of a simple perceptron trained with a training set of size @xmath54 and biased targets @xmath55 drawn with a probability @xmath56 given by ( [ proba ] ) .
the number @xmath3 of perceptrons necessary to correctly classify the initial training set satisfies @xcite : @xmath57 where @xmath58 stands for @xmath59 and the symbol @xmath60 for the composition of functions . in the limit of a large @xmath54 ,
the training error @xmath59 is close to @xmath14 and the number of simple perceptrons @xmath45 is large .
it is thus possible to use the continuum limit : @xmath61 with @xmath62 .
after integration of this differential equation we obtain @xcite the typical number of hidden units introduced by the tla : @xmath63 @xmath45 depends on the specific cost function ( [ cost ] ) used to train the perceptrons through @xmath64 , which is the training error of a simple perceptron learning a training set with biased targets .
in this section we determine the perceptron s training error @xmath64 for biased training sets , in the thermodynamic limit ( @xmath65 with @xmath46 fixed ) . for different possible choices of the potential @xmath66 in ( [ cost ] ) ,
we deduce the number of hidden units , @xmath67 and the algorithm storage capacity of the tla .
only the main results are presented here , the interested reader can find the details in @xcite .
the adatron potential is @xmath68 , where @xmath69 is a positive parameter called stability .
all the patterns with negative @xmath70 and those with @xmath71 but closer than @xmath69 to the hyperplane ( which are correctly classified ) contribute to the cost .
the training error @xmath64 is obtained through a replica calculation assuming replica symmetry ( rs ) , which can be shown to hold for all @xmath54 .
it turns out that for fixed @xmath69 , in the limit of large @xmath54 , @xmath72 . as a consequence ,
the constraint that the successive training errors are strictly decreasing , necessary for the convergence of the tla , is not satisfied .
this problem can be circumvented at the price of considering @xmath69 as a free parameter , and minimizing the training error with respect to it . in that case ,
@xmath73 and the algorithm storage capacity is @xmath74 in the limit of large @xmath54 .
the potential of the gardner - derrida ( gd ) cost function @xcite is @xmath75 .
the hypothesis of rs is incorrect for this potential , and the obtained value of @xmath76 is a lower bound to the true training error .
consequently , the replica calculation allows only to determine an upper bound to @xmath44 . if @xmath77 , the cost is nothing else but the number of misclassified patterns .
it gives the lowest bound to @xmath76 . in the limit of large training set size @xmath54 , we obtain : @xmath78 where @xmath79 satisfies @xmath80/[2 \ , ( \cosh a - a \sinh a - 1)]$ ] .
this leads to @xmath81 and @xmath82 , larger than @xmath83 , probably due to the failure of the rs hypothesis . in order to obtain a lower bound to @xmath83 we used the kuhn - tucker cavity method @xcite , which gives an upper bound to @xmath76 . as a result of both calculations ,
we can bound @xmath84 : @xmath85 on view of this result , we expect that the algorithm storage capacity of the tla behaves like @xmath86 with @xmath87 .
a calculation with one step of replica symmetry breaking would give an estimate of the exponent @xmath88 .
since the rs solution gives a better approximation of the training error than the kuhn - tucker cavity method , we expect the exponent @xmath88 to be close to @xmath89 , leading to an algorithm storage capacity close to the architecture s capacity .
this result shows that the tla may build a nearly optimal network . in order to improve the robustness against noise in the data , it is usually useful to impose some finite stability @xmath69 to the patterns .
the corresponding gd potential is @xmath90 . as with @xmath77 , here
also the rs solution is unstable .
the bounds on @xmath91 deduced from the results for @xmath76 obtained with the rs hypothesis and the kuhn - tucker cavity method @xcite , give : @xmath92 strikingly , imposing a finite stability @xmath69 has an important effect on the algorithm storage capacity , which in this case behaves as @xmath86 with @xmath93 .
the prefactors of the bounds of @xmath94 are @xmath69-dependent and they both diverge for @xmath95 . the exponent @xmath88 is independent of @xmath69 for finite @xmath69 but differs from the one corresponding to @xmath96 .
we determined analytically the storage capacity of the tilinglike learning algorithm for the parity machine , a constructive procedure generating a monolayer perceptron of binary hidden units .
a training set of input - output examples is used to determine the number of hidden units , which are introduced one after the other .
these are simple perceptrons that have to learn their weights using increasingly biased target distributions .
we have shown that the storage capacity of the tla depends crucially on the learning errors of the successively introduced perceptrons .
the properties of the algorithm used to train the latter have thus dramatic consequences on the size of the hidden layer generated by the tla , and may even hinder the convergence to a finite size network .
this arises , in particular , if the perceptrons are trained with the adatron algorithm unless the stability is adapted to the successive targets biases .
the smallest network is obtained using the gardner - derrida cost function with vanishing stability , which corresponds to minimizing the training error .
based on the results obtained within the replica symmetry hypothesis , and those using the kuhn - tucker cavity method , we expect a supra - linear storage capacity @xmath97 with @xmath98 , very close to the theoretical capacity corresponding to the architecture considered . | storage capacity is important neural network maximal number of random input - output patterns per entry with probability one quantity independent of algorithm depends on architecture refer as _ architecture storage capacity _ distinguish from algorithm - dependent algorithm storage capacity _ simplest neural network perceptron has inputs directly connected to output geometrical arguments statistical mechanics calculation determined perceptron learning algorithms like adatron achieve such storage capacity capacity can be increased using complicated architectures next increasing complexity is @xcite monolayer perceptron ( mlp ) has hidden perceptrons connected to output unit mlp can store any function inputs number hidden units adequate not worth to consider more complex architectures for storage problem hidden units define @xmath3-dimensional vector internal network s output is function of ir consider binary neurons focus on parity machine output is product of @xmath3 components of ir problem when training mlps irs of input patterns unknown proposed to build hidden layer using tilinglike learning algorithm hidden perceptrons included trained to correct learning errors preceding uniteach unit one error convergence ensured @xcite tla generates parity machine number hidden units depends on performance perceptron learning algorithm geometric arguments statistical mechanics calculation architecture storage capacity of parity machine in limit large hidden units @xmath3 is @xmath5 storage capacity achieved with learning algorithm tilinglike learning algorithm tla reach storage capacity close to architecture storage capacity hidden perceptrons trained with appropriate learning algorithm section [ sec.tla ] setting tla analytical expression algorithm storage capacity in limit large @xmath3 determined in section [ sec.alpha ] learning algorithm hidden units must satisfy stringent condition for with finite number hidden units results conditions rule out some perceptron learning algorithms like conclusions in section [ sec.concl ] training set @xmath6 of @xmath7 input - output patterns inputs @xmath8 random gaussian-dimensional vectors zero mean unit variance outputs @xmath10 are learning targets values @xmath11 randomly selected with probability @xmath13 bias @xmath14 probability of targets @xmath15 unbiased tla constructs parity machine perceptrons in hidden layerunit connected to input @xmath16 through weights @xmath18 threshold inputs classified @xmath19 perceptron separates input space with hyperplane orthogonal to @xmath20 distance @xmath18 to origin weights threshold learned through minimization cost function potential contribution to cost distance to hyperplane first perceptron @xmath25 trained to learn targets @xmath26 weights @xmath27 training error @xmath28 @xmath29 heaviside function @xmath30 perceptron output to pattern @xmath31 if training set classified tla stops with one perceptron new perceptron introduced successive perceptrons @xmath33 learn training sets @xmath34 targets @xmath35 @xmath36 if pattern @xmath37 classified learning algorithm successive training errors @xmath39 tla procedure converges to mlp with @xmath3 units @xmath40 perceptron first to meet condition @xmath41 product @xmath42 gives correct output to patterns training set @xmath43 algorithm storage capacity tla @xmath44 inverse function of @xmath45 average number of perceptrons training set size @xmath46determine @xmath45 consider @xmath47 hidden unit probability targets @xmath48 depends training error @xmath49 previous perceptron correlations outputs @xmath50 large training sets @xmath51 neglected assume targets @xmath50 independently drawn probability bias @xmath49 successive training errors satisfy recursive relation @xcite : @xmath52 @xmath53 training error simple perceptron trained set @xmath54 biased targets @xmath55 drawn probability @xmath56 number perceptrons classify initial training set satisfies @xcite @xmath57 @xmath58 stands @xmath59 symbol @xmath60 composition functions large @xmath54 training error @xmath59 close to @xmath14 number simple perceptrons @xmath45 large use continuum limit @xmath61 @xmath62 differential equation obtain typical number hidden units @xmath63 @xmath45 depends cost function perceptrons @xmath64 training error simple perceptron learning training set biased targets determine perceptron s training error @xmath64 for biased training sets thermodynamic limit @xmath65 @xmath46 fixedchoices potential @xmath66 cost deduce number hidden units @xmath67 algorithm storage capacity tla main results presented here details in @xcite adatron potential is @xmath68 @xmath69 positive parameter stability patterns with negative @xmath70 @xmath71 closer @xmath69 to hyperplane correctly classified contribute to cost training error @xmath64 obtained replica calculation replica symmetry rs ) all @xmath54 fixed @xmath69 limit large @xmath54 @xmath72 constraint successive training errors decreasing convergence tla not satisfied problem circumvented considering @xmath69 free parameter minimizing training error @xmath73 algorithm storage capacity is @xmath74 limit large @xmath54 potential gardner - derrida cost function @xcite is @xmath75 hypothesis of rs incorrect value @xmath76 lower bound to true training error replica calculation allows upper bound to @xmath44 @xmath77 cost number misclassified patterns lowest bound to @xmath76 limit of large training set size @xmath54 obtain @xmath78 where @xmath79 satisfies @xmath80/leads to @xmath81 larger than due to failure rs hypothesis lower bound to @xmath83 used kuhn - tucker cavity method upper bound to @xmath76 bound @xmath84 : @xmath85 algorithm storage capacity tla like @xmath86 with @xmath87 calculation replica symmetry breaking estimate exponent @xmath88 rs solution better training error expect exponent @xmath88 close to @xmath89 algorithm storage capacity close to architecture s capacity shows tla may build nearly optimal network improve robustness against noise impose finite stability @xmath69 to patterns corresponding gd potential is @xmath90 @xmath77 rs solution unstable bounds on @xmath91 deduced from results @xmath76 rs hypothesis - tucker method give : @xmath92 imposing finite stability @xmath69 algorithm storage capacity behaves as @xmath86 with @xmath93 bounds @xmath94 @xmath69-dependent diverge for @xmath95 exponent @xmath88 independent of @xmath69 differs from @xmath96determined storage capacity tilinglike learning algorithm parity machine generating monolayer perceptron binary hidden units training set input - output examples determine number hidden units introduced simple perceptrons learn weights using biased target distributions storage capacity tla depends on learning errors introduced perceptrons properties algorithm size hidden layer may hinder convergence to finite size network if perceptrons trained with adatron algorithm unless stability adapted to targets biases smallest network obtained using gardner - derrida cost function with vanishing stability minimizing training error based results replica symmetry hypothesis kuhn - tucker cavity method expect supra - linear storage capacity @xmath97 with @xmath98 close to theoretical capacity architecture | 0.437786 | 2,837 | 1,449 | , Saving $0.1 in GPT-4. |
67,380 | a large amount of population models use the concept of a carrying capacity .
simulated populations are bounded by invoking finite resources through a survival probability , commonly referred to as the verhulst factor .
the fact , however , that resources are not easily accounted for in actual biological systems makes the carrying capacity parameter ill - defined .
henceforth , we deem it essential to consider cases for which the parameter is unnecessary .
this work demonstrates the possibility of verhulst - free steady states using the penna aging model , with one semelparous birth per adult .
stable populations are obtained by setting a mutation threshold that is higher than the reproduction age .
population dynamics ; carrying capacity ; aging ; penna model | the evolution of a population can be described using the difference equation @xmath0 where @xmath1 is the population at time @xmath2 , @xmath3 is the intrinsic relative growth rate , and @xmath4 is the carrying capacity .
this logistic type of growth was first introduced by verhulst in the mid 1800s @xcite . in an environment where resources are finite
, populations are bounded .
the equivalent differential equation leads to a final state which is a fixed point .
discrete counterparts , such as the one presented above , result in bifurcations , limit cycles and chaos @xcite . an in - depth discussion of the more complex set of solutions obtained by considering various aspects of the verhulst equation may be found in @xcite .
the concept of a carrying capacity is commonly incorporated in population dynamics models .
limitations on food supply , space and other necessities suppress growth .
organisms compete in order to stay alive .
each subject to a survival probability , @xmath5 , also known as the verhulst factor .
the main motivation for its introduction is to make certain that population sizes are kept finite .
this parameter may be applied as a constraint for simulated populations in order to maintain their size within computer limits .
however , the carrying capacity is not constant @xcite .
an environment s actual ability to support life highly depends on prevailing knowledge and technologies @xcite .
it is not easily determined @xcite , neither is it well defined .
we choose to consider cases for which the carrying capacity is not essential .
because resources are not explicitly accounted for in many observed systems , we believe this case can be of great relevance to biological population studies .
the penna model @xcite is a popular technique for simulating aging populations @xcite .
individual characteristics are represented by a string of binary numbers of length @xmath6 . over time , mutations accumulate thereby modifying an organism s viability , survival and fertility . for each year
added to an individual s age , one bit in the genome is read .
active genetic traits , therefore , are located at bit numbers ( locus ) less than or equal to the current age .
zeroes correspond to healthy genes and ones are bad or mutated genes .
an individual dies a genetic death when it reaches old age ( @xmath7 ) or when the number of expressed harmful mutations equals the threshold value , @xmath8 . here ,
only bad mutations are considered because helpful ones , in comparison , are very rare @xcite .
reproductive maturity is achieved at age @xmath9 .
starting from this age until death , individuals generate @xmath10 offsprings per time interval , where @xmath10 is the birth rate . in asexually reproducing populations , newborns copy the genes of the parent ( both active and inactive ) and acquire @xmath11 additional mutations .
the mutation rate , @xmath11 , can take on any value but is usually set to one since expression of new mutations is not very frequent in nature @xcite . to include the effects of finite resources , individuals are subject to the same survival probability regardless of the quality of genome , fitness and age - va implementation .
it has been argued @xcite that older individuals should be more adaptable and , thus , have greater chances of survival .
furthermore , it was added that due to the gompertzian behavior of the original model , limiting the number of births is enough to avoid exponential growth .
hence , an alternative approach was proposed applying the verhulst factor to newborns only ( the vb implementation ) . for lack of a vital role in population dynamics @xcite
, efforts have been made to eliminate the random deaths imposed by the verhulst factor . in @xcite ,
a verhulst - free penna model was implemented on a square lattice , where each lattice site can accommodate only one individual at a time .
the abstract limit on the capacity of the environment is , thus , given by the size of the grid .
another study @xcite modified the model such that for every death , an adult is randomly chosen to produce one offspring ( a replacement ) .
this procedure keeps the total population size constant . in this work , we modify the 32-bit penna model , making it verhulst - free
. deaths are now due only to genetics - old age and mutation accumulation . to limit growth
, we allow individuals to breed only once in their lifetime ( at age @xmath9 ) .
this behavior is observed in semelparous populations such as the mayflies and the pacific salmon @xcite .
the simulation starts with @xmath12 perfect individuals ( no bad genes ) and runs for 3000 iterations .
the choice of evaluating deaths before births limits the maximum reproduction age value , @xmath13 .
in the @xmath14 limit , the behavior of the resulting population is dependent on the relative value of the threshold parameter and the reproduction age .
we observe exponential decrease when @xmath15 .
stability , on the other hand , is achieved by setting @xmath16 .
the populations presented hereafter are normalized about the maximum value obtained for each parameter set .
extinction has two causes - mutational meltdown and overpopulation @xcite .
the latter results from a sudden fluctuation in population size beyond the carrying capacity . but without the verhulst factor ( no stochasticity ) , the only possible causes of death are old age and mutation accumulation . at low @xmath8
, selection favors deaths at younger ages in order to keep the gene pool clean .
thus when @xmath15 , a good fraction of the population dies prior to reproductive maturity . with births limited to one offspring per individual , the number of newborns is not enough to compensate for losses due to bad mutations .
the population dies of mutational meltdown .
best - fits show delays in the onset of extinction for those that reproduce late in life , large @xmath9 ( fig .
[ fig1 ] ) .
the rate of decay is effectively slowed down by an increased tolerance for bad mutations .
regression analysis yields a power law relationship between the mutation threshold and the average decay rate ( fig .
[ fig2 ] ) .
stable populations were verified using first return maps @xcite .
consider the sequence @xmath17 and @xmath18 .
the first return map is a plot of consecutive entries of this sequence , i.e. , @xmath19 , @xmath20 , @xmath21 , @xmath22 . fig .
[ fig3 ] presents the return maps of the steady state populations associated with @xmath23 and 12 when @xmath14 and @xmath24 .
the number of fixed points reflect periodic variations in the time series . for these steady states ,
we find that the magnitude of the fluctuations in total population size increases with @xmath9 . to minimize fluctuations , running averages were obtained for the time series .
this was done using intervals of 50 .
variations in the transient part of the curves ( fig.[fig4 ] ) indicate differences in saturation time .
the slope from the highest value to the onset of steady state becomes steeper as @xmath9 increases .
hence , those with higher reproduction ages take longer to equilibriate .
similar to the original penna model results @xcite , steady state is achieved faster by allowing more mutations , that is , by increasing @xmath11 .
[ fig5 ] shows the variation of the mean population with threshold , for a given @xmath9 .
this corresponds to the average population size of the last 300 iterations of a particular simulation run . at these times , the simulated populations have already achieved steady state .
note that the mutation threshold can have values from 1 to @xmath6 , the bit - string length .
however , since stability is achieved only when @xmath16 ( for the @xmath14 case ) , the number of @xmath8 values that result to nonzero steady states decreases as @xmath9 increases .
we find that the total population scales with threshold .
considering all possible @xmath9 values ( from 1 to @xmath25 ) , we obtain linear fits that have slopes ranging from 0.015 to 0.035 . the steady state associated with @xmath26 ( or @xmath27 ) takes advantage of the finite bit - string characteristic of the penna model . following the derivation in @xcite ,
we let @xmath28 be the number of individuals with age @xmath29 at time @xmath2 .
the limit on the length of the string forces deaths at age @xmath6 . in the absence of the other death factors , the discrete time evolution of the population is described by @xmath30 and @xmath31 at steady state , the average number of newborns is constant , i.e. , @xmath32 . from the equations above ,
we have @xmath33 using ( [ parents ] ) in ( [ babies ] ) , it is straightforward to show that in the no mutation case ( @xmath26 or @xmath27 ) , stability is achieved only when @xmath14 .
no other nonzero steady states were found . beyond @xmath14 ,
populations either increase indefinitely or tend to zero ( fig.[fig6 ] ) .
it has been shown @xcite that asexual populations have at most @xmath34 mutations before the minimum reproductive age . for semelparous species ,
all bits above age @xmath9 are set to one . in @xcite , we have demonstrated that this is not always the case . as @xmath8 increases , the strength of selection decreases and more mutations are preserved in the genome . when @xmath16 , all bits at steady state contain defective genes .
the process of mutation accumulation is very fast in asexual populations . to slow down the process
, one may consider introducing recombination , as in sexual populations , or allowing positive mutations .
it is common practice in population models to attribute stability with the concept of a carrying capacity .
verhulst - free cases generally lead to malthus catastrophe - exponential growth or decay .
the carrying capacity , however , is an ill - defined parameter @xcite .
it is for this reason that we explore cases for which the concept is unnecessary , that is , to find stable populations without having to impose a verhulst factor . within the framework of the penna model ,
the artificial cap on the individual lifespan imposed by the bit - string length and the semelparous limit on reproduction make possible the verhulst - free implementation .
murray , mathematical biology : i. an introduction ( 3rd ed ) , springer - verlag , 2002 .
bernardes , et al .
j. b. 1 ( 1998 ) 393 .
m. ausloos and m. dirickx ( eds ) , the logistic map and the route to chaos , springer , berlin , 2006 .
s. moss de oliveira , p.m.c . de oliveira and j.s .
s@xmath35 martins , ijmp c 15 ( 2004 ) 301 .
k. malarz , m. sitarz , p. gronek and a. dydejczyk , lecture notes in computer science 3037 ( 2004 ) 638 .
s@xmath35 martins and s. cebrat , theory biosci .
119 ( 2000 ) 156 .
d. makowiec , physica a 289 ( 2001 ) 208 .
s. m. de oliveira , d. alves and j.s .
s@xmath35 martins , physica a 285 ( 2000 ) 77 .
a. pekalski and m. ausloos , physica a 387 ( 2008 ) 2526 .
heilig in the future population of the world : what can we assume today ?
w. lutz ) , earthscan , london , 1996 as cited in @xcite .
simon , h. kahn ( eds ) , the resourceful earth : a response to global 2000 , blackwell , oxford , 1984 as cited in @xcite .
cohen , how many people can the earth support ?
norton & company , new york & london , 1995 .
penna , j. stat .
78 ( 1995 ) 1629 .
s. moss de oliveira , physica a 257 ( 1998 ) 465 .
h. puhl , d. stauffer and s. roux , physica a 221 ( 1995 ) 445 .
k. malarz , ijmp c 11 ( 2000 ) 309 .
j. dabkowski , m. groth and d. makowiec , acta physica polonica b 31 ( 2000 ) 1027 .
de oliveira , s. moss de oliveira and j.s .
s@xmath35 martins , ijmp c 15 ( 2004 ) 301 .
penna and s. moss de oliveira , j. phys .
i france 5 ( 1995 ) 1697 .
penna and s. moss de oliveira , phys .
e 52 ( 1995 ) rr3309 .
meyer - ortmanns , ijmp c 12 ( 2001 ) 319 .
k. malarz , theory biosci . 125
( 2007 ) 147 .
g. p. williams , chaos theory tamed , taylor & francis limited , london , 1997 .
s@xmath35 martins and s. moss de oliveira , ijmp c 9 ( 1998 ) 421 .
piol and r.s .
banzon , e - print arxiv:1009.5095v1 | evolution of population described using difference equation @xmath0 where @xmath1 is population at time @xmath2 @xmath3 intrinsic relative growth rate @xmath4 carrying capacity logistic growth introduced by verhulst mid 1800s resources finite populations bounded equivalent differential equation leads to final state fixed point discrete counterparts result in bifurcations limit cycles chaos - discussion of complex solutions verhulst equation in @xcite concept carrying capacity incorporated in population dynamics models limitations on food supply space necessities suppress growth organisms compete to stay alive subject to survival probability verhulst factor motivation population sizes finite parameter constraint for simulated populations maintain size limits carrying capacity not constant environment s ability support life depends on knowledge technologies not easily determined well defined consider cases carrying capacity not essential resources not accounted for case relevance to biological population studies penna model for simulating aging populations individual characteristics represented by string binary numbers of length mutations accumulate modifying organism s viability survival fertility each year added to age one bit in genome readactive genetic traits located at bit numbers less than or equal to current age zeroes correspond to healthy genes ones bad or mutated genes individual dies genetic death old age ( @xmath7 ) or harmful mutations equals threshold value @xmath8 only bad mutations considered helpful ones rare reproductive maturity at age @xmath9 until death individuals generate @xmath10 offsprings per time interval @xmath10 is birth rate in asexually reproducing populations newborns copy genes parent acquire @xmath11 additional mutations mutation rate @xmath11 usually set to one new mutations not frequent individuals subject to same survival probability regardless of quality of genome fitness age - va implementation older individuals should be more adaptable have greater chances of survival limiting number of births to avoid exponential growth alternative approach proposed verhulst factor to newborns only efforts to eliminate random deaths verhulst factor verhulst - free penna model implemented on square lattice each one individual limit on capacity environment given by size of gridstudy @xcite modified model every death adult randomly one offspring procedure keeps population size constant 32-bit penna model verhulst - free deaths due to genetics old age mutation accumulation limit growth individuals breed once at age @xmath9 behavior observed in semelparous populations mayflies pacific salmon simulation starts with @xmath12 perfect individuals runs 3000 iterations evaluating deaths before births limits maximum reproduction age @xmath13 @xmath14 limit behavior population dependent on threshold parameter reproduction age exponential decrease when @xmath15 stability achieved by setting @xmath16 populations normalized about maximum value each parameter extinction causes mutational meltdown overpopulation results population size without verhulst factor only possible causes death old age mutation accumulation low @xmath8 selection favors deaths younger ages keep gene pool clean when @xmath15 fraction population dies prior reproductive maturity births limited to one offspring per individual newborns not enough compensate losses bad mutations population dies of mutational meltdown show delays extinction for late in life @xmath9decay slowed by increased tolerance for bad mutations regression analysis power law relationship between mutation threshold average decay rate stable populations verified using first return maps @xcite sequence @xmath17 and @xmath18 first return map of consecutive entries return maps of steady state populations with @xmath23 12 fixed points reflect variations in time series fluctuations in population size increases with @xmath9 running averages obtained for time series using intervals of 50 variations in curves indicate differences in saturation time slope from highest value to steady state steeper as @xmath9 increases higher reproduction ages take longer to equilibriate steady state achieved faster by more mutations increasing @xmath11 fig5 shows variation mean population with threshold for @xmath9 corresponds to average population size of last 300 iterations of simulation run populations achieved steady state mutation threshold values from 1 to @xmath6 bit - string lengthstability achieved when @xmath16 @xmath8 values nonzero steady states decreases as @xmath9 increases total population scales with threshold @xmath9 values to linear fits slopes 0.015 to 0.035 steady state @xmath26 or @xmath27 finite bit - string penna model @xmath28 be individuals with age @xmath29 at @xmath2 limit on length string forces deaths at age @xmath6 discrete time evolution population described by @xmath30 and @xmath31 at steady state average number newborns constant @xmath32 @xmath33 no mutation case @xmath26 or stability achieved when @xmath14 no other nonzero steady states found beyond @xmath14 populations increase indefinitely or tend to zero asexual populations have most @xmath34 mutations before minimum reproductive age semelparous species all bits above @xmath9 set to one not as @xmath8 increases strength of selection decreases more mutations preserved when @xmath16 all bits at steady state contain defective genes mutation accumulation fast in asexual populationsslow process consider introducing recombination allowing positive mutations common practice population models attribute stability carrying capacity verhulst - free cases lead growth decay carrying capacity ill - defined parameter explore cases concept unnecessary find stable populations without verhulst factor penna model artificial cap lifespan bit - string length semelparous limit reproduction verhulst - free implementation murray mathematical biology introduction 3rd ed springer - verlag 2002 bernardes j b. 1 ( 1998 ) 393 m ausloos dirickx logistic map route chaos springer berlin 2006 moss de oliveira martins ijmp c 15 ( 2004 ) 301 malarz sitarz gronek dydejczyk lecture notes computer science 3037 ( 2004 ) 638 martins cebrat theory biosci . 119 ( 2000 ) 156 makowiec physica a 289 ( 2001 ) 208 de oliveira alves martins physica 285 ( 2000 ) 77 pekalski ausloos physica 387 ( 2008 ) 2526 heilig future population worldlutz earthscan london 1996 @xcite simon kahn resourceful earth response global 2000 blackwell oxford 1984 cohen earth support norton company new york london 1995 penna stat 78 1995 ) 1629 moss oliveira physica 257 1998 ) 465 puhl stauffer roux physica 221 1995 ) 445 malarz ijmp 11 2000 ) 309 dabkowski groth makowiec acta polonica 31 2000 ) 1027 oliveira moss martins ijmp 15 2004 ) 301 penna moss oliveira france 5 1995 ) 1697 penna 52 1995 ) rr3309 meyer ortmanns ijmp 12 2001 ) 319 malarz theory biosci 125 2007 ) 147 williams chaos theory taylor francis london 1997 martins moss oliveira ijmp 9 1998 ) 421 piol banzon - print arxiv:1009.5095v1 | 0.441147 | 2,973 | 1,480 | , Saving $0.1 in GPT-4. |
67,397 | recent experimental works reported observation of the @xmath0 structure in steel . here , stability of the @xmath0 structure in steel is investigated based on first - principles with special interests in effects of interstitial c atoms .
the interstitial c atoms increase the energy of the @xmath0 structure compared with the ferromagnetic ( fm ) bcc .
the @xmath0 structure incorporating c atoms is also mechanically unstable unless the c concentration is 25 at.% .
it is concluded that the @xmath0 structure is mostly unstable in steel , and the @xmath0 structure in steel may be formed under special atomic constraints at twin boundaries or other interfaces .
first - principles calculations , @xmath0 structure , steel | nanometer - size domains of the @xmath0 structure have recently been reported in steel , i.e. , fe - c - based alloys , by detailed transmission electron microscopy observations @xcite .
this motivated us to investigate the stability of the @xmath0 structure in steel .
the present authors group has emphasized the importance of the @xmath0 structure on transformation between the bcc and the fcc structures in metallic systems .
togo and tanaka developed a search algorithm for transformation pathways based on a systematic set of first - principles calculations and revealed that the @xmath0 structure was located on a transformation pathway between the bcc and the fcc structures @xcite .
ikeda et al . suggested that the pressure - induced phase transition between the bcc and the fcc fe at high temperature occurred along this transformation pathway @xcite .
the present authors recently performed a systematic investigation into the @xmath0 structure of transition elements @xcite .
the elemental @xmath0 fe with antiparallel magnetic moments ( @xmath1 magnetic state ) was found to be the lowest in energy among the investigated magnetic states and to be mechanically stable .
the @xmath1 @xmath0 fe was , however , 170 mev / atom higher in energy than the ferromagnetic ( fm ) bcc fe .
this implies that the elemental @xmath0 fe should be difficult to be formed . in experiments , however , the nanometer - size @xmath0 structure was observed in steel @xcite .
the largest difference between elemental fe and steel may be the presence of c atoms , but little has been known about effects of the c atoms on the stability of the @xmath0 structure in steel .
here we report a first - principles study on the stability of the @xmath0 structure in steel with special interests in effects of interstitial c atoms .
four different magnetic states of the @xmath0 structure investigated in our previous study @xcite ( see fig .
1(c ) in ref .
@xcite for notations of the magnetic states ) are focused on .
possible interstitial sites for c atoms in the @xmath0 structure are first systematically searched , and then the energy of the @xmath0 structure is compared with that of the fm bcc at several c concentrations using supercell models .
finally , mechanical stability of the @xmath0 structure is analyzed based on phonon frequencies at the @xmath2 point of the supercell models .
the @xmath0 structure belongs to the hexagonal crystal system , and hence its primitive unit cell is specified by two lattice constants @xmath3 and @xmath4 .
the @xmath0 structure has three atoms inside the primitive unit cell , and their positions are @xmath5 , @xmath6 , and @xmath7 in fractional coordinates
. the @xmath0 structure can actually be obtained from the bcc structure by repeating to collapse a pair of neighboring @xmath8 } planes and to hold the next plane unaltered .
four magnetic states , which were focused on in our previous study @xcite , were investigated for the @xmath0 structure . the fm bcc structure and cementite fe@xmath9c@xmath10
were also calculated for comparison . the @xmath0-based bcc unit cell ( see fig .
1(b ) in ref .
@xcite for details ) was used for the calculation of the fm bcc structure so that computational conditions for the @xmath0 and the bcc structures are as similar to each other as possible . to systematically search possible interstitial sites for c atoms in the @xmath0 structure in fe
, we used the @xmath0 fe@xmath11c@xmath10 model composed of the @xmath12 supercell of the primitive @xmath0 unit cell for fe and an interstitial c atom .
the possible interstitial sites were searched in the following procedure .
first , we divided a primitive @xmath0 unit cell into the @xmath13 mesh and put a c atom on the mesh points that are symmetrically inequivalent to each other .
the numbers of the inequivalent points were 30 for the @xmath14 magnetic state and 21 for the other magnetic states .
then , we optimized lattice parameters and internal atomic positions of the structures of the supercell models .
some of the supercell models showed atomic positions largely deviated from those of the initial @xmath0 structure after the structural optimization .
these models were excluded from further consideration .
similarly , some models were excluded because the initially given magnetic state was broken after the structural optimization .
energies of the models were calculated based on the plane - wave basis projector augmented wave method @xcite in the framework of density - functional theory within the generalized gradient approximation of the perdew - burke - ernzerhof form @xcite as implemented in the vasp code @xcite .
a plane - wave energy cutoff of 400 ev was used .
the brillouin zones were sampled by the @xmath2-centered @xmath15 mesh per primitive @xmath0 unit cell , and the methfessel - paxton scheme @xcite with a smearing width of 0.4 ev was employed .
total energies were minimized until the energy convergences to be less than 10@xmath16 ev .
lattice parameters and internal atomic positions were optimized under zero external stress .
phonon frequencies were investigated within the harmonic approximation for a lattice hamiltonian using the finite - displacement method .
atomic displacements of 0.01 were used to obtain the second - order force constants .
phonon modes at the @xmath2 point of the supercell models were calculated to investigate their mechanical stability . when there are phonon modes with imaginary frequencies , the structure is considered to be mechanically unstable .
the phonopy code @xcite was used for these phonon calculations .
in our previous study @xcite , we made a systematic first - principles study on thermodynamical and mechanical stability of the @xmath0 structure in 27 transition elements ( sc to cu , y to ag , and lu to au ) .
only the @xmath0 structures of the group 4 elements ( ti , zr , and hf ) , the group 7 elements ( mn , tc , and re ) , and fe were found to be mechanically stable in their lowest - energy magnetic states . for the elemental @xmath0 fe ,
the @xmath1 magnetic state was the lowest in energy among the investigated magnetic states .
it was , however , 170 mev / atom higher in energy than the fm bcc fe .
this implies that the elemental @xmath0 fe is difficult to be formed from the viewpoint of thermodynamical stability . in the following ,
we investigate whether the presence of interstitial c atoms stabilizes the @xmath0 structure in fe . .
calculated energies of the @xmath0 fe@xmath11c@xmath10 models in mev/(fe atom ) .
the energies are relative to that for the fm bcc fe@xmath11c@xmath10 models where the c atom is located at an octahedral site .
the first column shows the initial position of the interstitial c atom in fractional coordinates for a primitive @xmath0 unit cell in the supercell models before the structural optimization .
`` na '' indicates that the corresponding structure could not be obtained because the optimized structure was largely deviated from the initial @xmath0 structure and/or because the initially given magnetic state was broken during the structural optimization .
[ tb : energies_c_sites ] [ cols="^,^,^,^,^",options="header " , ] table [ tb : imaginary_modes_concentration ] summarizes the number of imaginary phonon modes at the @xmath2 point of the fm and @xmath1 @xmath0 fe@xmath17c@xmath10 where the c atom is located at an octahedral site . except for the @xmath0 fe@xmath9c@xmath10 , which corresponds to 25 at.% c , all the models with interstitial c atoms have one or more imaginary phonon modes at the @xmath2 point and hence are mechanically unstable .
although the elemental @xmath1 @xmath0 fe is mechanically stable @xcite , it becomes mechanically unstable once it incorporates interstitial c atoms unless the c concentration is 25 at.% .
this result indicates that the interstitial c atoms mechanically destabilize the @xmath0 structure in fe .
although the @xmath0 fe@xmath9c@xmath10 shows no imaginary phonon modes at the @xmath2 point , the fm @xmath0 fe@xmath9c@xmath10 is still 25 mev/(fe atom ) higher in energy than the fm bcc fe@xmath9c@xmath10 and hence is thermodynamically less favorable .
furthermore , it is well known that cementite fe@xmath9c@xmath10 is thermodynamically much more favorable in such a high c concentration region .
our first - principles calculations actually show that the fm cementite fe@xmath9c@xmath10 is 111 and 135 mev/(fe atom ) lower in energy than the fm bcc fe@xmath9c@xmath10 and the fm @xmath0 fe@xmath9c@xmath10 , respectively .
the present calculations demonstrate that the @xmath0 structure is thermodynamically less favorable than the fm bcc irrespective of the c concentration up to 25 at.% .
furthermore , the fm and the @xmath1 @xmath0 structures are mechanically unstable once they incorporate c atoms unless the c concentration is 25 at.% .
these results imply the instability of the @xmath0 structure in fe - c alloys .
we conclude that the @xmath0 structure is mostly unstable and can not exist even as a metastable state in fe - c alloys . in experimental reports
@xcite , the @xmath0 structure was observed in steel as nanometer - size domains in the bcc structure or as nanometer - thick layers at twin boundaries of the bcc .
based on the present theoretical results , we suggest that the @xmath0 structure in steel is formed under special atomic constraints at twin boundaries or other interfaces .
it is actually known that twin boundaries in metallic systems sometimes show special structures such as the @xmath18 structure in cu and ag @xcite .
since the @xmath0 structure acts as the transition state of a transition pathway between the bcc and the fcc structures , the @xmath0 structure in steel may be a kind of residue of the martensitic transformation .
another possible factor may be enrichment of solute elements other than interstitial c atoms in the nanometer - scale regions .
the specimens used in the experiments actually contain si , mn , and/or cr as well as c @xcite .
finally we comment on theoretical calculations in ref .
two points are inconsistent with the present results .
first , the energy of the @xmath0 structure relative to that of the bcc structure is substantially different at 7.7 and 14.3 at.% c between ref .
@xcite and the present report .
the authors in ref .
@xcite reported that the relative energies are 142 and 78 mev/(fe atom ) at 7.7 and 14.3 at.% c , respectively .
in contrast , the present calculations show that the energies of the fm @xmath0 structure relative to those of the fm bcc are 174 and 112 mev/(fe atom ) at 7.7 and 14.3 at.% c , respectively , which are much larger than the values in ref . @xcite . since the authors in ref .
@xcite do not refer to magnetic states of the @xmath0 structure incorporating c atoms , it is difficult to discuss possible reasons for the discrepancies .
note that the present calculations demonstrate that the fm @xmath0 structure is the lowest in energy among the magnetic states above 4 at.% c , which was not reported in ref .
second , the authors of ref .
@xcite claimed that the bcc fe@xmath9c@xmath10 changed to the @xmath0 fe@xmath9c@xmath10 without energy barriers .
they did not , however , show the computational details for this issue .
contrary to their results , our fm bcc fe@xmath9c@xmath10 is mechanically stable and is 25 mev/(fe atom ) lower in energy than the fm @xmath0 fe@xmath9c@xmath10 .
stability of the @xmath0 structure in steel is investigated based on first - principles with special interests in effects of interstitial c atoms .
the energy of the @xmath0 structure is compared with that of the fm bcc , and mechanical stability is analyzed based on phonon frequencies at the @xmath2 point of supercell models .
possible interstitial sites for c atoms in the @xmath0 structure are systematically searched , and the octahedral site is found to be the most favorable among the interstitial sites .
the @xmath1 @xmath0 fe@xmath17c@xmath10 where the c atom is located at an octahedral site is the lowest in energy among the magnetic states below 4 at.% c , while the fm @xmath0 fe@xmath17c@xmath10 is the lowest above this concentration . even when c atoms are incorporated in the @xmath0 structure , it is thermodynamically less favorable than the fm bcc .
furthermore , the fm and @xmath1 @xmath0 structures are mechanically unstable once they incorporate c atoms unless the c concentration is 25 at.% .
these results indicate that interstitial c atoms destabilize the @xmath0 structure in fe - c alloys .
it is concluded that the @xmath0 structure is mostly unstable and can not exist even as a metastable state in fe - c alloys . the @xmath0 structure in steel observed in experiments
may be stabilized under special atomic constraints at twin boundaries or other interfaces .
the authors thank d. ping for providing valuable information on the observation of the @xmath0 structure in steel .
funding by the ministry of education , culture , sports , science and technology ( mext ) , japan , through elements strategy initiative for structural materials ( esism ) of kyoto university , is gratefully acknowledged . | nanometer - size domains @xmath0 structure reported in steel - - alloys by transmission electron microscopy observations motivated investigate stability in steel authors group emphasized importance @xmath0 structure on transformation between bcc fcc structures in metallic systems togo and tanaka developed search algorithm for transformation pathways revealed @xmath0 structure on transformation pathway between bcc fcc structures ikeda suggested pressure - induced phase transition between bcc fcc at high temperature along pathway authors performed systematic investigation @xmath0 structure transition elements elemental @xmath0 fe with antiparallel magnetic moments lowest in energy mechanically stable @xmath1 fe 170 mev / atom higher in energy than ferromagnetic bcc fe implies elemental @xmath0 fe difficult to be formed nanometer - size @xmath0 structure observed in steel largest difference between elemental fe steel presence of c atoms little known about effects on stability in steel report first - principles study on stability structure in steel effects interstitial c atoms four magnetic states of structure investigated. 1(c ) focused on interstitial sites for c atoms in @xmath0 structure searched energy compared with fm bcc at c concentrations using supercell models mechanical stability analyzed on phonon frequencies at @xmath2 point supercell models structure belongs to hexagonal crystal system primitive unit cell specified by lattice constants @xmath3 and @xmath4 has three atoms inside cell positions are @xmath5 @xmath6 @xmath7 in fractional coordinates structure obtained from bcc structure by neighboring @xmath8 planes hold next plane unaltered four magnetic states investigated for structure fm bcc structure and cementite fe@xmath9c@xmath10 calculated for comparison @xmath0-based bcc unit cell ( fig . 1(b ) used for calculation of fm bcc structure interstitial sites for c atoms in structure in fe used @xmath0 fe@xmath11c@xmath10 model @xmath12 supercell interstitial c atom interstitial sites searched divided @xmath0 unit cell into @xmath13 mesh put c atom on mesh points inequivalentinequivalent points 30 @xmath14 magnetic state 21 other states optimized lattice parameters internal atomic positions supercell models some models atomic positions deviated initial @xmath0 structure after optimization excluded excluded magnetic state broken energies calculated plane - wave projector augmented wave method @xcite density - functional theory gradient approximation perdew - burke - ernzerhof form plane - wave energy cutoff 400 ev brillouin zones sampled @xmath2-centered @xmath15 mesh per primitive @xmath0 unit cell methfessel - paxton scheme smearing width 0.4 ev total energies minimized until convergences less than 10@xmath16 ev lattice parameters internal atomic positions optimized under zero external stress phonon frequencies investigated harmonic lattice hamiltonian finite - displacement method atomic displacements 0.01 second - order force constants phonon modes at @xmath2 point calculated mechanical stability modes imaginary frequencies structure mechanically unstable phonopy code @xcite used phonon calculations previous study systematic study thermodynamical mechanical stability @xmath0 structure in 27 transition elements@xmath0 structures group 4 elements ti zr hf 7 elements mn tc re fe mechanically stable lowest - energy magnetic states elemental @xmath0 fe @xmath1 magnetic state lowest energy 170 mev / atom higher than fm bcc fe elemental @xmath0 fe difficult thermodynamical stability investigate interstitial c atoms stabilizes @xmath0 structure fe calculated energies @xmath0 fe@xmath11c@xmath10 models in mev/(fe atom ) energies relative fm bcc fe@xmath11c@xmath10 models c atom octahedral site first column shows initial position interstitial c atom coordinates primitive @xmath0 unit cell supercell models before structural optimization na '' indicates structure obtained deviated initial magnetic state broken optimization energies_c_sites table imaginary_modes_concentration summarizes imaginary phonon modes @xmath2 point fm @xmath1 @xmath0 fe@xmath17c@xmath10 c atom octahedral site except @xmath0 fe@xmath9c@xmath10 corresponds 25 at.models with interstitial c atoms have imaginary phonon modes at @xmath2 point mechanically unstable elemental @xmath1 @xmath0 fe mechanically stable becomes unstable interstitial c atoms unless c concentration 25 at% interstitial c atoms destabilize @xmath0 structure in fe @xmath0 fe@xmath9c@xmath10 no imaginary phonon modes at @xmath2 point @xmath0 fe 25 mev/ higher than bcc fe thermodynamically less favorable cementite fe@xmath9c more favorable in high c concentration region calculations show cementite fe 111 and 135 mev/(fe atom ) lower than fm bcc fe calculations demonstrate @xmath0 structure thermodynamically less favorable than fm bcc c concentration up to 25 at% fm @xmath1 @xmath0 structures unstable c atoms unless c concentration 25 at% imply instability of @xmath0 structure in fe - c alloys mostly unstable in fe - c alloys structure observed in steel as nanometer - size domains in layers at twin boundariestheoretical results suggest @xmath0 structure in steel formed under special atomic constraints at twin boundaries twin boundaries in metallic show special structures @xmath18 structure in cu ag @xcite @xmath0 structure transition between bcc fcc structures may residue of martensitic transformation possible factor enrichment of solute elements interstitial c atoms in nanometer - scale specimens contain si mn cr c @xcite theoretical calculations ref two points inconsistent with results energy @xmath0 structure bcc structure different at 7.7 14.3 at.% c between present report reported relative energies 142 78 mev/(fe atom ) at 14.3 present calculations show energies fm @xmath0 structure 174 112 mev/(fe atom ) at 7.7 14.3 larger than values in ref . @xcite refer to magnetic states structure incorporating c atoms difficult to discuss reasons discrepancies present calculations demonstrate fm @xmath0 structure lowest in energy among magnetic states above 4 at.% c not reported in ref authors of ref @xcite claimed bcc fe@xmath9c changed to @xmath0 without energy barriers show computational details contrary fm bcc fe@xmath9c mechanically stable 25 mev/ atom lower energy than fm @xmath0 stability @xmath0 structure in steel investigated effects interstitial c atoms energy compared with fm bcc mechanical stability analyzed phonon frequencies models interstitial sites for c atoms searched octahedral site most favorable @xmath1 @xmath0 fe@xmath17c@xmath10 c atom lowest energy states below 4 at.% c fm @xmath0 fe lowest above concentration c atoms incorporated in structure thermodynamically less favorable than fm bcc fm @xmath1 structures mechanically unstable c atoms unless c concentration 25 at.% interstitial c atoms destabilize @xmath0 structure in fe - c alloys structure mostly unstable exist metastable in fe - c alloys @xmath0 structure in steel may be stabilized under special atomic constraints at twin boundaries interfaces authors thank d.ping providing valuable information observation @xmath0 structure steel funding ministry education culture sports science technology japan elements strategy initiative structural materials kyoto university acknowledged | 0.502663 | 3,114 | 1,533 | , Saving $0.1 in GPT-4. |
67,417 | the light variations of first - overtone rr lyrae stars and contact eclipsing binaries can be difficult to distinguish .
the catalina periodic variable star catalog contains several misclassified objects , despite the classification efforts by @xcite .
they used metallicity and surface gravity derived from spectroscopic data ( from the sdss database ) to rule out binaries .
our aim is to further constrain the catalog using sdss colours to estimate physical parameters for stars that did not have spectroscopic data . | briefly , photo - met estimates the unknown physical parameters of stars by interpolating the known parameters of other stars that have very similar broad - band colours .
the method relies on two numerical algorithms : + - efficient @xmath0-nearest - neighbour finding in a four - dimensional metric colour - colour space and + - local linear regression . as with any empirical parameter estimation algorithm ,
the reliability of the entire process depends much more on the training set than on the actual numerical method . for this study
, we used an empirical training set based on sdss psf magnitudes of approximately 360000 stars .
the stellar parameters [ fe / h ] , @xmath1 and @xmath2 are the adopted weighted averages from sspp ( segue stellar parameter pipeline ) . for further details on the method
see @xcite .
the variable list of the catalina sky survey consists of 5467 stars marked as rrc .
the cross - match with the sdss dr10 cross - id tool using an 12 radius search resulted in 2762 stars . for delivering new results we took the stars with photometric measurements only ( 1732 objects ) and applied the photo - met method to estimate the surface gravity , the effective temperature and metallicity .
to distinguish between rrc and contact binary stars we applied two criteria : + - rrc stars are halo giants so they are generally expected to have lower metallicity ( [ fe / h ] @xmath3 ) and surface gravity ( @xmath4 ) than disk stars . + - furthermore according to the work of @xcite the rrc stars are concentrated between @xmath5 and @xmath6 on the @xmath7 plane , where @xmath8 statistic values is a measure of the fraction of time that an object spends below the mean magnitude .
we find 236 objects satisfied both , but 636 objects satisfied none of the criteria .
we could identify several contact binaries that were originally classified as rrc stars in the css periodic variable star catalog .
the photo - met method , however , clearly placed them outside the rrc domain set by @xcite .
we plotted in fig .
1 four light curves where the asymmetry of the minima can be clearly seen when folded with twice the assumed variation period
. however , despite our success in finding new contaminating binary stars in the sample , the large errors in the @xmath2 and [ fe / h ] determination ( up to @xmath9 dex ) and the low quality of the light curves of faint stars make this method uncertain , leaving a lot of ambiguous objects in the catalog .
this project has been supported by the lp2014 - 17 program of the hungarian academy of sciences , the nkfih k-115709 and the otka nn-114560 grants of the hungarian national research , development and innovation office .
l.m . was supported by the jnos bolyai research scholarship of the hungarian academy of sciences .
the css survey is funded by the national aeronautics and space administration under grant no .
nng05gf22 g issued through the science mission directorate near - earth objects observations program . | photo - met estimates unknown parameters stars interpolating known parameters similar colours method relies on two algorithms efficient @xmath0-nearest - neighbour finding local linear regression reliability depends on training set than numerical method used empirical training set sdss psf magnitudes of 360000 stars stellar parameters fe @xmath1 @xmath2 are adopted weighted averages from sspp pipeline details see @xcite catalina sky survey 5467 stars marked as rrc cross - match with sdss dr10 cross - id tool resulted in 2762 stars took stars with photometric measurements 1732 objects applied photo - met method to estimate surface gravity effective temperature metallicity distinguish between rrc and contact binary stars applied two criteria rrc stars are halo giants expected have lower metallicity surface gravity than disk stars rrc stars concentrated between @xmath5 and @xmath6 on @xmath7 plane @xmath8 below mean magnitude 236 objects satisfied both 636 objects satisfied none criteria several contact binaries originally classified as rrc stars in css periodic variable star catalogphoto - met method placed outside rrc domain @xcite plotted fig . 1 light curves asymmetry minima seen folded variation period success contaminating binary stars large errors in @xmath2 fe / h ] determination low quality light curves make method uncertain ambiguous objects catalog project supported by lp2014 - 17 program academy sciences nkfih k-115709 otka nn-114560 grants hungarian national research development innovation office supported jnos bolyai research scholarship academy survey funded national aeronautics space administration grant nng05gf22 science mission directorate near - earth objects observations program | 0.133483 | 656 | 333 | , Saving $0.0 in GPT-4. |
67,425 | the polarization of the cmbr scattered by galaxy clusters in the kinematic sunyaev zeldovich effect depends on the transverse velocity of the cluster .
this polarizing effect is proportional to the transverse velocity squared , and so weaker that the change in intensity due to the radial motion in the kinematic effect .
the value given by sunyaev and zeldovich , and which is frequently cited , underestimates the polarizing effect by a factor of ten .
we show furthermore that the polarization has a strong frequency dependence .
this means that the polarization should be detectable with the new generation of cmbr probes , at least for some clusters .
thus this effect offers , almost uniquely , a method of obtaining the vectorial velocity of clusters .
cosmology : theory cosmic microwave polarization | one of the main goals of observational cosmology in the coming years is the measurement of the cmbr and its fluctuations . before arriving to us ,
the cmbr photons traverse the entire observable universe .
it follows that if one can observe the effect of the interaction of these photons with matter they pass through , then one should be able to obtain crucial information about the distribution of matter in the universe . with this perspective
, sunyaev and zeldovich studied the various scattering processes that take place between black body photons and electrons in galaxy clusters .
these phenomena , which are referred to collectively as the sunyaev - zeldovich effect , are now being studied in great detail , and indeed are being used to probe cosmological structure formation ( @xcite and reference therein ) . in their paper of 1980
sunyaev and zeldovich @xcite discussed the possibility of measuring the radial velocity of a cluster of galaxies from the enhancement of the cmbr temperature in the direction of the cluster .
this effect is somewhat similar to the well known sunyaev zeldovich effect in which cmbr photons are compton scattered by the hot gas in the cluster , a process which gives rise to a distortion in the cmbr spectrum in the direction of the cluster . for this reason
the effect due to the bulk motion of the cluster has become known as the kinematic s - z effect , and the original effect as the thermal s - z effect .
the thermal effect leads to a variation in intensity of the radiation given by @xmath0 where @xmath1 is the comptonization factor ,
@xmath2 the temperature of the electrons in the cluster , @xmath3 , @xmath4 the optical depth , and @xmath5 the temperature of the cmbr . for small values of @xmath6 the change in intensity is negative , whilst for large values it is positive .
the kinematic effect gives rise to a change in intensity given by @xmath7 where @xmath8 is the radial velocity in units of the speed of light .
the sign of the fractional change in intensity does not change sign with frequency , and is positive if the peculiar radial motion is towards the observer .
the effective temperature change in independent of frequency , and given by @xmath9 in the same paper the question of the polarization that can be produced when the motion of the cluster has a component transverse to the line of sight is also discussed , and the degree of polarization induced in integrated light from the cluster is claimed to be of the order @xmath10 where @xmath11 is the transverse velocity of the cluster relative to the cmbr .
in fact , as we show below , the effect is exactly ten times larger , and has a strong frequency dependence .
( sunyaev and zeldovich also argued that there would be a contribution from higher order scattering given by @xmath12 , which could be of a comparable size for larger optical depths , but we shall not discuss this here as the latter effect is still more negligible than thought previously . )
this polarization will be in a direction perpendicular to the plane formed by the line of sight and the velocity of the cluster .
as sunyaev and zeldovich point out , potentially this polarizing effect of the kinematic s - z effect can yield a great deal of information about the peculiar velocities of galaxy clusters , and indeed in theory represents one of the few ways of determining their vectorial velocity .
peculiar velocities are thought to lie in the range 500 - 3000 km s@xmath13 .
cluster optical depths are badly known but appear to be in the range 0.01 to 0.05 , yielding a frequency integrated degree polarization due to single scattering of the order @xmath14 to @xmath15 .
such small values of polarization could be still be difficult to detect . however , the frequency dependence of this polarization is of crucial importance , firstly because this frequency dependence helps distinguish the signal from other polarizing mechanisms , and secondly because the polarization rapidly increases as a function of frequency .
although sunyaev and zeldovich did point out that the polarization should have a frequency dependence , they calculated only the frequency integrated polarization .
in this paper we look more closely at the dependence of the polarization on frequency and find that at values of @xmath16 the polarization will be orders of magnitude higher than the frequency integrated polarization .
furthermore this mechanism has a characteristic frequency dependence which should enable one to distinguish it from other polarizing mechanisms .
the origin of the polarization in the case of the kinematic s - z effect is the apparent anisotropy of the cmbr radiation seen by the cluster due to its motion relative to the cmbr frame .
although the degree of anisotropy in frequency integrated light for low values of @xmath17 is not very large , the monochromatic intensity displays much greater anisotropy .
figure [ [ anisotropy ] ] shows the variation of incident intensity as a function of the viewing direction , described by the polar angle , @xmath18 , with respect to the axis of symmetry of the radiation field ( defined by the cluster velocity ) , at different frequencies and for @xmath19 .
now consider the case where the cluster is moving perpendicularly to the line of sight .
photons incident from the forward direction and scattered towards the observer will be scattered through @xmath20 , and hence fully polarized perpendicular to the line of sight and velocity of the cluster .
similarly photons from the backward direction will be polarized in the same direction .
photons arriving perpendicular to this plane will be polarized in the opposite sense , and hence tend to cancel out the effect of the photons scattered from the forward and backward direction .
however the intensity of the incident radiation varies with the angle of incidence , @xmath18 .
it is in fact the quadrupole that produces the net polarization .
we also find that the degree of polarization depends not only on the transverse velocity , but also weakly on the absolute velocity of the cluster with respect to the cmbr in a non - degenerate way . in principle
this means that measurement of the degree of polarization at different frequencies provides a means of determining the vectorial velocity without reference either to the kinematic intensity sz effect , and independently of any knowledge of the optical depth . in practice , however , use of optical depth estimates from the thermal sz effect , and the radial component of the cluster velocity from the kinematic intensity sz effect would provide much greater precision on the vector velocity measurement . as in the case of the sunyaev zeldovich paper , we shall throughout use the thomson scattering approximation , using relativistic corrections for doppler and aberration effects .
of course , a complete treatment should consider a distribution of hot electrons , in which case the relativistic expression for electron scattering should be used .
for the points we make in this paper the simplified treatment is justified . moreover , by formulating the problem in the cluster rest frame and including all relativistic transformations and directional effects , this treatment makes itself amenable to generalisation to the full relativistic case .
the geometric situation is represented in figure [ [ geo ] ] .
the cluster has velocity @xmath21 relative to the rest frame of the cmbr , @xmath22 .
we shall assume throughout that the cmbr has a perfect isotropic black body spectrum .
the cluster will see an anisotropic but axisymmetric intensity distribution .
let @xmath23 be an observer at our position but at rest in the frame of the cluster , @xmath24 , and @xmath25 be the angle between the axis of symmetry and the line of sight .
we will first compute the stokes parameters as seen by @xmath23 .
( all the primed quantities correspond the the rest frame of the cmbr and the unprimed to the rest frame of the cluster ) .
the intensity of the cmbr in @xmath26 has a planck distribution given by : @xmath27 consider an incident photon in direction @xmath28 ( with polar and azimuthal angle @xmath18 and @xmath29 in the cluster frame @xmath24 ) and frequency @xmath30 also in the cluster frame , where the @xmath31 axis is defined by the axis of symmetry ( defined by @xmath21 ) .
the frequency of this photon in @xmath26 will be given by @xmath32 where @xmath33 .
due to relativistic abberation , the apparent direction of the photon will be different in @xmath34 and @xmath26 , but it does not matter since the radiation field is isotropic in @xmath26 .
writing the intensity in @xmath34 as @xmath35 and in @xmath36 as @xmath37 , and using the invariance property of the intensity , @xmath38 we have @xmath39 where @xmath40 . the frequency integrated intensity in the cluster frame in a direction @xmath41
is given by @xmath42 the incident and scattered photons will have the same frequency in the cluster frame .
the scattering angle is given by @xmath43 , and the unit vector perpendicular to the scattering plane by @xmath44 .
the right handed orthonormal basis is completed by @xmath45 .
the polarization is best described by the stokes parameters @xmath46 which represent the total intensity , the difference in intensity measured by polarimeter in two given perpendicular directions , the same difference but with the polarimeter rotated through @xmath47 , and the circular polarization respectively @xcite .
the incident beam will be unpolarized .
since thomson scattering produces no circular polarization , @xmath48 can be ignored .
we take an observer basis @xmath49 , where @xmath50 is equal to the direction @xmath51 of the scattered photon , and @xmath52 is perpendicular to the plane defined by @xmath21 and the line of sight ( @xmath50 ) .
this defines the orientation of the polarimeter . to calculate the contribution of all incident photons we have to rotate from the scattering plane to the observer frame , and integrate over all solid angles , @xmath53 , and along the line of sight .
we denote the angle of this rotation by @xmath54 ( see figure [ [ geo ] ] ) . using the phase function for thomson scattering , and introducing for convenience the abbreviations @xmath55 ( resp .
@xmath56 ) to denote the sine ( resp .
cosine ) of the variable @xmath6 , we may write @xmath57 in order to obtain the frequency integrated stokes parameters , @xmath58 should be replaced by @xmath59 in equations ( [ eq1 ] ) .
the optical thickness , @xmath4 , is given by an integral along the line of sight @xmath60 , where @xmath61 is the electron density and @xmath62 the thomson cross - section .
noting that @xmath63 , @xmath64 and @xmath65 and integrating over @xmath29 , equations ( [ eq1 ] ) reduce to @xmath66 the total intensity seen by @xmath67 is then given by @xmath68 where @xmath69 is the direct light , @xmath70 is the light scattered toward the observer and @xmath71 is the light scattered out of the line of sight .
the degree of polarization , @xmath72 , of the net light seen by @xmath67 is then given by @xmath73 all the above quantities are expressed for the observer @xmath67 .
for an observer @xmath74 at rest in the cmbr frame we have @xmath75 and @xmath76 where @xmath77 .
the fraction change in intensity given by @xmath78 describe the kinematic sz effect and , of course , coincide with the result of sunyaev and zeldovich @xcite in the case of radial motion @xmath79 .
we can now compute the frequency integrated degree of polarization by simply replacing @xmath80 by @xmath59 in equations ( [ eq2 ] ) . keeping only terms up to @xmath81 in the expansion of @xmath59 ( eq . [ temp ] ) and taking @xmath82 and
we obtain @xmath83 where we have introduced the transverse velocity @xmath11 .
this results has the same velocity and optical depth dependence as the one obtained by sunyaev and zeldovich , but is ten time greater . since @xmath72 is lorentz invariant this old also in the cmbr frame .
furthermore , as we shall see in the next section , the polarization has a strong frequency dependence . in the low frequency limit ,
the degree of polarization is independent of the frequency and is given by @xmath84 this results is similar to what was claimed by sunyaev and zeldovich for the frequency integrated effect .
in this section we present the degree of polarization that should be observed for a typical galaxy cluster .
all our results are given for @xmath85 but values for other optical depths can simply be scaled linearly .
figure [ [ pol_nu ] ] displays the degree of polarization against frequency for various transverse cluster speeds .
we can see that @xmath72 slightly depends on the inclination even for constant values of @xmath11 . as mentioned previously
, the polarization reaches a constant value in the raleigh - jeans ( low frequency ) limits according to formula [ [ lim_low ] ] .
the variation of @xmath72 with frequency is very strong . at @xmath86
the polarization is more than two orders of magnitude greater than the lower limit . in the planck higher frequency band , the expected polarization for @xmath87 and @xmath88
are of @xmath89 and @xmath90 respectively , although these need to be scaled according to the optical depth .
such degrees of polarization are above the detection limits foreseen by coming experiments .
figure [ [ pol_nu ] ] display the sensitivity of the individual polarimetric detector of the planck experiment .
since there are between 8 and 34 polarimetric detectors per band and since a cluster could have an angular size sensibly larger than the instrument resolution , it should be possible to get the polarimetric signal by proper processing of the data .
furthermore , new experiments could be designed especially for polarimetric measurement .
figure [ [ pol_beta ] ] displays the degree of polarization against @xmath17 for different frequencies . for all frequencies , as for the integreted effect
, the degree of polarization has a @xmath81 dependence .
we can see that the polarization due to thomson scattering in the cluster is much larger that the expected primordial polarization for a wide range of parameter and should therefore be measurable .
transverse velocities of galaxy clusters will produce polarized signal in the scattered cmbr
. this polarization will be perpendicular to the cluster velocity and the light of sight .
although in frequency integrated light the level of polarization might be difficult to detect with the techniques presently available , the polarization has a strong frequency dependence , increasing towards higher frequencies . for values of transverse velocity of 1000 km s@xmath13 , and an optical depth of 0.02 ,
the degree of polarization at 857 ghz is around @xmath91 .
since the polarization depends on the square of the transverse velocity , and linearly on the optical depth , certain clusters should show higher polarization .
other sources of polarization , in particular dust and synchrotron radiation , will contaminate the signal .
however filtering techniques that take into account the characteristic frequency dependence of this polarization should allow one to extract the signal due to the kinematic effect .
measured polarization could then be used to infer transverse velocities of clusters .
this question will be elaborated in future work .
birkinshaw , m. , 1998 , to appear in physics reports , astro - ph/9808050 chandrasekhar , s. , 1950 , radiative transfer , clarendon press , oxford hu , w. and white , m. , 1997 , new astronomy , * vol . 2 * , no 4 . , p323 sunyaev r.a . and zeldovich y.b . , 1980 , mnras , * 190 * , 413 - 420 | main goals of observational cosmology is measurement of cmbr fluctuations before cmbr photons traverse observable universe interaction photons with matter obtain crucial information about distribution of matter in sunyaev and zeldovich studied scattering processes between black body photons electrons in galaxy clusters phenomena as sunyaev - zeldovich effect now studied in detail used to probe cosmological structure formation in paper 1980 sunyaev zeldovich discussed possibility measuring radial velocity of cluster galaxies from enhancement cmbr temperature cluster effect similar to sunyaev zeldovich effect cmbr photons scattered by hot gas in cluster distortion in cmbr spectrum effect due to bulk motion known as kinematic s - z effect original effect thermal s - z effect thermal effect leads to variation in intensity of radiation given by @xmath0 comptonization factor temperature of electrons optical depth temperature of cmbr for small values of @xmath6 change in intensity negative large values positive kinematic effect change in intensity given by @xmath7 @xmath8 is radial velocity in speed of lightfractional change in intensity change with frequency positive if radial motion towards observer effective temperature change independent of frequency given by @xmath9 polarization when motion cluster transverse to line of sight discussed degree of polarization in integrated light from cluster order @xmath10 @xmath11 is transverse velocity cluster relative to cmbr effect ten times larger strong frequency dependence sunyaev zeldovich argued contribution from higher order scattering by @xmath12 comparable size for larger optical depths effect negligible polarization in direction perpendicular to plane by line of sight and velocity cluster polarizing effect of kinematic s - z effect can yield information about peculiar velocities of galaxy clusters determining vectorial velocity peculiar velocities in range 500 - 3000 km s@xmath13 cluster optical depths range 0.01 to 0.05 frequency integrated degree polarization due to single scattering @xmath14 to @xmath15 small values of polarization difficult to detect frequency dependence of polarization crucial importance signal from other polarizing mechanisms polarization increases as function of frequencysunyaev zeldovich polarization frequency dependence calculated frequency integrated polarization this paper dependence polarization on frequency at values @xmath16 polarization higher than frequency integrated polarization mechanism characteristic frequency dependence from other polarizing mechanisms origin of polarization kinematic s - z effect is apparent anisotropy of cmbr radiation cluster due to motion cmbr frame anisotropy in frequency integrated light for low values of @xmath17 not large monochromatic intensity displays greater anisotropy figure anisotropy shows variation of incident intensity viewing direction polar angle @xmath18 axis symmetry radiation field cluster velocity at different frequencies @xmath19 cluster moving perpendicularly to line of sight photons from forward direction observer through @xmath20 fully polarized perpendicular to line of sight velocity cluster photons backward direction polarized same direction photons perpendicular polarized opposite cancel out effect of photons intensity incident radiation varies with angle of incidence @xmath18 quadrupole produces net polarizationpolarization depends on transverse velocity absolute velocity cluster cmbr measurement of polarization at different frequencies vectorial velocity without reference kinematic intensity sz effect independently optical depth use of optical depth estimates from thermal sz effect radial component cluster velocity from kinematic intensity sz effect greater precision on vector velocity measurement use thomson scattering approximation relativistic corrections for doppler aberration effects complete treatment distribution of hot electrons relativistic expression for electron scattering used simplified treatment justified formulating problem in cluster rest frame including relativistic transformations directional effects treatment amenable to generalisation to full relativistic case geometric situation represented in figure [ geo ] ] cluster has velocity @xmath21 relative to rest frame cmbr @xmath22 cmbr has isotropic black body spectrum cluster anisotropic axisymmetric intensity distribution @xmath23 be observer rest frame @xmath24 @xmath25 be angle between axis of symmetry line of sight compute stokes parameters as seen by @xmath23 primed quantities correspond rest frame cmbr unprimed to rest frame clusterintensity cmbr @xmath26 planck distribution @xmath27 incident photon @xmath28 polar azimuthal angle @xmath18 @xmath29 cluster frame @xmath24 frequency @xmath30 @xmath31 axis defined by axis symmetry @xmath21 frequency photon @xmath26 by @xmath32 @xmath33 relativistic abberation apparent direction photon different @xmath34 @xmath26 radiation field isotropic @xmath26 intensity @xmath34 @xmath35 @xmath36 @xmath37 invariance property intensity @xmath38 @xmath39 @xmath40 frequency intensity cluster frame @xmath41 by @xmath42 incident scattered photons same frequency scattering angle @xmath43 unit vector perpendicular scattering @xmath44 right handed orthonormal basis @xmath45 polarization stokes parameters @xmath46 total intensity difference intensity measured polarimeter perpendicular directions polarimeter rotated @xmath47 circular polarization incident beam unpolarized thomson scattering no circular polarization @xmath48 ignoredobserver basis @xmath49 @xmath50 equal to direction scattered photon @xmath52 perpendicular to plane @xmath21 line of sight @xmath50 defines orientation polarimeter calculate contribution incident photons rotate from scattering plane to observer frame integrate over solid angles @xmath53 line of sight denote angle rotation by @xmath54 figure phase function for thomson scattering abbreviations @xmath55 @xmath56 ) sine variable @xmath6 write @xmath57 frequency integrated stokes parameters @xmath58 replaced by @xmath59 equations optical thickness @xmath4 by integral line of sight @xmath60 @xmath61 electron density @xmath62 thomson cross - section @xmath63 @xmath65 integrating over @xmath29 equations reduce to @xmath66 total intensity @xmath67 given by @xmath68 @xmath69 direct light light scattered observer light scattered out of line sight degree polarization net light by @xmath73 quantities expressed for observer @xmath67 observer @xmath74 rest frame @xmath75 @xmath76 @xmath77fraction change in intensity @xmath78 kinematic sz effect with sunyaev and zeldovich radial motion @xmath79 compute frequency integrated degree polarization replacing @xmath80 by @xmath59 in equations terms up to @xmath81 expansion @xmath59 @xmath82 @xmath83 introduced transverse velocity @xmath11 same velocity optical depth dependence as sunyaev and zeldovich ten time greater @xmath72 is lorentz invariant in cmbr frame polarization strong frequency dependence low frequency limit polarization independent of frequency by @xmath84 similar to sunyaev and zeldovich frequency integrated effect degree of polarization for typical galaxy cluster results for @xmath85 other optical depths scaled linearly figure displays polarization against frequency for transverse cluster speeds @xmath72 depends on inclination for constant values of @xmath11 polarization reaches constant value in raleigh - jeans low frequency limits variation of @xmath72 with frequency strong at @xmath86 polarization more than two orders of magnitude greater than lower limitplanck higher frequency band expected polarization for @xmath87 @xmath88 @xmath89 @xmath90 need scaled optical depth degrees polarization above detection limits figure pol_nu sensitivity polarimetric detector planck experiment 8 34 polarimetric detectors per band cluster angular size larger than instrument resolution possible to get polarimetric signal by processing data new experiments designed for polarimetric measurement figure pol_beta displays polarization against @xmath17 different frequencies polarization @xmath81 dependence polarization due to thomson scattering cluster larger that expected primordial polarization measurable transverse velocities of galaxy clusters produce polarized signal scattered cmbr polarization perpendicular to cluster velocity light of sight frequency light polarization difficult detect strong frequency dependence increasing towards higher frequencies transverse velocity 1000 km s@xmath13 optical depth 0.02 polarization at 857 ghz around @xmath91 polarization depends on transverse velocity optical depth certain clusters show higher polarization other sources polarization dust synchrotron radiation contaminate signalfiltering techniques frequency dependence polarization extract signal kinematic effect measured polarization infer transverse velocities clusters question elaborated future work birkinshaw 1998 physics reports astro - ph/9808050 chandrasekhar 1950 radiative transfer clarendon press oxford white 1997 new astronomy vol . 2 no 4 p323 sunyaev r zeldovich y.b 1980 mnras 190 413 - 420 | 0.569678 | 3,437 | 1,760 | , Saving $0.1 in GPT-4. |
67,432 | we review recent advances concerning helioseismology , solar models and solar neutrinos .
particularly we address the following points : i ) helioseismic tests of recent ssms ; ii ) predictions of the beryllium neutrino flux based on helioseismology ; iii ) helioseismic tests regarding the screening of nuclear reactions in the sun . | helioseismology has added important data on the solar structure which provide severe constraint and tests of ssm calculations .
for instance , helioseismology accurately determines the depth of the convective zone @xmath0 and the photospheric helium abundance @xmath1 . with these additional constraints
there are essentially no free parameters for ssm builders .
in this paper we review recent advances concerning helioseismology , solar models and solar neutrinos .
particularly we shall address the following points : i ) helioseismic tests of recent ssms ; ii ) prediction of the beryllium neutrino flux based on helioseismology ; iii ) helioseismic tests concerning the screening of nuclear reactions in the sun .
while we refer to e.g. @xcite for a review of the method and to @xcite for the data , we recall that by measurements of thousands of solar frequencies ( p - modes ) with a typical accuracy of @xmath2 , one derives : * _ properties of the present convective envelope _ , such as depth and helium abundance : @xmath3 the quoted errors , mostly resulting from systematic uncertainties in the inversion technique , have been estimated conservatively by adding linearly all known individual uncertainties , see @xcite . if uncertainties are added in quadrature , the global error is about one third of that indicated in eqs .
( [ eqprop1],[eqprop2 ] ) , see again @xcite .
this latter procedure was also used by bahcall et al .
@xcite with similar results .
this yields the so called `` @xmath4 '' errors .
we shall refer to the conservative estimate as the `` @xmath5 '' determination .
we remark however that this terminology is part of a slang , and it does not correspond to well defined confidence levels , as one has to combine several essentially systematic errors . * _ sound speed profile . _ by inversion of helioseismic data one can determine the sound speed in the solar interior . this analysis can be performed in terms of either the isothermal sound speed squared , @xmath6 , or in terms of the adiabatic sound speed squared @xmath7 , as the coefficient @xmath8 is extremely well determined by means of the equation of state of the stellar plasma . + in fig .
[ fig1 ] we show the helioseismic value of @xmath9 as a function of the radial coordinate @xmath10 .
the typical `` @xmath5 '' errors are of order @xmath11 in the intermediate solar region , @xmath12 , and increase up to @xmath13 near the solar center .
fig . [ fig2 ] compares the results of six different observational determinations of the sound speed with the results of the best solar model of ref .
@xcite , hereafter bp2000 .
this figure suggests several comments : i)different measurements yield quite consistent value of the sound speed , to the level 0.1% ; \ii ) the solar model bp2000 is in agreeement with helioseismic data to better than 0.5% at any depth in the sun .
we remark that also the properties of the convective envelope predicted by bp2000 ( @xmath14 ) are in agreement with helioseismic determinations , see eqs .
( [ eqprop1],[eqprop2 ] ) .
\iii ) on the other hand , the predicted sound speed differs from the helioseismic determination at the level of 0.3 - 0.4% _ just below the convective envelope_. concerning this last point , we remark that the difference is however at the level of the `` @xmath5 '' uncertainty of the helioseismic determination , as it is shown in fig . [ fig3 ] .
nevertheless it can be taken as an indication of some imperfection of the ssm .
a marginally better agreement can be obtained in solar models including mixing induced by rotation , see e.g. @xcite and in models including macroscopic transport term , see e.g. @xcite . in summary
all this means that ssm predictions are accurate to the level of one per cent or better , although there are indications of some deficiencies at the level of per mille .
concernig neutrino physics , the properties of the central solar region , where nuclear reactions are efficient , are relevant .
specifically , boron and beryllium neutrinos are produced very near to the solar center with maximal production rates respectively at @xmath15 and @xmath16 .
since the p - modes which are observed do not propagate ( actually are exponentially damped ) so deeply in the sun the question often arises if present helioseismic data can determine the sound speed in region of beryllium and boron production . from an extensive analysis of the inversion method and of data available at that time we already concluded in @xcite that @xmath17 is determined with a `` @xmath4 '' accuracy of 1%
this point was further elucidated in @xcite where a simplified analysis was presented in order to produce convincing evidence that helioseismology fixes the sound speed near the solar center with such an accuracy .
as well known , the production of neutrinos from @xmath18 is an important item in the context of the so called `` solar neutrino puzzle '' for several reasons : + i ) the result of gallium experiments would be ( partially ) consistent with the hypothesis of standard neutrinos if beryllium neutrino ( @xmath19 ) production rate , @xmath20 , is suppressed by an order of magnitude with respect to the prediction of the standard solar model ( ssm ) , see e.g. @xcite .
+ ii ) if one accepts neutrino oscillations as the solution of the solar neutrino puzzle , the determination of the neutrino mass matrix depends however on the predicted value of @xmath20 .
+ iii ) direct experiments aiming to the determination of the @xmath19 signal are in preparation and of course the interpretation of their result will rely on @xmath20 @xcite .
the ssm prediction for beryllium neutrinos @xcite , @xmath21 is very robust , much more than that of boron neutrinos . however any additional information which does not rely on ssm are clearly welcome . in ref .
@xcite a lower limit on the beryllium neutrino flux on earth was found , @xmath22 @xmath23 s@xmath24 , in the absence of oscillations , by using helioseismic data , the b - neutrino flux measured by superkamiokande and the hydrogen abundance at the solar center @xmath25 predicted by standard solar model ( ssm ) calculations .
we remark that this abundance is the only result of ssms needed for getting @xmath26 .
lower bounds for the gallium signal , @xmath27 snu , and for the chlorine signal , @xmath28 snu , in the absence of oscillations have also been derived .
they are about @xmath5 above the corresponding experimental values , @xmath29 snu @xcite and @xmath30 snu @xcite .
we remark that predictions for @xmath25 are very stable among different ( standard and non standard ) solar models , see @xcite .
in fact @xmath25 is essentially an indicator of how much hydrogen has been burnt so far .
the stability of @xmath25 corresponds to the fact that any solar model has to account for the same present and time integrated solar luminosity . in ref .
@xcite a step forward was made ; a determination of @xmath20 was directly obtained by means of helioseismology without using additional assumptions .
the basic idea is the following . in the ssm the pp - ii termination ( which is the beryllium producting branch )
accounts for an appreciable fraction of the @xmath31he produced near the solar center .
if beryllium production is suppressed now and in the past less @xmath31he would have been produced near the center . as a consequence ,
the molecular weight should decrease and the sound speed should increase in this region . in other words , we know that ssm calculations are in good agreement with helioseismology and we can expect that this agreement is spoiled if @xmath19 production is substantially altered . in order to test this idea , solar models with artifically changed @xmath19 production were constructed . an efficient way for producing arbitrary variations of @xmath20 is to vary the zero energy astrophysical s - factor ( @xmath32 ) for the reaction @xmath33 , since , as well known ( see e.g. @xcite ) , the production rate @xmath20 is directly proportional to @xmath32 : @xmath34 we remind that @xmath32 is measured with an accuracy of about ten per cent , ( @xmath35 kevb @xcite ) .
it has been varied , however , well beyond its experimental uncertainty in order to simulate several effects which have been claimed to suppress @xmath19 production , e.g. hypothetical plasma effects which could alter nuclear reaction rates .
the resulting values for the quantities which can be tested by means of helioseismology are shown in fig .
[ fig4 ] and fig .
[ fig5 ] , for several assumed values of the parameter @xmath36 .
the photospheric helium abundance @xmath1 is weekly sensitive to the value of @xmath32
whereas the depth of the convective envelope @xmath0 is altered by more than @xmath4 if @xmath32 is reduced below one half of the ssm value , i.e. : @xmath37 as previously mentioned , one expects that the sound speed is altered , particularly near the solar center .
in fact , stringent constraints arise from the sound speed profile , particularly near @xmath38 .
the requirement that the sound speed is not changed by more than @xmath4 yields : @xmath39 in conclusion , helioseismology directly confirms the production rate of beryllium neutrinos as predicted by ssms to within @xmath40 ( @xmath4 error ) .
this constraint is somehow weaker than that estimated from uncertainties of the ssm , see eq.[eqlbessm ] , however it relies on direct observational data .
the study of screened nuclear reaction rates was started with the pioneering work of salpeter @xcite , who discussed both the extreme cases of `` weak '' and `` strong '' screening , providing suitable expressions for the screening factors @xmath41 the solar core is not far from the weak screening case , however it does not satisfy the usual conditions under which the weak screening approximation holds .
this is the reason why the possibilities of large deviations from weak screening have been investigated by several authors , expecially as an attemp to avoid or mitigate the `` solar neutrino problem '' , see e.g. @xcite and references therein .
we remind that solar models are built by using stellar evolutionary codes which include specific expressions for the nuclear reaction rates . if one uses different formulas for the screening factors @xmath42 one obtains different solar models . on the other hand
, helioseismology provides precise information on the sound speed profile and on the properties of the convective envelope which have to be reproduced by the correct solar model .
therefore helioseismology can potentially provide a test of screening models . in ref .
@xcite , solar models using different screening assumptions were constructed and compared with helioseismic data . specifically , four different model were considered : \i ) the weak screening approximation ( wes ) .
the screening factors @xmath42 are given by : @xmath43 where @xmath44 are the charges of the interacting nuclei , @xmath45 is the temperature and @xmath46 is the debye radius . as clear from equation above ,
the screening factors are always larger than unity , i.e. the plasma provides enhancement of the thermonuclear reaction rates .
\ii ) the mitler result @xcite ( mit ) , obtained with an analytical method which goes beyond the linearized approach and which correctly reproduces both the limits of weak and strong screening .
\iii ) neglect completely any screening effect ( nos ) , i.e. nuclear reactions occur with rates @xmath47 .
this case is considered in connection with the suggestions that screening can be much smaller than salpeter s estimate , see e.g. @xcite .
\iv ) the tsytovich model ( tsy ) @xcite , which provides a decrease of all the thermonuclear reaction rates with respect to the case of bare nuclei . .screening factors in solar center , for weak screening ( wes ) @xcite , mitler model ( mit ) @xcite , no screening ( nos ) and tsytovitch model ( tsy ) @xcite .
[ cols="<,^,^,^,^",options="header " , ] [ tabmod ] \i ) the difference between the tsytovitch model ( tsy ) and the weak screening model ( wes ) exceeds the conservative " uncertainty on @xmath9 in a significant portion of the solar profile .
we remark that also the depth of the convective envelope is significantly altered . in other words the anti - screening predictions of ref .
@xcite can be excluded by means of helioseismology .
\ii ) also the difference between the no - screening model ( nos ) and wes is significant for both @xmath9 and @xmath0 in comparison with helioseismic uncertainty . in other words
the existence of a screening effect can be proved by means of helioseismology .
\iii ) the mitler model of screening ( mit ) can not be distinguished from the weak screening model within the present accuracy of helioseismology .
we summarize here that main points of our discussion : \i ) the comparison of recent ssms with helioseismic measurements shows that ssms prediction are accurate to the level of one per cent or better , although there are indications of some deficiencies at the level of per mille .
\iii ) models for screening of nuclear reactions in the sun can be tested by means of helioseismology .
in particular , the anti - screening predictions of @xcite can be excluded .
in addition , the existence of a screening effect can be proved by helioseismology .
lowl1+bison : s. basu et al . ,
_ mnras _ * 292 * ( 1997 ) 234 ; golf : s. turck - chieze et al . , _ solar phys . _
* 175 * ( 1997 ) 247 ; gong : j.harvey et al . ,
_ science _ * 272 * ( 1996 ) 1284 ; mdi : e.j .
rhodes et al . _
solar phys . _
* 175 * ( 1997 ) 287 ; lowl1 : s. basu et al .
_ * 460 * ( 1996 ) 1064 ; golf2 : thierry et al . _
a&a _ * 355 * ( 2000 ) 743 . | helioseismology added data on solar structure constraint tests ssm calculations determines depth convective zone photospheric helium abundance constraints no free parameters for ssm builders paper review advances helioseismology solar models solar neutrinos address points helioseismic tests of recent ssms prediction beryllium neutrino flux helioseismic tests screening nuclear reactions sun refer to @xcite for review method data measurements of solar frequencies accuracy @xmath2 derives properties of convective envelope depth helium abundance quoted errors from systematic uncertainties in inversion technique estimated conservatively by adding linearly all known individual uncertainties if uncertainties added in quadrature global error about one third of indicated in eqs . eqprop1] procedure used by bahcall et al . @xcite similar results yields `` @xmath4 '' errors refer to conservative estimate as `` @xmath5 '' determination terminology slang not correspond to confidence levels systematic errors sound speed profile by inversion of helioseismic data determine sound speed in solar interioranalysis performed isothermal sound speed squared @xmath6 or adiabatic sound speed squared @xmath7 coefficient @xmath8 determined by equation of state of stellar plasma fig helioseismic value of @xmath9 function of radial coordinate @xmath10 typical @xmath5 '' errors @xmath11 in intermediate solar region @xmath12 increase to @xmath13 near solar center fig fig2 compares six observational determinations sound speed with best solar model of @xcite bp2000 measurements yield consistent value sound speed 0.1% solar model bp2000 with helioseismic data better than 0.5% at any in sun properties convective envelope predicted by bp2000 ( @xmath14 ) agreement with helioseismic determinations predicted sound speed differs from helioseismic determination 0.3 - 0.4% below convective envelope difference at level `` @xmath5 '' uncertainty of helioseismic determination shown in fig fig3 indication of imperfection of ssm better agreement in solar models including mixing induced by rotation models macroscopic transport termssm predictions accurate to one per cent or deficiencies at per mille neutrino physics properties central solar region nuclear reactions efficient relevant boron beryllium neutrinos produced near to solar center maximal production rates at @xmath15 and @xmath16 p - modes not propagate damped deeply in sun question arises if helioseismic data determine sound speed in beryllium boron production analysis concluded @xmath17 determined with @xmath4 '' accuracy of 1% elucidated in @xcite analysis evidence helioseismology fixes sound speed near solar center production of neutrinos from @xmath18 important solar neutrino puzzle '' result of gallium experiments partially consistent with hypothesis standard neutrinos if beryllium neutrino ( @xmath19 ) production rate @xmath20 suppressed by order of magnitude standard solar model if neutrino oscillations as solution solar neutrino puzzle determination of neutrino mass matrix depends on predicted value of @xmath20 direct experiments determination @xmath19 signal in preparation interpretation result on @xmath20 @xcite ssm prediction for beryllium neutrinos @xcite @xmath21 robust more than boron neutrinos additional information ssm welcome in @xcite lower limit on beryllium neutrino flux found, @xmath22 @xmath23@xmath24 helioseismic data b - neutrino flux superkamiokande hydrogen abundance at solar center @xmath25 predicted by standard solar model ssm calculations abundance only result of ssms for @xmath26 lower bounds for gallium signal @xmath27 chlorine signal @xmath28 snu derived about @xmath5 above experimental values @xmath29 @xmath30 snu predictions for @xmath25 stable among solar models indicator of hydrogen burnt stability of corresponds to solar model for same present time integrated solar luminosity in @xcite step forward determination of @xmath20 obtained by helioseismology without additional assumptions pp - ii termination beryllium producting branch accounts for of @xmath31he produced near solar center if beryllium production suppressed less @xmath31he produced near center molecular weight should decrease sound speed increasessm calculations with helioseismology spoiled if @xmath19 production altered test solar models with changed @xmath19 production constructed variations of @xmath20 vary zero energy astrophysical s - factor ( @xmath32 ) for reaction @xmath33 production rate @xmath20 proportional to @xmath32 @xmath32 measured with accuracy ten per cent varied beyond uncertainty to simulate effects suppress @xmath19 production plasma effects nuclear reaction rates resulting values for quantities tested by helioseismology shown in fig and fig5 for assumed values parameter @xmath36 photospheric helium abundance @xmath1 weekly sensitive to value @xmath32 depth of convective envelope @xmath0 altered by more than @xmath4 if @xmath32 reduced below half ssm value sound speed altered particularly near solar center stringent constraints from sound speed profile near @xmath38 sound speed not changed by more than @xmath4 yields helioseismology confirms production rate of beryllium neutrinos to within @xmath40 ( @xmath4 error )constraint weaker than estimated from uncertainties ssm relies on observational data study of screened nuclear reaction rates started with work salpeter @xcite discussed extreme cases of weak and strong screening suitable expressions for screening factors @xmath41 solar core not far from weak screening case satisfy usual conditions screening large deviations from weak screening investigated by avoid solar neutrino problem solar models built stellar evolutionary codes include specific expressions for nuclear reaction rates different formulas for screening factors different solar models helioseismology provides precise information on sound speed profile properties convective envelope reproduced by solar model can test of screening models in @xcite solar models using different screening assumptions constructed compared with helioseismic data four model considered weak screening approximation screening factors @xmath42 given by @xmath43 @xmath44 charges of interacting nuclei temperature debye radius screening factors always larger than unity plasma provides of thermonuclear reaction rates mitler result @xcite ( obtained with analytical method correctly reproduces limits of weak and strong screeningneglect screening effect ( nos ) nuclear reactions occur with rates @xmath47 case considered with suggestions screening smaller than salpeter s estimate see @xcite tsytovich model ( tsy ) @xcite decrease of thermonuclear reaction rates bare nuclei.screening factors in solar center weak screening mitler model no screening nos tsytovitch model tsy difference between tsytovitch model and weak screening model exceeds conservative uncertainty on @xmath9 in solar profile depth convective envelope altered anti - screening predictions of @xcite excluded by helioseismology difference between no - screening model nos and wes significant for @xmath9 and @xmath0 helioseismic uncertainty existence screening effect proved by helioseismology mitler model of screening mit not distinguished from weak screening model accuracy helioseismology comparison of recent ssms with helioseismic measurements shows ssms prediction accurate to one per cent or better deficiencies at per mille models for screening of nuclear reactions in sun tested by helioseismology anti - screening predictions of @xcite excludedscreening effect proved helioseismology lowl1+bison s. basu mnras 292 1997 ) 234 golf : s. turck - chieze solar phys. 175 1997 ) 247 gong : j.harvey science 272 1996 ) 1284 mdi e.j rhodes solar 175 1997 ) 287 lowl1 s. basu 460 1996 ) 1064 golf2 : thierry a&a 355 ( 2000 ) 743 | 0.506626 | 3,228 | 1,635 | , Saving $0.1 in GPT-4. |
67,433 | the study of scattering and extinction properties of possible nanodiamond grains in the ism are reported .
calculations using discrete dipole approximation ( dda ) for varying ellipsoidal shapes and sizes from @xmath0 to @xmath1 are considered .
nanodiamonds show negligible extinction from ir to near - uv and very sharp far - uv rise .
comparison with observations rule out possibility of independent nanodiamond dust but point towards possibility of nanodiamonds as a component in the ism .
radiation induced transformations may lead to carbonaceous grains with different core and mantles .
so calculations are also performed for a core - mantle target model with nanodiamond core in graphite mantles .
the graphite extinction features get modified with the peak at 2175 being lowered , broadened , blue shifted and accompanied by enhanced extinction in the far - uv .
such variations in the 2175 band and simultaneous far - uv rise are observed along some sources . a three component dust model incorporating silicate , graphite and graphite with nanodiamond core is also considered .
the model extinction compares very well with the average galactic extinction in the complete range from @xmath2 to @xmath3 .
the best fit requires small size and small number of nanodiamonds .
[ firstpage ] interstellar medium
nanodiamonds , extinction , dda . | to account for the ultraviolet extinction in the diffuse interstellar medium @xcite first proposed the possibility of diamonds in interstellar medium ( ism ) .
the 3.43 and 3.53 @xmath4 emission bands in circumstellar media of ae / be herbig stars hd 97048 and elias 1 show convincing presence of nanodiamonds @xcite .
@xcite attribute these bands to c - h stretching in hydrogenated nanodiamond , as distinct from the 3.3 @xmath5 m pah feature .
the 3.47 @xmath4 feature in absorption toward a number of proto - stars is attributed to the tertiary c - h stretching mode in diamond like structures @xcite .
spectra of diamondoid molecules @xcite show that nanodiamonds a few nanometre in size could be responsible for the 3.53 and 3.43 @xmath5 m emission lines and smaller diamondoids give the 3.47 @xmath5 m absorption feature . cosmic nanodiamonds are also detected in primitive carbonaceous meteorites of presolar origin @xcite .
in fact , nanodiamonds are considered to be the most abundant presolar component in meteorites @xcite .
@xcite studied @xmath6 stretching mode of nanodiamonds extracted from orgueil meteorite and observed the above mentioned ir bands .
if the ism nanodiamonds are not hydrogenated their detection is hard @xcite and they could be more abundant than estimated by the observations of infrared @xmath6 bands @xcite .
it is , therefore , important to incorporate nanodiamonds in uv extinction models for which their independent scattering and extinction properties need to be understood .
the extinction properties of nanodiamond dust have been reported @xcite for spherical shape and the far - uv break of spectral energy in quasars is attributed to nanodiamond dust @xcite . in this communication extinction properties of nanodiamond grains of different ellipsoidal shapes and sizes are reported . to study the effect of nanodiamond within graphite extinction efficiencies
are also reported for nanodiamond - graphite as core - mantle grains .
an extinction curve modeling is proposed including nanodiamond as a component .
various aspects of evolution of carbon in the ism have been discussed in literature @xcite . it is generally assumed that the ism carbon is primarily hydrogenated amorphous carbon ( hac ) that transforms to graphite and diamond like structures in the presence of uv radiation field @xcite .
regions with low uv radiations , such as dense clouds and giant star atmospheres , graphitic particles may be absent @xcite .
the estimated time - scale of graphitization varies from few weeks for circumstellar grains @xcite to @xmath7 years for interstellar grains @xcite .
nanodiamonds can result from annealing of the carbonaceous material by uv radiation e.g. in supernova ejecta @xcite .
observations show that the hydrogenated nanodiamond features arise in the inner regions close to the central star , having higher uv flux @xcite .
surface energy studies show that small diamonds are more stable against thermal evaporation and chemical attack compared to graphite particles @xcite .
formation of nanodiamonds has been studied in laboratory giving insight to possible routes of their formation in ism .
nanodiamond formation is seen to result from carbon vapour deposition ( cvd ) @xcite , detonation of carbon based explosives @xcite , electron irradiation of graphitic shells ( carbon onions ) @xcite and ion beam irradiation of amorphous carbon films @xcite .
@xcite on the basis hrtem studies suggest that diamonds and graphite few nanometre in size are formed by nucleation in organic ice mixture subjected to uv photolysis .
structural rearrangement between diamond and graphite is also possible .
graphite subjected to annealing / irradiation may go to shelled onion like form @xcite or transform into nanodiamond @xcite .
laboratory experiments also show that the @xmath8 ratio changes by different doses of uv radiation @xcite .
chemical conversion of pah clusters to nanodiamonds is proposed by @xcite .
partial graphitization of nanodiamond due to heat @xcite or pressure conditions @xcite is suggested on the basis of some experimental studies .
theoretical analysis by @xcite show that surface graphitization of nanodiamond under strong radiation field can lead to core - mantle like shell structure with upto 80% graphitization .
thus nanodiamonds are possible within ism carbonaceous matter and would modify the optical properties of dust .
graphitic mantle will make the nanodiamond crystals chemically less active and hard to detect spectroscopically .
to calculate the extinction , scattering , absorption and polarization efficiencies the discrete dipole approximation ( dda ) is used @xcite .
the ddscat 6.1 ( http://www.astro.princeton.edu/draine/ddscat.htm ) is utilized for the calculations . in dda
the continuous target is discretized by an array of dipoles placed at cubic lattices and the dielectric constant of the target material as a function of wavelength at ism temperatures is required .
in general , optical properties of nano - sized particles are different from that of the bulk due to additional collisions of conducting electrons with the grain boundary .
the imaginary part of the dielectric constant is changed to accommodate size effects and expressed as : @xmath9 @xcite , where @xmath10 is the electron velocity at the fermi surface , @xmath11 is the plasma frequency , @xmath12 is a constant of order unity and @xmath13 is the frequency of incident electromagnetic radiation . the second term increases with decreasing size and decreases rapidly with increase in frequency .
as a result the dielectric constant does not vary appreciably with size in high frequency visible and uv regions data for @xmath10 , @xmath11 and @xmath12 from @xcite the imaginary part of dielectric constant becomes @xmath14 , where a is in nanometer . at this wavelength @xmath15
is @xmath16 3.2 and will be negligibly affected by particle size ( second term ) . the free electron density in diamond is smaller than graphite so the corresponding @xmath10 and @xmath17 will be smaller . thus at high frequencies size affect in the case of nanodiamonds will be even smaller . ] whereas , it varies slightly in the ir .
the extinction , absorption and scattering efficiencies for nano - sized particles are very small in ir compared to visible and uv region , so small change in refractive index in the ir will not affect the result in general .
thus the refractive index data of bulk graphite and diamond are taken for all nano - sized particles .
meteoritic nanodiamonds , with their density lower by 30% and relatively high hydrogen content @xcite , differ from terrestrial diamonds .
while terrestrial diamonds are cubic , the meteoritic diamonds are of allende type @xcite ( small nanodiamonds found within allende meteorites ) or carbonado diamonds @xcite ( relatively large , black and porous ) .
ism nanodiamonds can be assumed to have properties more like the meteoritic nanodiamonds .
for the present work we consider allende nanodiamond to be the candidate material and take the dielectric constants reported by @xcite .
the size of meteoritic nanodiamonds are reported to be @xmath18 in allende type @xcite and between @xmath19 to @xmath20 in orgueil meteorite @xcite .
theoretical study of nanodiamonds with fullerene like surface @xcite suggests particle sizes from @xmath19 to @xmath20 .
@xcite report that graphite particles with radius @xmath21 are thermodynamically unstable while the diamond structure is stable .
ir spectroscopic study of diamondoids @xcite points towards possibility of @xmath22 nanodiamond in ism while @xcite suggest that the 3.43 and 3.53 @xmath4 emission features are observable for nanodiamond @xmath23 .
@xcite suggests log - normal particle size distribution between @xmath19 to @xmath1 with peak diameter at @xmath24 . for the present work nanodiamond size
is taken to be between @xmath19 to @xmath1 .
the target nanodiamonds are taken as oblate and prolate spheroids and ellipsoids and their different axial ratios as given in table 1 .
the size of these ellipsoids are taken such that the effective volume for each is equal to a sphere of radius @xmath25 , @xmath26 , @xmath27 and @xmath28 , so that nearly all possible nanodiamond sizes are taken into account .
table 1 also gives , for @xmath1 nanodiamonds , the largest semi - axes and the number of dipoles used in dda calculations .
the validity criteria for dda i.e. @xmath29 is well considered and in none of the calculation it is more than @xmath30 . to reduce calculation time 9 orientations for nanodiamond are considered as calculation for all 27 orientations done for some cases makes a difference of less than @xmath31 .
the core - mantle model consisting of spherical nanodiamond core in different ellipsoidal graphitic mantle is represented in fig.[fig1 ] with the ellipsoid semi axes ` a ' , ` b ' and ` c ' and spherical core of radius ` r ' .
the last column in table 1 gives nanodiamond percentage in total grain volume for ` a ' and ` r ' in the ratio @xmath32 .
all nanodiamond - graphite ( core - mantle ) calculations are done for 27 orientations .
calculations are done for 64 wavelengths from far - uv to near ir ( @xmath33 ) and linear in @xmath34 .
.ellipsoid axial ratios , largest axis , no . of dipoles considered & nanodiamond % in core - mantle calculations [ cols="^,^,^,^ " , ] for comparison table 2 also shows the @xmath35 nanodiamond , two component model , that gives 0.40 silicate and 0.49 graphite fractions .
this is well within estimated silicate and graphite grain abundances @xcite . on adding nanodiamond - graphite core - mantle particles as the third component the fit improves .
this very small grain component is a small fraction of carbonaceous particles having even smaller nanodiamond core .
that is much lower than the @xmath36 upper limit of ism nanodiamond reported by @xcite .
the core - mantle component also reduces the graphite fraction slightly and changes negligibly the relative fraction of silicates .
the @xmath37 values increase with increasing percentage of nanodiamond in graphite mantle but its overall contribution ( fraction ` s ' ) decreases .
the fraction of nanodiamond - graphite is also much smaller than used by @xcite for @xmath38 nanodiamond grains .
considering nanodiamond density of 3.52 gm / cc the nanodiamonds in our models are @xmath39 fraction of carbonaceous particles .
the best fit , in the complete 0.2 to 10 @xmath40 wavelength range , stands for smaller nanodiamond in larger amount . for effective grain size of @xmath38 , the @xmath41 nanodiamond core implies @xmath42 size , which is about the size of nanodiamond in meteorites @xcite .
the above results assume only allende type nanodiamonds in ism .
alternatively ism nanodiamonds could be like terrestrial cubic diamonds , sp3 hybridized polymeric carbon or a complex mixture of these forms .
different types of diamonds have different dielectric functions and will lead to slightly different extinction profiles . while @xcite consider sp3 polymeric carbon coatings , @xcite consider both cubic and allende type nanodiamonds .
allende type nanodiamonds show far - uv steepening and peaking at lower frequencies than cubic diamonds and the far - uv break in quasars is better represented when greater proportion of allende type nanodiamonds are considered @xcite .
various experimental , theoretical and observational studies suggest the possibility of carbon in diamond form in the ism . besides enhanced far - uv extinction @xcite , luminescence from diamond nano - crystals could be responsible for the extended red emission ( ere ) @xcite .
the study of extinction properties of possible nanodiamond grains of different shapes and sizes is done to understand their role in far - uv extinction and in modification of the overall extinction curve .
the extinction due to nanodiamonds of all shapes and sizes is essentially similar with negligible extinction from ir to near uv range and sharp rise in extinction in the far - uv .
the extinction due to non - spherical shape is higher than that due to spherical . for nanosized particles the extinction efficiency increases linearly with effective radius .
considering nanodiamond - graphite core - mantle target , modification in the 2175 peak and graphite extinction is studied . in general
the 2175 peak gets lowered , broadened and there is enhanced far - uv extinction .
increase in % of @xmath43 character in graphite gives rise to higher far - uv extinction .
the three component modeling of average galactic extinction gives best results for small sized and low % of nanodiamond .
this enables us to explain far - uv rise and simultaneous modification of the 2175 bump without putting constraints on the interstellar abundances .
extinction from specific objects that show enhanced far - uv rise can be attempted by incorporating a nanodiamond component .
this will enhance understanding of radiation induced transformations of carbonaceous matter in the ism .
the use of high performance computing and library facilities at iucaa , pune is acknowledged .
valuable discussions with dr .
r. gupta and dr .
vaidya are greatly appreciated . | ultraviolet extinction interstellar medium @xcite proposed diamonds interstellar medium 3.43 3.53 @xmath4 emission bands circumstellar stars elias 1 show nanodiamonds bands c - h stretching hydrogenated nanodiamond distinct 3.3 @xmath5 m pah feature 3.47 @xmath4 feature absorption proto - stars tertiary c - h stretching diamond structures spectra diamondoid molecules nanodiamonds nanometre 3.53 3.43 @xmath5 m emission lines smaller diamondoids 3.47 @xmath5 m absorption feature cosmic nanodiamonds detected primitive carbonaceous meteorites most abundant presolar component meteorites studied @xmath6 stretching mode nanodiamonds orgueil meteorite observed bands nanodiamonds not hydrogenated detection hard more abundant infrared bands important incorporate nanodiamonds in uv extinction models scattering extinction properties extinction properties nanodiamond dust reported spherical shape far - uv break spectral energy quasars attributed nanodiamond dust extinction properties nanodiamond grains ellipsoidal shapes sizes reportedstudy effect nanodiamond graphite extinction efficiencies reported for nanodiamond - graphite core mantle grains extinction curve modeling proposed including nanodiamond component evolution carbon in ism discussed assumed ism carbon primarily hydrogenated amorphous carbon transforms to graphite diamond structures uv radiation low uv radiations dense clouds star atmospheres graphitic particles may absent estimated time graphitization varies few weeks circumstellar to years interstellar grains nanodiamonds result from annealing carbonaceous material by uv radiation supernova ejecta hydrogenated nanodiamond features arise inner regions close central star higher uv flux studies small diamonds stable against thermal evaporation chemical attack graphite particles formation nanodiamonds studied laboratory from carbon vapour deposition detonation carbon based explosives electron irradiation graphitic shells ion beam irradiation of amorphous carbon films suggest diamonds graphite formed by nucleation organic ice mixture uv photolysis structural rearrangement between diamond graphite possiblegraphite subjected to may shelled onion form or transform into nanodiamond show @xmath8 ratio changes by doses uv radiation chemical conversion of pah clusters to nanodiamonds proposed by partial graphitization of nanodiamond due to heat pressure suggested experimental studies surface graphitization under strong radiation to core - mantle shell structure 80% graphitization nanodiamonds possible within carbonaceous matter modify optical properties of dust graphitic mantle nanodiamond crystals less active hard to detect extinction scattering absorption polarization efficiencies discrete dipole approximation used ddscat 6.1 for calculations target discretized by dipoles at cubic lattices dielectric constant material function of wavelength at temperatures optical properties of nano - sized particles different from bulk due to collisions electrons with grain boundary dielectric constant changed size effects expressed as @xmath9 electron velocity plasma frequency constant order unity frequency of electromagnetic radiation second term increases with size decreases with frequencydielectric constant vary with size in high frequency visible uv regions for @xmath10 imaginary part becomes @xmath14 a nanometer wavelength @xmath15 is @xmath16 3.2 negligibly affected by particle size free electron density in diamond smaller than graphite @xmath10 @xmath17 smaller high frequencies size affect nanodiamonds smaller varies slightly in ir extinction absorption scattering efficiencies for nano - sized particles small in ir change in refractive index affect result refractive index data of graphite diamond taken for all nano - sized particles meteoritic nanodiamonds density lower 30% high hydrogen content differ from terrestrial diamonds terrestrial diamonds cubic meteoritic diamonds allende type carbonado nanodiamonds properties like meteoritic nanodiamonds allende nanodiamond candidate material dielectric constants by @xcite size meteoritic nanodiamonds @xmath18 in allende type @xcite @xmath19 to @xmath20 in meteorite @xcite study suggests particle sizes from @xmath19 to @xmath20@xcite graphite particles radius @xmath21 thermodynamically unstable diamond structure stable spectroscopic study @xmath22 nanodiamond 3.43 3.53 @xmath4 emission features nanodiamond @xmath23 suggests log - normal particle size distribution between @xmath19 to @xmath1 peak diameter @xmath24 present work nanodiamond size between @xmath19 to @xmath1 target nanodiamonds oblate prolate spheroids ellipsoids axial ratios table 1 effective volume equal sphere radius @xmath25 nanodiamond sizes table 1 @xmath1 nanodiamonds largest semi - axes number dipoles dda calculations validity criteria @xmath29 more than @xmath30 9 orientations nanodiamond 27 orientations difference less than @xmath31 core - mantle model spherical nanodiamond core ellipsoidal graphitic mantle represented fig ellipsoid semi axes ` a b c ' spherical core radius ` r ' last column table 1 nanodiamond percentage total grain volume ` a ' r ' ratio @xmath32 nanodiamond - graphite calculations 27 orientationscalculations 64 wavelengths ir @xmath33 @xmath34 axial ratios largest axis dipoles nanodiamond % core - mantle calculations table 2 shows @xmath35 nanodiamond 0.40 silicate 0.49 graphite fractions within estimated silicate graphite abundances adding nanodiamond - core - mantle particles fit improves small grain component fraction carbonaceous particles smaller nanodiamond core lower than @xmath36 upper limit ism nanodiamond core - mantle reduces graphite fraction changes fraction silicates @xmath37 values increase nanodiamond graphite mantle contribution decreases fraction nanodiamond - graphite smaller @xmath38 nanodiamond grains nanodiamond density 3.52 gm / cc @xmath39 fraction carbonaceous particles best fit 0.2 to 10 @xmath40 wavelength range smaller nanodiamond larger amount grain size @xmath38 @xmath41 nanodiamond core implies @xmath42 size nanodiamond meteorites results assume allende type nanodiamonds ism nanodiamonds terrestrial cubic diamonds sp3 hybridized polymeric carbondiamonds different dielectric functions different extinction profiles @xcite sp3 polymeric carbon coatings cubic allende nanodiamonds show far - uv lower frequencies far - uv break in quasars better represented studies suggest carbon in diamond form luminescence from diamond nano - crystals extended red emission study extinction properties nanodiamond grains shapes sizes role far - uv extinction extinction curve extinction due nanodiamonds similar with negligible extinction from ir near uv range rise extinction far - uv extinction due to non - spherical shape higher spherical nanosized particles extinction efficiency increases with effective radius nanodiamond - core - mantle target modification 2175 peak graphite extinction studied 2175 peak lowered broadened enhanced far - uv extinction increase % @xmath43 character in graphite higher far - uv extinction three component modeling galactic extinction results for small sized low % nanodiamond far - uv rise modification 2175 bump without interstellar abundancesextinction objects far - uv rise incorporating nanodiamond component understanding radiation transformations carbonaceous matter use high performance computing library facilities iucaa pune acknowledged discussions with dr gupta vaidya appreciated | 0.497726 | 2,975 | 1,481 | , Saving $0.1 in GPT-4. |
67,437 | h.e.s.s . is one of the most sensitive instruments in the very high energy ( vhe ; @xmath0 100 gev ) gamma - ray domain and has revealed many new sources along the galactic plane . after the successful first vhe galactic plane survey of 2004 , h.e.s.s .
has continued and extended that survey in 20052008 , discovering a number of new sources , many of which are unidentified . + some of the unidentified h.e.s.s .
sources have several positional counterparts and hence several different possible scenarios for the origin of the vhe gamma - ray emission ; their identification remains unclear .
others have so far no counterparts at any other wavelength .
particularly , the lack of an x - ray counterpart puts serious constraints on emission models .
+ several newly discovered and still unidentified vhe sources are reported here . | very high energy ( vhe , @xmath1 ev ) particles can be traced within our galaxy by a combination of non - thermal x - ray emission and vhe gamma - ray emission via leptonic ( i.e. inverse compton scattering of electrons , bremsstrahlung and synchrotron radiation ) or hadronic ( i.e. the decay of charged and neutral pions , due to interactions of energetic hadrons ) processes .
the high energy stereoscopic system ( h.e.s.s . ) detects vhe @xmath2-rays above an energy threshold of @xmath3100 gev and up to @xmath3100 tev with a typical energy resolution of 15% per photon .
the angular resolution is @xmath4 per event , allowing a positional error better than 40 for a point source detected with a statistical significance 6 @xmath5 .
the h.e.s.s .
field of view is almost @xmath6 in diameter with a point source sensitivity of @xmath7 ergs @xmath8 s@xmath9 ( @xmath31% of the crab nebula ) for a 5 @xmath5 detection in 25 hours of observations @xcite .
after the successful first survey of 2004 @xcite , h.e.s.s . extended the survey in 2005 - 2008 @xcite , leading to the discovery of several new vhe gamma - ray sources .
of these , several have been associated with supenovae remnants ( snrs ; such as ctb 37a , ctb 37b , rcw 86 , kes 75 ) , some of them are candidate pwne ( such as hess j1356 - 654 and hess j1849 - 000 ) , however the rest remain unidentified .
the understanding of the nature of the unidentified sources is a crucial point , since they represent a large fraction of the high energy and very high energy source population : almost 50% of the h.e.s.s .
galactic sources are still unidentified ; more than 60% of egret sources were unidentified @xcite and also almost 50% of fermi lat sources are unidentified @xcite .
+ unidentified sources could point to unknown and potentially new phenomena . and also in the case that they would represent particular unknown stages of known phenomena ,
their study will allow us to put serious constraints on the current astrophysical models and hence on their physical emission models .
e.g. in the case that the vhe unidentified sources would be , as suggested by @xcite , ancient pwne , a new stage of evolution of pwne is discovered and studied , putting serious constraints on the model of emission .
here five recently discovered vhe gamma - ray sources , that are still unidentified , will be discussed , showing their morphology , their spectrum and providing evidence for possible counterparts : * hess j1507 - 622 * ( discussed separately in @xcite ) , * hess j1848 - 018 * , * hess j1503 - 582 * and a particular emphasis is dedicated to the new sources in the galactic center region , which is very important for fermi lat studies of diffuse emission or dark matter ( * hess j1741 - 302 * and * hess j1745 - 303 * ) .
the vhe gamma - ray emitter hess j1503 - 582 ( figure [ 1503 ] ) has been recently discovered by h.e.s.s .
@xcite , it is still unidentified , i.e. it does not have any of the typical counterparts at lower energies ( e.g. snrs , energetic pulsars , or bright _ fermi _ gamma - ray sources ) .
however , it is now considered unique , since it is the first vhe gamma - ray source that appears to be associated with a _ forbidden - velocity wing _ ( fvw ) .
fvws are faint hi 21 cm emission line structures which are visible at velocities which deviate from the canonical galactic rotation curve , suggesting associated dynamical phenomena @xcite
. deeper observations of hess j1503 - 582 are scheduled in order to further investigate this new source of vhe gamma - rays . ) , and is not powerful enough to explain a pwn scenario . ]
hess j1503 - 582 , visible in figure [ 1503 ] , shows significance of @xmath36@xmath5 in @xmath324 hours of effective exposure .
the spectrum can be fit well by a power law with index @xmath10 = 2.4 @xmath11 0.4@xmath12 @xmath11 0.2@xmath13 with a flux above 1 tev of @xmath14 @xmath8 s@xmath9 ( @xmath36 % of the crab flux ) .
hess j1848 - 018 is an extended vhe gamma - ray source which was recently detected in the h.e.s.s .
galactic plane survey with a statistical significance of @xmath39@xmath5 in @xmath350 hours of effective exposure @xcite .
figure [ 1848 ] shows the excess map of the recently discovered source hess j1848 - 018 @xcite .
its differential energy spectrum is well fit by a power law with index @xmath10 = 2.8 @xmath11 0.2@xmath12 @xmath11 0.2@xmath13 and it has an integrated flux above 1 tev of @xmath32 @xmath15 10@xmath16 erg @xmath8 s@xmath9 , corresponding to @xmath38% of the crab nebula .
@xmath11 0.02@xmath17 .
the position of wr 121a is marked in green and in magenta is shown the position of the fermi bright source ofgl j1848.6 - 0138 and its radius of 95% confidence region @xcite . ]
hess j1848 - 018 is located along the scutum - crux spiral arm tangent and is in the direction of , but slightly offset from , the star - forming region w 43 , which hosts a giant hii region ( g30.8@xmath180.2 ) , a giant molecular cloud , and the wolf - rayet ( wr ) star wr 121a in the central stellar cluster .
if hess j1848 - 018 is indeed associated with w 43 , it would be only the second known case , after westerlund 2 @xcite , of vhe gamma - ray emission associated with a star - forming region and wr star .
+ the complex , multi - wavelength morphology of hess j1848 - 018 is currently under investigation .
the discovery of new vhe sources close to the galactic center is very important for the studies about diffuse galactic emission and dark matter and will play a crucial role in the out - coming fermi lat studies .
hess j1745 - 303 is not a new source @xcite , but it has been recently observed in vhe gamma - rays in detail by h.e.s.s .
and in x - rays by _ xmm - newton _ @xcite
. radius ) and significant contours of hess j1745 - 303 @xcite .
the white dashed line represents the galactic plane .
superimposed there are the two snrs g359.1 - 0.5 and g359.0 - 0.9 , the two pulsars psr b1742 - 30 and psr j1747 - 2958 , and the egret source 3eg 1744 - 3011 . ]
figure [ 1745fig ] shows the image of gamma - ray excess counts of hess j1745 - 303 smoothed with a radius of @xmath19 .
overlaid on the image are the contours of 4@xmath5 , 5@xmath5 , 6@xmath5 , 7@xmath5 of significance .
hess j1745 - 303 is detected with a significance of 10.2@xmath5 in 79 hours of observation .
the spectrum can be fit well by a power law with index @xmath20 and a flux above 1 tev of @xmath21 @xmath8 s@xmath9 .
+ the region labeled a in figure [ 1745fig ] is thought to be associated with the interaction of the snr g359.1 - 0.5 @xcite with @xmath22co molecular clouds @xcite spatially coincident with the peak of the vhe gamma - ray emission .
the part labeled b in figure [ 1745fig ] is more likely to be a pwn powered by the pulsar psr b1742 - 30 ( @xmath23 erg s@xmath9 kpc@xmath24 ) , which requires a conversion efficiency from rotational kinetic energy to gamma - rays of @xmath25 to produce the entire vhe emission .
hess j1745 - 303 is also spatially coincident with the egret source 3eg 1744 - 3011 @xcite .
the discovery of a faint ( @xmath31% of the crab flux ) source , hess j1741 - 302 , has been recently announced @xcite , one of the faintest newly discovered sources , at the lower end of the h.e.s.s .
sensitivity . , with superimposed the positions of psr b1737 - 30 @xcite and psr j1741 - 3016 @xcite . ]
the discovery peak significance is 8.1 @xmath5 in 143.5 hours of observations . with _
standard analysis cuts _
@xcite , hess j1741 - 302 appears as an irregular blob . with
_ hard analysis cuts _ ( 200 p.e . ) @xcite resulting in improved angular resolution , improved background rejection and higher energy threshold , two apparent hot spots start to appear in the map ( figure [ 1741fig ] ) ; however figure [ nostat ] shows that current statistics do not allow detailed statements about source morphology , and that the hot spots are consistent with statistical fluctuations within the source .
+ the preliminary spectrum can be fit well by a power law with index @xmath26 with a flux above 1 tev of @xmath27 @xmath8 s@xmath9 .
in other sources in the galactic center region ( e.g. hess j1745 - 303 @xcite and the gc diffuse emission @xcite ) , there is evidence that interactions of cosmic rays with molecular clouds ( mcs ) play a role in vhe gamma - ray production .
such an association in this case is not clear , though in some velocity ranges there is some weak molecular emission coincident with the h.e.s.s . source @xcite .
another suitable scenario could be an offset pulsar wind nebula powered by the somewhat offset but relatively powerful pulsar psr b1737 - 30 @xcite ( spin - down luminosity @xmath28 ) .
this pulsar has a spin - down luminosity that is almost one order of magnitude fainter than other vhe gamma - ray emitting pulsars ( e.g. like hess j1825 - 137 @xcite ) , which have typical vhe fluxes of @xmath310% of the crab .
given that the estimated flux of hess j1741 - 303 is approximately 1% of the crab , a weak pwn scenario is plausible .
the brightest hot spot is spatially coincident with the pulsar psr j1741 - 3016 @xcite , which is , however , probably too faint ( spin - down luminosity @xmath29 ) to sustain a pwn scenario .
five recently discovered h.e.s.s .
unidentified sources and their possible counterparts have been discussed .
all the sources are located in the galactic plane ( or slightly off the plane ) and appear extended in vhe gamma rays . since these sources have so far no clear counterpart in lower - energy wavebands , further multi - wavelength studies are required in order to understand the emission mechanism powering them .
the support of the namibian authorities and of the university of namibia in facilitating the construction and operation of hess is gratefully acknowledged , as is the support by the german ministry for education and research ( bmbf ) , the max planck society , the french ministry for research , the cnrs - in2p3 and the astroparticle interdisciplinary programme of the cnrs , the u.k .
science and technology facilities council ( stfc ) , the ipnp of the charles university , the polish ministry of science and higher education , the south african department of science and technology and national research foundation , and by the university of namibia .
we appreciate the excellent work of the technical support staff in berlin , durham , hamburg , heidelberg , palaiseau , paris , saclay , and in namibia in the construction and operation of the equipment . | high energy particles traced galaxy by non - thermal x - ray vhe gamma - ray emission via leptonic or hadronic processes high energy stereoscopic system detects vhe @xmath2-rays above energy threshold @xmath3100 gev up to @xmath3100 tev energy resolution 15% per photon angular resolution @xmath4 per event positional error better than 40 for point source detected 6 @xmath5 field of view almost @xmath6 in diameter point source sensitivity @xmath7 ergs @xmath8@xmath9 ( @xmath31% of crab nebula for 5 @xmath5 detection in 25 hours observations after successful first survey 2004 extended survey 2005 - 2008 several new vhe gamma - ray sources several associated with supenovae remnants some candidate pwne j1356 - 654 j1849 - 000 rest remain unidentified understanding of nature unidentified sources represent large fraction of high energy source population almost 50% of h.e.s.s .galactic sources unidentified 60% egret sources unidentified @xcite 50% fermi lat sources unidentified @xcite unidentified sources could point to unknown new phenomena represent unknown stages known phenomena study constraints on current astrophysical models physical emission models vhe unidentified sources ancient pwne new stage evolution pwne discovered constraints model emission five recently discovered vhe gamma - ray sources unidentified discussed morphology spectrum evidence for possible counterparts : * hess j1507 - 622 * hess j1848 - 018 * hess j1503 - 582 * emphasis to new sources galactic center region important for fermi lat studies diffuse emission dark matter ( * hess j1741 - 302 * hess j1745 - 303 * vhe gamma - ray emitter hess j1503 - 582 figure 1503 ) recently discovered h.e.s.s @xcite unidentified typical counterparts at lower energies considered unique first vhe gamma - ray source associated with forbidden - velocity wing _fvws 21 cm emission line structures visible galactic rotation curve suggesting dynamical phenomena observations hess j1503 - 582 scheduled new source gamma - rays not powerful pwn scenario hess j1503 - 582 figure 1503 shows @xmath36@xmath5 in @xmath324 hours exposure spectrum power law index @xmath10 = 2.4 0.4 0.2 flux above 1 tev @xmath14 % crab flux hess j1848 - 018 extended vhe gamma - ray source detected galactic plane survey significance @xmath39@xmath5 in hours exposure figure 1848 excess map hess j1848 - 018 differential energy spectrum fit power law index @xmath10 = 2.8 0.2 0.2 integrated flux above 1 tev @xmath32 @xmath38% crab nebula @xmath11 0.02@xmath17 position of wr 121a marked green magenta fermi bright source j1848.6 - 0138 radius 95% confidence regionhess j1848 - 018 scutum - crux spiral arm tangent offset star - forming region w 43 giant hii region giant molecular cloud wolf - rayet star wr 121a central stellar cluster if associated with w 43 second case after westerlund 2 gamma - ray emission star - forming region star complex morphology hess j1848 - 018 under investigation discovery new vhe sources galactic important for diffuse galactic emission dark matter fermi lat studies hess j1745 - 303 not new source observed in gamma - rays x - rays - newton radius contours hess j1745 - 303 white dashed line galactic plane two snrs g359.1 - 0.5 g359.0 - 0.9 pulsars psr b1742 - 30 j1747 - 2958 egret source 3eg 1744 - 3011 figure gamma - ray excess counts hess j1745 - 303 radius @xmath19 contours 4 5 6 7 significance j1745 - 303 detected significance 10.2@xmath5 in 79 hours observationspectrum power law index @xmath20 flux above 1 tev @xmath21 region a 1745fig associated interaction snr g359.1 - 0.5 @xcite @xmath22co molecular clouds coincident peak vhe gamma - ray emission part b likely pwn powered pulsar psr b1742 - 30 requires conversion rotational kinetic energy to gamma - rays @xmath25 vhe emission hess j1745 - 303 coincident with egret source 3eg 1744 - 3011 @xcite discovery faint @xmath31% crab flux source hess j1741 - 302 announced faintest lower end h.e.s.s . sensitivity positions psr b1737 - 30 @xcite psr j1741 - 3016 @xcite discovery peak significance 8.1 @xmath5 in 143.5 hours observations standard analysis hess j1741 - 302 irregular blob hard analysis improved angular resolution background rejection higher energy threshold two hot spots appear map figure 1741fig current statistics allow detailed source morphology hot spots consistent with statistical fluctuations sourcepreliminary spectrum power law index @xmath26 flux above 1 tev @xmath27 sources galactic center region hess j1745 - 303 @xcite gc diffuse emission evidence interactions cosmic rays molecular clouds gamma - ray production association not clear weak molecular emission coincident h.e.s.s . source @xcite scenario offset pulsar wind nebula powerful pulsar psr b1737 - 30 @xcite spin - down luminosity @xmath28 ) pulsar luminosity fainter other gamma - ray emitting pulsars hess j1825 - 137 @xcite fluxes @xmath310% crab estimated flux hess j1741 - 303 1% crab weak pwn scenario plausible brightest hot spot coincident with pulsar psr j1741 - 3016 @xcite too faint spin - down luminosity @xmath29 ) pwn five h.e.s.s . unidentified sources counterparts discussed galactic plane extended in gamma rays no counterpart lower - energy wavebands further multi - wavelength studies required understand emission mechanismsupport namibian authorities university of namibia construction operation hess acknowledged support german ministry education max planck society french ministry research cnrs - in2p3 astroparticle interdisciplinary programme u.k . science technology facilities council ipnp charles university polish ministry science education south african department science technology national research foundation university of namibia appreciate work technical support staff berlin durham hamburg heidelberg palaiseau paris saclay namibia construction operation equipment | 0.434362 | 2,798 | 1,414 | , Saving $0.1 in GPT-4. |
67,439 | we apply strong - coupling perturbation theory to the qcd lattice hamiltonian .
we begin with naive , nearest - neighbor fermions and subsequently break the doubling symmetry with next - nearest - neighbor terms .
the effective hamiltonian is that of an antiferromagnet with an added kinetic term for baryonic `` impurities , '' reminiscent of the @xmath0@xmath1 model of high-@xmath2 superconductivity . as a first step , we fix the locations of the baryons and make them static . following analyses of the @xmath0@xmath1 model , we apply large-@xmath3 methods to obtain a phase diagram in the @xmath4 plane at zero temperature and baryon density .
next we study a simplified @xmath5 toy model , in which we add baryons to the vacuum .
we use a coherent state formalism to write a path integral which we analyze with mean field theory , obtaining a phase diagram in the @xmath6 plane .
color superconductivity @xcite at high density is so far a prediction only of weak - coupling analysis , valid ( if at all ) only at very high densities .
we seek confirmation from methods that do not depend on weak coupling , as well as an extension to the regime of moderate densities .
since euclidean monte carlo methods are unavailable when the chemical potential is non - zero , we turn to the strong - coupling limit of qcd ; harking back to the early days of lattice gauge theory , we work in the hamiltonian formulation . we derive an effective hamiltonian for color - singlet states that takes the form of an antiferromagnet with a kinetic term for baryons .
this effective hamiltonian is very difficult to study . as a first step , we fix the position of the baryons and study mesonic excitations in the baryonic background .
coherent - state methods then enable us to derive equivalent models that are tractable in various limits of large @xmath7 and/or @xmath8 .
we benefit from the considerable work done on antiferromagnets in the context of large-@xmath2 superconductors @xcite , as well as from the early work of the slac group @xcite and of smit @xcite on strong - coupling qcd . | the lattice gauge hamiltonian is composed of electric , magnetic , and fermion terms , @xmath9 where the first term is the unperturbed hamiltonian in the strong coupling limit , @xmath10 the ground state sector of @xmath11 is highly degenerate , consisting of of all states with zero electric flux , whatever their fermion content , @xmath12|\chi \rangle_f.\ ] ] neglecting the magnetic term , which only contributes in high order , we perturb with the fermion kinetic term , @xmath13 we use four - component fermions with a general ( diagonal ) kernel @xmath14 .
this is the best we can do , since domain - wall fermions are unavailable to us at strong coupling @xcite and there is no hamiltonian overlap formalism @xcite .
we will discuss the properties of these fermions in a moment .
the degeneracy is lifted in second order via diagonalization of an effective hamiltonian , @xmath15 where @xmath16 is a sign factor and @xmath17 is a new kernel .
the effective hamiltonian contains fermion bilinears at each site , @xmath18 where the @xmath19 matrices act on the dirac and the flavor indices , @xmath20 .
the operators @xmath21 generate a @xmath22 algebra .
if we fix the baryon number @xmath23 on a site , the color - singlet states on that site make up an irreducible representation of this algebra , whose young tableau has @xmath24 rows and @xmath7 columns ( see fig . 1 ) .
] expressed in terms of @xmath21 , the effective hamiltonian is an antiferromagnet .
if the kernel @xmath14 is chosen to give nearest - neighbor couplings only , @xmath25 will take the form @xmath26 this antiferromagnet has the accidental @xmath22 symmetry of naive fermions , which is responsible for part of the doubling problem .
the even-@xmath27 terms in ( [ heff2 ] ) break this symmetry to @xmath28 , which [ apart from the unbreakable axial @xmath29 is the desired continuum symmetry .
since we are interested in strong coupling where the free fermion dispersion relation is of no interest , this might be a good - enough partial solution of the doubling problem .
the theory of @xmath25 contains only static baryons , which make their presence felt through fixing the rep of @xmath22 on each site .
these baryons can move in the next order in perturbation theory ( only if @xmath30a fortunate special case ) .
the effective hamiltonian in third order is @xmath31_{ii'}b_{\bn+j\hat\mu}^{i ' } \nonumber\\ & & \quad + h.c.,\end{aligned}\ ] ] where the baryon operators are the color singlets @xmath32 that belong to the rep of @xmath22 .
the simple form of @xmath33 is deceptive , for the @xmath34 s are composite and hence do not obey canonical anticommutation relations . for nearest - neighbor fermions
, the usual spin diagonalization @xcite gives a simplified hamiltonian , @xmath35 the complete hamiltonian @xmath36 resembles that of the @xmath0@xmath1 model , which represents the strong - binding limit of the hubbard model and is much studied in connection with high-@xmath2 superconductivity @xcite .
but our hamiltonian is much more complex .
let us beat a strategic retreat to the second - order theory , where baryons constitute a static background .
if we begin with the nearest - neighbor model , in a state with _ no _ baryons , then the effective hamiltonian ( [ hnn ] ) is that of a @xmath22 antiferromagnet with spins in a rep specified by @xmath7 and by @xmath37 ( which can vary from site to site ) . this can be studied in the limits of large @xmath38 by various transformations @xcite , and the result still for @xmath39is the phase diagram in fig . 2 .
] ( note that the coupling constant @xmath40 is just a scale , which has no effect at @xmath41 . )
the large-@xmath8 phase is disordered , and will not concern us .
the location of the phase boundary can be established by studying a schwinger boson representation of the @xmath22 spins , and its slope turns out to be @xmath42 ; this means that the qcd vacuum is safely in the _ ordered _ phase for any reasonable number of flavors . the ordered phase is conveniently studied in a @xmath43 model representation , which comes from rewriting ( [ hnn ] ) in a basis of spin coherent states @xcite .
( this is valid for any @xmath4 but proves soluble in the @xmath44 limit . )
the degrees of freedom of the @xmath43 model are the @xmath45 matrices ( with @xmath46 ) @xmath47 where @xmath48 the field @xmath49 runs over the group @xmath50 , and the manifold covered by @xmath51 is the coset space @xmath52 $ ] .
the action of the @xmath43 model ( in continuous time ) is @xmath53,&&\nonumber\\\ \label{action}\end{aligned}\ ] ] where , in terms of the matrices @xmath51 , the nearest - neighbor hamiltonian takes the form @xmath54 clearly as @xmath44 the ground state is the classical minimum of @xmath55 , in which on all sites @xmath56 , which can be rotated to @xmath57 .
thus the symmetry is spontaneously broken as @xmath58 , with @xmath59 goldstone bosons @xcite .
we now restore next - nearest - neighbor couplings , viz .
@xmath60 where @xmath61 .
the symmetry of the theory , as discussed above , is @xmath62 ; the classical minimum is at @xmath63 , which breaks all the axial generators and leaves the vector @xmath64 unbroken .
this is what we would expect for the ground state in the vacuum sector .
the @xmath39 states considered above were specified by choosing the @xmath37 rep on each site . choosing a different rep adds ( or subtracts ) baryons on a site - by - site basis .
for instance , we can add a single baryon by adding a row to the young tableau ( fig .
theory [ a @xmath65 antiferromagnet ] ] one can similarly add baryons on an entire sublattice .
the @xmath44 limit directs us to find the classical ground state of the hamiltonian , which always breaks the symmetry spontaneously along the lines shown in the section above .
we can study the effects of the fluctuations by doing mean field theory for finite @xmath7 . to do this
, we drop the kinetic term in ( [ action ] ) and go to @xmath66 by calculating the resulting classical partition function .
in mf theory , we write down a trial hamiltonian and calculate a variational free energy @xmath67 .
the simplest trial hamiltonian , containing no site site correlations , is @xmath68 where the magnetic fields @xmath69 are variational parameters .
( we write @xmath70 for the vector whose @xmath71 components are @xmath21 . ) the free energy obeys @xmath72 with @xmath73 here @xmath74 and @xmath75 is the magnetization , @xmath76 note that the integration measure @xmath77 depends on the @xmath50 representation chosen for site @xmath78 .
we minimize @xmath67 with respect to @xmath79 and get the mean field equations , @xmath80 if all sites are in the same rep of @xmath50 , then the bipartite nature of the antiferromagnetic system gives two coupled sets of mf equations .
otherwise one gets as many coupled mf equations as there are inequivalent sites . after solving for @xmath81 we evaluate @xmath67 via @xmath82 we seek the global minimum of @xmath67 .
once we find it , we can examine its symmetry properties and identify the phase favored at temperature @xmath83 .
the @xmath84 theory , the @xmath65 antiferromagnet ( with impurities ) , contains many degrees of freedom in which to do mean field theory .
a simpler non - trivial model is a toy model with @xmath5 symmetry , which does not correspond to an actual value of @xmath8 .
the most symmetric @xmath39 state in this model can not have @xmath39 on every site , but must alternate between the reps corresponding to @xmath85 ( see fig .
4 ) .
we in fact create a non - zero density of baryons by adding a row to one or more sublattices , forming a lattice with a unit cell of @xmath87 sites .
the unit cell may contain sites with @xmath88 , 2 , or even 3 . for @xmath89
the manifold of @xmath90 is @xmath91 $ ] , while for @xmath92 , the singlet state , we have @xmath93 . the form of @xmath94 for a site with @xmath95 is @xmath96 , with @xmath97
we write @xmath98 with @xmath99 and @xmath100 . with these definitions , @xmath101 is given by @xmath102 here @xmath103 and @xmath104 , with @xmath105 and @xmath106 .
the induced measure on the four - dimensional manifold turns out to be @xmath107 an example : for the @xmath39 case shown in fig . 4 , the inequivalent sites are just the even and odd sublattices .
the corresponding mf equations are @xmath108 cases with @xmath86 will give more coupled sets of mf equations , according to the number of inequivalent sites in the unit cell . for each baryonic configuration
we obtain a phase transition as a function of temperature . in some cases
it is first order and in others , second order . in all cases
the symmetry breakdown at low temperature is @xmath109 $ ] . as we increase the baryon density the transition temperature decreases , but it never vanishes ; mf theory always breaks symmetry at @xmath41 , even in one dimension .
we summarize our findings in the phase diagram in the temperature density plane shown in fig
. 6 .
the future holds , we hope , the removal of the various approximations that led from qcd to the @xmath5 toy model . to begin with
, we must do better than classical mean field theory and include the quantum kinetic term in the @xmath43 model ( [ action ] ) .
the @xmath5 model must be generalized to the @xmath22 symmetry group of naive fermions , which should be broken to @xmath62 by the nnn coupling .
going beyond the static baryon picture , we can disorder the baryon background by a replica method ; eventually dynamical baryons should be included with the third - order kinetic term .
once the theory becomes realistic enough , we can compare the results at each step to the weak - coupling predictions of color superconductivity with various values of @xmath8 @xcite .
there is , however , a limitation inherent in the strong - coupling theory . as we saw in the mean - field analysis above , one
is easily misled into filing the lattice with baryons to saturation .
the baryon density is limited according to @xmath110 and strong coupling means large lattice spacing @xmath111 .
it is possible that color superconductivity will not show up at any density short of saturation . in that case
we will have to content ourselves with a new effective theory for baryonic matter , short of the transition to quark matter .
9 k. rajagopal and f. wilczek , arxiv : hep - ph/0011333 .
i. affleck , nucl .
b * 257 * ( 1985 ) 397 ; j. b. marston and i. affleck , phys . rev .
b * 39 * ( 1989 ) 11538 .
d. p. arovas and a. auerbach , phys . rev .
b * 38 * ( 1988 ) 316 . n. read and s. sachdev , nucl .
b * 316 * ( 1989 ) 609 ; phys .
b * 42 * ( 1990 ) 4568 .
b. svetitsky , s. d. drell , h. r. quinn and m. weinstein , phys .
d * 22 * ( 1980 ) 490 .
j. smit , nucl .
b * 175 * ( 1980 ) 307 .
r. c. brower and b. svetitsky , phys .
d * 61 * ( 2000 ) 114511 [ arxiv : hep - lat/9912019 ] .
m. creutz , i. horvath and h. neuberger , nucl .
* 106 * ( 2002 ) 760 [ arxiv : hep - lat/0110009 ] .
a. auerbach , _ interacting electrons and quantum magnetism _ ( springer - verlag , 1994 ) .
t. schfer , nucl .
b * 575 * ( 2000 ) 269 [ arxiv : hep - ph/9909574 ] . | lattice gauge hamiltonian electric magnetic fermion terms @xmath9 first term unperturbed hamiltonian strong coupling limit ground state sector @xmath11 degenerate states zero electric flux fermion content @xmath12 neglecting magnetic term perturb fermion kinetic term @xmath13 four - component fermions general diagonal kernel @xmath14 domain - wall fermions unavailable strong coupling no hamiltonian overlap formalism discuss properties degeneracy lifted second diagonalization effective hamiltonian @xmath15 @xmath16 sign factor @xmath17 new kernel effective hamiltonian contains fermion bilinears each site @xmath19 matrices act dirac flavor indices operators @xmath21 generate @xmath22 algebra fix baryon number @xmath23 site color - singlet states irreducible representation algebra tableau @xmath24 rows @xmath7 columns fig . 1 effective hamiltonian antiferromagnet kernel @xmath14 nearest - neighbor couplings @xmath25 form @xmath26 antiferromagnet accidental @xmath22 symmetry naive fermions responsible doubling problemeven-@xmath27 terms break symmetry to @xmath28 desired continuum symmetry interested in strong coupling free fermion dispersion relation no interest partial solution doubling problem theory @xmath25 contains static baryons fixing rep @xmath22 each site baryons move next order in perturbation theory @xmath30a special case effective hamiltonian third order is @xmath31_ baryon operators are color singlets @xmath32 belong to rep @xmath22 simple form of @xmath33 deceptive @xmath34 s composite obey anticommutation relations nearest - neighbor fermions spin diagonalization gives simplified hamiltonian complete hamiltonian @xmath36 resembles @xmath0@xmath1 model strong - binding limit of hubbard model studied high-@xmath2 superconductivity hamiltonian more complex retreat to second - order theory baryons constitute static background begin with nearest - neighbor model no baryons effective hamiltonian @xmath22 antiferromagnet with spins in rep specified by @xmath7 and @xmath37 vary from site to sitestudied limits @xmath38 transformations @xcite result @xmath39is phase diagram fig . 2 coupling constant @xmath40 scale no effect at @xmath41 large-@xmath8 phase disordered not concern location phase boundary schwinger boson representation @xmath22 spins slope @xmath42 qcd vacuum in ordered phase for flavors ordered phase studied in @xmath43 model rewriting spin coherent states @xcite valid for @xmath4 soluble in @xmath44 limit degrees of freedom @xmath43 model @xmath45 matrices @xmath46 ) @xmath47 @xmath48 field @xmath49 runs over group @xmath50 manifold covered @xmath51 is coset space @xmath52 action @xmath43 model is @xmath53 @xmath51 nearest - neighbor hamiltonian form @xmath54 @xmath44 ground state classical minimum of @xmath55 sites @xmath56 rotated to @xmath57 symmetry broken as @xmath58 @xmath59 goldstone bosons @xcite restore next - nearest - neighbor couplings @xmath60 where @xmath61symmetry theory is @xmath62 classical minimum @xmath63 breaks axial generators leaves vector @xmath64 unbroken ground state in vacuum sector @xmath39 states specified choosing @xmath37 rep each site different rep adds or subtracts baryons site - by - site add baryon row young tableau @xmath65 baryons sublattice @xmath44 limit classical ground state hamiltonian breaks symmetry study effects fluctuations mean field theory for finite @xmath7 drop kinetic term @xmath66 classical partition function mf theory trial hamiltonian calculate variational free energy @xmath67 simplest trial hamiltonian no correlations @xmath68 magnetic fields @xmath69 are variational parameters write @xmath70 for vector @xmath71 components @xmath21 free energy obeys @xmath72 @xmath73 @xmath74 and @xmath75 magnetization integration measure @xmath77 depends on @xmath50 representation for site @xmath78 minimize @xmath67 @xmath79 get mean field equations if sites in same rep @xmath50 bipartite antiferromagnetic system gives two coupled sets mf equationscoupled mf equations inequivalent sites solving @xmath81 evaluate @xmath67 @xmath82 seek global minimum @xmath67 examine symmetry identify phase favored at temperature @xmath83 @xmath84 theory @xmath65 antiferromagnet contains degrees freedom field theory model toy model with @xmath5 symmetry correspond value @xmath8 @xmath39 state every site alternate between reps @xmath85 fig . 4 create non - zero density baryons adding row to sublattices forming lattice unit cell @xmath87 sites unit cell contain sites @xmath88 2 3 @xmath89 manifold of @xmath90 is @xmath91 @xmath92 @xmath93 form @xmath94 site @xmath95 @xmath96 @xmath97 @xmath98 with @xmath99 @xmath100 @xmath101 given by @xmath102 @xmath103 @xmath104 @xmath105 @xmath106 induced measure on four - dimensional manifold @xmath107 @xmath39 case fig . 4 inequivalent sites even odd sublatticesmf equations are @xmath108 @xmath86 more coupled sets mf inequivalent sites in unit cell each baryonic configuration phase transition function of temperature some first order second order symmetry breakdown at low temperature is @xmath109 baryon density transition temperature decreases never vanishes mf theory breaks symmetry at @xmath41 summarize findings in phase diagram temperature density plane fig . 6 future holds removal of approximations qcd to @xmath5 toy model include quantum kinetic term in @xmath43 model @xmath5 model to @xmath22 symmetry group of naive fermions broken to @xmath62 by nnn coupling disorder baryon background by replica method dynamical baryons included with third - order kinetic term compare results to weak - coupling predictions of color superconductivity with values @xmath8 @xcite limitation in strong - coupling theory filing lattice with baryons to saturation baryon density limited to @xmath110 strong coupling means large lattice spacing @xmath111 color superconductivity up at density short of saturationnew theory baryonic matter transition quark matter rajagopal f. wilczek hep - ph/0011333 affleck nucl b 257 1985 ) 397 marston affleck . rev b 39 1989 ) 11538 arovas auerbach rev 38 1988 ) 316 read sachdev b 316 1989 ) 609 b 42 1990 ) 4568 svetitsky drell quinn weinstein phys . d 22 1980 ) 490 smit nucl b 175 1980 ) 307 brower b svetitsky . d 61 2000 ) 114511 hep - lat/9912019 creutz horvath h. neuberger nucl 106 2002 ) 760 hep - lat/0110009 auerbach interacting electrons quantum magnetism springer - verlag 1994 ) schfer nucl b * 575 2000 ) 269 arxiv hep - ph/9909574 | 0.436599 | 3,012 | 1,602 | , Saving $0.1 in GPT-4. |
67,461 | we investigate isolated sub- and super - chandrasekhar white dwarfs which lose angular momentum through magnetic dipole braking .
we construct constant rest mass sequences by fulfilling all stability criteria of rotating configurations and show how the main structure of white dwarfs such as the central density , mean radius and angular velocity change with time .
we explicitly demonstrate that all isolated white dwarfs regardless of their masses , by angular momentum loss , shrink and increase their central density .
we also analyze the effects of the structure parameters on the evolution timescale both in the case of constant magnetic field and constant magnetic flux . | in accordance with the current understanding , white dwarf ( wd ) stars represent the final product of the evolution of progenitor stars with the masses ranging from roughly 0.07 to 8 @xmath0 ( solar mass ) .
@xcite most of the observed wds masses are clustered around ( 0.5 - 0.7)@xmath0,@xcite whereas their size is typically to that of a planet .
hence they possess large average densities , huge surface gravities and low luminosities .
@xcite in order to explain the formation of low mass ( sub - chandrasekhar ) wds it is believed that their progenitors were main sequence stars of low or medium mass.@xcite however to explain the formation of massive white dwarfs ( super - chandrasekhar ) there exist several scenarios : a single degenerate scenario,@xcite where a wd increases its mass through accretion of matter from a normal stellar companion ; a double degenerate scenario,@xcite where two average mass wds merge after losing energy and angular momentum through the radiation of gravitational waves ; and a core degenerate scenario,@xcite where the merger occurs in a common envelop with a massive asymptotic branch star.@xcite in this regard we investigated general relativistic configurations of uniformly rotating sub- and super - chandrasekhar wds in our recent work@xcite within hartle s formalism.@xcite we used the relativistic feynman - metropolis - teller ( rfmt ) equation of state@xcite ( eos ) for wd matter , which generalizes the traditionally used equations of state of chandrasekhar@xcite and salpeter.@xcite the stability of rotating wds was analyzed taking into account the mass - shedding limit , inverse @xmath1-decay instability , pynonuclear instability , and secular axisymmetric instability , with the last being determined by using the turning point method of friedman et al.@xcite in this work , we explore the compression of isolated rotating sub- and super - chandrasekhar wds by angular momentum loss based on the results of ref . .
eventually , by fulfilling all the stability criteria for rotating wds in general relativity ( gr ) we calculate the change in time of the basic parameters of rotating wds .
we consider two particular cases : a ) we assume that dipolar magnetic is constant throughout the entire evolution of wds ; b ) we adopt the magnetic flux conservation , relaxing the constancy of the magnetic field , for the sake of comparison .
in order to examine the compression of isolated white dwarfs with time , first , taking into account all the stability criteria , we constructed constant mass sequences for selected sub- and super - chandrasekhar wds .
second , knowing how the main parameters change along the sequences we find their dependence on each other .
third , using the equation for the rotational energy loss of pulsars via magnetic dipole braking,@xcite we estimate the evolution time that is required to go from one instability point to another.@xcite fourth , depending on what parameter we are interested in , we modify the equation and analyze how this parameter changes in time . for example , if we want to examine how the central density evolves with time we need to rewrite the equation as follows @xmath2 where we choose all parameters as a function of the central density @xmath3 @xmath4 and @xmath5 is the evolution time , @xmath6 is the speed of light in vacuum , @xmath7 is the dipole magnetic field , @xmath8 is the mean spherical radius , @xmath9 is the polar radius , @xmath10 is the equatorial radius , @xmath11 is the angular velocity and @xmath12 is the angular momentum of the white dwarf .
we adopt the maximum brake , so that the angle between the rotation axis and the magnetic field is ninety degrees .
the values of @xmath13 and @xmath14 are calculated along a given constant baryon ( rest ) mass sequence as mentioned above . here
we choose carbon white dwarfs using the rfmt eos as in ref . .
according to ref .
and there are limiting values of all the parameters of rotating wds on the border of the stability region .
these limiting values determine the range of integration of the central density @xmath3 . performing numerical integration of eq .
[ eq : rhot ] for each moment of time we obtain the evolution of the central density with time .
we use the following expression to account for the magnetic flux conservation , relaxing the constancy of the magnetic field @xmath15 where @xmath16 is the surface dipole magnetic field corresponding to the initial value of @xmath7 at @xmath17 , @xmath18 is the mean radius corresponding to the initial values of @xmath19 at @xmath17 . substituting @xmath7 in eq .
[ eq : rhot ] we obtain an equation that describes the evolution of @xmath3 with time accounting for the magnetic flux conservation . clearly , in fig .
[ fig : rhoti ] we see the main difference between two cases . in both cases
wds will increase their central density and will be compressed by losing angular momentum through magnetic dipole braking .
however in the case with the magnetic flux conservation the value of @xmath7 will be increasing due to the compression thus causing more torque and evolving faster with respect to the @xmath7=constant case . to calculate how the mean radius of the wd evolves with time we need to rewrite eq .
[ eq : rhot ] as follows @xmath20 where now @xmath21 the values of @xmath22 and @xmath23 are calculated along a given constant rest mass sequence like in the previous case . here
the range of integration of @xmath19 is also defined along the constant rest mass sequence on the border of the stability region . in fig .
[ fig : rmeanti ] the mean ( average ) radius is plotted as a function of time . over the course of time the mean radius decreases , hence wds shrink with time .
for the case of the conserved magnetic flux the mean radius decreases faster than for the case with constant surface magnetic field . the procedure to estimate the evolution of other parameters of rotating wds is analogous to the central density with eq .
[ eq : rhot ] and to the mean radius with eq .
[ eq : rmeant ] . to take into consideration the magnetic flux conservation we need to use eq .
[ eq : flux ] .
consequently , in fig . [
fig : oti ] we plot angular velocity of wds as a function of time . here
we see that a wd of mass @xmath24 spins - up at the beginning and spins - down at the end , whereas super - chandrasekhar wds spin - up only .
this is another confirmation of our previous calculations ( see e.g. refs . ) .
isolated wds regardless of their masses will always lose angular momentum via magnetic dipole braking . by losing angular momentum
wds tend to get more stable configuration by increasing their central density and by decreasing their mean radius .
eventually super - chandrasekhar wds will spin - up and sub - chandrasekhar wds spin - down .
however wds having masses close to the chandrasekhar mass limit at certain time of their evolution will experience both spin up and spin down epochs .
[ fig : oti ] is another illustration of the spin - up and spin - down epochs.@xcite
in this work , in view of our previous computations@xcite we adopted that we have various stable rotating wds at different times in their evolutionary track without involving the details and the routes of their entire evolution . in this regard
, we calculated constant rest mass sequences for selected masses and constructed tables showing the dependence of the wd s parameters on each other along the sequences .
furthermore , by means of the equation for the rotational energy loss of pulsars through magnetic dipole braking , we showed how the main parameters of rotating wds evolve with time . for the sake of comparison we considered two cases with constant magnetic field and constant magnetic flux .
for the magnetic flux conservation case the evolution time ( life time ) turned out to be shorter than for the constant magnetic field .
consequently , we showed that all uniformly rotating wds will be compressed by angular momentum loss via magnetic dipole braking in order to pass from one equilibrium state to another .
we performed all computations in gr by using the hartle formalism for uniformly - rotating configurations for the sake of generality . by exploiting the relativistic feynman - metropolis - teller equation of state
, we considered mainly wds consisting of carbon , although the typical white dwarfs are known not to consist of a pure chemical element , but a mixture of elements such as carbon , oxygen , neon , magnesium , etc
. the consideration of different chemical compositions will be the issue of our future investigations .
this work was supported by grant no .
3101/gf4 ipc-11 of the ministry of education and science of the republic of kazakhstan .
k.b . acknowledges icranet for its hospitality and support .
100 g. fontaine , p.brassard , p. bergeron , _ pasp _ * 113 * ( 782 ) , 409 - 435 ( 2001 ) .
s. o. kepler , i. pelisoli , d. koester , g. ourique , a. d. romero , n. reindl , s. j. kleinman , d. j. eisenstein , a. d. m. valois , l. a. amaral , _ mon . not .
soc . _ * 455 * , 3413 ( 2016 ) . s. l. shapiro , s. a. teukolsky and a.p .
lightman , black holes , white dwarfs , and neutron stars : the physics of compact objects .
new york : wiley - interscience , 1983 .
672 p. j. whelan and i. iben , jr .
, _ astrophys .
j. _ * 186 * , 1007 ( 1973 ) . k. nomoto , _ astrophys .
j. _ * 253 * , 798 ( 1982 ) . z. han and p. podsiadlowski , _ mon . not .
soc . _ * 350 * , 1301 ( 2004 ) .
i. iben , jr . , and a. v. tutukov , _ astrophys . j. suppl
. ser . _ * 54 * , 335 ( 1984 ) .
r. f. webbink , _ astrophys .
j. _ * 277 * , 355 ( 1984 ) .
m. h. van kerkwijk , p. chang , and s. justham , _ astrophys.j.lett . _
* 722 * , l157 ( 2010 ) .
a. kashi and n. soker , _ mon . not .
soc . _ * 417 * , 1466 ( 2011 ) .
m. ilkov and n. soker _
mon . not .
soc . _ * 419*,1695 ( 2012 ) .
k. a. boshkayev , j. a. rueda , r. ruffini and i. siutsou , _ astrophys . j. _ * 762 * , 1 ( 2013 ) .
j. b. hartle , _ astrophys .
j. _ * 150 * 1005 ( 1967 ) .
j. b. hartle and k. s. thorne , _ astrophys . j. _ * 153 * 807 ( 1968 ) .
m. rotondo , j. a. rueda , r. ruffini and s .- s .
c _ * 83 * , 045805 ( 2011 ) .
s. chandrasekhar , _ astrophys .
j. _ * 74 * , 81 ( 1931 ) .
e. e. salpeter , _ astrophys .
j. _ * 134 * , 669 ( 1961 ) .
j. l. friedman , j. r. ipser and r. d. sorkin _ astrophys . j. _ * 325*,722 ( 1988 ) .
m. malheiro , j. a. rueda and r. ruffini _ pasj _ * 64 * , 56 ( 2012 ) . k. a. boshkayev , l. izzo , j. a. rueda and r. ruffini , _ astronomy and astrophys . _ * 555 * , 9 ( 2013 ) .
m. rotondo , j. a. rueda , r. ruffini and s .- s .
xue , _ phys .
d _ * 84 * , 084007 ( 2011 ) .
k. a. boshkayev , j. a. rueda and r. ruffini , _
int.j.m.p.e . _ * 20 * , 136 - 140 ( 2011 ) .
k. a. boshkayev , j. a. rueda and r. ruffini , _
int.j.m.p.conf.ser . _
* 23 * , 193 - 197 ( 2013 ) .
k. a. boshkayev , j. a. rueda , r. ruffini and i. siutsou , _ journal of the korean physical society _ * 65 * , 855 - 860 ( 2014 ) . k. a. boshkayev , j. a. rueda , r. ruffini and i. siutsou , _ proceedings of the mg13 meeting on general relativity . edited by rosquist k. et al . published by wspc _
isbn 9789814623995 , 2468 - 2474 ( 2015 ) . | understanding white dwarf stars final product evolution masses 0.07 to 8 @xmath0 solar mass observed wds masses clustered around 0.5 - 0.7)@xmath0 size planet large densities huge surface gravities low luminosities formation of low mass ( sub - chandrasekhar ) wds progenitors were main sequence stars low or medium mass formation massive white dwarfs super - chandrasekhar ) scenarios single degenerate scenario increases mass through accretion matter double degenerate scenario two mass wds merge losing energy core degenerate scenario merger in common envelop with massive asymptotic branch star investigated relativistic configurations of uniformly rotating sub- super - chandrasekhar wds used relativistic feynman - metropolis - teller equation of state for wd matter equations state stability of rotating wds analyzed mass - shedding limit inverse @xmath1-decay instability pynonuclear instability secular axisymmetric instability turning point method of friedman et al.work explore compression of isolated rotating sub- super - chandrasekhar wds by angular momentum loss based on results ref . fulfilling stability criteria for rotating wds calculate change in time of basic parameters rotating wds consider two cases assume dipolar magnetic constant evolution wds adopt magnetic flux conservation relaxing constancy magnetic field for comparison examine compression of isolated white dwarfs with time constructed constant mass sequences for sub- super - chandrasekhar wds main parameters find dependence on each other using equation for rotational energy loss of pulsars via magnetic dipole braking estimate evolution time from one instability point to another depending parameter modify equation analyze parameter changes in time examine central density with time rewrite equation choose all parameters as function of central density evolution time speed of light in vacuum dipole magnetic field mean spherical radius polar radius equatorial radius angular velocity angular momentum of white dwarf adopt maximum brake angle between rotation axis and magnetic field is ninety degrees values @xmath13 @xmath14 calculated constant baryon rest mass sequence choose carbon white dwarfs using rfmt eos ref . limiting values parameters rotating wds on border stability region determine range integration central density @xmath3 numerical integration of eq . rhot evolution central density with time expression magnetic flux conservation relaxing constancy magnetic field @xmath15 @xmath16 surface dipole magnetic field initial value @xmath7 @xmath18 mean radius initial values @xmath19 at @xmath17 substituting @xmath7 in eq rhot equation evolution of @xmath3 with time magnetic flux conservation fig . difference between two cases both wds increase central density compressed losing angular momentum magnetic dipole braking magnetic flux conservation value @xmath7 due compression causing more torque evolving faster calculate mean radius wd with time rewrite eq . [ : rhot ] @xmath20 values of @xmath22 @xmath23 calculated along constant rest mass sequence range of integration of @xmath19 defined along constant rest mass sequence border stability region fig .fig : rmeanti mean radius function of time radius decreases wds shrink with time conserved magnetic flux mean radius decreases faster than constant surface magnetic field procedure estimate evolution of parameters rotating wds analogous to central density mean radius with magnetic flux conservation use eq . flux ] fig . angular velocity of wds function of time wd of mass @xmath24 spins - up beginning spins - down end super - chandrasekhar wds spin - up only confirmation of previous calculations isolated wds lose angular momentum via magnetic dipole braking losing angular momentum wds stable configuration increasing central density decreasing mean radius super - chandrasekhar wds spin - up sub - chandrasekhar wds spin - down wds masses close to chandrasekhar mass limit experience spin up and spin down epochs fig : oti ] illustration of spin - up - down epochs adopted stable rotating wds at different times in evolutionary track calculated constant rest mass sequences for selected masses constructed tables showing dependence of wd s parametersequation rotational energy loss of pulsars through magnetic dipole braking showed parameters rotating wds evolve with time considered cases constant magnetic field magnetic flux magnetic flux conservation case evolution time shorter constant magnetic field showed uniformly rotating wds compressed by angular momentum loss via magnetic dipole braking equilibrium state performed computations hartle formalism for uniformly - rotating configurations exploiting relativistic feynman - metropolis - equation state considered wds carbon typical white dwarfs not pure mixture of carbon oxygen neon magnesium consideration different chemical compositions issue future investigations work supported by grant 3101/gf4 ipc-11 ministry of education and science kazakhstan acknowledges icranet hospitality support g. fontaine p.brassard. bergeron pasp * 113 * ( 782 ), 409 - 435 ( 2001 ) s. kepler pelisoli koester ourique romero reindl kleinman eisenstein. valois l. amaral 455 *, 3413 ( 2016 ). shapiro. teukolsky a.p lightman black holes white dwarfs neutron stars compact objects new york wiley interscience 1983 672 j. whelan i. iben jr astrophys 186 1007 ( 1973 ) nomoto astrophys 253 798 ( 1982 ) z. han p. podsiadlowski 350 1301 ( 2004 ). iben v. tutukov astrophys 54 335 ( 1984 ) f. webbink astrophys 277 355 ( 1984 ) m h. van kerkwijk p. chang s. justham astrophys l157 ( 2010 ) a. kashi n. soker 417 1466 ( 2011 ) m. ilkov n. soker 419,1695 ( 2012 ). boshkayev j a. rueda r. ruffini i. siutsou astrophys 762 1 ( 2013 ) j b. hartle 150 1005 ( 1967 ) b. hartle k s. thorne 153 807 ( 1968 ) m. rotondo j a. rueda r. ruffini 83 045805 ( 2011 )chandrasekhar astrophys 74 81 ( 1931 ) salpeter astrophys 134 669 ( 1961 ) friedman ipser sorkin astrophys 325*,722 ( 1988 ) malheiro j a. rueda r. ruffini 64 , 56 ( 2012 ) boshkayev l izzo j a. rueda ruffini astronomy astrophys 555 9 ( 2013 ) rotondo j a. rueda r ruffini xue 84 * 084007 ( 2011 ) boshkayev j rueda r. ruffini int.j.m.p 20 136 - 140 ( 2011 ) boshkayev rueda int.j.m.p.conf 23 193 - 197 ( 2013 ) boshkayev rueda ruffini siutsou journal korean physical society 65 - 860 ( 2014 ) boshkayev rueda ruffini siutsou proceedings mg13 meeting general relativity edited rosquist published wspc 9789814623995 2468 - 2474 ( 2015 ) | 0.440875 | 3,050 | 1,541 | , Saving $0.1 in GPT-4. |
67,477 | the measurements performed at lep and slc have substantially improved the precision of the test of the minimal standard model .
the precision is such that there is sensitivity to pure weak radiative corrections .
this allows to indirectly determine the top mass ( = 161@xmath08 gev ) , the w - boson mass ( = 80.37@xmath00.03 gev ) , and to set an upper limit on the the higgs boson mass of 262 gev at 95% confidence level .
= cmr8 1.5pt # 1#2#3#4#1 * # 2 * , # 3 ( # 4 ) 0@xmath1
0@xmath2 * frederic teubert + * + | in the context of the minimal standard model ( msm ) , any electroweak ( ew ) process can be computed at tree level from @xmath3 ( the fine structure constant measured at values of @xmath4 close to zero ) , ( the w - boson mass ) , ( the z - boson mass ) , and ( the cabbibo - kobayashi - maskawa flavour - mixing matrix elements ) . when higher order corrections are included , any observable can be predicted in the `` on - shell '' renormalization scheme as a function of : o_i & = & f_i ( , _ , _ , _ , _ , _ , v_jk ) and contrary to what happens with `` exact gauge symmetry theories '' , like qed or qcd , the effects of heavy particles do not decouple .
therefore , the msm predictions depend on the top mass ( @xmath5 ) and to less extend to the higgs mass ( log(@xmath6 ) ) , or to any kind of `` heavy new physics '' .
the subject of this letter is to show how the high precision achieved in the ew measurements from 0 physics allows to test the msm beyond the tree level predictions and , therefore , how this measurements are able to indirectly determine the value of and , to constrain the unknown value of , and at the same time to test the consistency between measurements and theory . at present
the uncertainties in the theoretical predictions are dominated by the precision on the input parameters .
the w mass is one of the input parameters in the `` on - shell '' renormalization scheme .
it is known with a precision of about 0.07% , although the usual procedure is to take ( the fermi constant measured in the muon decay ) to predict as a function of the rest of the input parameters and use this more precise value .
therefore , the input parameters are chosen to be : input parameter & value & relative uncertainty + ^-1(^2 _ ) = & 128.896 ( 90 ) & 0.07 % + _ ( ^2 _ ) = & 0.119 ( 2 ) & 1.7 % + g _ = & 1.16639 ( 2 ) 10 ^ -5 & 0.0017 % + _ = & 91.1867 ( 21 ) & 0.0023 % + _ = & 173.8 ( 50 ) & 2.9 % + _ > & 89.8 @95% c.l .
& notice that the less well known parameters are , and , of course , the unknown value of .
the next less well known parameter is @xmath7 , even though its value at @xmath8 is known with an amazing relative precision of @xmath9 , ( @xmath10 ) .
the reason for this loss of precision when one computes the running of @xmath3 , ^-1(^2 _ ) & = & is the large contribution from the light fermion loops to the photon vacuum polarisation , @xmath11 .
the contribution from leptons and top quark loops is well calculated but for the light quarks non - perturbative qcd corrections are large at low energy scales . the method so far has been to use the measurement of the hadronic cross section through one - photon exchange , normalised to the point - like muon cross - section , r(s ) , and evaluate the dispersion integral : ( ^ _ ) & = & ( ds ) giving @xcite @xmath12 , the error being dominated by the experimental uncertainty in the cross section measurements .
recently , several new _
`` theory driven '' _
calculations @xcite @xcite have reduced this error by a factor of about 4.5 , by extending the regime of applicability of perturbative qcd ( pqcd ) .
this needs to be confirmed using precision measurements of the hadronic cross section at @xmath13 1 to 7 gev .
the very preliminary first results from bess ii @xcite seem to validate this procedure , being in agreement with the predictions from pqcd .
from the point of view of radiative corrections we can divide the 0 measurements into three different groups : the 0 total and partial widths , the partial width into b - quarks ( @xmath14 ) , and the 0 asymmetries ( ) . for instance , the leptonic width ( ) is mostly sensitive to isospin - breaking loop corrections ( @xmath15 ) , the asymmetries are specially sensitive to radiative corrections to the 0 self - energy , and is mostly sensitive to vertex corrections in the decay @xmath16 .
one more parameter , @xmath17 , is necessary to describe the radiative corrections to the relation between and .
the sensitivity of these three 0 observables and to the input parameters is shown in table [ tab : sensitiv ] .
the most sensitive observable to the unknown value of are the 0 asymmetries parametrised via .
however also the sensitivity of the rest of the observables is very relevant compared to the achieved experimental precision . .
relative error in units of per - mil on the msm predictions induced by the uncertainties on the input parameters .
the second column shows the present experimental errors.[tab : sensitiv ] [ cols="<,^,^,^,^,^",options="header " , ] although the most precise determination of the higgs mass is still coming from the 0 asymmetries , it is clear from table [ tab : consistency ] that any future improvement will be limited by the uncertainty in @xmath18 . if @xmath19 , then the error on @xmath20 is reduced to about 0.23 ( @xmath21 \oplus [ \delta\mathrm{m}_{\mathrm{t}}]$ ] ) , coming only from the 0 asymmetries measurements . the accuracy of the w - boson mass is going to improve by a significant factor in the near future .
however , even if the w mass is measured with a precision of 30 mev , the error on @xmath20 is going to be dominated by @xmath22 and will not be better than 0.23 obtained with the 0 asymmetries only , both determinations being highly correlated by the uncertainty on the top mass . therefore , the error on @xmath20 is not going to improve significantly until a precise measurement of the top mass ( 2 gev ) becomes available .
in such a case , one can easily obtain a precision in @xmath20 close to 0.15 .
this is what is shown in figure [ fig : chi2mh ] .
also shown in figure [ fig : chi2mh ] is the expected direct search limit from lep2 .
if the tendency to prefer a very low value of continues with the new or updated measurements , and the accuracy on the top mass and w - boson mass are improved significantly , consistent with the indirect determinations , we may be able to constrain severely the value of .
the measurements performed at lep and slc have substantially improved the precision of the tests of the msm , at the level of ( 0.1% ) .
the effects of pure weak corrections are visible with a significance larger than three standard deviations from 0 observables and about seven standard deviations from the w - boson mass .
the top mass predicted by the msm fits , ( = @xmath23 gev ) is compatible ( about @xmath24 ) with the direct measurement ( = @xmath25 gev ) and of similar precision .
the w - boson mass predicted by the msm fits , ( @xmath26 ) is in very good agreement with the direct measurement ( @xmath27 ) .
the mass of the higgs boson is predicted to be low , ( _ / ) = & 1.88^+0.33_-0.41 & ( _ = 76^+85_-47 ) + _ & < & 262 @ 95 % this uncertainty is reduced to @xmath28 when the uncertainty from @xmath29 is negligible , and will be further reduced to @xmath30 when is known with a 2 gev precision and is known with a 30 mev precision .
i would like to thank prof .
joan sol and all the organizing committee for the excellent organization of the workshop .
i m very grateful to martin w. gr@xmath31newald and gunter quast for his help in the preparation of the numbers and plots shown in this paper .
i also thank guenther dissertori for reading the paper and giving constructive criticisms . | context minimal standard model electroweak ew process computed tree level from @xmath3 fine structure constant measured at values @xmath4 close to zero w - boson mass z - boson mass cabbibo - kobayashi - maskawa flavour - mixing matrix elements higher order corrections included observable predicted in on - shell '' renormalization scheme function of o_i & = & f_i ( , _ , _ v_jk ) contrary effects heavy particles not decouple msm predictions depend on top mass ( @xmath5 ) less higgs mass ( log(@xmath6 ) heavy new physics letter high precision in ew measurements from 0 physics test msm beyond tree level predictions determine value of constrain unknown value test consistency between measurements theory uncertainties in theoretical predictions dominated by precision on input parameters w mass on - shell renormalization scheme known with precision about 0.07% usual procedure fermi constant measured in muon decay predict function input parameters more precise valueinput parameters chosen value relative uncertainty + ^-1(^2 _ ) 128.896 ( 90 ) 0.07 % + ( ^2 _ ) 0.119 ( 2 ) 1.7 % + g _ 1.16639 ( 2 ) 10 ^ -5 0.0017 % + _ 91.1867 ( 21 ) 0.0023 % + _ 173.8 ( 50 ) 2.9 % + _ 89.8 @95% c.l less known parameters unknown value @xmath7 value @xmath8 known precision @xmath9 loss of precision @xmath3 large contribution light fermion loops photon vacuum polarisation contribution leptons top quark loops calculated light quarks non - perturbative qcd corrections large at low energy scales method measurement hadronic cross section one - photon exchange point - like muon cross - section evaluate dispersion integral @xmath12 error dominated experimental uncertainty cross section measurements new calculations reduced error factor 4.5 extending regime applicability perturbative qcd precision measurements hadronic cross section @xmath13 1 to 7 gev preliminary results ii @xcite validate procedure agreement with predictions pqcdradiative corrections 0 measurements into three groups 0 total partial widths partial width b - quarks @xmath14 0 asymmetries leptonic width sensitive to isospin - breaking loop corrections @xmath15 asymmetries sensitive to radiative corrections 0 self - energy vertex corrections in decay @xmath16 more parameter @xmath17 necessary radiative corrections relation between sensitivity of three 0 observables input parameters shown in table sensitiv most sensitive observable to unknown value 0 asymmetries parametrised sensitivity rest observables relevant compared to experimental precision relative error in units per - mil on msm predictions induced by uncertainties on input parameters second column shows present experimental errors most precise determination of higgs mass from 0 asymmetries future improvement limited by uncertainty in @xmath18 @xmath19 error on @xmath20 reduced to 0.23 ( @xmath21 from 0 asymmetries measurements accuracy of w - boson mass improve futureeven if w mass measured precision 30 mev error on @xmath20 dominated by @xmath22 not better than 0.23 with 0 asymmetries correlated by uncertainty top mass error @xmath20 not until precise measurement top mass ( 2 gev ) available obtain precision in @xmath20 close to 0.15 shown in figure fig : chi2mh ] expected direct search limit from lep2 if low value continues measurements accuracy on top mass w - boson mass improved may constrain value of measurements at lep and slc improved precision of tests msm ( 0.1% ) effects of weak corrections visible larger than three standard deviations from 0 observables seven standard deviations from w - boson mass top mass predicted by msm @xmath23 gev compatible with direct measurement @xmath25 gev ) similar precision w - boson mass @xmath26 ) with direct measurement @xmath27 ) mass higgs boson predicted low = 1.88^+0.33_-0.41 _ = 76^+85_-47 ) 262 @ 95 % uncertainty reduced to @xmath28 uncertainty @xmath29 negligible reduced @xmath30 2 gev precision 30 mev precision thank prof . joan sol organizing committee excellent organization workshop grateful martin w. gr@xmath31newald gunter quast help preparation numbers plots paper thank guenther dissertori constructive criticisms | 0.314449 | 1,872 | 943 | , Saving $0.1 in GPT-4. |
67,479 | this short piece celebrates haldane s seminal j. phys .
c * 14 * , 2585 ( 1981 ) paper laying the foundations of the modern theory of luttinger liquids in one - dimensional systems . | in a pioneering paper published in j. phys . c in 1981 @xcite ,
duncan haldane taught us that one - dimensional fermionic systems behave in a fundamentally different manner than those in the universality class described by landau s fermi liquid theory @xcite-@xcite . at the simplest level
, one - dimensionality alters the kinematics of simple excitations around a fermi sea ( fig .
[ fig:1d_fermi_sea ] ) .
the spectrum is formed by particle / hole excitations in the vicinity of the two fermi points ( fig .
[ fig:1d_1ph_exc ] ) , and their ( multiple ) umklapp modes . in 1d , in contrast to higher dimensions , forbidden regions of the frequency @xmath0-momentum @xmath1 plane exist .
low - energy modes around integer multiples of the fermi momentum @xmath2 are always present ( fig .
[ fig:1d_multiph_continuum ] ) , but there exist no low - energy excitations in the lobes between points @xmath3 and @xmath4 for any integer @xmath5 . for energies ( much ) below the height of the lobes ( so @xmath6 in which @xmath7 is the fermi velocity ) , we can thus explicitly separate the excitations into different sectors labelled by an even integer @xmath8 .
is the one - particle dispersion relation .
the chemical potential @xmath9 sets the value of the fermi wavevector @xmath2 .
filled single - particle states are represented by black dots , unfilled ones by open circles .
right : one particle - hole excitation continuum .
the a , b labels refer to the location within the continuum of the particular single particle - hole excitations sketched in fig .
[ fig:1d_1ph_exc ] . ,
title="fig:",width=302 ] is the one - particle dispersion relation .
the chemical potential @xmath9 sets the value of the fermi wavevector @xmath2 .
filled single - particle states are represented by black dots , unfilled ones by open circles .
right : one particle - hole excitation continuum .
the a , b labels refer to the location within the continuum of the particular single particle - hole excitations sketched in fig .
[ fig:1d_1ph_exc ] .
, title="fig:",width=302 ] the lowest - energy state of sector @xmath10 is the @xmath5-th umklapp state at momentum @xmath4 .
these states are ` persistent current ' modes obtained by a galilean transformation of the ground state , giving momentum @xmath11 to the whole system , and carrying a quadratic ( in @xmath5 ) energy shift . above the persistent current umklapp states , but for energies still low on the scale of @xmath12 , the spectrum of the theory is expected to be given by a linear in momentum boson ( sound wave ) spectrum ( up to nonlinear corrections ) .
similarly , adding @xmath13 particles shifts the fermi momentum linearly ( in @xmath13 ) , and the energy quadratically .
we can thus immediately guess , following haldane s reasoning in @xcite , that an effective theory for our fermions universally takes the form @xmath14 where @xmath15 is the ground - state energy and @xmath16 and @xmath17 respectively denote sound , charge and current velocities .
it was haldane s great insight to realize that in the presence of interactions , the effective description based on eq .
( [ eq : haldaneguess ] ) remains valid , albeit with renormalized parameters @xmath16 and @xmath17 ( constrained by @xmath18 ) .
this theory , which is thus universally valid for gapless one - dimensional systems ( be they based on underlying fermionic , bosonic or spin degrees of freedom ) , should be viewed as a ` theory of everything ' in 1d , similarly to the status that landau s fermi liquid theory has achieved in higher dimensions .
this is the point of view so eloquently put forward in haldane s paper @xcite .
many important contributions paved the way for this synthesis .
the development of bosonization goes back almost to the very beginnings of quantum mechanics . in 1934 , bloch @xcite used the fact that 1d fermions have the same type of low - energy excitations as a harmonic chain in his study of incoherent x - ray diffraction .
some years later , in 1950 , tomonaga @xcite applied bloch s sound wave method to interacting fermions in 1d .
his main contribution was probably to realize that the physical density operator splits up into left- and right - moving modes , obeying a bosonic hamiltonian .
he however missed the @xmath19 contribution to the density correlation function , and did not notice the anomalous decay of correlations .
the fundamental paper of luttinger @xcite follows in 1963 , in which ( perhaps unaware of tomonaga s 1950 paper ) he formulated his ( part)namesake model .
using toeplitz determinants , he found that the average occupation in the ground state behaves as a power - law with an interaction - dependent anomalous dimension .
he thus realized the crucial fact that the fermi surface discontinuity is destroyed by interactions in 1d .
this work was however incorrect in its treatment of the commutation relations of the density operators .
later , mattis and lieb offered a correct treatment in their seminal paper @xcite .
the idea of bosonization , namely that bosons could be used to construct a complete set of states of a 1d fermionic system , appeared in 1965 in @xcite .
subsequently , early computations of correlation functions appeared in @xcite and @xcite .
in particular , in the first of these , theumann noticed the absence of single - particle poles in the green s function .
she thus correctly concluded that single - particle excitations are absent in such theories . in their famous 1974 paper @xcite ,
dzyaloshinskii and larkin recovered the absence of single - particle pole and of fermi surface discontinuity , and offered an interpretation of mattis and lieb s solution starting from conventional diagrammatic perturbation theory .
an early version of the actual bosonization operator identity appeared in @xcite .
this was refined by mattis in 1974 @xcite , rendering calculation of correlation functions straightforward .
similar results appeared in the work of luther and peschel @xcite .
the power - law form for correlations was also recovered from equations of motion techniques in @xcite .
bosonization was then applied to spin chains and vertex models in @xcite .
the first precise field - theoretical bosonization formula ( as an operator identity ) including the ( until that point neglected ) particle number raising / lowering operators ( klein factors ) is in general attributed to @xcite . an early formulation of luttinger liquid concepts appeared in 1975 in @xcite .
but the fact remains that it is haldane , in a remarkable series of papers , who gave luttinger liquid theory the form it has today .
starting in @xcite , he gave the first explicit construction of charge - raising operatos ( klein factors ) .
subsequently , in @xcite , @xcite and most notably in @xcite he offered the complete and explicit construction of the bosonization operator identities , and cross - checked results with exactly solvable models .
most importantly , he proposed the concept of the luttinger liquid ( as he so defined it ) as the proper replacement for the fermi liquid in one dimension , and showed that many different types of systems of fermions , bosons and spins belong to this new universality class .
this realization sparked much of the revolutionary advances achieved in low - dimensional quantum systems over the last 35 years . | paper j. phys . c 1981 duncan haldane taught one - dimensional fermionic systems behave landau s fermi liquid theory one - dimensionality alters kinematics excitations around fermi sea spectrum formed by particle / hole excitations two fermi points_exc multiple umklapp modes 1d forbidden regions frequency-momentum plane exist low - energy modes around integer multiples fermi momentum @xmath2 present_multiph_continuum no low - energy excitations in lobes between points @xmath3 and @xmath4 integer @xmath5 energies below height lobes separate excitations into sectors labelled by even integer one - particle dispersion relation chemical potential @xmath9 sets value fermi wavevector @xmath2 filled single - particle states represented by black dots unfilled by open circles one particle - hole excitation continuum a b labels refer to location continuum of single particle hole excitations fig fig:1d_1ph_exc one - particle dispersion relation chemical potential @xmath9 sets value fermi wavevector @xmath2filled single - particle states represented by black dots unfilled open circles one particle - hole excitation continuum a b labels refer to location continuum single particle - hole excitations in fig . lowest - energy state of sector @xmath10 is @xmath5-th umklapp state at momentum @xmath4 states persistent current modes by galilean transformation ground state giving momentum @xmath11 system quadratic energy shift persistent current energies low scale @xmath12 spectrum theory by linear momentum boson spectrum adding @xmath13 particles shifts fermi momentum linearly energy quadratically effective theory for fermions takes form @xmath14 @xmath15 ground - state energy @xmath16 @xmath17 denote sound charge current velocities haldane s interactions effective description based eq remains valid with renormalized parameters @xmath16 and @xmath17 constrained by @xmath18 theory universally valid for gapless one - dimensional systems fermionic bosonic spin freedom viewed as theory of everything ' in 1d landau s fermi liquid theory in higher dimensionspoint of view in haldane s paper @xcite contributions paved for synthesis development of bosonization beginnings quantum mechanics 1934 bloch @xcite used 1d fermions low - energy excitations harmonic chain incoherent x - ray diffraction 1950 tomonaga applied s sound wave method to interacting fermions in 1d contribution physical density operator splits into left- and right - moving modes bosonic hamiltonian missed @xmath19 contribution to density correlation function notice decay of correlations fundamental paper luttinger @xcite 1963 formulated part)namesake model average occupation in ground state as power - law with interaction - dependent dimension fermi surface discontinuity destroyed by interactions in 1d work incorrect in commutation relations of density operators mattis and lieb correct treatment in paper @xcite idea of bosonization construct states 1d fermionic system appeared in 1965 in @xcite early computations of correlation functions in @xcite theumann noticed absence of single - particle poles in green s function concluded excitations absent1974 paper @xcite dzyaloshinskii larkin recovered single - particle pole fermi surface discontinuity interpretation mattis lieb s solution diagrammatic perturbation theory early bosonization operator identity @xcite refined by mattis 1974 @xcite calculation correlation functions straightforward similar results luther peschel @xcite power - law form correlations recovered from equations motion techniques @xcite bosonization applied to spin chains vertex models @xcite first precise field - theoretical bosonization formula operator identity particle number raising / lowering operators klein factors attributed @xcite early formulation luttinger liquid concepts 1975 @xcite haldane gave luttinger liquid theory form @xcite first construction charge - raising operatos @xcite offered complete construction bosonization operator identities cross - checked results with solvable models proposed concept luttinger liquid replacement for fermi liquid showed systems fermions bosons spins belong new universality class sparked revolutionary advances low - dimensional quantum systems last 35 years | 0.271144 | 1,702 | 855 | , Saving $0.1 in GPT-4. |
67,486 | it has been proposed that the geometry of an extra dimension could automatically adjust itself to compensate for an arbitrary energy density on the 3-d brane which we are presumed to inhabit , such that a static solution to einstein s equation results .
this would solve the long - standing cosmological constant problem , of why our universe is not overwhelmed by the enormous energy of the quantum vacuum fluctuations predicted by quantum field theory .
i will review some of the attempts along these lines , and present a no - go theorem showing that these attempts are doomed , at least within one of the most promising classes of models .
address = physics department , mcgill university , montral , qubec , canada h2a 2t8 , altaddress = presenter address = physics department , mcgill university , montral , qubec , canada h2a | one of the most vexing problems in theoretical particle physics is the magnitude of the vacuum energy density , @xmath0 .
ironically , einstein introduced it into general relativity through the field equation g _
r_-12 g_r = 8g(t _ + g _ ) in order to find static cosmological solutions .
he called it his `` biggest blunder '' when the universe was subsequently found to be expanding .
generically the presence of a positive cosmological constant leads to a universe which is accelerating , not static , but from the friedmann equation we see that the scale factor can be static for a fine - tuned value of @xmath0 , ( aa)^2 = 8g3(+ ) - ka^2 ( in fact one must also tune @xmath1 to be zero ) .
the irony of einstein s supposed blunder is three - fold .
first , of course , the universe is not static ; second , particle theorists came to realize that @xmath0 should be present due to quantum fluctuations of the vacuum , which can be visualized as spontaneous creation and annihilation of particle / antiparticle pairs .
third , there is now strong evidence from cosmology that @xmath0 is nonzero , ( 2.410 ^ -3)^4 since the universe appears to be accelerating and at critical density , even though there is not enough dark matter to account for most of the energy density .
the big problem is that naive computations of the theoretical value of @xmath0 from quantum field theory give a value which is many orders of magnitude greater , _ theo . ~(10
^ 28)^4 there must be some mechanism for explaining the difference between the observed and expected values , but so far no really convincing idea has been proposed .
one might suspect some kind of adjustment mechanism is at work , which somehow nullifies the effect of @xmath0 no matter what its underlying value might be .
however weinberg has given a no - go theorem against such ideas @xcite .
for example , one might imagine that the effective physical value of @xmath2 $ ] depends on the underlying value @xmath3 and upon a scalar field @xmath4 which automatically adjusts itself so that @xmath2=0 $ ] . to be concrete ,
consider @xmath5 as shown in fig .
[ fig1](a ) .
although there are values of @xmath4 where @xmath6 , they are not generally stationary , so fine tuning is still required .
although this is very obvious in the present example , weinberg s theorem shows that essentially the same problem will plague even more clever attempts to implement such an idea .
[ fig1 ] weinberg s theorem assumes that the universe is 4-dimensional , so one might hope that recent brane - world ideas might provide a loophole . in the braneworld picture illustrated in fig.1(b ) , we have an extra dimension , @xmath7 , surrounding our 4-d universe which is presumed to be the brane at @xmath8 .
one could imagine that the effects of the vacuum energy density on the brane are counteracted by some new physics in the bulk .
if the 5-d nature of this setup ca nt be described by an approximate 4-d picture , then it might be possible to evade the no - go theorem .
some attempts at constructing a self - tuning solution to the cosmological constant problem were presented in references @xcite .
the idea is to add a scalar field @xmath4 in the bulk , which is described by the metric ds^2 = a^2(y ) dx_dx^+ dy^2 the scalar is presumed to couple to the bare energy density of the brane , @xmath3 through a potential @xmath9 .
the coupled field equations for the scalar and the metric give rise to a solution which is singular at some position @xmath10 in the bulk , as illustrated in fig . 2 .
the value of @xmath10 depends on @xmath3 , which is what constitutes the self tuning . in other words , for any value of @xmath3 , a static solution can be found with some value of @xmath10 .
it is precisely because the solution is static rather than exponentially expanding that an observer would infer that the physical value of @xmath0 is zero .
[ fig2 ] there are a number of problems with this idea @xcite .
for one , the static solution is not unique : one can also find expanding or contracting ones @xcite , which shows that fine tuning of the initial conditions are needed to get the static solution .
moreover we do nt like the presence of naked singularities .
the cosmic censorship hypothesis asserts that singularities should be hidden behind an event horizon , as is the case with a black hole .
_ @xcite have proposed a modification to the previous self - tuning model which includes a horizon , as shown in fig .
2(b ) . the points @xmath11 are identified with each other by imposing a @xmath12 orbifolding . in this solution ,
the metric becomes the ads - reissner - nrdstrom solution shown in fig . 3 , ds^2 = -h(y ) dt^2 + a^2(y )
dx^2 + h^-1(y ) dy^2 with @xmath13 and @xmath14 ; @xmath15 is the radius of curvature of the 5-d anti - desitter space , @xmath16 is mass of the black brane and @xmath17 is its charge under a u(1 ) gauge symmetry .
the self - tuning is now accomplished by @xmath16 and @xmath17 taking on the appropriate values to cancel the effect of the energy density @xmath18 which is on the brane .
note that these quantities appear not as inputs to the lagrangian , but rather as integration constants in the bulk solution .
however , this @xmath18 can not be quite the same thing as a cosmological constant because the self - tuning solution works only for a rather bizarre equation of state for the energy density on the brane : p < -this is in contrast to a vacuum energy density which obeys @xmath19 , and is in violation of the weak energy condition .
such violations may not always be bad , but at least in the case of a classical scalar field theory they are problematic .
there we have @xmath20 and @xmath21 so that @xmath22 implies @xmath23 : negative kinetic energy .
this would correspond to a ghost which leads to nonconservation of probability in quantum field theory .
our goal in the present work was to try to find a self - tuning solution with a horizon and without violating the weak energy condition .
unfortunately , we found instead another no - go theorem : one must have @xmath24 either on the brane or in the bulk , as we will now describe @xcite . [ fig3 ] for the 5-d ads - reissner - nrdstrom spacetime.,title="fig : " ] before discovering our no - go result ,
we searched for a generalization of the original solution which would hopefully give us more leeway . a natural possibility is to try adding a scalar field in the bulk , so the lagrangian becomes = ( 12 g^ _ _ -v ( ) - v_0()(y ) ) we obtained numerical solutions using various potentials and found empirically that a horizon could be obtained only when the black hole charge @xmath25 became negative .
interestingly , @xmath26 implies that @xmath24 for the bulk stress energy tensor contributed by the gauge field .
it is not difficult to show that the violation of the weak energy condition is actually necessary for obtaining a horizon . by adding together the ( 00 ) and one of the spatial ( @xmath27 ) components of the einstein equations , and integrating from the outer horizon ( @xmath28 ) to the brane ( @xmath29 ) , we obtain [ nogo ] h(y_h ) = - ^2a^3(2_y_h^y_b ( t^0 _ 0-t^1 _ 1 ) a^3 dy + ( + p)|_y_b ) where @xmath30 is the 5-d gravitational constant and @xmath31 is the bulk stress energy tensor . in particular , @xmath32 is the bulk contribution to @xmath33 .
notice the minus sign in front of everything , and the fact that @xmath34 must be positive at the outer horizon ( see fig .
these two facts conspire to need @xmath35 to be negative somewhere , in the bulk , the brane or both .
this conclusion is independent of any details of the theory like the choice of scalar potentials .
we have also found it to be robust against other generalizations : arbitrary field content in the bulk , addition of higher derivative terms to the gravitational action , such as gauss - bonnet , and also relaxation of the @xmath12 symmetry we assumed around the brane .
for example , the gb action is s = 12 ^ 2d^5x(r + ( r^2 - 4r_abr^ab + r_abcdr^abcd ) ) we find that the no - go result ( [ nogo ] ) gets modified by the factor .(1-(a/a)^2 h ) h |_y_h = but since @xmath36 at the horizon by definition , the left - hand - side is in effect unmodified . relaxing the @xmath12 symmetry amounts to having two different black holes with different values of @xmath16 and @xmath17 on the two sides of the branes , with horizons at different distances .
this does not change the conclusion .
there is one apparent loophole , which is to allow for spatial curvature on the brane .
the friedmann equation becomes @xmath37 in 4-d , where @xmath38 for positive , zero or negative spatial curvature , respectively . in our no - go theorem
we assumed @xmath39 , which is consistent with observations of the cosmic microwave background and expectations from inflation . with this modification
we obtain h(y_h ) = + 4ka^3_y_h^y_b a dy this term allows us to obtain a positive value for @xmath40 even if @xmath41 everywhere , provided that @xmath42 .
however , this does not solve the cosmological constant problem because it requires that @xmath3 already be of order @xmath43 ev@xmath44 , which is the original fine - tuning problem all over again .
larger values of @xmath3 would require larger curvatures , which are observationally excluded . as a final attempt
, one might try the same trick except with a very large curvature in some hidden extra dimensions rather than the large ones .
however , the new terms in the einstein equations which depend on the extra - dimensional curvature exactly cancel out of the combination @xmath45 which gave us the no - go result , so this does not help .
it seems that nature abhors the self - tuning idea .
each time we attempt to cure one problem , it introduces a new pathology .
ref.@xcite tried to cure the problem of the naked singularity by hiding it behind a horizon , but had to violate the weak energy condition to do so .
we find that allowing for positive spatial curvature can cure this problem , but only at the price of reintroducing the original fine - tuning problem .
it is a good illustration of why the cosmological constant problem is one of the most daunting in theoretical physics .
some progress with self - tuning seems to have been made recently in ref.@xcite .
however , it remains a challenge to understand how constants of integration which correspond to physical sources of stress energy can dynamically adjust in a situation where the vacuum energy is changing , as in a phase transition , or alternatively , how they can `` know '' which value is the ultimate one that must be canceled .
it would seem to be a very intelligent mechanism that could fully allow for the effects of vacuum energy during inflation but not in the present .
99 s. weinberg , rev .
phys . * 61 * , 1 ( 1989 ) .
n. arkani - hamed , s. dimopoulos , n. kaloper and r. sundrum , phys .
b * 480 * , 193 ( 2000 ) [ hep - th/0001197 ] .
s. kachru , m. schulz and e. silverstein , phys .
d * 62 * , 045021 ( 2000 ) [ hep - th/0001206 ] . s. forste , z. lalak , s. lavignac and h. p. nilles , jhep * 0009 * , 034 ( 2000 ) [ hep - th/0006139 ] .
p. binetruy , j. m. cline and c. grojean , phys .
b * 489 * , 403 ( 2000 ) [ hep - th/0007029 ] . c. csaki , j. erlich and c. grojean , nucl .
b * 604 * , 312 ( 2001 ) [ hep - th/0012143 ] . c. csaki , j. erlich and c. grojean , `` the cosmological constant problem in brane - worlds and gravitational lorentz violations , '' gr - qc/0105114 .
j. m. cline and h. firouzjahi , phys .
d * 65 * , 043501 ( 2002 ) [ arxiv : hep - th/0107198 ] .
p. binetruy , c. charmousis , s. c. davis and j. f. dufaux , arxiv : hep - th/0206089 . | vexing in particle physics is vacuum energy density @xmath0 einstein introduced it into relativity through equation g _ r_-12 g_r = 8g(t _ + g _ ) find static cosmological solutions called biggest blunder when universe found expanding positive cosmological constant leads to universe accelerating not static friedmann equation scale factor can be static for fine - tuned value @xmath0 (^2 = 8g3(+ ) - ka^2 must tune @xmath1 to be zero irony of einstein s blunder three - fold universe not static particle theorists @xmath0 should be present due to quantum fluctuations vacuum strong evidence that @xmath0 is nonzero ^4 universe appears accelerating at critical density not enough dark matter for energy density naive computations of theoretical value @xmath0 from give value orders magnitude greater must mechanism for explaining difference between observed and expected values no convincing idea proposed might suspect adjustment mechanism effect of @xmath0 weinberg no - go theorem against such ideasimagine effective physical value of @xmath2 depends on value @xmath3 scalar field @xmath4 adjusts so @xmath2=0 $ ] consider @xmath5 in fig .(a values of @xmath4 @xmath6 not stationary fine tuning required weinberg s theorem shows same problem attempts theorem assumes universe 4-dimensional recent brane - world ideas provide loophole braneworld picture.1(b ) extra dimension @xmath7 surrounding 4-d universe presumed brane at @xmath8 effects of vacuum energy density on brane counteracted by new physics in bulk if 5-d evade no - go theorem attempts self - tuning solution cosmological constant problem presented in references idea add scalar field @xmath4 in bulk described by metric ds^2 = a^2(y ) dx_dx^+ dy^2 scalar to couple to bare energy density of brane @xmath3 potential @xmath9 coupled field equations for scalar metric solution singular at position @xmath10 in bulk illustrated in fig . 2 value of @xmath10 depends on @xmath3 self tuning value @xmath3 static solution with @xmath10 physical value @xmath0 zero problems with idea @xcite static solution not unique expanding or contracting ones fine tuning conditions needed static solution naked singularities cosmic censorship hypothesis singularities hidden behind event horizon black hole @xcite proposed modification self - tuning model horizon . 2(b ) points @xmath11 identified @xmath12 orbifolding metric becomes ads - reissner - nrdstrom solution fig . 3 , ds^2 = -h(y ) dt^2 + a^2(y ) dx^2 + h^-1(y ) dy^2 with @xmath13 and @xmath14 ; @xmath15 radius of curvature of 5-d anti - desitter space @xmath16 mass black brane @xmath17 charge under u(1 ) gauge symmetry self - tuning by @xmath16 and @xmath17 values to cancel effect energy density @xmath18 on brane quantities not inputs lagrangian integration constants in bulk solution@xmath18 as cosmological constant self - tuning solution works for bizarre equation for energy density on brane contrast to vacuum energy density obeys @xmath19 violation of weak energy condition violations classical scalar field theory problematic @xmath20 and @xmath21 @xmath22 implies @xmath23 negative kinetic energy to ghost leads to nonconservation of probability in quantum field theory goal self - tuning solution with horizon without violating weak energy condition found no - go theorem must have @xmath24 on brane or in bulk fig3 nrdstrom before no - go result searched for generalization of original solution more leeway possibility adding scalar field in bulk lagrangian becomes = ( 12 g^ _ _ -v ( - v_0()(y obtained numerical solutions horizon obtained when black hole charge @xmath25 negative @xmath26 implies @xmath24 for bulk stress energy tensor by gauge field violation of weak energy condition necessary for obtaining horizonadding 00 spatial @xmath27 components einstein equations integrating outer horizon @xmath28 to brane @xmath29 [ nogo ] h(y_h ) = - ^2a^3(2_y_h^y_b ( t^0 _ 0-t^1 _ 1 ) a^3 dy + ( + p)|_y_b ) @xmath30 5-d gravitational constant @xmath31 bulk stress energy tensor @xmath32 bulk contribution to @xmath33 minus sign @xmath34 positive at outer horizon @xmath35 negative bulk brane conclusion independent of theory scalar potentials robust against generalizations arbitrary field content bulk addition higher derivative terms gravitational action relaxation @xmath12 symmetry brane gb action s = 12 ^ 2d^5x(r + r^2 - 4r^ab no - go result modified by factor .(1-(a/a)^2 h ) h |_y_h = @xmath36 at horizon left - hand - side unmodified relaxing @xmath12 symmetry two different black holes different values @xmath16 @xmath17 sides branes different distances change conclusion loophole spatial curvature on branefriedmann equation becomes @xmath37 in 4-d @xmath38 for positive zero negative spatial curvature no - go theorem assumed @xmath39 consistent with cosmic microwave background expectations inflation modification h(y_h ) = + 4ka^3_y_h^y_b allows positive value for @xmath40 @xmath41 everywhere @xmath42 solve cosmological constant problem requires @xmath3 order @xmath43 ev@xmath44 original fine - tuning problem larger values @xmath3 require larger curvatures observationally excluded final attempt same trick large curvature in hidden extra dimensions new terms einstein equations extra - dimensional curvature cancel combination @xmath45 no - go result help nature abhors self - tuning idea problem introduces new pathology ref.@xcite tried cure problem naked singularity behind horizon weak energy condition allowing positive spatial curvature cure problem reintroducing original fine - tuning problem cosmological constant problem daunting in theoretical physics progress with self - tuning made in ref.@xcitechallenge understand constants integration stress adjust vacuum energy phase transition value ultimate intelligent mechanism effects vacuum energy inflation not present weinberg phys . 61 * 1 1989 ) arkani - hamed dimopoulos kaloper sundrum phys 480 * 193 ( 2000 ) hep - th/0001197 kachru schulz silverstein d 62 * 045021 ( 2000 ) hep - th/0001206 s forste z. lalak lavignac. nilles 0009 * 034 ( 2000 ) hep - th/0006139 binetruy j m. cline c. grojean b 489 403 ( 2000 ) hep - th/0007029. csaki j erlich grojean 604 * 312 ( 2001 ) hep - th/0012143 csaki erlich grojean cosmological constant problem brane - worlds gravitational lorentz violations - qc/0105114 m cline firouzjahi phys . d * 65 * 043501 ( 2002 ) hep - th/0107198 p. binetruy. charmousis. davis f. dufaux hep - th/0206089 | 0.617275 | 2,979 | 1,543 | , Saving $0.1 in GPT-4. |
67,497 | alice is the dedicated heavy - ion experiment at the lhc . due to the unique particle identification capabilities of the central barrel detectors ( @xmath0 )
, @xmath1 can be measured in the di - electron channel in the very demanding environment of central pb
pb collisions at the lhc . in addition @xmath1 are measured at forward rapidity ( @xmath2 ) with a dedicated muon spectrometer .
alice is the only lhc experiment with an acceptance for @xmath1 that reaches down to @xmath3 at both , mid- and forward rapidity .
preliminary results on the nuclear modification factor of the inclusive @xmath1 production at mid- and forward rapidity in pb
pb collisions at @xmath4tev are presented .
= 1 alice , quarkonia , nuclear modification | heavy quarkonium states , such as the @xmath1 , are expected to provide essential information on the properties of high - energy heavy - ion collisions where the formation of a quark - gluon plasma ( qgp ) is expected .
the impact on the @xmath1 production of such a hot and dense medium formed in the early times of the collision has been extensively studied at sps and rhic energies .
it is expected that due to colour screening mechanisms @xmath1 production is suppressed in a plasma of quarks and gluons @xcite and thus provides a unique probe for qgp formation .
@xmath1 suppression , however , is also induced by other effects that have to be taken into account , such as nuclear shadowing , cold nuclear matter effects and comover absorption scenarios . at lhc energies charm is produced abundantly in central pb
pb collisions and scenarios where originally uncorrelated charm and anti - charm quarks ( re)combine gain importance . such scenarios are described in several statistical and transport models ( e.g. @xcite ) . measuring the @xmath1 production at lhc will help to disentangle between the different mechanisms .
alice @xcite measures the @xmath1 production at mid - rapidity ( @xmath5 ) in the di - electron channel , as well as at forward rapidity ( @xmath2 ) in the di - muon decay channel .
the di - electron analysis is based on a dataset of pb pb collisions at @xmath6tev , taken with a minimum bias trigger in 2010 . in total 12.8 m ( 0 - 80% most central ) events
were analysed corresponding to an integrated luminosity of 2.1@xmath7 .
fig.[fig1 ] ( left ) shows the invariant mass spectrum of the selected @xmath8 candidates . in the top panel
the opposite sign spectrum of the same event ( black ) is displayed together with the background distribution obtained from event mixing ( red ) .
the background distribution is scaled to match the integral of the signal distribution in an invariant mass range between 3.2 and 4gev/@xmath9 .
the bottom panel shows the background subtracted distribution .
signal extraction is performed by integrating the signal counts in an invariant mass range of 2.92 - 3.16gev/@xmath9 . in total
about 2000 @xmath1 are available for the analysis .
this statistics allowed a signal extraction in three centrality intervals . in case of the di - muon analysis
a dataset at the same collision energy , measured in 2011 , was analysed .
a dedicated di - muon trigger allowed for recording 70@xmath7 of integrated luminosity , corresponding to 17.7 m triggered events .
this corresponds to a factor 20 higher statistics than previously reported @xcite .
fig.[fig1 ] ( right ) shows the invariant mass spectrum of @xmath10 candidates for 0 - 90% most central events .
the signal is extracted by performing a combined fit for the background and signal contribution .
the background is described by a gaussian with a width ( @xmath11 ) which varies as a function of the mass value . for the signal description
a modified crystal ball function was used which is a convolution of a gaussian and power law functions that can fit the tails of the measured signal . in total about 40k @xmath1 have been found .
this statistics allows for a differential analysis of the nuclear modification factor in either nine bins of centrality , seven bins in transverse momentum or six bins in rapidity .
the raw yield has been corrected for acceptance and efficiency .
for the di - electron analysis hijing simulations enriched with primary @xmath1 were used . in case of the di - muon analysis monte carlo ( mc )
@xmath1 were embedded into real pb
pb events . for the primary @xmath1 sample in both cases
a parametrisation was used for the transverse momentum and rapidity shape , obtained from an interpolation of rhic , fermilab and lhc data @xcite .
in addition the effects of shadowing were taken into account using eks98 calculations @xcite .
the polarisation of the simulated @xmath1 was assumed to be zero .
first measurements of the @xmath1 polarisation in pp collisions at @xmath12tev @xcite ( @xmath13 , @xmath14gev/@xmath15 ) , show results compatible with zero . in both cases
there is only a weak dependence of the overall efficiency on the collision centrality . for the di - electron channel the total acceptance times efficiency is about 8% , in case of the di - muon channel it is about 14% . ) , the right figure the di - muon channel at forward rapidity ( @xmath2 ) .
for details see text.,title="fig:",scaledwidth=40.0% ] ) , the right figure the di - muon channel at forward rapidity ( @xmath2 ) . for details see text.,title="fig:",scaledwidth=40.0% ]
the nuclear modification factor @xmath16 is defined as : @xmath17 where the @xmath1 yield , @xmath18 , is the number of @xmath1 ( @xmath19 ) corrected for acceptance times efficiency ( @xmath20 ) , the branching ratio ( @xmath21 ) and normalised to the number of analysed minimum bias interactions @xmath22 .
@xmath23 is the nuclear overlap function , obtained from a glauber mc , @xmath24 is the @xmath1 production cross - section in pp collisions at the same centre of mass energy per nucleon pair , which was measured by alice @xcite . as a function of the average number of participant nucleons at mid- ( blue ) and forward ( red ) rapidity.,scaledwidth=45.0% ]
fig.[fig2 ] shows the preliminary @xmath16 at mid- and forward rapidity as a function of the collision centrality , expressed by the number of participant nucleons .
the error bars correspond to the statistical uncertainties , the boxes to the uncorrelated systematic uncertainties of the pb
pb measurement .
the main contribution to the systematic uncertainty in case of the di - electron channel results from the imperfections of the description of the combinatorial background .
other contributions are the uncertainty in @xmath23 , uncertainties in the mc description of the detector and possible variations of the mc input spectrum used for the @xmath1 . in case of the di - muon channel
@xmath23 carries the main uncertainty , followed by the signal extraction and the trigger and tracking efficiency .
global systematic uncertainties are quoted in the figure legend : 26% for the di - electron and 14% for the di - muon channel , where the main contribution is due to the respective pp reference .
it has to be remarked that the uncertainties in the di - electron measurement are rather large , due to the low statistics that was available for the analysis .
fig.[fig3 ] ( left ) shows the centrality integrated @xmath16 as a function of rapidity .
the mid - rapidity point is for 0 - 80% most central events , the forward region corresponds to 0 - 90% most central events .
the open symbols are the forward results reflected at mid - rapidity .
it can be clearly seen that @xmath16 decreases with rapidity .
fig.[fig3 ] ( right ) shows @xmath16 as a function of @xmath25 as measured by alice ( red ) compared to cms data @xcite ( blue ) as well as results from phenix @xcite ( black ) at a lower collision energy .
the @xmath16 is decreasing from 0.6 at low @xmath25 to about 0.4 at higher @xmath25 .
the cms results ( 0 - 100% centrality , @xmath26 , @xmath27gev/@xmath15 ) are in agreement with the alice measurements ( 0 - 90% centrality , @xmath13 , @xmath28 ) in the overlapping transverse momentum range .
the lower energy results from phenix ( 0 - 20% centrality , @xmath29 ) show a significantly smaller @xmath16 .
fig.[fig4 ] compares the alice results at mid- ( left ) and forward ( right ) rapidity with results from a statistical hadronization model @xcite , as well as two different transport models @xcite .
all models take into account ( re)combination of @xmath30-pairs . within the uncertainties
the models describe the data for @xmath31 larger than 50 . as a function of the average number of participants at mid- ( left ) and forward ( right ) rapidity in comparison with model calculations that take into account ( re)combination of @xmath15 and @xmath32 quarks.,title="fig:",scaledwidth=45.0% ] as a function of the average number of participants at mid- ( left ) and forward ( right ) rapidity in comparison with model calculations that take into account ( re)combination of @xmath15 and @xmath32 quarks.,title="fig:",scaledwidth=45.0% ]
alice has measured the nuclear modification factor ( @xmath16 ) of inclusive @xmath1 production in pb
pb collisions at @xmath33tev at mid- and forward rapidity . in comparison with results at rhic energies ,
the @xmath16 at lhc energies is significantly larger .
the cms results , measured at slightly lower rapidity , are in agreement with the alice data in the overlapping transverse momentum range ( @xmath34gev/@xmath15 ) .
it was shown that @xmath16 decreases with increasing @xmath25 and towards larger rapidity .
a comparison with several models that take into account a ( re)combination of @xmath30-pairs are in agreement with the alice results .
a. andronic , p. braun - munzinger , k. redlich and j. stachel , j. phys .
g * 38 * ( 2011 ) 124081 [ arxiv:1106.6321 [ nucl - th ] ] .
x. zhao and r. rapp , nucl .
a * 859 * ( 2011 ) 114 [ arxiv:1102.2194 [ hep - ph ] ] .
y. -p .
liu , z. qu , n. xu and p. -f .
zhuang , phys .
b * 678 * ( 2009 ) 72 [ arxiv:0901.2757 [ nucl - th ] ] .
k. aamodt _ et al . _
[ alice collaboration ] , jinst * 3 * ( 2008 ) s08002 .
b. abelev _ et al . _
[ alice collaboration ] , phys .
rev . lett .
* 109 * ( 2012 ) 072301 [ arxiv:1202.1383 [ hep - ex ] ] .
f. bossu , z. c. del valle , a. de falco , m. gagliardi , s. grigoryan and g. martinez garcia , arxiv:1103.2394 [ nucl - ex ] .
k. j. eskola , v. j. kolhinen and c. a. salgado , eur .
j. c * 9 * ( 1999 ) 61 [ hep - ph/9807297 ] .
b. abelev _ et al . _
[ alice collaboration ] , phys .
lett . * 108 * ( 2012 ) 082001 [ arxiv:1111.1630 [ hep - ex ] ] . b. abelev _ et al . _
[ alice collaboration ] , arxiv:1203.3641 [ hep - ex ] .
s. chatrchyan _ et al . _
[ cms collaboration ] , jhep * 1205 * ( 2012 ) 063 [ arxiv:1201.5069 [ nucl - ex ] ] .
a. adare _ et al .
_ [ phenix collaboration ] , phys .
c * 84 * ( 2011 ) 054912 [ arxiv:1103.6269 [ nucl - ex ] ] . | quarkonium states @xmath1 provide information on high - energy heavy - ion collisions formation quark - gluon plasma impact on @xmath1 production studied at sps rhic energies colour screening @xmath1 production suppressed in plasma quarks gluons @xcite provides unique probe for qgp formation suppression induced by effects nuclear shadowing cold nuclear matter effects comover absorption scenarios at lhc energies charm produced in central pb pb collisions uncorrelated charm anti - charm quarks re)combine gain importance scenarios described in statistical transport models measuring @xmath1 production at lhc disentangle mechanisms @xcite measures @xmath1 production at mid - rapidity forward rapidity analysis based on dataset pb pb collisions at @xmath6tev minimum 2010 12.8 m ( 0 - 80% central events analysed integrated luminosity of 2.1@xmath7 fig shows invariant mass spectrum of @xmath8 candidates opposite sign spectrum event displayed background distribution from event mixing background distribution scaled to match signal distribution in invariant mass range between 3.2 and 4gev/@xmath9 bottom panel shows background subtracted distribution signal extraction integrating counts invariant mass range 2.92 - 3.16gev/@xmath9 2000 @xmath1 available extraction three centrality intervals di - muon analysis dataset same collision energy measured 2011 analysed di - muon trigger 70@xmath7 integrated luminosity 17.7 m triggered events factor 20 higher statistics than @xcite fig shows invariant mass spectrum @xmath10 candidates for 0 - 90% central events signal extracted combined fit background signal contribution background described by gaussian width @xmath11 ) varies mass value signal description modified crystal ball function used 40k @xmath1 found differential analysis nuclear modification factor in nine bins centrality seven transverse momentum or six rapidity raw yield corrected for acceptance efficiency di - electron analysis simulations enriched with primary @xmath1 used di - muon analysis @xmath1 embedded into real events parametrisation used for transverse momentum rapidity shape from interpolation rhic fermilab lhc data effects of shadowing using eks98 calculations polarisation of simulated @xmath1 assumed zeromeasurements @xmath1 polarisation in collisions at @xmath12tev @xcite show results compatible with zero weak dependence efficiency on collision centrality di - electron channel acceptance times efficiency 8% di - muon channel 14% di - muon channel at forward rapidity - muon channel forward rapidity nuclear modification factor @xmath16 defined as @xmath17 @xmath1 yield number of @xmath1 ) corrected for acceptance times efficiency branching ratio normalised to analysed minimum bias interactions @xmath22 @xmath23 nuclear overlap function @xmath24 @xmath1 production cross - section in collisions at same centre mass energy per nucleon pair measured by @xcite function of average number of participant nucleons at mid- forward rapidity.0% preliminary @xmath16 at mid- forward rapidity function of collision centrality expressed by number of participant nucleons error bars correspond statistical uncertainties boxes uncorrelated systematic uncertainties measurement main contribution to systematic uncertainty di - electron channel from imperfections combinatorial backgroundcontributions are uncertainty @xmath23 mc description detector variations mc input spectrum @xmath1 di - muon channel @xmath23 main uncertainty followed by signal extraction trigger tracking efficiency global systematic uncertainties figure 26% di - electron 14% di - muon channel main contribution due to pp reference uncertainties di - electron measurement large due to low statistics fig shows centrality @xmath16 function of rapidity mid - rapidity point 0 - 80% central events forward region 0 - 90% central events open symbols forward results mid - rapidity @xmath16 decreases with rapidity @xmath16 function of @xmath25 measured by alice compared to cms data @xcite results phenix lower collision energy @xmath16 decreasing from 0.6 low @xmath25 to 0.4 at higher @xmath25 cms results 0 - 100% agreement with alice measurements 0 - 90% momentum range lower energy results from phenix 0 - 20% centrality show smaller @xmath16 fig[fig4 compares alice results at mid- forward rapidity with statistical hadronization model @xcite two transport models models re)combination of @xmath30-pairs models describe data for @xmath31 larger than 50 function of average number participants at mid- forward rapidity with calculations re)combination @xmath15 quarks alice measured nuclear modification factor @xmath16 ) of @xmath1 production in collisions at @xmath33tev mid- forward rapidity @xmath16 at lhc energies larger cms results at lower rapidity agreement with alice data in overlapping transverse momentum range/@xmath15 @xmath16 decreases with increasing @xmath25 towards larger rapidity comparison with models re)combination of @xmath30-pairs with alice results a. andronic p. braun - munzinger k. redlich j. stachel j. phys . 38 * ( 2011 ) 124081 [ arxiv:1106.6321 nucl - th x. zhao r. rapp nucl 859 * ( 2011 ) 114 [ arxiv:1102.2194 hep - ph liu zxu zhuang . 678 2009 ) 72 arxiv:0901.2757 nucl - th aamodt alice collaboration jinst 3 2008 ) s08002 abelev alice collaboration phys . 109 2012 ) 072301 arxiv:1202.1383 hep - ex f. bossu z. del valle. de falco. gagliardi grigoryan. martinez garcia arxiv:1103.2394 nucl - ex j. eskola v j. kolhinen a. salgado c 9 1999 ) 61 hep - ph/9807297 abelev alice collaboration phys . 108 2012 ) 082001 arxiv:1111.1630 hep - ex abelev collaboration arxiv:1203.3641 hep - ex chatrchyan cms collaboration jhep 1205 2012 ) 063 arxiv:1201.5069 nucl - ex adare collaboration . 84 ( 2011 ) 054912 arxiv:1103.6269 nucl - ex | 0.360228 | 2,710 | 1,365 | , Saving $0.1 in GPT-4. |
67,503 | a coupled array of @xmath0 identical cavities , each of which contains a five - level atom is investigated .
the results show that the atoms via the exchange of virtual photons can be effectively equal to spin 1 heisenberg model under certain conditions . by tuning the laser fields ,
the parameters of the effective hamiltonian can be controlled individually . | spin chain has been played an important role in quantum information field as well as in condensed matter physics .
the interaction between on site spin can offer us entanglement in solid and realistic way [ 1 - 7 ] .
it has been found that spin chain with an open boundary condition can be applied in quantum communication @xcite , which can translate information with high fidelity from one end to another .
perfect state transfer can be realized by changing the interaction between the qubits in spin networks @xcite . however , because of the microscopic properties of solid - state materials , it is very hard to address individual spins while it is the prerequisite for quantum information processing .
single spin addressability can also be very helpful to obtain deeper and more detail insight into condensed matter physics . in order to do this ,
it has been show that the arrays of josephson junctions @xcite , quantum dots @xcite , optical lattices @xcite , can provide effective spin - chain hamiltonian where spin - coupling constants can be controlled .
recently under active investigation is to use the array of coupled cavities , which are ideally suited to addressing individual spins [ 14 - 23 ] .
intense interest has arisen from the demonstration that a polaritonic mott transition and a bose - hubbard interaction can be generated in these structures @xcite .
hartmann @xcite have shown that single atoms in interacting cavities that are operated in a strong coupling regime can form a heisenberg spin @xmath1 hamiltonian in which all parameters of the effective hamiltonian can be tuned individually .
however , people are increasedly interested in the multilevel systems .
a lots of papers treat with multilevel systems in different areas of physics such as in condensed matter physics@xcite , statistical mechanics yip , pire , as well as in quantum information @xcite .
now multilevel systems can be considered as a important field . in this paper , we consider a coupled array of @xmath0 identical cavities , each of which contain a five - level atom . we show that under large detuning case the atoms via the exchange of virtual photons can be effectively equal to spin 1 model .
we use atomic bare basis as spin - level
. it should be easy to manipulate the individual atom when we need to project measure etc .
on the individual qutrit .
we consider an array of cavities which are coupled via exchange of photons .
each of the cavities contains one five - level atom .
the atomic configuration is shown in fig.1 .
the three long - lived levels @xmath2 , @xmath3 , @xmath4 , and @xmath5 while four lasers drive the atom with rabi frequencies @xmath6 @xmath7 , respectively .
the same atomic configurations has been used in @xcite for addressing individual atoms in optical lattices with standing - wave driving fields . in the interaction picture
, the hamiltonian reads @xmath8 \notag \\ & & + \sum_{j=1}^{n}j(a_{j}^{\dagger } a_{j+1}+a_{j}a_{j+1}^{\dagger } ) , \notag\end{aligned}\ ] ] where @xmath9 is the coupling strength between the cavity and the atom , and @xmath10 express the detuning shown in fig.1 , and @xmath11 means the annihilation operator of the cavity field . here ,
for simplicity , we assume that the coupling strength @xmath9 and laser rabi frequency @xmath6 do not change with cavity sequence @xmath12 , which means that the identical atoms are in the identical cavities and driven by identical laser fields .
the first sum denote the interaction within the cavities , and the second sum is the inter - cavity hopping between nearest neighbor with hopping rate @xmath13 .
now we consider the large detuning case with @xmath14 @xmath15,@xmath16 , @xmath17 .
we adiabatic eliminate atomic excited state @xmath18 and @xmath19 and have , @xmath20 , @xmath21 represent the three spin states .
the cavity mode couples to the transtions @xmath22 , @xmath23 , @xmath24 @xmath7 , respectively .
, width=302,height=188 ] @xmath25a_{j}^{\dagger } a_{j } \notag \\ & & -\sum_{j=1}^{n}[\frac{\omega _ { 2}^{2}}{\delta _ { 2}}|a_{j}\rangle \langle a_{j}|+(\frac{\omega _ { 1}^{2}}{\delta _ { 1}}+\frac{\omega _ { 4}^{2}}{\delta _ { 4}})|b_{j}\rangle \langle b_{j}|+\frac{\omega _ { 3}^{2}}{\delta _ { 3}}% & & -\sum_{j=1}^{n}[\frac{\omega _ { 1}g_{1}}{\delta _ { 1}}a_{j}|b_{j}\rangle \langle a_{j}|+\frac{\omega _ { 2}g_{2}}{\delta _ { 2}}a_{j}|a_{j}\rangle \langle b_{j}| \notag \\ & & + \frac{\omega _ { 3}g_{3}}{\delta _ { 3}}a_{j}|c_{j}\rangle \langle b_{j}|+% \frac{\omega _ { 4}g_{4}}{\delta _ { 4}}a_{j}|b_{j}\rangle \langle c_{j}|+h.c . ]
\notag \\ & & + \sum_{j=1}^{n}j(a_{j}^{\dagger } a_{j+1}+a_{j}a_{j+1}^{\dagger } ) .\end{aligned}\]]we assume @xmath26the effective hamiltonian can be classified as @xmath27with@xmath28 \\ & & -\sum_{j=1}^{n}[\frac{\omega _ { 2}^{2}}{\delta _ { 2}}|a_{j}\rangle \langle a_{j}|+(\frac{\omega _ { 1}^{2}}{\delta _ { 1}}+\frac{\omega _ { 4}^{2}}{\delta _ { 4}})|b_{j}\rangle \langle b_{j}|+\frac{\omega _ { 3}^{2}}{\delta _ { 3}}% @xmath29.\end{aligned}\ ] ] noting that the every term in @xmath30 commutes each other , when we perform unitary transformation we can do it separately . diagonalize the cavity - hopping terms by employing a fourier transformed basis as @xmath31 , where @xmath32 @xmath33 and @xmath34 the diagonalized form reads @xmath35 @xmath36 with @xmath37 .
now , we goes into the new frame rotating with @xmath30 , under the conditions@xmath38we have the hamiltonian @xmath39 \notag \\ & & e^{-i\nu _ { j}t}f_{jk}b_{k}+h.c.\ } \notag\end{aligned}\]]with @xmath40 .
we define that the eigenstates of @xmath41 for @xmath42 are atomic bare level @xmath2 , @xmath43 , @xmath44 . using the spin operator , we rewrite the hamiltonian eq.(7 )
as @xmath45.\end{aligned}\]]if the rabi frequency \{@xmath46,@xmath47}@xmath48 , this allows us to make use of the adiabatic elimination once more .
considering the subspace without real photons , we deduce the effective hamiltonian as@xmath49f_{jk}f_{lk}^{\ast } .\]]in view of @xmath50 @xmath51 , we expand @xmath52 as@xmath53this expand demand the condition @xmath54 .
substitute the relation eq.(10 ) into eq.(9 ) , we finally obtain the effective hamiltonian [ @xmath55 $ ] between the hamiltonian eq.(2)(blue - dashed line ) with the effective spin chain eq.(11)(red - solid line ) .
( a ) : the initial atomic state is @xmath56 .
( b ) : the initial atomic state is @xmath57 , @xmath58 , @xmath59,@xmath60 , @xmath61 , @xmath62 , @xmath63 , @xmath64.,width=264,height=188 ] @xmath65+bs_{jz}+c(s_{xj}s_{xj+1}+s_{yj}s_{yj+1})\ ] ] with @xmath66we see clearly that it is spin 1 xy antiferromagnetic hamiltonian ( @xmath67 ) . because it is spin 1 , the term @xmath68 can not be omitted , which is of essential importance in high - spin cases .
although individual control of the coefficients @xmath69 , @xmath70 , @xmath71 is limited owing to their mutual dependency , we still can change them ; because @xmath72 ( @xmath73 ) meeting with the condition eqs.(3),(6 ) , we still have seven variable so as to adjust the coefficients @xmath69 , @xmath70 , @xmath74 to confirm the validity of our approximations , we numerically simulate the dynamics generated by hamiltonian eq.(2 ) and compare it to the dynamics generated by effective spin model eq.(11 ) . in fig .
2a , we consider two atoms in two cavities initially in the state @xmath56 ( atom one in @xmath20 and atom two in @xmath21 ) and calculate the occupation probability @xmath75 ( atom 2 in state @xmath76 which corresponding to the probability of spin 2 to point down .
fig.2a shows the time evolution of @xmath75 for hamiltonian eq.(2 ) and @xmath77 for effective spin mode eq.(11 ) . for our choice of the parameters
, one can easy check that they satisfy all kinds of approximation condition such as eqs.(3 ) , ( 6 ) and the two times adiabatic elimination condition @xmath78\{@xmath79,@xmath80},@xmath81 , @xmath17 , and @xmath82,@xmath83\{@xmath84}. from the fig .
2a , we know that the interaction can be effectively equal to spin 1 heisenberg chain because the two curves almost merge into one . for
the initial atomic state latexmath:[$% between the two level , and level @xmath2 do not participate the interaction so that we have the reduced two - level system ( level @xmath2 has no population ) .
fig.2b plots the population @xmath75 for initial state @xmath86 interaction .
we still employ the atomic level configuration but now we only need the two laser beams working between @xmath5 and @xmath22 shown in fig.3 . the hamiltonian is as interaction .
two classical fields are deleted , width=264,height=188 ] @xmath87+\sum_{j=1}^{n}j(a_{j}^{\dagger } a_{j+1}+a_{j}a_{j+1}^{\dagger } ) \end{aligned}\ ] ] the other two cavity fields denoting with @xmath88 and @xmath89 in eq .
( 1 ) will only induce a stark shift on the level @xmath20 and do not affect the effective interaction , so we do not write it in eq .
( 13 ) . under larger detuning
@xmath90,@xmath91 @xmath92 , @xmath93 , @xmath94 @xmath95 , @xmath96 , we adiabatic eliminate atomic excited state and have@xmath97a_{j}^{\dagger } a_{j}-\nu _ { j}b_{j}^{\dagger } b_{j}\ } \notag \\ & & -\sum_{j=1}^{n}[\frac{\omega _ { 3}^{2}}{\delta _ { 3}^{\prime } } |c_{j}\rangle \langle c_{j}|+\frac{\omega _ { 2}^{2}}{\delta _ { 1}^{\prime } } |a_{j}\rangle \langle a_{j}| ] \notag \\ & & -\sum_{j=1}^{n}[\frac{\omega _
{ 2}g_{1}}{\delta _ { 1}^{\prime } } % ( a_{j}s_{zj}^{{}}+h.c.),\end{aligned}\]]where we need the condition @xmath98 , @xmath99 .
switch into interaction picture and then adiabatic eliminate once more under the condition @xmath100 .
finally , we have the effective hamiltonian @xmath101where @xmath102 .
now , the effective hamiltonian is @xmath103 interaction . during the process of deduction of
effective hamiltonian eqs.(11 ) and ( 15 ) , we have changed working picture for two times .
but the atomic population probability do not change with the changing of the picture .
so , we can compare the atomic population probability in different picture , for example we have done in fig .
2 . however , the hamiltonian eqs .
( 13 ) and ( 15 ) do not affect atomic population probability .
therefore , we do not plot it again . the two hamiltonians eqs . ( 11 ) and ( 15 ) can be combined into one effective hamiltonian if we employ the method proposed by @xcite .
the lasers that generate the hamiltonian ( 11 ) are turned on for a short time interval @xmath104 ( @xmath105 ) followed by another time interval @xmath106 ( @xmath107 ) with the lasers that generate the hamiltonian ( 15 ) turned on . by repeating this sequence until the total time range to be simulated is covered .
the effective hamiltonian @xmath108 finally can be obtained .
the decoherence of the system mainly results from the decay mechanisms via the photons or the excited state @xmath19 ( @xmath18 ) . to overcome the decoherence
, the coefficients of effective hamiltonian @xmath69 , @xmath70 , and @xmath71 should be much larger than the decays rates of photons and the excited state @xmath19 ( @xmath18 ) . using the group of the parameters of fig .
2 ( all the parameters are scaled in @xmath109 ) , we have @xmath110 , @xmath111 , @xmath112 . for a strongly coupled single quantum dot cavity system , @xmath113 ( @xmath114 ) , @xmath115 ( @xmath116 )
in which @xmath117 ( @xmath118 ) means the decay of the cavity ( the excited state ) have been achieved for off resonance nature .
therefore @xmath119 , @xmath120 can be realized in experiment .
we consider a coupled array of @xmath0 identical cavities , each of which contains a five - level atom .
we show that under large detuning case the atoms via the exchange of virtual photons can be effectively equal to spin 1 heisenberg model .
although the coefficients are related one another , we still can tune them by controlling the laser fields so that our system is a good simulator for spin 1 heisenberg model . when operated in a two - dimensional array of cavities
the device is thus able to simulate spin lattices . | spin chain important in quantum information field condensed matter physics interaction between on site spin entanglement solid realistic way [ 1 - 7 ] found spin chain open boundary condition applied in quantum communication @xcite information high fidelity perfect state transfer by changing interaction between qubits in spin networks microscopic properties of solid - state materials hard to address individual spins prerequisite for quantum information processing single spin addressability deeper insight into condensed matter physics arrays of josephson junctions quantum dots optical lattices provide effective spin - chain hamiltonian spin - coupling constants controlled investigation array of coupled cavities addressing individual spins [ 14 - 23 ] interest from polaritonic mott transition bose - hubbard interaction generated in structures single atoms in interacting cavities strong coupling regime form heisenberg spin hamiltonian parameters tuned individually people interested in multilevel systems papers treat multilevel systems in different areas physics condensed matter physics statistical mechanics quantum information important paper consider coupled array of @xmath0 identical cavities each five - level atomlarge detuning case atoms exchange virtual photons equal to spin 1 model use atomic bare basis as spin - level easy to manipulate individual atom consider array cavities coupled via exchange photons each contains one five - level atom atomic configuration shown in fig.1 three long - lived levels @xmath2 four lasers drive atom with rabi frequencies @xmath6 @xmath7 same configurations used in @xcite atoms optical lattices with standing - wave driving fields interaction picture hamiltonian reads @xmath8 @xmath9 coupling strength between cavity atom @xmath10 detuning.1 @xmath11 annihilation operator cavity field coupling strength @xmath9 laser rabi frequency @xmath6 change with cavity sequence @xmath12 identical atoms in identical cavities driven by identical laser fields first interaction within cavities second inter - cavity hopping between nearest neighbor with hopping rate @xmath13 consider large detuning case with @xmath14adiabatic eliminate atomic state @xmath18 represent three spin states cavity mode couples transtions width=302,height=188 @xmath25a_{j}\dagger\notag -\sum_{j=1}^{n}\frac{\omega _ 2}^{2}}{\delta 2}}\rangle+\frac{\omega 1}^{2}}{\delta 1}}+\frac\omega 4}^{2}}{\delta 4}}|b+\frac{\omega 3}^{2}}{\delta _ 3}} -\sum_{j=1}^{n}\frac{\omega 1}g_{1}}{\delta 1}}a_{j\rangle+\frac{\omega 2}g_{2}}{\delta 2}}a_{j b\frac{\omega _ 3}g_{3}}{\delta _ 3}}a_{j\rangle\langle b+ \frac{\omega _ 4}g_{4}}{\delta _ 4}}a_{j} c+h\notag \sum_{j=1}^{n}j(a_{j}{\dagger }{j+1}{aligned assume effective hamiltonian @xmath27with@xmath28 -\sum_{j=1}^{n}\frac{\omega _ 2}^{2}}{\delta{j}\rangle\omega 1\delta\omega 4\delta{j}\frac{\omega 3}^{2}}{\delta 3}} @xmath29.{aligned} term @xmath30 commutes unitary transformation separately diagonalize cavity - hopping terms fourier transformed basis @xmath31 @xmath32 @xmath33 @xmath34 diagonalized form reads @xmath35 @xmath36 @xmath37 new frame rotating @xmath30 conditions hamiltonian @xmath39 \notag{-i\nu j}t}f_{jk}b_{k}+h.c.{aligned @xmath40 eigenstates of @xmath41 @xmath42 are atomic bare level @xmath2 @xmath44 spin operator rewrite hamiltonian eq.(7 ) as @xmath45.{aligned rabi frequency@xmath46 allows adiabatic elimination subspace without photons deduce effective hamiltonian as@xmath49f_{jk}f_{lk} @xmath50 @xmath51 expand @xmath52 as@xmath53this condition @xmath54 substitute relation eq.(10 ) into eq.(9 ) obtain effective hamiltonian @xmath55 between hamiltonian eq.(2)(blue - dashed line effective spin chain eq.(11)(red - solid line initial atomic state @xmath56 state @xmath57=264,height=188 @xmath65+bs_{jz}+c(s{xj}s{xj+1}_{yj}s{yj+1} @xmath66we spin 1 xy antiferromagnetic hamiltonian ( @xmath67 ) spin 1 term @xmath68 omitted essential importance high - spin cases control coefficients @xmath69 @xmath71 limited mutual dependency change @xmath72 @xmath73 ) condition eqs.(3),(6 ) seven variable adjust coefficients @xmath69 @xmath70 @xmath74 approximations numerically simulate dynamics hamiltonian eqcompare to dynamics effective spin model eq.(11 ) fig . 2a consider two atoms in two cavities @xmath56 @xmath20 @xmath21 calculate occupation probability @xmath75 2 @xmath76 probability spin 2 point down fig.2a shows time evolution of @xmath75 for hamiltonian eq.(2 ) @xmath77 effective spin mode eq.(11 ) parameters check satisfy approximation condition.(3 two times adiabatic elimination condition @xmath78} interaction equal to spin 1 heisenberg chain two curves merge into one initial atomic state latexmath @xmath2 participate reduced two - level system @xmath2 no population fig.2b plots population @xmath75 for initial state @xmath86 interaction employ atomic level configuration need two laser beams between @xmath5 and @xmath22 fig.3 hamiltonian interaction two classical fields deleted width=264=188 @xmath87+_{j=1{aligned other two cavity fields with @xmath88 and @xmath89 in eq .1 ) induce stark shift level @xmath20 affect effective interaction write eq . ( 13 ) larger detuning @xmath90 eliminate atomic excited state@xmath97a_{j}^{\dagger }-\nu}b_{j\dagger{j}\notag\sum{j=1}^{n}\frac{\omega _ 3}^{2}}{\delta 3}^{\prime }_{j}\rangle_{j}{\omega 2}^{2}}{\delta 1}^{\prime }{j{j}\sum{j=1}^{n}\frac{\omega 2_{1}}{\delta 1}^{\prime } } a_{j}s_{zj}+h.c{aligned need condition @xmath98 @xmath99 switch interaction picture adiabatic eliminate condition @xmath100 effective hamiltonian @xmath101where @xmath102 effective hamiltonian @xmath103 interaction deduction effective hamiltonian eqs.(11 ) 15 ) changed working picture two times atomic population probability change changing picture compare atomic population probability different picture fig . 2 . hamiltonian eqs .13 ) 15 ) affect atomic population probability plot again hamiltonians 11 ) 15 ) combined into one effective hamiltonian method proposed by @xcite lasers hamiltonian ( 11 ) turned on @xmath104 ) followed @xmath106 ) lasers 15 ) turned on repeating sequence until total time range covered effective hamiltonian @xmath108 obtained decoherence results from decay mechanisms via photons excited state @xmath19 @xmath18 ) overcome coefficients of effective hamiltonian @xmath69 @xmath71 larger than decays rates photons excited state parameters of fig . 2 @xmath109 @xmath110 @xmath112 strongly coupled single quantum dot cavity system @xmath113 @xmath115 ) @xmath117 @xmath118 ) means decay cavity achieved off resonance @xmath119 @xmath120 realized in experiment consider coupled array of @xmath0 identical cavities each contains five - level atom under large detuning atoms via virtual photons equal to spin 1 heisenberg modelcoefficients related tune controlling laser fields system good simulator for spin 1 heisenberg model operated in two - dimensional cavities device spin lattices | 0.56135 | 3,476 | 1,801 | , Saving $0.1 in GPT-4. |
67,504 | the power spectrum of the cosmic microwave background from both the _ planck _ and _ wmap _ data exhibits a slight dip for multipoles in the range of @xmath0 .
we show that such a dip could be the result of the resonant creation of massive particles that couple to the inflaton field . for our best - fit models , the epoch of resonant particle creation reenters the horizon at a wave number of @xmath1 ( @xmath2 mpc@xmath3 ) .
the amplitude and location of this feature corresponds to the creation of a number of degenerate fermion species of mass @xmath4 @xmath5 during inflation where @xmath6 is the coupling constant between the inflaton field and the created fermion species , while @xmath7 is the number of degenerate species .
although the evidence is of marginal statistical significance , this could constitute new observational hints of unexplored physics beyond the planck scale . | in this paper we summarize an analysis@xcite of a peculiar feature visible in the observed power spectrum near multipoles @xmath8 .
this is an interesting region in the cmb power spectrum because it corresponds to angular scales that are not yet in causal contact , so that the observed power spectrum is close to the true primordial power spectrum .
the _ planck_@xcite observed power spectrum in this region is shown in figure [ fig:1 ] from ref .
[ ] . although the error bars are large , there is a noticeable systematic deviation in the range @xmath8 below the best fit based upon the standard @xmath9cdm cosmology with a power - law primordial power spectrum .
there is also a well - known suppression of the quadrupole moment in the cmb .
these same features are visible in the cmb power spectrum from the wilkinson microwave anisotropy probe ( _ wmap _ ) @xcite , and hence , are likely a true feature in the cmb power spectrum , although it should be noted that in the planck cosmological parameters paper,@xcite the deviation from a simple power law in the range @xmath8 was deduced to be of weak statistical significance due to the large cosmic variance at low @xmath10 .
points with error bars are from the _ planck _ data release @xcite .
the dashed line shows the best standard @xmath9cdm fit to the _ planck _ cmb power spectrum based upon a power - law primordial power spectrum .
the solid line shows the best fit for a model with resonant particle creation during inflation.,width=288 ] a number of mechanisms have been proposed ( see summary in ref . [ ] ) to deal with the suppression of the power spectrum on large scales and low multipoles .
most of these works , however , were concerned with the suppression of the lowest moments via changes in the inflation generating potential . in the present work@xcite
, however , we are concerned specifically with suppression of the power spectrum in the range @xmath8 due to the possibility that new trans - planckian physics occurs near the end of the inflation epoch corresponding to the resonant creation @xcite of planck - scale particles that couple to the inflaton field .
our best fit is shown by the solid line in figure [ fig:1 ] which we describe in detail in the following sections .
this interpretation has the intriguing aspect that , if correct , an opportunity emerges to use the cmb to probe properties of new particle species that existed at and above the planck scale ( @xmath11 gev ) . that is the goal of the present work . indeed
, string theory compactification schemes generically postulate the existence of massive particles at or above the planck scale from the kaluza - klein states , winding modes , string excitations , etc . moreover
, the coupling of the inflaton to other particle species near the end of inflation is not only natural , but probably required .
this is because the energy density in the inflaton must be converted to entropy in light or heavy particle species at the end of inflation as a means to reheat the universe .
hence , the existence of planck - scale mass particles that couple to the inflaton near the end of inflation is a scenario that is both natural and even required .
moreover , this provides a possible opportunity to uncover new physics in the trans - planckian regime .
the details of the resonant particle creation paradigm during inflation have been explained in refs . .
indeed , the idea was originally introduced @xcite as a means for reheating after inflation . since @xcite subsequent work @xcite has elaborated on the basic scheme into a model with coupling between two scalar fields . in this minimal extension from the basic picture ,
the inflaton @xmath12 is postulated to couple to particles whose mass is of order the inflaton field value .
these particles are then resonantly produced as the field obtains a critical value during inflation .
if even a small fraction of the inflaton field is affected in this way , it can produce an observable feature in the primordial power spectrum .
in particular , there can be either an excess in the power spectrum as noted in @xcite , or a dip in the power spectrum as described in this paper .
such a dip offers important new clues to the trans - planckian physics of the early universe . in the simplest slow roll approximation @xcite , the generation of density perturbations of amplitude , @xmath13 ,
when crossing the hubble radius is just , @xmath14 where @xmath15 is the expansion rate , and @xmath16 is the rate of change of the inflaton field when the comoving wave number @xmath17 crosses the hubble radius during inflation . for the application here,@xcite we adopt a positive yukawa coupling of strength @xmath18 between an inflaton field @xmath12 and a field @xmath19 of @xmath7 degenerate fermion species.the total lagrangian density including the inflaton scalar field @xmath12 , the dirac fermion field , and the yukawa coupling term is simply , @xmath20 for this lagrangian , it is obvious that the fermions have an effective mass of @xmath21 this vanishes for a critical value of the inflaton field , @xmath22 .
resonant fermion production will then occur@xcite in a narrow range of the inflaton field amplitude around @xmath23 .
the cosmic scale factor is labeled @xmath24 at the time @xmath25 at which resonant particle production occurs . considering a small interval around this epoch
, one can treat @xmath26 as approximately constant ( slow roll inflation ) .
the number density @xmath27 of particles can be taken as zero before @xmath25 and afterwards as @xmath28^{3}$ ] .
the fermion vacuum expectation value can then be written , @xmath29 } ~~.\ ] ] where @xmath30 is a step function
. then , one obtains@xcite the perturbation in the primordial power spectrum as it exits the horizon : @xmath31_{n \lambda = 0}}{1 + \theta ( a - a _ * ) ( n \lambda n_*/\vert \dot \phi_*\vert h _ * ) ( a_*/a)^3 \ln{(a / a _ * ) } } ~~. \label{deltahnew}\ ] ] it is clear that the power in the fluctuation of the inflaton field will diminish as the particles are resonantly created when the universe grows to some critical scale factor @xmath24 . using @xmath32 , then the perturbation spectrum eq .
( [ deltahnew ] ) can be reduced @xcite to a simple two - parameter function .
@xmath33_{n \lambda = 0}}{1 + \theta ( k - k_*)a ( k_*/k)^3 \ln{(k / k _ * ) } } ~~. \label{perturb}\ ] ] where the amplitude @xmath34 and characteristic wave number @xmath35 ( @xmath36 ) can be fit to the observed power spectrum from the relation : @xmath37 , where @xmath38 is the comoving distance to the last scattering surface , taken here to be 14 gpc .
the values of @xmath34 and @xmath35 determined from from the cmb power spectrum relate to the inflaton coupling @xmath18 and fermion mass @xmath39 , for a given inflation model via eqs .
( [ deltahnew ] ) and ( [ perturb ] ) .
@xmath40 the connection between resonant particle creation and the cmb temperature fluctuations is straightforward .
we have made a multi - dimensional markov chain monte - carlo analysis @xcite of the cmb using the _ planck _
data @xcite and the _ cosmomc _ code @xcite . for simplicity and speed in the present study we only marginalized over parameters which do not alter the matter or cmb transfer functions
hence , we only varied @xmath34 and @xmath35 , along with the six parameters , @xmath41 . here , @xmath42 is the baryon content , @xmath43 is the cold dark matter content , @xmath44 is the acoustic peak angular scale , @xmath45 is the optical depth , @xmath46 is the power - law spectral index , and @xmath47 is the normalization . adding this perturbation to the primordial power spectrum improves the total @xmath48 for the fit from 9803 to 9798 .
one expects that the effect of interest here would only make a small change ( @xmath49 ) in the overall fit because it only affects a limited range of @xmath50 values with large error bars .
nevertheless , from the likelihood contours we can deduce a mean value of @xmath51 with a maximum likelihood value of @xmath52 , and a mean value of @xmath53
the coefficient @xmath34 can be related@xcite directly to the coupling constant @xmath18 this gives @xmath54 hence , for the maximum likelihood value of @xmath55 , we have @xmath56 so , @xmath57 requires @xmath58 as expected .
the fermion particle mass @xmath39 can then be deduced from @xmath59 . from eq .
( [ nconst ] ) then we have @xmath60 . for this purpose , we adopt a general monomial potential for which : @xmath61 for @xmath53 , and @xmath62 mpc@xmath63 we have @xmath64 .
typically one expects @xmath65 .
we note , however , that one can have the number of e - folds as low as @xmath66 in the case of thermal inflation @xcite . for standard inflation a monomial potential with @xmath67 would have @xmath68 . however , the limits on the tensor to scalar ration from the _ planck _ analysis @xcite rule out @xmath67 at the 95% confidence level .
monomial potentials are more consistent with @xmath69 ( @xmath70 ) , or even @xmath71 ( @xmath72 ) .
hence , we have roughly the constraint , @xmath73 so , one can deduce@xcite a family of possible properties of the resonantly produced particle ( i.e. its mass and coupling strength ) in terms of a single parameter , the degeneracy @xmath7 .
we have analyzed the @xmath8 dip in the _ planck _ cmb power spectrum in the context of a model for the creation of @xmath7 nearly degenerate trans - planckian massive fermions during inflation .
the best fit to the cmb power spectrum implies an optimum feature at @xmath74mpc@xmath3 and @xmath75 . for monomial inflation potentials consistent with the _ planck _ tensor - to - scalar ratio
, this feature would correspond to the resonant creation of nearly degenerate particles with @xmath76 @xmath77 and a yukawa coupling constant @xmath18 between the fermion species and the inflaton field of @xmath78 for @xmath7 degenerate fermion species .
we conclude that if the present analysis is correct , this may be one of the first hints at observational evidence of new particle physics at the planck scale .
indeed , one expects a plethora of particles at the planck scale , particularly in the context of string theory .
perhaps , the presently observed cmb power spectrum contains the first suggestion that a subset of such particles may have coupled to the inflaton field leaving a relic signature of their existence in the cmb primordial power spectrum .
work at the university of notre dame is supported by the u.s .
department of energy under nuclear theory grant de - fg02 - 95-er40934 .
work at naoj was supported in part by grants - in - aid for scientific research of jsps ( 26105517 , 24340060 ) .
work at nagoya university supported by jsps research grant number 24340048 .
0 g. j. mathews , m. r. gangopadhyay , k. ichiki , and t. kajino , phys .
* d92 * 123519 ( 2015 ) .
collaboration , astron .
& astrophys . submitted ( 2015 )
arxive:1502.01589 , astro - ph g. hinshaw , et al .
( _ wmap collaboration _ ) astrophys .
* 208 * , 19 ( 2013 ) .
collaboration , astron .
& astrophys . submitted ( 2015 )
arxive:1502.02114 , astro - ph d. j. h. chung , e. w. kolb , a. riotto , and i. i. tkachev , phys .
d * 62 * , 043508 ( 2000 ) .
g. j. mathews , d. chung , k. ichiki , t. kajino , and m. orito , phys . rev . *
d70 * , 083505 ( 2004 ) .
l. kofman , a. d. linde , and a. a. starobinsky , phys .
lett . * 73 * 3195 ( 1994 ) .
o. elgaroy , s. hannestad , and t. haugboelle , jcap , 09 , 008 ( 2003 ) .
a. e. romano and m. sasaki , phys .
d * 78 * , 103522 ( 2008 ) .
n. barnaby , z. huang , l. kofman , and d. pogosyan , phys .
d * 80 * , 043501 ( 2009 ) .
m. a. fedderke , e. w. kolb , m. wyman , phys .
, d * 91 * , 063505 ( 2015 ) .
a. r. liddle and d. h. lyth , _
cosmological inflation and large scale structure _ , ( cambridge university press : cambridge , uk ) , ( 1998 ) .
e. w. kolb and m. s. turner , _ the early universe _ , ( addison - wesley , menlo park , ca . , 1990 ) .
n. christensen and r. meyer , l. knox , and b. luey , class . and quant .
grav . , * 18 * , 2677 ( 2001 )
. a. lewis and s. bridle , phys .
d * 66 * , 103511 ( 2002 ) .
n. d. birrell and p. c. w. davies , quantum fields in curved space , ( cambridge univ . press , cambridge , 1982 ) . | paper summarize analysis peculiar feature observed power spectrum near multipoles @xmath8 interesting region corresponds to angular scales not causal contact close to true primordial power spectrum planck_@xcite observed power spectrum shown in figure error bars large noticeable systematic deviation range @xmath8 below best fit standard @xmath9cdm cosmology with power - law primordial power spectrum known suppression of quadrupole moment in cmb features visible wilkinson microwave anisotropy probe likely true feature planck cosmological parameters paper deviation from power law range @xmath8 weak statistical significance due to large cosmic variance at low @xmath10 points error bars from planck data release dashed line shows best standard @xmath9cdm fit to planck cmb power spectrum power - law primordial power spectrum solid line shows best fit for model with resonant particle creation during inflation mechanisms proposed deal suppression power spectrum on large scales low multipoles concerned suppression lowest moments via changes inflation generating potentialpresent work concerned with suppression power spectrum range @xmath8 due to new trans - planckian physics end inflation epoch resonant creation of planck - scale particles to inflaton field best fit shown by solid line in figure ] interpretation if correct opportunity to use cmb probe properties new particle species at above planck scale @xmath11 gev goal present work string theory compactification schemes postulate massive particles at or above planck scale from kaluza - klein states winding modes string excitations coupling inflaton to other particle species end inflation natural required energy density inflaton converted to entropy light particle species reheat universe existence of planck - scale mass particles couple to inflaton end inflation natural required opportunity uncover new physics in trans - planckian regime details resonant particle creation paradigm during inflation explained in refs idea originally introduced reheating after inflation work elaborated basic scheme into model coupling between two scalar fields inflaton @xmath12 postulated to couple to particles mass inflaton field value particles resonantly produced field obtains critical value during inflationsmall fraction inflaton field affected observable feature primordial power spectrum excess or dip dip offers clues trans - planckian physics early universe slow roll approximation generation density perturbations amplitude crossing hubble radius @xmath14 @xmath15 expansion rate @xmath16 rate change inflaton field when comoving wave number crosses hubble radius inflation positive yukawa coupling strength @xmath18 between inflaton field @xmath12 field @xmath19 degenerate fermion species total lagrangian density including inflaton scalar field dirac fermion field yukawa coupling term , @xmath20 fermions have effective mass @xmath21 vanishes for critical value inflaton field @xmath22 resonant fermion production in narrow range inflaton field amplitude around @xmath23 cosmic scale factor labeled @xmath24 at resonant particle production occurs @xmath26 as constant slow roll inflation number density of particles taken zero before @xmath25 afterwards @xmath28^{3}$ ] fermion vacuum expectation value written @xmath29 } ~~ @xmath30 step functionobtains perturbation primordial power spectrum exits horizon @xmath31\lambda = + \theta ( a - a _ * ) h _ * ) a_*/a)^3 {(a / a _ * ){deltahnew power fluctuation inflaton field particles created universe grows critical scale @xmath24 @xmath32 perturbation spectrum reduced two - parameter function @xmath33\lambda = 0}}{1 + \theta ( k - k_*)a k_*/k)^3 {(k / k _ * ){perturb amplitude @xmath34 characteristic wave number @xmath35 @xmath36 fit observed power spectrum @xmath37 @xmath38 comoving distance last scattering surface 14 gpc values @xmath34 @xmath35 cmb power spectrum relate inflaton coupling @xmath18 fermion mass @xmath39 inflation model connection resonant particle creation cmb temperature fluctuations straightforward multi - dimensional markov chain monte - carlo analysis cmb using planck data cosmomc codesimplicity speed study marginalized parameters alter matter cmb transfer functions varied @xmath34 @xmath35 six parameters @xmath41 @xmath42 baryon content cold dark matter content acoustic peak angular scale optical depth power - law spectral index normalization perturbation primordial power spectrum improves total @xmath48 fit from 9803 to 9798 effect interest small change @xmath49 ) fit affects limited range @xmath50 values large error bars likelihood contours deduce mean value @xmath51 maximum likelihood value @xmath52 mean value @xmath53 coefficient @xmath34 related to coupling constant @xmath18 gives @xmath54 maximum likelihood value @xmath55 @xmath56 @xmath57 requires @xmath58 fermion particle mass @xmath39 deduced from @xmath59 have @xmath60 adopt general monomial potential @xmath61 @xmath53 @xmath62@xmath63 @xmath64 expects @xmath65 number e - folds low as @xmath66 thermal inflation standard inflation monomial potential with @xmath67 @xmath68 limits tensor to scalar ration planck analysis rule out @xmath67 95% confidence monomial potentials consistent with @xmath69 @xmath71 constraint @xmath73 deduce properties resonantly produced particle mass coupling strength single parameter degeneracy @xmath7 analyzed @xmath8 dip planck cmb power spectrum model creation @xmath7 degenerate trans - planckian massive fermions during inflation best fit optimum feature at @xmath74mpc@xmath3 @xmath75 monomial inflation potentials consistent tensor - to - scalar ratio resonant creation nearly degenerate particles with @xmath76 @xmath77 yukawa coupling constant @xmath18 between fermion species inflaton field @xmath78 for @xmath7 degenerate fermion species analysis correct new particle physics at planck scale plethora particles planck scale string theory observed cmb power spectrum first suggestion particles coupled to inflaton field leaving signature primordial power spectrum work university of notre dame supported by u.s department of energy nuclear theory grant - - 95-er40934naoj supported grants research jsps 26105517 24340060 nagoya university supported jsps grant 24340048 g j. mathews gangopadhyay k. ichiki t kajino d92 123519 2015 collaboration arxive:1502.01589 g. hinshaw astrophys 208 19 2013 collaboration 2015 arxive:1502.02114 h chung e w. kolb a. riotto i i. tkachev phys 62 043508 ( 2000 ) j mathews d. chung ichiki t. kajino m. orito phys d70 083505 ( 2004 ) l. kofman d. linde a starobinsky 73 * 3195 ( 1994 ) o. elgaroy s. hannestad t. haugboelle 008 ( 2003 ) e. romano m. sasaki phys 78 103522 ( 2008 ) n. barnaby z. huang l. kofman d. pogosyan phys 80 043501 ( 2009 ). fedderke e w. kolb m. wyman phys 91 063505 ( 2015 )liddle lyth cosmological inflation large scale structure cambridge university press uk 1998 kolb m. s. turner early universe addison - wesley menlo park ca 1990 christensen r. meyer l. knox b. luey class quant . grav. 18 2677 2001 a lewis s. bridle . 66 103511 2002 d birrell w davies quantum fields curved space cambridge univ press 1982 | 0.505671 | 3,211 | 1,652 | , Saving $0.1 in GPT-4. |
67,509 | we present some early results from our deep imaging survey of galaxy clusters intended to detect and study intracluster light ( icl ) . from our observations to date , we find that icl is common in galaxy clusters , and that substructure in the icl also appears to be common as well .
we also discuss some initial comparisons of our imaging results to high - resolution numerical simulations of galaxy clusters , and give avenues for future research .
[ 1996/06/01 ] a&a | the concept of intracluster starlight ( icl ) , or stars between the galaxies in galaxy clusters is not a new one : it was first proposed over 50 years ago ( zwicky 1951 ) .
however , progress in studying icl has been slow due to its low surface brightness , which is less than 1% of the brightness of the night sky ( see vlchez - gmez 1999 , feldmeier 2000 for reviews ) .
this is unfortunate , because icl is a powerful probe of the evolution of galaxies in clusters ( dressler 1984 ) , and of cluster evolution overall . in the last six years
however , the study of icl has increased dramatically .
numerous individual intracluster stars have been detected in nearby and distant galaxy clusters ( arnaboldi et al .
1996 ; theuns & warren 1997 ; feldmeier , ciardullo & jacoby 1998 ; ferguson , tanvir & von hippel 1998 ; durrell et al .
2002 ; feldmeier et al . 2003 ; arnaboldi et al . 2003
; gal - yam et al .
2003 ) . at the same time , deep imaging of clusters with ccd detectors have also detected the intracluster light ( uson , boughn , & kuhn 1991 ; vchez - gomez , pell , & sanahuja , 1994 ; bernstein et al .
1995 ; gonzalez et al .
detections of tidal debris arcs ( trentham & mobasher 1998 ; gregg & west 1998 ; calcaeo - roldn et al .
2000 ) in clusters have shown that at there is significant substructure in the icl , and that the production of icl is ongoing . encouraged by these results , we have undertaken a deep imaging survey of galaxy clusters , intended to quantify the properties of icl as a function of environment , and overall galaxy cluster properties . from our deep imaging , with careful attention to systematic errors ( e.g. , morrison , boroson , & harding 1994 ; morrison et al .
1997 ) , we are able to measure the icl to faint surface brightnesses many magnitudes below that of the night sky ( @xmath0 @xmath1 2628 ) . in tandem with the observations
, we are constructing numerical simulations of galaxy clusters in a cosmological context , similar to those of dubinski ( 1998 ) . here , we present the results of our deep imaging survey to date .
we note that there are several other searches for icl underway that have complementary goals ( see gonzalez et al .
this volume ; krick et al .
this volume ) . for our initial survey ,
we focus on observing abell clusters ( abell , corwin , & olowin 1989 ) at large distance classes ( 56 , corresponding to z @xmath1 0.10.2 ) .
this is to ensure that the entire cluster is contained in our field - of - view for good sky subtraction , but the clusters are not so distant that cosmological @xmath2 surface brightness dimming is a major effect .
we are also observing a few nearby clusters from the mkw / awm ( morgan , kayser , white 1975 , albert , white , & morgan 1977 ) as a comparison sample .
our goal is to gather a _ representative _ sample of clusters with different richness , bautz - morgan and rood - sastry types .
we have observed ten clusters so far ( aco 84 , 98 , 545 , 801 , 1234 , 1413 , 1553 , 1914 , 2443 and mkw 7 ) , and plan to observe 45 more before completing this initial survey .
we observe using the kpno 2.1 m , and image through the washington _
filter , which is similar to _ v _ but contains fewer night sky emission lines .
we use the ultra - deep surface photometry techniques of morrison , boroson , & harding ( 1994 ) for our observations and data reduction .
we spend half of our telescope time constructing our dark sky flats ( which are flat to 0.3% on all scales in the worst case ) , and the other half observing the clusters .
we then carefully mask out all objects in our frames , using a combination of the daophot ( stetson 1987 ) , sextractor ( bertin & arnouts 1996 ) and our own software .
after detecting the icl , we construct an error model for each cluster that includes all sources of error , both random and systematic .
our data has a signal - to - noise of five ( including systematic errors ; see feldmeier et al .
2002 for an example of the error model ) at a surface brightness of 26.5 mag arcsec@xmath3 .
the signal - to - noise ratio approaches unity at = 28.3 mag arsec@xmath3 .
the results of the first two clusters have been published in feldmeier et al .
( 2002 ) : work is ongoing on the other clusters . a mosaic of the first five clusters is shown in figure [ fig : mosaic ] . once this initial survey is complete , we plan to observe nearby clusters using the cwru burrell schmidt , which is currently being optimized for ultra - deep surface photometry . from the clusters we have observed so far
, it seems that diffuse intracluster light and intracluster tidal debris is common in galaxy clusters .
fig [ fig : mkw7 ] shows an example of a tidal plume superimposed over the brightest cluster galaxy in the cluster mkw 7 .
similar tidal features can be seen in abell 1914 ( see the boxed regions in figure [ fig : mosaic ] ) , and in several other clusters .
although our sample is still small , it seems that the clusters that are cd - dominated ( abell 1413 , mkw 7 ) have smoother , more regular icl than those in clusters that do not have a cd galaxy ( aco 1234 , 1553 & 1914 ) . in these non cd clusters
, the icl follows an irregular distribution , and is not well correlated with galaxy density .
tidal features of the kind we have detected can be seen in high - resolution cluster n - body simulations ( e.g. , moore et al .
1996 ; dubinski 1998 ; dubinski , murali , & ouyed 2001 ; see also willman , this volume ) , but thus far we have seen few long tidal debris arcs .
plume - like intracluster debris structures seem to be more common in the clusters we have surveyed thus far .
however , it is clear from the simulations that the vast majority of tidal debris seen in cluster simulations has a surface brightness much lower than our limit of = 26.5 mag arsec@xmath3 for the initial observations .
the structures that we have observed are likely to be the brightest features in each cluster .
another interesting facet of our observations is the nature of cd galaxy envelopes , which are characterized by an excess of diffuse light ( compared to an r@xmath4 law ) at large radius .
the origin of cd envelopes remains unclear : are they formed in the initial stages of cluster collapse ( merritt 1983 , 1984 ) , or later , as galaxies continue to fall in the cluster and become tidally stripped ?
the accepted view of cd envelopes ( schombert 1992 ) is that cds in rich galaxy clusters have large extended cd envelopes , while brightest cluster galaxies in poorer clusters do not . in our observations of abell 1413 and mkw 7 ,
we have found the opposite behavior : mkw 7 , a poor cluster has a strong cd envelope , while abell 1413 has a weaker cd envelope than previously measured . since other researchers have also found cds in rich clusters with pure r@xmath4 laws ( e.g. , gonzalez 2000 ) , this may signify that the earlier photographic surface photometry of cds may need to be re - examined .
our initial survey will be completed by the end of 2003 : we then plan to begin observations on the burrell schmidt .
the wide field of view of the schmidt ( @xmath1 1.5 degrees ) will allow us to observe more nearby galaxy clusters , where much more is known about their properties .
we also plan to run additional large - scale galaxy cluster simulations over a range of cosmologies and initial conditions so that we can make detailed comparisons between our imaging survey and theoretical results .
a final goal is to create quantitative metrics for the icl that can be used on simulations and observations . in figure
[ fig : lmu ] we show one potential metric : the amount of stellar luminosity as a function of surface brightness , applied to the cluster simulation of dubinski ( 1998 ) at two different redshifts . in this relation , dynamically evolved clusters have a shallower slope to their `` surface brightness distribution function '' than dynamically younger clusters . from comparison of multiple simulations ,
the intrinsic scatter in this slope at fixed redshift is low ( @xmath1 4% ) .
more testing is needed , but the results seem promising . | concept of intracluster starlight ( icl ) stars between galaxies in clusters not new first proposed 50 years ago ( zwicky 1951 ) progress studying icl slow due to low surface brightness less than 1% brightness night sky see vlchez - gmez 1999 feldmeier 2000 icl powerful probe evolution of galaxies in clusters ( dressler 1984 ) cluster evolution last six years study of icl increased dramatically individual intracluster stars detected in nearby distant galaxy clusters ( arnaboldi 1996 theuns warren 1997 1998 durrell 2002 feldmeier 2003 gal - yam 2003 deep imaging of clusters with ccd detectors detected intracluster light uson 1991 1994 bernstein 1995 gonzalez detections of tidal debris arcs trentham 1998 gregg west 1998 calcaeo - roldn 2000 in clusters shown significant substructure in icl production icl ongoing encouraged results undertaken deep imaging survey of galaxy clusters to quantify properties of icl function environment galaxy cluster properties deep imaging attention to systematic errors morrison 19941997 icl faint surface brightnesses below night sky ( @xmath0 @xmath1 2628 ) constructing numerical simulations of galaxy clusters cosmological context similar dubinski ( 1998 ) present results deep imaging survey other searches for icl complementary goals gonzalez et al krick et al initial survey focus observing abell clusters at large distance classes ( 56 z @xmath1 0.10.2 ) entire cluster field view good sky subtraction not distant cosmological @xmath2 surface brightness dimming observing nearby clusters from mkw / awm morgan 1975 1977 comparison sample goal representative sample clusters different richness bautz - morgan rood - sastry types observed ten clusters ( aco 84 98 545 801 1234 1413 1553 1914 2443 mkw 7 ) plan to observe 45 more before survey observe kpno 2.1 m washington _ filter fewer night sky emission lines use ultra - deep surface photometry techniques of morrison boroson harding 1994 for observations data reduction half telescope time constructing dark sky flats 0.3% other half observing clustersmask objects in frames using daophot sextractor our software after detecting icl construct error model for each cluster includes sources error random systematic data has signal - to - noise of five including systematic errors feldmeier et al 2002 at surface brightness 26.5 mag arcsec@xmath3 signal - to - noise ratio approaches unity at = 28.3 mag arsec@xmath3 results of first two clusters published in feldmeier et al . 2002 work ongoing on other clusters mosaic of first five clusters shown in figure ] initial survey complete plan to observe nearby clusters using cwru burrell schmidt optimized for ultra - deep surface photometry diffuse intracluster light tidal debris common in galaxy clusters fig mkw7 shows tidal plume over brightest cluster galaxy similar tidal features in abell 1914 other clusters sample small clusters cd - dominated ( 1413 mkw 7 ) have smoother regular icl than in non cd clusters icl follows irregular distribution not correlated with galaxy density tidal features seen in high - resolution cluster n - body simulations1996 dubinski 1998 murali ouyed 2001 willman volume seen few long tidal debris arcs plume - like intracluster debris structures common in clusters surveyed majority of tidal debris surface brightness lower than limit = 26.5 mag arsec@xmath3 initial observations structures observed likely brightest features in each cluster nature cd galaxy envelopes excess diffuse light r@xmath4 law at large radius origin of cd envelopes unclear formed cluster collapse or galaxies tidally stripped ? accepted view schombert 1992 cds in rich galaxy clusters have large extended cd envelopes brightest galaxies in poorer clusters not observations of abell 1413 mkw 7 found opposite behavior strong cd envelope 1413 weaker cd envelope other researchers found cds in rich clusters with pure r@xmath4 laws gonzalez 2000 earlier photographic surface photometry of cds may need re - initial survey completed end of 2003 plan to begin observations on burrell schmidt wide field of view @xmath1 1.5 degrees ) observe more nearby galaxy clusters more known propertiesplan run additional large - scale galaxy cluster simulations cosmologies initial conditions detailed comparisons between imaging survey theoretical results final goal create quantitative metrics for icl simulations observations figure potential metric stellar luminosity function surface brightness applied cluster simulation dubinski 1998 two redshifts dynamically evolved clusters shallower slope surface brightness distribution function younger clusters intrinsic scatter slope at fixed redshift low ( @xmath1 4% ) more testing needed results promising | 0.322548 | 2,040 | 1,021 | , Saving $0.1 in GPT-4. |
67,519 | the time interval of successive water - drips from a faucet was examined over a wide range of the flow rate .
the dripping interval alternately exhibits a stable state and a chaotic state as the flow rate increases . in the stable state ,
the volume of the drip is kept constant at fixed flow rates , and the constant volume increases with the flow rate . in the chaotic state , in addition to a mechanics that the drip is torn by its own weight , the vibration of the drip on the faucet takes part in the strange behavior of the interval .
_ department of physics , tokyo metropolitan university , 1 - 1 minami - ohsawa , hachioji - shi , tokyo , 193 - 0397 _ | it has been experimentally shown that drips from a faucet have chaotic time - intervals .
@xcite shaw @xcite found several types of chaotic attractor in a time series of dripping intervals.wu and schelly @xcite showed that the dependence of the surface tension and of the flow rate on the attractor is dramatic unlike that of temperature .
sartorelli , gonalves , and pinto @xcite in detail investigated a time series of dripping intervals within a narrow range of the flow rate , and they revealed that the time series has boundary crisis and tangent bifurcation for a change in the flow rate .
a theoretical approach to the dripping phenomenon is still in an early stage .
a mass - spring model by shaw @xcite generated a few types of attractor he obtained experimentally .
recently , oliveira and penna @xcite carried out a numerical simulation on the ising model , and they showed that the dripping dynamics of tiny - sized drops has discrete attractors .
@xcite the dripping interval of pure water is affected by the diameter of a nozzle , the temperature of the water , the pressure of the water , and the flow rate of the water .
all the studies mentioned above have been performed for a narrow range of the flow rate . in our experiment ,
the temperature and the flow rate are control parameters .
the behavior of the dripping interval was examined over a wide range of the flow rate .
figure [ fig:1 ] shows the schematic diagram of the apparatus used . distilled water with a constant depth of 90.5 cm , which fills
the inner tube with an inside diameter of 20 mm , applies a constant pressure to the needle valve connected to the inner tube .
the drips fall from the glass - tube nozzle of the needle valve .
the nozzle has an inside diameter of 5 mm and an outside diameter of 7 mm .
the temperature of the water in the tank ( inner tube ) was kept constant by circulating water at a constant temperature of 15 through the outer tube .
the temperature of the water flowing through the needle valve was elevated with an electric heater winded around the needle valve and the nozzle .
the elevated temperature was controlled by the adjustment of a constant current applied to the heater .
the temperature of the drip , @xmath0 , was determined by that of the needle valve . in view of disturbance free ,
the apparatus was installed on a vibration - proof stand .
the flow rate of the water dripping from the nozzle , @xmath1 , was coarsely controlled by the adjustment of the needle valve .
the fine adjustment of @xmath2 was made through the control of the temperature @xmath0 .
the drops falling successively from the nozzle intersected a he - ne laser beam , which ran at a distance of 15 cm under the edge of the nozzle and was detected by a photodiode .
the dripping time - intervals , denoted by @xmath3 for @xmath4 were recorded by a pc with an accuracy of 0.1 ms .
on the temperature @xmath0 for three different adjustments of the valve .
the straight lines have a slope of @xmath5 ) . ]
the measurement of @xmath3 was made for the flow rates ranging from @xmath6 to @xmath7 .
this range was covered by 50 different adjustments of the valve .
the drip temperature @xmath0 was changed at an interval of some 1 from 15 up to 27 for each adjustment of the valve .
figure [ fig:2 ] shows the dependence of the flow rate @xmath2 on the temperature @xmath0 for three adjustments of the valve .
the flow rate increases linearly with @xmath0 .
this result suggests that the flow passing through the needle valve is not affected by a very weak convective flow occurring in the tank .
the change in the flow rate was about @xmath8 per 1 , irrespective of the adjustment of the valve .
the number of data for obtaining the return map of @xmath3 was about two thousands .
at the same value of @xmath9 ; ( a ) @xmath10 and @xmath11 and ( b ) @xmath12 and @xmath13 . ] figures [ fig:3 ] ( a ) and [ fig:3 ] ( b ) show that , for a change in @xmath0 , the same return map of @xmath3 is obtained at slightly different flow rates from each other . here
, let @xmath14 be the small difference in @xmath2 , and let @xmath15 be the change in @xmath0 .
the ratio of @xmath14 to @xmath15 , @xmath16 , was @xmath17 ) on average for differed values of @xmath2 .
we introduced a reduced flow rate @xmath9 defined as @xmath18 where @xmath19 . at the same values of @xmath9 ,
the time series of @xmath3 drew the same maps .
figures [ fig:3 ] ( a ) and [ fig:3 ] ( b ) illustrate the maps obtained when @xmath20 . the temperature @xmath0 was changed for each of 50 different adjustments of the valve .
the experimental results obtained thus were rearranged with the reduced flow rates @xmath9 at @xmath21 . ;
( a ) a low range from @xmath22 to @xmath23 ; ( b ) a middle range from @xmath24 to @xmath25 ; and ( c ) a high range from @xmath26 to @xmath27 .
the arrows indicate transition points from stable states to chaotic states . ]
figure [ fig:4 ] shows the dripping spectrum of @xmath3 , i.e. , the bifurcation diagram of @xmath3 against @xmath9 , in three ranges of @xmath9 . here
, the dripping spectrum was obtained with the number of data of several hundreds .
the dripping spectrum has a stable state where @xmath3 is constant for @xmath28 and a chaotic state where it fluctuates .
these two states appear alternately with increasing @xmath9 . in the stable state ,
the interval @xmath3 increases with @xmath9 ; however , in the chaotic state , the interval @xmath3 decreases on average over @xmath29 with increasing @xmath9 .
the arrows in fig .
[ fig:4 ] indicate the transition points from the stable state to the chaotic state .
let @xmath30 be the reduced flow rate at the transition point at which @xmath3 takes a maximum , where @xmath31 is a transition - point number being from 1 to 43 toward a large value of @xmath30 .
vs. @xmath30 .
the straight line has a slope of -0.92 . ] as a function of the transition - point number @xmath31 . ]
let @xmath32 be the time interval at the transition point @xmath30 .
figure [ fig:5 ] shows the log - log plot of @xmath32 vs. @xmath30 . from the plot
, we have @xmath33 since @xmath34 has the exponent above -1 , the volume of the drip , which is defined as @xmath35 , increases with @xmath30 : @xmath36 . figure [ fig:6 ] shows the plot of @xmath32 against @xmath31 .
the interval @xmath32 is a linearly decreasing function of @xmath31 , having a constant decrement of @xmath37 . in a chaotic state ranging from @xmath38 to @xmath39 . ]
figure [ fig:7 ] shows four return maps of @xmath3 , which are typical of ones obtained in the chaotic state ranging from @xmath40 to @xmath41 since the values of @xmath9 are different from each other , the return maps are quite different in the form .
the number of the data is much smaller in the dripping spectrum than in the maps , so that very sparsely distributed dots of the map shown in fig .
[ fig:7 ] ( d ) do not appear in the spectrum in fig .
[ fig:4 ] .
the mapping functions @xmath42 are well - defined .
although some of the maps are multi - valued functions of @xmath3 , the chaotic attractors probably have low dimensions .
the attractors obtained were not such high - dimensional ones as obtained by wu and schelly .
since , in the stable states , the interval @xmath3 increases with @xmath9 as seen in fig .
[ fig:4 ] , the volume of the drop increases .
the chaotic state , however , has a decrease in the average volume , because the average of @xmath3 decreases more rapidly than @xmath43 with an increase in @xmath9 , which is seen in fig .
[ fig:4 ] .
the volume alternately repeats the increase and decrease with an increase in @xmath9 and therefore the increased volume in the stable state is partially compensated by the decreased volume in the chaotic states .
if the increase and decrease is ignored , then the volume can be assumed to increase monotonicaly with @xmath9 in the form of @xmath44 .
we observed with the naked eye that the drip vibrates when it stays on the edge of the nozzle , and also that the amplitude of the vibration depends on the flow rate .
the drip on the edge , therefore , grows with the vibration into the volume of the falling drop .
if the drip is torn off by its own weight irrespective of the vibration , the increase and decrease in the volume and the chaotic fluctuations in @xmath3 can not occur .
our experimental results suggest that , in addition to a mechanics that the drip is torn off by its own weight , the vibration affects the behavior of @xmath3 . in the stable state ,
the volume of the drip is kept constant for the drip number @xmath29 at fixed flow rates .
this suggests that the steady behavior of @xmath3 is chiefly dominated by the mechanics that the drip is torn off by its own weight .
however , the behavior that the constant volume abruptly increases with @xmath9 is not explained only by the mechanics .
moreover , in the chaotic state , the mechanics no longer dominates the behavior of @xmath3 .
in addition to the mechanics , some factors may take part in the strange behavior of @xmath3 .
the vibration of the drip on the edge is the most important factor . to have a clear understanding of the chaotic behavior of the dripping interval
, one would need a quantitative analysis of the vibration .
99 r. shaw : _ the dripping faucet as a model chaotic system _ ( aerial press , santa cruz , ca , 1984 ) .
x. wu and z. a. schelly : physica d * 40 * ( 1989 ) 433 .
k. dreyer and f. r. hickey : am .
* 59 * ( 1991 ) 619 .
j. c. sartorelli , w. m. gonalves , and r. d. pinto : phys .
e * 49 * ( 1994 ) 3963 .
p. m. c. de oliveira and t. j. p. penna : j. stat .
* 73 * ( 1993 ) 789 , and int .
phys . * 5*(1994 ) 997 .
very recently , n. fuchikami , s. ishioka and k. kiyono ( submitted to j. phys . soc .
jpn ) made a simulation based on a new algorithm with lagrangian description , which showed a good agreement with our experimental results . | experimentally shown drips from faucet have chaotic time - intervals shaw found types chaotic attractor in dripping intervals.wu schelly showed dependence of surface tension flow rate on attractor dramatic unlike temperature sartorelli gonalves pinto investigated time series dripping intervals narrow range flow rate revealed time series has boundary crisis tangent bifurcation for change flow rate theoretical approach to dripping early stage mass - spring model by shaw generated types attractor oliveira penna numerical simulation showed dripping dynamics of tiny - sized drops discrete attractors dripping interval of pure water affected by diameter nozzle temperature pressure flow rate studies performed for narrow range flow rate in our experiment temperature flow rate are control parameters behavior dripping interval examined over wide range flow rate figure shows apparatus distilled water constant depth 90.5 cm fills inner tube applies constant pressure to needle valve drips fall from glass - tube nozzle inside diameter 5 mm outside diameter 7 mm temperature water kept constant by circulating water at constant temperature 15 through outer tubetemperature water through needle valve elevated with electric heater temperature controlled by adjustment constant current heater temperature drip @xmath0 determined by needle valve apparatus installed on vibration - proof stand flow rate water from nozzle @xmath1 controlled by adjustment needle valve fine adjustment @xmath2 through control temperature @xmath0 drops from nozzle intersected laser beam 15 cm under edge detected by photodiode dripping time - intervals recorded by pc with accuracy 0.1 ms on temperature for three adjustments straight lines have slope of @xmath5 measurement @xmath3 for flow rates from @xmath6 to @xmath7 covered by 50 adjustments drip temperature changed at interval 1 from 15 to 27 for each adjustment figure shows dependence flow rate @xmath2 on temperature @xmath0 for three adjustments valve flow rate increases linearly with @xmath0 flow not affected by weak convective flow tank change flow rate about @xmath8 per 1 irrespective of adjustment valve data for return map of @xmath3 about two thousands same value @xmath9 @xmath12 figures fig:3 show change @xmath0 same return map @xmath3 different flow rates let @xmath14 small difference @xmath2 @xmath15 change @xmath0 ratio @xmath14 to @xmath15 @xmath16 @xmath17 average for differed values @xmath2 introduced reduced flow rate @xmath9 @xmath18 @xmath19 same values @xmath9 time series @xmath3 drew same maps figures fig:3 illustrate maps obtained @xmath20 temperature @xmath0 changed 50 adjustments valve experimental results rearranged with reduced flow rates @xmath9 @xmath21 low range @xmath22 to @xmath23 middle range @xmath24 to @xmath25 high range @xmath26 to @xmath27 arrows indicate transition points stable to chaotic states figure fig:4 shows dripping spectrum @xmath3 bifurcation diagram against @xmath9 three ranges dripping spectrum obtained number data several hundreds stable state @xmath3 constant @xmath28 chaotic state fluctuates two states appear alternately increasing @xmath9 stable state interval @xmath3 increases with @xmath9 chaotic state @xmath3 decreases over @xmath29 @xmath9 arrows indicate transition points from stable to chaotic @xmath30 reduced flow rate transition point @xmath3 maximum @xmath31 transition - point number from 1 to 43 toward large value @xmath30 straight line slope -0.92 function transition - point number @xmath31 @xmath32 time interval at transition point @xmath30 log - plot of @xmath32 vs. @xmath30 @xmath33 @xmath34 exponent above -1 volume drip @xmath35 increases with @xmath30 @xmath36 plot @xmath32 against @xmath31 interval @xmath32 linearly decreasing function of @xmath31 constant decrement of @xmath37 chaotic state from @xmath38 to @xmath39:7 shows four return maps of @xmath3 chaotic state @xmath40 to @xmath41 values @xmath9 different return maps different form number data smaller in dripping spectrum maps sparsely distributed dots map appear in spectrum mapping functions @xmath42 well - definedmaps are multi - valued functions of @xmath3 chaotic attractors have low dimensions attractors not high - dimensional by wu and schelly in stable states interval @xmath3 increases with @xmath9 volume drop increases chaotic state in average volume average @xmath3 decreases rapidly than @xmath43 with increase in @xmath9 in fig volume repeats increase and decrease with @xmath9 increased volume in stable state partially compensated by decreased volume in chaotic states if increase decrease ignored volume increase monotonicaly with @xmath9 form of @xmath44 drip vibrates on edge nozzle amplitude vibration depends on flow rate drip on edge grows with vibration into volume falling drop if drip torn off by weight increase and decrease in volume chaotic fluctuations in @xmath3 occur experimental results suggest vibration affects behavior @xmath3 in stable state volume drip constant for drip number @xmath29 at fixed flow rates steady behavior of @xmath3 dominated by mechanics by weight constant volume increases with @xmath9 not explained only by mechanicschaotic state mechanics behavior @xmath3 factors behavior vibration drip edge important chaotic behavior dripping interval quantitative analysis vibration r. shaw dripping faucet model chaotic system aerial press santa cruz ca 1984 ) x. wu z a. schelly physica 40 * 1989 ) 433. dreyer f r. hickey 59 * ( 1991 ) 619 . sartorelli w m. gonalves r. d. pinto phys . 49 * 1994 ) 3963 . de oliveira p. penna j. stat . 73 1993 ) 789 int . . 5*(1994 ) 997 n. fuchikami s. ishioka k. kiyono j phys . soc jpn simulation new algorithm lagrangian description agreement experimental results | 0.358176 | 2,560 | 1,275 | , Saving $0.1 in GPT-4. |
67,522 | we investigate the posibility of improving the @xmath0mass measurement at atlas . given the high statistics of both @xmath0and @xmath1bosons expected at the lhc , we estimate that a precision of 7 @xmath2per channel can be reached with @xmath3 . | the standard model ( sm ) is a very predictive framework . given precise measurements of @xmath4 , @xmath5 , and @xmath6 , the @xmath0 boson mass plays a central role , as it allows for both a sm cross check , confronting predictions of the @xmath0 and top quark masses @xcite with measurements @xcite , and limits on the sm higgs boson mass @xcite .
finally , constraints on the contributions of other heavy particles , like supersymmetric particles @xcite can be obtained .
the @xmath0 mass precision has continually improved with statistics , yielding the current world average of @xmath7 @xcite .
further improvement will translate into more precise indirect predictions of the sm higgs mass .
the simulated @xmath0 and @xmath1 boson signal and associated background samples used in this study are computed using the pythia general purpose event generator @xcite , with photon radiation in @xmath0and @xmath1decays treated via an interface to photos @xcite .
the size of the expected samples are computed assuming the nlo @xmath0and @xmath1cross - sections , as obtained from resbos @xcite , and simulated with complete simulation of the atlas detector using geant4 @xcite . at hadron colliders ,
@xmath0and @xmath1events can be detected and reconstructed in the @xmath8 , @xmath9 , @xmath10 , and @xmath11 final states . in the following
, the term lepton ( @xmath12 ) will refer to either an electron or muon .
electrons are measured using the inner detector ( i d ) and electromagnetic calorimeter ( emc ) .
they are reconstructed and identified with an efficiency of about 65% , while rejecting background from jets up to one part in @xmath13 .
the transition region from barrel to endcap in the emc ( @xmath14 ) is not used . for muons ,
the i d is used together with the muon spectrometer with a reconstruction efficiency of about 95% .
backgrounds are less than for electrons , and diminished using isolation .
the transverse momentum of the neutrino is inferred from the transverse energy imbalance as determined by the calorimeters .
the relative energy resolution is typically 1.5% for electrons and 2.0% for muons , while the missing transverse momentum ( met ) has a resolution of 15 - 25% @xcite .
the @xmath0signal is extracted by selecting events with one isolated lepton ( @xmath15 and @xmath16 ) along with significant met due to the undetected neutrino ( @xmath17 ) .
these selections have a total efficiency ( trigger and selection ) of about 20% ( 40% ) for the electron ( muon ) channel , providing a sample of about @xmath18 ( @xmath19 ) events .
the backgrounds are at the 3% ( 6% ) percent level .
likewise , the @xmath1signal is required to have two opposite sign leptons ( @xmath15 and @xmath20 ) .
the efficiency of this selection is about 10% ( 30% ) in the electron ( muon ) channel , yielding samples of about @xmath21 ( @xmath22 ) events .
g. aad _ et al . _ [ atlas collaboration ] , jinst * 3 * ( 2008 ) s08003 b. abbott _ et al .
_ [ d0 collaboration ] , phys . rev .
d * 61 * ( 2000 ) 032004 [ arxiv : hep - ex/9907009 ] .
s. berge , p. m. nadolsky , f. i. olness and c. p. yuan , aip conf .
* 792 * ( 2005 ) 722 [ arxiv : hep - ph/0508215 ] .
n. besson , m. boonekamp , e. klinkby , t. petersen and s. mehlhase , arxiv:0805.2093 [ hep - ex ] . | standard model predictive precise measurements @xmath4 @xmath0 boson mass central allows sm cross check confronting predictions @xmath0 top quark masses limits sm higgs boson mass constraints on contributions heavy particles supersymmetric obtained @xmath0 mass precision improved world average of @xmath7 @xcite improvement precise predictions sm higgs mass simulated @xmath0 @xmath1 boson signal background samples computed using pythia event generator @xcite photon radiation treated via interface photos @xcite size expected samples computed nlo @xmath0and @xmath1cross - sections obtained simulated with atlas detector geant4 @xcite hadron colliders @xmath0and @xmath1events detected reconstructed in @xmath8 final states term lepton ( @xmath12 ) electron or muon electrons measured using inner detector electromagnetic calorimeter reconstructed identified efficiency 65% rejecting background from jets one part in @xmath13 transition region from barrel to endcap in emc @xmath14 ) not usedmuons i d used muon spectrometer reconstruction efficiency 95% backgrounds less electrons diminished isolation transverse momentum neutrino inferred from energy imbalance calorimeters relative energy resolution 1.5% electrons 2.0% muons missing transverse momentum resolution 15 - 25% @xcite @xmath0signal extracted selecting events one isolated lepton ) significant met undetected neutrino ) selections efficiency 20% ( 40% ) electron muon channel sample @xmath18 ( @xmath19 ) events backgrounds 3% ( 6% ) percent level @xmath1signal two opposite sign leptons ) efficiency 10% ( 30% ) electron muon channel samples @xmath21 ( @xmath22 ) events. aad et al . atlas collaboration jinst 3 ( 2008 ) s08003 abbott al collaboration phys . rev . 61 2000 ) 032004 hep - ex/9907009 berge nadolsky olness yuan conf. 792 2005 ) 722 hep - ph/0508215 besson boonekamp klinkby petersen mehlhase arxiv:0805.2093 hep - ex | 0.136116 | 918 | 460 | , Saving $0.0 in GPT-4. |
67,528 | we investigate curvature effects on geometric parameters , energetics and electronic structure of zigzag nanotubes with fully optimized geometries from first - principle calculations . the calculated curvature energies , which are inversely proportional to the square of radius , are in good agreement with the classical elasticity theory .
the variation of the band gap with radius is found to differ from simple rules based on the zone folded graphene bands .
large discrepancies between tight binding and first principles calculations of the band gap values of small nanotubes are discussed in detail . | single wall carbon nanotubes ( swnts ) are basically rolled graphite sheets , which are characterized by two integers @xmath0 defining the rolling vector of graphite.@xcite therefore , electronic properties of swnts , at first order , can be deduced from that of graphene by mapping the band structure of 2d hexagonal lattice on a cylinder.@xcite such analysis indicates that the @xmath1 armchair nanotubes are always metal and exhibit one dimensional quantum conduction @xcite .
the @xmath2 zigzag nanotubes are generally semiconductor and only are metal if @xmath3 is an integer multiple of three .
however , recent experiments@xcite indicate much more complicated structural dependence of the band gap and electronic properties of swnts . the semiconducting behavior of swnts has been of particular interest ,
since the electronic properties can be controlled by doping or implementing defects in a nanotube - based optoelectronic devices.@xcite it is therefore desirable to have a good understanding of electronic and structural properties of swnts and the interrelations between them .
band calculations of swnts were initially performed by using a oneband @xmath4orbital tight binding model.@xcite subsequently , experimental data @xcite on the band gaps were extrapolated to confirm the inverse proportionality with the radius of the nanotube.@xcite later , first principles calculation @xcite within local density approximation ( lda ) showed that the @xmath5@xmath6 hybridization becomes significant at small @xmath7 ( or at high curvature ) .
such an effect were not revealed by the @xmath4orbital tight - binding bands .
recent analytical studies @xcite showed the importance of curvature effects in carbon nanotubes
. nonetheless , band calculations performed by using different methods have been at variance on the values of the band gap .
while recent studies predict interesting effects , such as strongly local curvature dependent chemical reactivity @xcite , an extensive theoretical analysis of the curvature effects on geometric and electronic structure has not been carried out so far . in this paper , we present a systematic ab - initio analysis of the band structure of zigzag swnts showing interesting curvature effects .
our analysis includes a large number of zigzag swnts with @xmath3 ranging from 4 to 15 .
the fully optimized structural and electronic properties of swnts are obtained from extensive first - principle calculations within the generalized gradient approximation@xcite ( gga ) by using pseudopotential planewave method @xcite .
we used plane waves up to an energy of 500 ev and ultrasoft pseudopotentials @xcite .
the calculated total energies converged within 0.5 mev / atom .
more details about the calculations can be found in ref s [ ] .
, @xmath8 , @xmath9 and @xmath10 .
radius dependence of these variables are important in tight - binding description of swnts as discussed in the text . ]
first , we discuss effects of curvature on structural parameters such as bond lengths and angles . figure [ fig : zigzagstr ] shows a schematic side view of a zigzag swnt which indicates two types of cc bonds and ccc bond angles , respectively .
the curvature dependence of the fully optimized structural parameters of zigzag swnts are summarized in fig .
[ fig : idstr ] .
the variation of the normalized bond lengths ( _ i.e. _ @xmath11 where @xmath12 is the optimized cc bond length in graphene ) and the bond angles with tube radius @xmath7 ( or @xmath3 ) are shown in fig .
[ fig : idstr]a and b , respectively . both the bond lengths and
the bond angles display a monotonic variation and approach the graphene values as the radius increases .
as pointed out earlier for the armchair swnts @xcite , the curvature effects , however , become significant at small radii . the zigzag bond angle ( @xmath9 )
decreases with decreasing radius .
it is about @xmath13 less than @xmath14 , namely the bond angle between @xmath15 bonds of the graphene , for the @xmath16 swnt , the smallest tube we studied .
the length of the corresponding zigzag bonds ( @xmath8 ) , on the other hand , increases with decreasing @xmath7 . on the other hand ,
the length of the parallel bond ( @xmath17 ) decreases to a lesser extent with decreasing @xmath7 , and the angle involving this bond ( @xmath10 ) is almost constant . and
@xmath18 ) versus the tube radius @xmath7 .
( @xmath19 ) .
( b ) the bond angles ( @xmath9 and @xmath10 ) versus @xmath7 .
( c ) the curvature energy , @xmath20 per carbon atom with respect to graphene as a function of tube radius .
the solid lines are the fit to the data as @xmath21 . ]
an internal strain is implemented upon the formation of tubular structure from the graphene sheet .
the associated strain energy , which is specified as the curvature energy , @xmath20 , is calculated as the difference of total energy per carbon atom between the bare swnt and the graphene ( _ i.e. _ @xmath22 ) for @xmath23 . the calculated curvature energies are shown in fig .
[ fig : idstr]c .
as expected @xmath24 is positive and increases with increasing curvature .
consequently , the binding ( or cohesive ) energy of carbon atom in a swnt decrease with increasing curvature . we note that in the classical theory of elasticity the curvature energy is given by the following expression @xcite @xmath25 here @xmath26 is the young s modulus , @xmath27 is the thickness of the tube , and @xmath28 is the atomic volume .
interestingly , the ab - initio curvature energies yield a perfect fit to the relation @xmath29 as seen in fig . [
fig : idstr]c .
this situation suggests that the classical theory of elasticity can be used to deduce the elastic properties of swnts . in this fit @xmath30
is found to be 2.14 ev @xmath31/atom , wherefrom @xmath26 can be calculated with an appropriate choice of @xmath27 .
[ cols="<,^,^,^,^,^,^,^,^,^,^,^,^",options="header " , ] an overall behavior of the electronic band structures of swnts has been revealed from zone folding of the graphene bands.@xcite accordingly , all @xmath2 zigzag swnt were predicted to be metallic when @xmath3 is multiple of 3 , since the double degenerate @xmath4 and @xmath6 states , which overlap at the @xmath32point of the hexagonal brillouin zone ( bz ) of graphene folds to the @xmath33 point of the tube @xcite .
this simple picture provides a qualitative understanding , but fails to describe some important features , in particular for small radius or _
metallic _ nanotubes .
this is clearly shown in table [ table : ideg ] , where the band gaps calculated in the present study are summarized and compared with results obtained from other methods in the literature .
for example , our calculations result in small but non - zero energy band gaps of 93 , 78 and 28 mev for @xmath34 , @xmath35 and @xmath36 swnts , respectively ( see table 1 ) .
recently , these gaps are measured by scanning tunneling spectroscopy ( sts ) experiments @xcite as 80 , 42 and 29 mev , in the same order .
the biggest discrepancy noted in table [ table : ideg ] is between the tight - binding and the first - principles values of the gaps for small radius tubes such as @xmath37 .
these results indicate that curvature effects are important and the simple zone folding picture has to be improved . moreover , the analysis of the lda bands of the @xmath38 swnt calculated by blase _
@xcite brought another important effect of the curvature .
the antibonding singlet @xmath6 and @xmath5 states mix and repel each other in curved graphene . as a result
, the purely @xmath6 state of planar graphene is lowered with increasing curvature . for zigzag swnts ,
the energy of this singlet @xmath6state is shifted downwards with decreasing @xmath7 ( or increasing curvature ) . here , we extended the analysis of blase _
@xcite to the @xmath2 swnts with @xmath23 by performing gga calculations .
states ( vb ) , the double degenerate @xmath6states ( cb ) and the singlet @xmath6state as a function of nanotube radius .
each data point corresponds to @xmath3 ranging from 4 to 15 consecutively .
( b ) the calculated band gaps as a function of the tube radius shown by filled symbols .
solid ( dashed ) lines are the plots of eq .
[ gap1r2 ] ( eq . [ gap1r ] ) .
the experimental data are shown by open diamonds @xcite . ] in fig .
[ fig : idband]a , we show the double degenerate @xmath4states ( which are the valence band edge at the @xmath33point ) , the double degenerate @xmath6states ( which become the conduction band edge at @xmath33 for large @xmath7 ) , and the singlet @xmath6state ( which is in the conduction band for large @xmath7 ) .
as seen , the shift of the singlet @xmath6state is curvature dependent , and below a certain radius determines the band gap . for tubes with radius greater than 3.3 ( _
i.e. _ @xmath39 ) , the energy of the singlet @xmath6state at the @xmath33point of the bz is above the doubly degenerate @xmath6 states ( _ i.e. _ bottom of the conduction band ) , while it falls between the valence and conduction band edges for @xmath40 , and eventually dips even below the double degenerate valence band @xmath4states for the zigzag swnt with radius less than 2.7 ( _ i.e. _ @xmath41 ) .
therefore , all the zigzag tubes with radius less than 2.7 are metallic . for @xmath40 , the edge of the conduction band
is made by the singlet @xmath6state , but not by the double degenerate @xmath6state .
the band gap derived from the zone folding scheme is reduced by the shift of this singlet @xmath6state as a result of curvature induced @xmath42 mixing .
this explains why the tight binding calculations predict band gaps around 1 ev for @xmath40 tubes while the self - consistent calculations predict much smaller value .
another issue we next address is the variation of the band gap , @xmath43 , as a function of tube radius .
based on @xmath4orbital tight binding model , it was proposed @xcite that @xmath43 behaves as @xmath44 which is independent from helicity . within the simple @xmath4orbital tight binding model , @xmath45 is taken to be equal to the hopping matrix element @xmath46 .
( @xmath12 is the bond length in graphene ) .
however , as seen in fig .
[ fig : idband]b , the band gap displays a rather oscillatory behavior up to radius 6.0 . the relation given in eq .
[ gap1r ] was obtained by a second order taylor expansion of one - electron eigenvalues of @xmath4-orbital tight binding model @xcite around the @xmath32point of the bz , and hence it fails to represent the effect of the helicity . by extending the taylor expansion to the next higher order , yorikawa and
muramatsu @xcite included another term in the empirical expression of the band gap variation , @xmath47 , \label{gap1r2}\ ] ] which depends on the chiral angle , @xmath48 , as well as an index @xmath49 . here
@xmath50 is a constant and the index @xmath49 is defined as the integer from @xmath51 .
the factor @xmath52 comes from the fact that the allowed @xmath53 is nearest to either the @xmath32- or @xmath54- point of the hexagonal brillouin zone . for zigzag
nanotubes studied here , the chiral angle is zero , so the second term just gives @xmath55 dependence as @xmath56 .
hence , the solid lines in fig . [ fig : idband]b are fits to the empirical expression , @xmath57 , obtained from eq . [ gap1r2 ] for @xmath580 by using the parameters @xmath59 ev and @xmath60 .
the experimental data obtained by sts @xcite are shown by open diamonds in the same figure .
the agreement between our calculations and the experimental data is very good considering the fact that there might be some uncertainties in identifying the nanotube ( _ i.e. _ assignment of @xmath0 indices ) in the experiment .
the fit of this data to the empirical expression given by eq .
[ gap1r ] are also presented by a dashed line for comparison .
the situation displayed in fig .
[ fig : idband ] indicates that the variation of the band gap with the radius is not simply @xmath61 , but additional terms incorporating the chirality dependence are required .
most importantly , the mixing of the singlet @xmath6state with the the singlet @xmath5state due to the curvature , and its shift towards the valence band with increasing curvature is not included in neither the @xmath4orbital tight binding model , nor the empirical relations expressed by eq . [ gap1r ] and [ gap1r2 ] .
this behavior of the singlet @xmath6states is of particular importance for the applied radial deformation that modifies the curvature and in turn induces metallization @xcite . in conclusion
, we investigated structural and electronic properties that result from the tubular nature of the swnts .
the first - principles total energy calculations indicated that significant amount of strain energy is implemented in a swnt when the radius is small .
however , the elastic properties can be still described by the classical theory of elasticity .
we showed how the singlet @xmath62state in the conduction band of a zigzag tube moves and eventually enters in the band gap between the doubly degenerate @xmath62-conduction and @xmath4-valence bands . as a result , the energy band structure and
the variation of the gap with radius ( or @xmath3 ) differs from what one derived from the zone folded band structure of graphene based on the simple tight binding calculations .
dresselhaus , g. dresselhaus and p.c .
eklund , _ science of fullerenes and carbon nanotubes _ ; academic press , san diego ( 1996 ) ; r. saito , g. dresselhaus and m. s. dresselhaus , _ physical properties of carbon nanotubes _ ; imperial college press , london ( 1998 ) . | single wall carbon nanotubes are rolled graphite sheets characterized by two integers defining rolling vector graphite electronic properties deduced from graphene by mapping band structure of 2d hexagonal lattice on cylinder @xmath1 armchair nanotubes are metal exhibit one dimensional quantum conduction @xmath2 zigzag nanotubes semiconductor metal if @xmath3 is multiple of three recent experiments indicate complicated structural dependence of band gap electronic properties semiconducting behavior electronic properties controlled by doping implementing defects in nanotube optoelectronic devices desirable to understanding of electronic structural properties interrelations band calculations performed oneband @xmath4orbital tight binding model data on band gaps confirm inverse proportionality with radius nanotube @xmath5@xmath6 hybridization significant at small high curvature effect not revealed by @xmath4orbital tight - binding bands studies showed importance of curvature effects in carbon nanotubes band calculations on values band gaprecent studies predict effects local curvature dependent chemical reactivity extensive theoretical analysis of curvature effects on geometric electronic structure not carried out paper present systematic analysis of band structure of zigzag swnts showing curvature effects analysis includes zigzag swnts with @xmath3 4 to 15 optimized structural electronic properties obtained from first - principle calculations generalized gradient approximation pseudopotential planewave method used plane waves to 500 ev ultrasoft pseudopotentials calculated total energies converged within 0.5 mev / atom details calculations in ref s @xmath8 @xmath10 radius dependence variables important in description of swnts discuss effects curvature on structural parameters bond lengths angles figure shows view of zigzag swnt two types of cc bonds bond angles curvature dependence of structural parameters summarized in fig variation of normalized bond lengths optimized bond angles with tube radius @xmath7 shown in fig bond lengths bond angles display monotonic variation approach graphene values as radius increasespointed out armchair swnts curvature effects significant small radii zigzag bond angle @xmath9 decreases decreasing radius @xmath13 less than @xmath14 bond angle between @xmath15 bonds graphene @xmath16 swnt smallest tube studied length zigzag bonds @xmath8 ) increases decreasing @xmath7 length parallel bond @xmath17 decreases decreasing @xmath7 angle bond @xmath10 almost constant @xmath18 versus tube radius @xmath7 @xmath19 bond angles @xmath9 @xmath10 ) versus @xmath7 curvature energy @xmath20 per carbon atom graphene function tube radius solid lines data @xmath21 internal strain implemented formation tubular structure graphene sheet strain energy curvature energy @xmath20 calculated difference total energy per carbon atom between bare swnt @xmath22 @xmath23 calculated curvature energies shown in fig . @xmath24 positive increases increasing curvature binding energy carbon atom swnt decrease increasing curvatureclassical theory elasticity curvature energy by expression @xcite @xmath25 @xmath26 young s modulus @xmath27 thickness tube @xmath28 atomic volume ab - initio curvature energies fit to relation @xmath29 in fig . classical theory elasticity deduce elastic properties of swnts @xmath30 2.14 ev @xmath31/atom @xmath26 calculated with choice @xmath27 behavior electronic band structures of swnts revealed from zone folding bands @xmath2 zigzag swnt predicted metallic when @xmath3 multiple of 3 double degenerate @xmath4 @xmath6 states overlap at @xmath32point hexagonal brillouin zone graphene folds to @xmath33 point tube picture qualitative understanding fails describe important features for small radius metallic nanotubes shown in table band gaps calculated summarized compared with other methods calculations result in small non - zero energy band gaps of 93 78 28 mev for @xmath34 @xmath35 @xmath36 swnts table 1gaps measured by scanning spectroscopy experiments @xcite 80 42 29 mev discrepancy between tight - binding first - principles values gaps for small radius tubes @xmath37 curvature effects important zone folding picture analysis lda bands @xmath38 swnt effect curvature antibonding singlet @xmath6 @xmath5 states mix repel in curved graphene @xmath6 state graphene lowered with increasing curvature zigzag swnts energy singlet @xmath6state shifted downwards decreasing @xmath7 increasing curvature extended analysis to @xmath2 swnts @xmath23 gga calculations double degenerate @xmath6states singlet @xmath6state function of nanotube radius each data point corresponds to @xmath3 4 to 15 consecutively calculated band gaps function of tube radius by symbols solid lines plots of eq gap1r2 ] experimental data shown by open diamonds @xcite in fig .fig idband double degenerate @xmath4states valence band edge @xmath33point @xmath6states conduction band edge @xmath33 large @xmath7 singlet @xmath6state conduction band large @xmath7 shift singlet @xmath6state curvature dependent below radius determines band gap tubes radius greater than 3.3 @xmath39 energy singlet @xmath6state @xmath33point above @xmath6 states bottom conduction band falls between valence conduction band edges @xmath40 dips below valence band @xmath4states zigzag radius less than 2.7 @xmath41 zigzag tubes radius less than 2.7 are metallic @xmath40 edge conduction band made by singlet @xmath6state not double degenerate @xmath6state band gap reduced by shift singlet @xmath6state curvature induced @xmath42 mixing tight binding calculations predict band gaps around 1 ev for @xmath40 tubes self - consistent calculations smaller issue variation band gap @xmath43 function of tube radius @xmath43 behaves as @xmath44 independent from helicity@xmath4orbital model @xmath45 equal to hopping element @xmath46 @xmath12 bond length in graphene fig band gap oscillatory behavior up to radius 6.0 relation in eq gap1r obtained second order taylor expansion - electron eigenvalues around @xmath32point fails represent effect helicity extending taylor expansion order yorikawa muramatsu @xcite included term band gap variation @xmath47{gap1r2 depends on chiral angle @xmath48 index @xmath49 @xmath50 constant index @xmath49 integer from @xmath51 factor @xmath52 from @xmath53 nearest to @xmath32- or @xmath54- point hexagonal brillouin zone zigzag nanotubes chiral angle zero second term gives @xmath55 dependence as @xmath56 solid lines in fig . empirical expression @xmath57 from eq gap1r2 for @xmath580 parameters @xmath59 ev and @xmath60 experimental data sts @xcite shown by open diamonds figure agreement between calculations experimental data good uncertainties in identifying nanotubeassignment @xmath0 indices in experiment fit data to empirical expression eq gap1r ] presented by dashed line situation in fig . indicates variation band gap with radius not simply @xmath61 additional terms chirality dependence required mixing of singlet @xmath6state with @xmath5state due to curvature shift towards valence band not included in @xmath4orbital tight binding model empirical relations eq [ gap1r ] and gap1r2 ] behavior singlet @xmath6states for radial deformation curvature metallization investigated structural electronic properties from tubular nature of swnts energy calculations significant strain energy implemented in swnt when radius small elastic properties described by classical theory of elasticity singlet @xmath62state in moves enters in band gap between @xmath62-conduction and @xmath4-valence bands energy band structure variation of gap with radius @xmath3 ) differs from zone folded band structure of graphene simple tight binding calculations dresselhaus g eklund science of fullerenes carbon nanotubes academic press san diego 1996dresselhaus carbon nanotubes imperial college press london 1998 | 0.559875 | 3,317 | 1,692 | , Saving $0.1 in GPT-4. |
67,541 | correlated _ ab - initio _ ground - state calculations , using relativistic energy - consistent pseudopotentials , are performed for six ii - vi semiconductors .
valence ( @xmath0 ) correlations are evaluated using the coupled cluster approach with single and double excitations . an incremental scheme is applied based on correlation contributions of localized bond orbitals and of pairs and triples of such bonds . in view of the high polarity of the bonds in ii - vi compounds , we examine both
, ionic and covalent embedding schemes for the calculation of individual bond increments . also , a partitioning of the correlation energy according to local ionic increments is tested .
core - valence ( @xmath1 ) correlation effects are taken into account via a core - polarization potential . combining the results at the correlated level with corresponding hartree - fock data we recover about 94% of the experimental cohesive energies ; lattice constants are accurate to @xmath21% ; bulk moduli are on average 10% too large compared with experiment . | _ ab initio _ calculations for solids
are mostly performed nowadays formally within a one - particle picture using density - functional theory ( dft ) , i.e.describing the non - local exchange together with many - particle correlation effects within the local - density approximation ( lda)@xcite or within more sophisticated generalized - gradient approximations ( gga)@xcite . as an alternative , it is possible at the one - particle level to calculate the non - local exchange of a solid exactly , using a periodic hartree - fock ( hf ) scheme@xcite .
+ however , for a proper microscopic treatment of electron correlations it is necessary to go beyond the one - particle framework and to deal with many - body wave functions . in the case of finite systems ,
accurate quantum - chemical methods , of the configuration interaction ( ci ) or coupled - cluster ( cc ) type , have been developed for the determination of many - body wave functions and the corresponding correlation energies . in the case of solids ,
the latter methods are not directly applicable , but it is often possible , to a very good approximation , to cast the correlation energy of the infinite system into a rapidly converging expansion in terms of increments from localized orbital subsystems@xcite . due to the local character of the correlation hole these increments may be determined for finite fragments of the solid and , as a consequence , the correlation energy of the solid becomes accessible to a quantum - chemical treatment . + the method of local increments has been applied to elementary@xcite and iii - v semiconductors@xcite , with an expansion in terms of correlation contributions related to localized bond orbitals and to pairs and triples of such bonds . on the other hand ,
the same method has also successfully been used for ionic compounds such as mgo , cao , and nio@xcite ; there , the local increments refer to groups of atomic orbitals which can be assigned to given anionic or cationic centers . in an attempt to bridge the gap between covalent and ionic solids we now investigate the use of incremental schemes for correlation effects in ii - vi semiconductors . + in section ii ,
we present the results of hartree - fock self - consistent field ( scf ) calculations for the solids just mentioned .
the inclusion of correlations is discussed in section iii ; we first focus on the various correlation effects which have to be considered , and then elaborate on embedding procedures and the computational details of the method .
results follow in section iv .
finally , pertinent conclusions are given in section v.
a starting point for treating many - body correlation effects in solids are reliable hartree - fock self - consistent field results for the infinite system .
we performed hf ground - state calculations for six ii - vi semiconductors , i.e. , zns , znse and znte and the corresponding cd compounds in the zinc - blende structure , using the program package crystal92@xcite .
+ the large number of electrons and the relativistic effects occurring in these materials can efficiently be handled by means of pseudopotentials .
for the group vi elements we use the scalar - relativistic energy - consistent pseudopotentials of bergner et al . , together with the corresponding @xmath3 $ ] atomic valence basis sets@xcite . for their application in the crystal calculations the basis sets have to be modified , however : very diffuse exponents which are necessary to properly describe the tails of the free - atom wavefunctions cause numerical problems in crystal . in the solid the basis functions of neighbouring atoms take over their role due to close - packing
. therefore we leave out the most diffuse @xmath4 exponent and use a reduced atomic @xmath5 $ ] basis set supplemented by a @xmath6 polarization function , which has been optimized for the zn - compounds in the solid ( @xmath7=0.45 , @xmath8=0.35 , @xmath9=0.25 ) . + for the group iib elements zn and cd we use two different sets of pseudopotentials : first 20-valence - electron pseudopotentials , where the outer - core @xmath10 shell and the corresponding @xmath11 shells are explicitly treated together with the @xmath12 valence electrons@xcite . again , optimized @xmath13 $ ] basis sets are available ( which we use for the calculation of the free atoms ) , but a reoptimization is needed for the solid . leaving out the most diffuse @xmath14 and @xmath4 exponents of the atomic basis sets ( for cd the outer @xmath6 exponent , too ) , we determined , for each angular momentum independently , the smallest exponent which still leads to a stable solution in the crystal calculation . starting with these ( fixed ) outer exponents we reoptimized the inner ones yielding @xmath15 or @xmath16 $ ] basis sets ( see table [ basis ] ) .
+ furthermore , we use large - core 2-valence electron pseudopotentials for zn @xcite(see table [ pp ] ) and cd @xcite ( which will also be used at the correlated level , cf .
sect . 3 ) .
.[pp]the zn 2-valence - elelctron pseudopotential is represented in the semilocal form @xmath17 , where @xmath18 is the projection operator onto the hilbert subspace of angular symmetry @xmath19 .
the exponents @xmath20 and the coefficients @xmath21 are listed below . [ cols="^,^,^",options="header " , ] @xmath22 reference [ ] + @xmath23 reference [ ] the influence of the valence correlations , which are calculated at five points around the experimental lattice constant , is much smaller . with basis
a , they reduce the bulk moduli of the zn - compounds by about 4% . for the cd - compounds the reduction is below 1% with basis a ;
for cdte , where we also applied basis b , there is an reduction of about 4% .
overall , the calculated bulk moduli are about 10% too large compared with mean experimental values .
the error bars in our calculations can be roughly estimated to @xmath247% , all in all ; if we define experimental error bars through the spread of the various experimental values , they are of the same order of magnitude as our theoretical ones .
lda calculations reach the same level of accuracy .
the advantage of our approach is that we can discuss the influence of various electron - correlation effects on the lattice constant and the bulk modulus at a microscopic level .
we have determined ground - state properties ( cohesive energies , lattice constants , bulk moduli ) of six ii - vi semiconductors , at various theoretical levels .
the hartree - fock results have been obtained using the solid - state program crystal92 , explicitly treating the outer - core @xmath6 shell of zn and cd with small - core pseudopotentials and crystal - optimized basis sets . for the cohesive energies
, we recover between 60% and 70% of the experimental values .
the lattice constants are overestimated by up to 4.5% and the bulk moduli , evaluated at the experimental lattice constants , are too large .
accounting for core - valence correlations by means of core - polarization potentials has only a slight influence on cohesive energies but yields a large reduction of the lattice constants and the bulk moduli .
correlations of the valence electrons are determined with the help of an incremental expansion in terms of localized bonds , and pairs and triples of such bonds ; individual increments are evaluated at the coupled - cluster ( ccsd ) level , for fragments of the solid in a point - charge embedding .
including valence correlations , we reach 85% of the correlation contributions to the cohesive energies , or @xmath2 95% of the total experimental values .
valence correlations reduce the lattice constants only slightly , leading to final values within 1% of the experimental data ; there is also moderate decrease of the bulk moduli .
+ the present calculations show that a reliable microscopic description of electron - correlation effects in ii - vi semiconductors can be obtained in terms of local excitations , and that a unified treatment ( with uniform accuracy ) is possible this way for a variety of ionic as well as covalently bonded solids .
b. paulus , p. fulde , and h. stoll , phys .
b * 54 * , 2556 ( 1996 ) ; s. kalvoda , b. paulus , p. fulde , and h. stoll , phys . rev .
b * 55 * , 4027 ( 1997 ) ; b. paulus , f .- j . shi , and h. stoll , j. phys .
matter * 9 * , 2745 ( 1997 ) ; k. doll , m. dolg , p. fulde and h. stoll , phys .
b * 52 * , 4842 ( 1995 ) ; k. doll , m. dolg , and h. stoll , phys . rev .
b * 54 * , 13529 ( 1996 ) ; k. doll , m. dolg , p. fulde and h. stoll , phys . rev .
b * 55 * , 10282 ( 1997 ) ; r. dovesi , v. r. saunders , and c. roetti , _
computer code crystal 92 _ , gruppo di chimica teorica , university of torino and united kingdom science and engineering research council laboratory , daresbury , 1992 . is a package of _ ab initio _ programs written by h .- j .
werner and p.j .
knowles , with contributions from j. almlf , r.d .
amos , a. berning , c. hampel , r. lindh , w. meyer , a. nicklass , p. palmieri , k.a .
peterson , r.m .
pitzer , h. stoll , a.j .
stone , and p.r .
taylor ; c. hampel , k. peterson and h .- j .
werner , chem .
. lett . * 190 * , 1 ( 1992 ) ; p.j .
knowles , c. hampel and h .- j .
werner , j. chem .
* 99 * , 5219 ( 1993 ) .
w. mller , j. flesch , and w. meyer , j. chem .
* 80 * , 3297 ( 1984 ) ; p. fuentealba , h. preuss , h. stoll and l. v. szentply , chem .
phys . lett . *
89 * , 418 ( 1982 ) ; a. nicklass and h. stoll , mol . phys . *
86 * , 317 ( 1995 ) . | calculations for solids performed within one - particle picture using density - functional theory non - local exchange many - particle correlation effects local - density approximation or generalized - gradient approximations possible at one - particle level to calculate non - local exchange solid using periodic hartree - fock scheme for microscopic treatment of electron correlations necessary beyond one - particle deal with many - body wave functions finite systems accurate quantum - chemical methods coupled - cluster developed for determination of many - body wave functions correlation energies solids methods not directly applicable possible to cast correlation energy infinite system into converging expansion increments from localized orbital subsystems local character correlation increments determined for finite fragments solid correlation energy accessible to quantum - chemical treatment method of local increments applied to elementary@xcite iii - v semiconductors expansion correlation contributions related to localized bond orbitals pairs triples of bonds same method used for ionic compounds local increments refer to groups atomic orbitals assigned to anionic or cationic centersbridge gap between covalent ionic solids investigate incremental schemes for correlation effects in ii - vi semiconductors section ii present results hartree - fock - field calculations for solids inclusion correlations discussed section iii focus correlation effects embedding procedures computational details results follow section iv conclusions section v starting point for many - body correlation effects in solids hartree - fock self - consistent field results for infinite system performed calculations for six ii - vi semiconductors zns znse znte cd compounds in zinc - blende structure using program package crystal92@xcite large number electrons relativistic effects handled pseudopotentials group vi elements use scalar - relativistic energy - consistent pseudopotentials of bergner et al . @xmath3 $ atomic valence basis sets@xcite crystal calculations basis sets modified diffuse exponents cause numerical problems crystal solid functions leave diffuse @xmath4 exponent use reduced atomic @xmath5 $ basis set supplemented @xmath6 polarization function optimized for zn - compounds in solid.45.35.25group iib elements zn cd use two sets pseudopotentials first 20-valence - electron pseudopotentials outer - core @xmath10 @xmath11 shells treated with @xmath12 valence electrons@xcite optimized @xmath13 $ basis sets available free atoms reoptimization needed for solid leaving diffuse @xmath14 @xmath4 exponents @xmath6 exponent determined each angular momentum smallest exponent stable solution crystal calculation outer exponents reoptimized inner ones @xmath15 or @xmath16 $ ] basis sets use large - core 2-valence electron pseudopotentials for zn @xcite cd @xcite correlated level zn 2-valence - elelctron pseudopotential semilocal form @xmath17 @xmath18 projection operator hilbert subspace angular symmetry @xmath19 exponents @xmath20 coefficients @xmath21 listed below @xmath22 @xmath23 reference influence valence correlations calculated at five points around experimental lattice constant smaller reduce bulk moduli zn - compounds by 4%cd - compounds reduction below 1% with basis a cdte basis b reduction about 4% calculated bulk moduli 10% too large experimental values error bars calculations estimated to @xmath247% experimental error bars same as theoretical ones lda calculations reach same accuracy influence electron - correlation effects on lattice constant bulk modulus microscopic level determined ground - state properties cohesive energies lattice constants bulk moduli of six ii - vi semiconductors theoretical levels hartree - fock results obtained using solid - state program crystal92 treating outer - core shell zn cd with small - core pseudopotentials crystal - optimized basis sets cohesive energies recover 60% and 70% experimental values lattice constants overestimated by 4.5% bulk moduli too large core - valence correlations core - polarization potentials slight cohesive energies large reduction lattice constants bulk moduli correlations valence electrons determined incremental expansion localized bonds individual increments evaluated at coupled - cluster level including valence correlations reach 85% correlation contributions to cohesive energies 95% total experimental valuesvalence correlations reduce lattice constants slightly final values within 1% experimental data moderate decrease bulk moduli calculations show reliable microscopic description electron - correlation effects ii - vi semiconductors local excitations unified treatment uniform accuracy possible for ionic covalently bonded solids b. paulus p. fulde h. stoll . b * 54 *, 2556 ( 1996 ) kalvoda stoll 55 * 4027 ( 1997 ) paulus shi h stoll phys . 9 *, 2745 ( 1997 ) k. doll m. dolg p. fulde h stoll phys b 52 *, 4842 ( 1995 ) stoll 54 * 13529 ( 1996 ) 55 * 10282 ( 1997 ) r. dovesi v saunders c. roetti computer code crystal 92 gruppo chimica teorica university of torino united kingdom science engineering research council laboratory daresbury 1992 package programs by h j werner p.j knowles contributions j almlf amos berning hampel lindh meyer nicklass palmieripeterson pitzer h stoll stone taylor c hampel peterson werner chem 190 1 1992 ) knowles c hampel werner j chem 99 5219 ( 1993 ) w. mller j. flesch w. meyer j chem 80 3297 ( 1984 ) p. fuentealba h. preuss h. stoll v szentply chem 89 418 ( 1982 ) a. nicklass h. stoll mol . 86 317 ( 1995 ) | 0.35674 | 2,326 | 1,174 | , Saving $0.1 in GPT-4. |
67,542 | we report about recent progress in computing four - loop massive correlators .
the expansion of these correlators in the external momentum leads to vacuum integrals .
the calculation of these vacuum integrals can be used to determine taylor expansion coefficients of the vacuum polarization function and decoupling functions in perturbative quantum chromodynamics .
new results at four - loop order for the lowest taylor expansion coefficient of the vacuum polarization function and for the decoupling relation are presented . | two - point correlators have been studied in great detail in the framework of perturbative quantum field theory . due to simple kinematics ( only one external momentum )
even multi - loop calculations can be performed .
the results for all physically interesting diagonal and non - diagonal correlators and including _ full _ quark mass dependence are available up to @xmath0 @xcite .
+ at four - loop order the two - point correlators can be considered in two limits . in the high energy limit
massless propagators need to be calculated and in the low energy limit vacuum diagrams ( tadpole integrals without dependence on the external momentum ) arise .
the evaluation of these massive tadpoles in three - loop approximation has been pioneered in ref .
@xcite and automated in ref .
+ similar to the three - loop case , the analytical evaluation of four - loop tadpole integrals is based on the traditional integration - by - parts ( ibp ) method .
in contrast to the three - loop case the manual construction of algorithms to reduce arbitrary diagrams to a small set of master integrals is replaced by laporta s algorithm @xcite . in this context
the ibp identities are generated with numerical values for the powers of the propagators and the irreducible scalar products . in the next step ,
the resulting system of linear equations is then solved in the next step by expressing systematically complicated integrals in terms of simpler ones .
the resulting solutions are then substituted into all the other equations . + this reduction has been implemented in an automated @xcite based program in which partially ideas described in ref . @xcite
have been implemented .
the rational functions in the space - time dimension @xmath1 , which arise in this procedure , are simplified with the program @xcite .
the automated exploitation of _ all _ symmetries of the diagrams by reshuffling the powers of the propagators of a given topology in a unique way strongly reduces the number of equations which need to be solved . in general
, the tadpole diagrams encountered during our calculation contain both massive and massless lines .
in contrast , the computation of the four - loop @xmath2-functions can be reduced to the evaluation of four - loop tadpoles composed of completely massive propagators .
these special cases have been considered in @xcite .
the outline of this paper is as follows . in section [ vacpol ]
we discuss the calculation of the lowest expansion coefficient of the vacuum polarization function and present the results at four - loop order using methods as described above . in section [ decoupling ]
we present new results for the decoupling relation at four - loop order in perturbative qcd .
our conclusions are presented in section [ conclusion ] .
the vacuum polarization tensor @xmath3 is defined as @xmath4 where @xmath5 is the external momentum and @xmath6 is the electromagnetic current of a heavy quark with mass @xmath7 .
the tensor @xmath3 can be expressed by a scalar function , the vacuum polarization function @xmath8 through @xmath9 the longitudinal part @xmath10 vanishes due to the ward identity .
the polarization function @xmath8 is related to the experimentally measurable @xmath11-ratio @xmath12 through the dispersion relation : @xmath13 performing the @xmath14-th derivative of eq .
( [ disrel ] ) with respect to @xmath15 at @xmath16 one obtains the moments @xmath17 , which can be determined experimentally : @xmath18 the derivatives of the polarization function on the rhs are related to the taylor expansion coefficients @xmath19 : @xmath20 ( @xmath21 ) which can be calculated in perturbative qcd .
the first and higher derivatives are important for a precise determination of the charm- and bottom - quark mass ( see e.g. @xcite ) .
but also the lowest expansion coefficient @xmath22 has an interesting physical meaning : it relates the coupling of electromagnetic interaction in different renormalization schemes . in the case of
qed - on - shell - renormalization the residue of the photon propagator is one and the electrical charge @xmath23 coincides with the classical limit .
if one performs renormalization in the @xmath24-scheme one obtains a relation between the coupling constant of the electromagnetic interaction @xmath25 in qed - on - shell - renormalization and the coupling constant @xmath26 in the @xmath24-scheme : @xmath27 for massive quarks , interacting through gluons , @xmath28 has been evaluated in ref .
@xcite . at three - loop order in perturbative qcd
this relation has already been determined in ref .
@xcite . for the qed case
the corresponding result was calculated in ref .
@xcite . the first taylor coefficient @xmath22 has been calculated using the methods described in section [ intro ] .
all tadpole diagrams were expressed through the set of 13 master integrals shown in figure [ mastertopologies ] .
, title="fig:",height=56 ] @xmath29 , title="fig:",height=56 ] @xmath30 , title="fig:",height=56 ] @xmath31 , title="fig:",height=56 ] @xmath32 , title="fig:",height=56 ] @xmath33 , title="fig:",height=56 ] @xmath34 , title="fig:",height=56 ] @xmath35 , title="fig:",height=56 ] @xmath36 , title="fig:",height=56 ] @xmath37 , title="fig:",height=56 ] @xmath38 , title="fig:",height=56 ] @xmath39 , title="fig:",height=56 ] @xmath40 , title="fig:",height=56 ] @xmath41 these master integrals have been calculated in refs .
@xcite . inserting the master integrals into the lowest taylor coefficient of the polarization function and performing
the renormalization in the @xmath24-scheme one obtains the following result for the expansion coefficient @xmath42 : @xmath43 where @xmath44 is the strong coupling constant in the @xmath24-scheme and the symbol @xmath45 denotes the number of light quarks considered as massless .
the part of the four - loop contribution proportional to the number @xmath46 of active quarks has been calculated previously up to the first physical moment @xmath47 in ref .
the terms proportional to @xmath48 are even known to all orders @xmath49 @xcite .
@xcite the terms being proportional to @xmath50 have been calculated through an independent numerical method , which is based on the pad approximation @xcite .
these results are in complete agreement with those of eq .
( [ c0numerik ] ) .
the masses of the known quark species differ vastly in their magnitude .
often the mass of a heavy quark @xmath51 is much larger than the characteristic momentum scale @xmath52 of the reaction under consideration .
in such a two interrelated problems appear when using an ms - like renormalization scheme .
+ first , two large but quite different mass scales , @xmath52 and @xmath7 , lead to two different types of potentially dangerously large logarithms of @xmath53 and @xmath54 and the standard trick of a proper choice of the renormalization scale @xmath55 is no longer applicable ; + second , according to the appelquist - carazzone theorem @xcite the effects due to heavy particles eventually should in general ` decouple ' from low - energy physics .
however , a peculiarity of mass - independent renormalization schemes is that the decoupling theorem does not hold in its naive form for theories renormalized in such schemes .
the effective qcd action that appears will not be not canonically normalized .
large mass logarithms appear , when one calculates a physical observable .
+ fortunately , both problems are controlled once the expansion parameters are properly choosen and renormalization group improvement is performed @xcite . to be specific ,
consider qcd with @xmath56 light quarks and one heavy quark @xmath51 with mass @xmath7 .
the effective coupling constant @xmath57 is then expressed in terms of the one of the full theory through : @xmath58 where @xmath59 is the decoupling function and @xmath60 is the @xmath24 running mass of the heavy quark .
the decoupling function is known up to three - loop order @xcite .
its calculation can be reduced to the solution of vacuum integrals @xcite .
thus the methods described in section [ intro ] can be applied to calculate the decoupling function at four - loop order .
the same master integrals as those shown in figure [ mastertopologies ] appear . for a renormalization scale @xmath61 one
obtains the following numerical result at the four - loop level has been recently finished by y. schrder and m. steinhauser .
the same is true for @xmath22 of eq .
( [ c0numerik ] ) .
their results are in full agreement with ours .
both calculations have been made reecntly available in @xcite . ] : @xmath62 the knowledge of this decoupling function is of phenomenological importance , because it allows the determination of @xmath63 at the @xmath64-boson scale through evolution of the measured value of @xmath65 at the @xmath66-lepton scale .
a careful analysis of the effects of four - loop running and three - loop matching on the extraction of @xmath67 defined for 5 active quark flavors from @xmath68 defined for 3 active quark flavors has been recently performed @xcite .
we have checked that the inclusion of the newly computed four - loop matching condition leads to a further reduction of the theoretical error from the evolution .
we have calculated the lowest taylor expansion coefficient of the vacuum polarization function and the decoupling relation at four - loop order in perturbative qcd .
the lowest taylor expansion coefficient relates the electromagnetic coupling in the on - shell and in the @xmath24 renormalization schemes .
the decoupling relation is important for the determination of @xmath67 through evolution from the measured value of @xmath68 . +
* acknowledgments : * + we would like to thank m. faisst for discussions about the pad approach and m. tentyukov for great help in dealing with form3 and fermat .
k. ch . is grateful to m. steinhauser for useful discussions of the decoupling relations .
c.s . would like to thank the graduierten kolleg `` _ hochenergiephysik und teilchenastrophysik _ '' for financial support .
this work was supported by the sonderforschungsbereich transregio 9 . | two - point correlators studied perturbative quantum field theory simple kinematics one external momentum multi - loop calculations performed results for diagonal non - diagonal correlators quark mass dependence available up to @xmath0 @xcite at four - loop order two - point correlators considered in two limits high energy limit massless propagators low energy limit vacuum diagrams tadpole integrals without dependence external momentum arise evaluation of massive tadpoles in three - loop pioneered in ref . @xcite automated in ref . analytical evaluation of four - loop tadpole integrals based on traditional integration - by - parts method manual integrals replaced by laporta s algorithm @xcite ibp identities generated with numerical values for powers propagators irreducible scalar products linear equations solved by expressing complicated integrals in simpler ones resulting solutions substituted into equations reduction implemented in automated @xcite program rational functions in space - time dimension @xmath1 simplified with @xcite automated exploitation of symmetries reshuffling powers propagators reduces number equations tadpole diagrams contain massive and massless linescomputation four - loop @xmath2-functions evaluation four - tadpoles massive propagators special cases considered in @xcite outline paper section vacpol calculation lowest expansion coefficient vacuum polarization function results four - loop order decoupling new results decoupling four - perturbative qcd conclusions section conclusion vacuum polarization tensor @xmath3 defined @xmath4 @xmath5 external momentum @xmath6 electromagnetic current heavy quark mass @xmath7 tensor @xmath3 expressed scalar function vacuum polarization function @xmath8 @xmath9 longitudinal part @xmath10 vanishes due ward identity polarization function @xmath8 related to experimentally measurable @xmath11-ratio @xmath12 dispersion relation @xmath14-th derivative obtains moments @xmath17 determined experimentally derivatives polarization function related taylor expansion coefficients @xmath19 @xmath20 calculated in perturbative qcd first higher derivatives important for precise determination charm- bottom - quark mass lowest expansion coefficient @xmath22 coupling electromagnetic interaction renormalization schemesqed - on - shell - renormalization residue photon propagator one electrical charge @xmath23 coincides with classical limit renormalization in @xmath24-scheme obtains relation between coupling constant electromagnetic interaction @xmath25 coupling constant @xmath26 @xmath24-scheme @xmath27 for massive quarks interacting through gluons @xmath28 evaluated in ref . @xcite three - loop order perturbative qcd relation determined in ref . @xcite . qed case result calculated in ref . @xcite . first taylor coefficient @xmath22 calculated methods section [ intro ] tadpole diagrams expressed through 13 master integrals figure [ mastertopologies ] title="fig:",height=56 @xmath29 @xmath30 @xmath36 @xmath37 @xmath40 @xmath41 master integrals calculated in refs . @xcite .master integrals into lowest taylor coefficient polarization function renormalization in @xmath24-scheme obtains result for expansion coefficient @xmath42 : @xmath43 @xmath44 strong coupling constant @xmath45 light quarks massless four - loop contribution proportional to @xmath46 active quarks calculated to first physical moment @xmath47 terms proportional to @xmath48 known to all orders @xmath49 @xcite terms proportional to @xmath50 calculated independent numerical method based on pad approximation results agreement with eq masses quark species differ mass heavy quark @xmath51 larger than characteristic momentum scale @xmath52 reaction problems appear ms - like renormalization scheme different mass scales @xmath52 and @xmath7 lead to dangerously large logarithms of @xmath53 and @xmath54 choice renormalization scale @xmath55 no applicable appelquist - carazzone theorem effects heavy particles decouple from low - energy physics mass - independent renormalization schemes decoupling theorem hold for renormalized effective qcd action canonically normalized large mass logarithms appear physical observableproblems controlled expansion parameters choosen renormalization group improvement performed @xcite consider qcd with @xmath56 light quarks one heavy quark @xmath51 mass @xmath7 effective coupling constant @xmath57 expressed theory @xmath58 @xmath59 decoupling function @xmath60 @xmath24 running mass heavy quark decoupling function known three - loop order @xcite calculation reduced to solution vacuum integrals methods intro calculate decoupling function four - loop order same master integrals figure mastertopologies appear renormalization scale @xmath61 obtains numerical result four - loop level finished by y. schrder m. steinhauser same @xmath22 eq results agreement calculations available in @xcite knowledge decoupling function allows determination of @xmath63 at @xmath64-boson scale evolution measured value @xmath65 @xmath66-lepton scale analysis effects four - loop running three - loop matching on extraction @xmath67 5 flavors @xmath68 3 performed @xcite inclusion computed four - loop matching condition theoretical error evolutioncalculated lowest taylor expansion coefficient vacuum polarization function decoupling relation four - loop order perturbative qcd lowest taylor expansion coefficient relates electromagnetic coupling on - shell @xmath24 renormalization schemes decoupling relation important @xmath67 evolution @xmath68 m. faisst pad approach m. tentyukov form3 fermat grateful m. steinhauser decoupling relations graduierten kolleg hochenergiephysik teilchenastrophysik financial support work supported sonderforschungsbereich transregio 9 | 0.353802 | 2,344 | 1,177 | , Saving $0.1 in GPT-4. |
67,551 | rainfall in kerala state , the southern part of indian peninsula in particular is caused by the two monsoons and the two cyclones every year . in general ,
climate and rainfall are highly nonlinear phenomena in nature giving rise to what is known as the ` butterfly effect ' .
we however attempt to train an abf neural network on the time series rainfall data and show for the first time that in spite of the fluctuations resulting from the nonlinearity in the system , the trends in the rainfall pattern in this corner of the globe have remained unaffected over the past 87 years from 1893 to 1980 .
we also successfully filter out the chaotic part of the system and illustrate that its effects are marginal over long term predictions . | although technology has taken us a long way towards better living standards , we still have a significant dependence on nature .
rain is one of nature s greatest gifts and in third world countries like india , the entire agriculture depends upon rain .
it is thus a major concern to identify any trends for rainfall to deviate from its periodicity , which would disrupt the economy of the country .
this fear has been aggravated due to the threat by the global warming and green house effect .
the present study has a soothing effect since it concludes that in spite of short term fluctuations , the general pattern of rainfall in kerala has not undergone major deviations from its pattern in the past .
the geographical configuration of india with the three oceans , namely indian ocean , bay of bengal and the arabian sea bordering the peninsula gives her a climate system with two monsoon seasons and two cyclones inter - spersed with hot and cold weather seasons .
the parameters that are required to predict the rainfall are enormously complex and subtle so that the uncertainty in a prediction using all these parameters is enormous even for a short period . the period over which a prediction may be made is generally termed the event horizon and in best results , this is not more than a week s time .
thus it is generally said that the fluttering wings of a butterfly at one corner of the globe may cause it to produce a tornado at another place geographically far away .
this phenomenon is known as the _ butterfly effect .
_ the objective of this study is to find out how well the periodicity in these patterns may be understood using a neural network so that long term predictions can be made .
this would help one to anticipate with some degree of confidence the general pattern of rainfall for the coming years . to evaluate the performance of the network ,
we train it on the rainfall data corresponding to a certain period in the past and cross validate the prediction made by the network over some other period .
a difference diagram is plotted to estimate the extent of deviation between the predicted and actual rainfall .
however , in some cases , the cyclone may be either delayed or rushed along due to some hidden perturbations on the system , for example an increase in solar activity .
( see figure [ figure2 ] . )
these effects would appear as spikes in the difference diagram .
the information from the difference diagram is insufficient to identify the exact source of such spike formation .
it might have resulted from slight perturbations from unknown sources or could be due to an inaccurate modeling of the system using the neural network .
we thus use a standard procedure in statistical mechanics that can quantitatively estimate the fluctuations@xcite .
this is to estimate the difference in the fourier power spectra ( fps ) of the predicted and actual sequences . in the fps ,
the power corresponding to each frequency interval , referred to as the _ spectral density _ gives a quantitative estimate of the deviations of the model from reality .
rapid variations would contribute to high frequency terms and slowly varying quantities would correspond to low frequency terms in the power spectra .
the degree of information that may be extracted from the fps is of great significance .
if the model agrees with reality , the difference of the power spectra ( _ hereafter referred to as the residual fps _ ) should enclose minimum power in the entire frequency range .
an exact model would produce no difference and thus no residual fps pattern . if there are some prominent frequency components in the residue , that could indicate two possibilities ; either the network has failed to comprehend the periodicity , or that there is a new trend in the real world which did not exist in the past .
one can test whether the same pattern exists in the residual fps produced on the training set and confirm whether it is a new trend or is the drawback of the model .
a random fluctuation would be indicated in the residual fps by amplitudes at all frequency values as in the case of ` white ' noise spectra . here again , how much power is enclosed within the fps gives a quantitative estimate of the perturbations .
a low amplitude fluctuation can happen due to so many reasons .
but its effect on the overall predictability of the system would be minimal .
if however , the residual fps encloses a substantial power from the actual rainfall spectrum , the fluctuations could be catastrophic . in this study ,
the results indicate that the perturbations produced by the environment on the rainfall pattern of kerala state in india is minimal and that there is no evidence to envisage a significant deviation from the rainfall patterns prevailing here .
the general outline of the paper is as follows . in section [ abfnn ]
we present a brief outline of the adaptive basis function neural network ( abfnn ) , a variant of the popular back - propagation algorithm . in section
[ setups ] , the experimental set up is explained followed by the results and concluding remarks in section [ results ] .
it was shown in @xcite that a variant of the back - propagation algorithm ( backprop ) known as the adaptive basis function neural network performs better than the standard backprop networks in complex problems .
the abfnn works on the principle that the neural network always attempts to map the target space in terms of its basis functions or node functions . in standard backprop networks , this function is a fixed sigmoid function that can map between zero and plus one or between minus one and plus one the input applied to it from minus infinity to plus infinity .
it has many attractive properties that make the backprop an efficient tool in a wide variety of applications .
however serious studies conducted on the backprop algorithm have shown that in spite of its widespread acceptance , it systematically outperforms other classification procedures _ only _ when the targeted space has a sigmoidal shape @xcite .
this implies that one should _
choose _ a basis function such that the network may represent the target space as a nested sum of products of the input parameters in terms of the basis function .
the abfnn thus starts with the standard sigmoid basis function and alters its nonlinearity by an algorithm similar to the weight update algorithm used in backprop . instead of the standard sigmoid function ,
abfnn opts for a variable sigmoid function defined as @xmath0 here @xmath1 is a control parameter that is initially set to unity and is modified along with the connection weights along the negative gradient of the error function . it is claimed in @xcite that such a modification could improve the speed and accuracy with which the network could approximate the target space .
the error function is computed as : @xmath2 with @xmath3 representing the network output and @xmath4 representing the target output value . with the introduction of the control parameter
, the learning algorithm may be summarized by the following update rules .
it is assumed that each node , @xmath5 , has an independent node function represented by @xmath6 . for the output layer nodes ,
the updating is done by means of the equation : @xmath7 where @xmath8 is a constant which is identical to the learning parameter in the weight update procedure used by backprop .
for the first hidden layer nodes , the updating is done in accordance with the equation : @xmath9 \frac{\partial o_{k}}{\partial o_{k-1}}\ ] ] here @xmath10 is the connection weight for the propagation of the output from node @xmath11 to node @xmath12 in the subsequent layer in the network . the introduction of the control parameter results in a slight modification to the weight update rule ( @xmath13 ) in the computation of the second partial derivative term @xmath14 in : @xmath15 as : @xmath16 \left ( 1-o_{j}\right)\ ] ] the algorithm does not impose any limiting values for the parameter @xmath1 and it is assumed that care is taken to avoid division by zero .
in pace with the global interest in climatology , there has been a rapid updating of resources in india also to access and process climatological database .
there are various data acquisition centers in the country that record daily rainfall along with other measures such as sea surface pressure , temperature etc . that are of interest to climatological processing .
these centers are also associated to the world meteorological organization ( wmo ) .
the database used for this study was provided by the department of atmospheric sciences of cochin university of science and technology , a leading partner in the nation wide climatological study centers .
the database consists of the rainfall data from trivandrum in kerala , situated at latitude - longitude pairs ( @xmath17 n - @xmath18 e ) .
although the rainfall data from 1842 were recorded , there were many missing values in between that we had to restrict to periods for which a continuous time series was available .
this was obtained for the period from 1893 to 1980 . for training
our network , the trivandrum database from 1893 to 1933 was used .
since rainfall has an yearly periodicity , we started with a network having 12 input nodes .
it was observed that the network accuracy would systematically improve as we increased the number of input nodes from 12 to 48 covering the data corresponding to 4 years .
any increase in input nodes resulted in poorer representations .
further experimentation showed that it was not necessary to include information corresponding to the whole year , but a 3 month information centered over the predicted month of the fifth year in each of the 4 previous years would give good generalization properties .
we thus finalized our network with 12 input nodes each for the 3 months input data over 4 years , 7 hidden nodes and one output node .
thus , based on the information from the four previous years , the network would predict the amount of rain to be expected in each month of the fifth year .
the training was carried out until the root mean - square error stabilized to around 0.085 over the training data(40 years from 1893 - 1933 ) .
the abfnn converged to this value in around 1000 iterations starting from random weight values and took less than 10 minutes on a 300mhz pc using celeron processor and linux operating system .
after training , the data corresponding to the entire 87 years of rainfall in trivandrum city was presented to the network .
the output from the network was compared with the actual data in the time series .
in addition to visual comparison , the spectral analysis was also done on the test dataset to obtain a quantitative appreciation of the performance of the network .
in figure [ figure1 ] we show the difference pattern produced by the network over the entire 87 year period on the trivandrum database . the training period is also shown .
it may be noted that the deviations of the predicted signal from the actual in the entire dataset falls within the same magnitude range .
the root mean square error on the independent test dataset was found to be around 0.09 .
the two positive spikes visible in the plot corresponding to the two instances where the rainfall was delayed with reference to the month for which it was predicted are shown in figure [ figure2 ] .
the cause for the negative spike in figure [ figure1 ] is also seen in figure [ figure2 ] as due to the deviation of the time series from the predicted upward trend .
factors such as the el - nino southern oscillations ( enso ) resulting from the pressure oscillations between the tropical indian ocean and the tropical pacific ocean and their quasi periodic oscillations are observed to be the major cause of these deviations in the rainfall patterns@xcite . to obtain a quantitative appreciation of the learning algorithm , we resort to the spectroscopic analysis of the predicted and actual time series . the average power per sample distributed in the two sequence is shown in figure [ figure3 ] .
it is seen that the two spectra compare very well . to identify the differences ,
the deviation of the fps ( residual fps ) of the predicted sequence from the actual in the months following the training period was plotted . for comparison ,
the deviation of the fps of the actual rainfall patterns in this period from that of the training period was also plotted .
the resulting graphs are shown in figure [ figure4 ] . in attempting to predict the evolution of the system in the months following the training period , it is to be noted that the deviation of the spectra of the predicted sequence for this period is less than that of the fourier spectra corresponding to the training period .
this is because the network attempts to _ learn _ the behavior of a system where as fourier analysis means only a transformation from time domain to frequency domain .
also fourier analysis assumes the system to be stationary , which is not true in the case of rainfall phenomena . comparing the fps of the network output corresponding to the training period showed the existence of residual frequency components at 1 hz and 3 hz , as seen in the residual fps of figure [ figure4 ] .
they represent the error bars or the inherent uncertainty to be expected in the prediction of the quantum of rainfall due to those oscillations .
the information for _ learning _ such fluctuations is not in the time series data .
future studies shall include other climatic parameters such as temperature , sea surface pressure , solar activity etc . in conjunction with the time series information in an attempt to reduce the error bars in the prediction .
the rest of the spectra is spread over the entire frequency range producing random fluctuations in the rainfall phenomena .
it is also to be noted that the abf neural network is a better tool than the popular fourier analysis methods for predicting long term rainfall behavior .
the learning ability of the network can give a concise picture of the actual system where as the fourier spectra is only a transformation from time domain to frequency domain , assuming that the time domain sequence is stationary .
it thus fails to represent the dynamics of the system which is inherent in most natural phenomena and rainfall in particular .
the authors wish to express their sincere thanks to professor k. mohankumar of the department of atmospheric sciences of the cochin university of science and technology for providing us with the rainfall database and for useful discussions .
the first author would like to thank dr .
moncy v john of the department of physics , kozhencherri st .
thomas college for long hours of discussions .
n. s. philip and k. b. joseph , _ adaptive basis function for artificial neural networks _ , ( _ in press _ ) neurocomputing journal , also available at http://www.geocities.com/sajithphilip/research.htm m. a. kraaijveld _ small sample behaviour of multi - layer feed forward network classifiers : theoretical and practical aspects _
( delft university press , 1993 ) .
a. chowdhury and s. v mhasawade , _ variations in meteorological floods during summer monsoon over india _ , mausam , * 42 * , 2 , 167 - 170 , 1991 . | technology towards better living standards significant dependence on nature rain is nature s greatest gifts in third world countries like india agriculture depends upon rain major concern to identify trends for rainfall to deviate from periodicity disrupt economy fear aggravated due to threat global warming green house effect present study concludes short term fluctuations general pattern of rainfall in kerala not major deviations from past geographical configuration india with three oceans indian ocean bay of bengal arabian sea climate system with two monsoon seasons two cyclones hot and cold weather seasons parameters to predict rainfall complex subtle uncertainty in prediction enormous even for short period period prediction event horizon best results not more than a week s time fluttering wings of butterfly one may cause produce tornado known as _ butterfly effect objective study is to find periodicity in patterns using neural network long term predictions anticipate confidence general pattern of rainfall for coming years to evaluate performance network train on rainfall data past cross validate prediction other period difference diagram plotted to estimate extent deviation between predicted and actual rainfallcyclone may be delayed or rushed due to hidden perturbations for example increase in solar activity see figure these effects appear as spikes in difference diagram information insufficient to identify exact source spike formation might from perturbations unknown or to inaccurate modeling standard procedure in statistical mechanics fluctuations@xcite difference in fourier power spectra fps ) of predicted and actual sequences power to each frequency interval spectral density gives deviations model from reality rapid variations contribute to high frequency terms slowly varying correspond to low frequency terms degree of information extracted from fps of significance if model agrees with reality difference of power spectra residual fps should enclose minimum power in entire frequency range exact model no difference no residual fps pattern if prominent frequency components in residue could indicate two possibilities network failed to comprehend periodicity or new trend can test same pattern in residual fps on training set confirm new trend or drawback of model random fluctuation indicated in residual fps by amplitudes at all frequency values power within fps gives quantitative estimate of perturbationslow amplitude fluctuation due to many reasons effect on predictability system minimal if residual fps encloses substantial power from rainfall spectrum fluctuations could be catastrophic study results indicate perturbations environment on rainfall pattern of kerala state india minimal no evidence significant deviation from rainfall patterns outline of paper section [ abfnn ] brief outline of adaptive basis function neural network variant of popular back - propagation algorithm section setups experimental set up explained results concluding remarks results variant of back - propagation algorithm adaptive basis function neural network performs better than standard networks in complex problems abfnn works neural network map target space in basis functions standard backprop networks function fixed sigmoid function map between zero and plus one or minus one plus one input from to attractive properties make backprop efficient tool in applications studies backprop outperforms other classification procedures only when targeted space has sigmoidal shape implies choose basis function network represent target space as nested sum of input parameters abfnn starts with standard sigmoid basis function alters nonlinearity by algorithm similar to weight update algorithm in backpropstandard sigmoid abfnn opts for variable sigmoid function @xmath0 @xmath1 is control parameter initially set to unity modified connection weights along negative gradient of error function modification could improve speed accuracy network target space error function computed as @xmath2 @xmath3 network output @xmath4 target output value introduction control parameter learning algorithm summarized by update rules each node @xmath5 has independent node function by @xmath6 output layer nodes updating equation : @xmath7 @xmath8 constant identical to learning parameter weight update procedure first hidden layer nodes updating equation : @xmath9\partial o @xmath10 is connection weight for propagation output from node @xmath11 to node @xmath12 introduction control parameter to weight update rule ( @xmath13 ) computation second partial derivative term @xmath14 @xmath15 @xmath16 algorithm impose limiting values for parameter @xmath1 avoid division by zero global interest climatology rapid updating of resources in india to access process climatological database various data acquisition centers record daily rainfall measures sea surface pressure temperature interest climatological processingcenters associated to world meteorological organization wmo database for study provided by department of atmospheric sciences of cochin university science partner in nation climatological study centers database rainfall data from trivandrum kerala at latitude - longitude pairs ( @xmath17 n - @xmath18 e ) rainfall data from 1842 recorded missing values to continuous time series obtained for 1893 to 1980 training network trivandrum database from 1893 to 1933 used rainfall yearly started with network 12 input nodes network accuracy increased input nodes from 12 to 48 data 4 years increase nodes poorer representations experimentation not information whole year 3 month information over predicted month fifth year 4 good generalization finalized network with 12 input nodes for 3 4 years 7 hidden nodes one output node based on network rain each month fifth year training until root mean - square error stabilized to around 0.085 over(40 years 1893 - 1933 abfnn converged to value in 1000 iterations less than 10 minutes on 300mhz pc after training data to 87 years of rainfall in trivandrum presented to network output compared with actual data time seriesvisual comparison spectral analysis done on test dataset performance network figure figure1 ] difference pattern network over 87 year period on trivandrum database training period shown deviations of predicted signal from actual in entire dataset same magnitude range root mean square error on independent test dataset around 0.09 two positive spikes plot to rainfall delayed predicted shown in figure figure2 ] negative spike in due to deviation of time series from predicted upward trend el - nino southern oscillations pressure oscillations between tropical indian ocean tropical pacific ocean quasi periodic oscillations major cause of deviations in rainfall patterns learning algorithm spectroscopic analysis of predicted and actual time series average power per sample in two sequence shown in figure figure3 ] two spectra compare well differences deviation of fps of predicted sequence from actual in months following training period plotted for comparison deviation of fps of actual rainfall patterns from training period also plotted resulting graphs shown in figure [ figure4 ] predict evolution system in months following training period deviation of spectra of predicted sequence for this period less than fourier spectra to training periodnetwork learn behavior system fourier analysis transformation from time to frequency fourier assumes system stationary not true in rainfall phenomena fps network output training period residual frequency components at 1 hz 3 hz in residual fps figure figure4 represent error bars uncertainty in prediction rainfall due oscillations information for learning fluctuations not in time series data future studies include other climatic parameters temperature sea surface pressure solar activity series information reduce error bars spectra spread over frequency range producing random fluctuations in rainfall abf neural network better tool than fourier analysis for predicting long term rainfall behavior learning ability network concise picture of system fourier spectra transformation from time to frequency domain time domain sequence stationary fails to represent dynamics system in natural phenomena rainfall authors thanks to professor k. mohankumar atmospheric sciences cochin university science technology for rainfall database discussions first author dr . moncy v john department of physics kozhencherri st thomas college for discussions n s. philip k. b. joseph adaptive basis function for artificial neural networks neurocomputing journal available at http://www.geocities kraaijveld small sample multi - layer feed network classifiers theoretical practical aspects delft university press 1993 mhasawade variations meteorological floods summer monsoon india mausam 167 - 170 1991 | 0.511776 | 3,110 | 1,512 | , Saving $0.1 in GPT-4. |
67,552 | let @xmath0 be a simply connected exponential solvable lie group , @xmath1 a closed connected subgroup , and let @xmath2 be a representation of @xmath0 induced from a unitary character @xmath3 of @xmath1 .
the spectrum of @xmath2 corresponds via the orbit method to the set @xmath4 of coadjoint orbits that meet the spectral variety @xmath5 .
we prove that the spectral measure of @xmath6 is absolutely continuous with respect to the plancherel measure if and only if @xmath1 acts freely on some point of @xmath7 . as a corollary
we show that if @xmath0 is nonunimodular , then @xmath2 has admissible vectors if and only if the preceding orbital condition holds . | at the intersection of abstract harmonic analysis and wavelet theory lies the fundamental notion of admissibility . given a unitary representation @xmath2 of a locally compact topological group @xmath0 , a vector @xmath8 is admissible if the mapping @xmath9 is an isometry of @xmath10 into @xmath11 .
which representations have admissible vectors ?
this classical question is answered in a variety of contexts and for various classes of representations , for example when @xmath0 is type i and @xmath2 is irreducible @xcite , or when @xmath2 is the left regular representation of @xmath0 @xcite .
the monograph @xcite , in addition to containing numerous other references for admissibility , describes the relation between this question and plancherel theory . in this paper
we consider the following class of representations .
let @xmath0 be an exponential solvable lie group , and let @xmath2 be the unitary representation of @xmath0 induced from a unitary character of @xmath1 .
a description of the irreducible decomposition of @xmath2 is given in terms of the coadjoint orbit picture in @xcite .
on the other hand an explicit plancherel formula for @xmath0 is given in @xcite using coadjoint orbit parameters .
using these results , we give a simple necessary and sufficient condition that @xmath2 be a subrepresentation of the left regular representation of @xmath0 in terms of the orbit picture .
specifically , let @xmath2 be induced from the character @xmath12 of @xmath1 , let @xmath13 be the linear functional on @xmath14 corresponding to @xmath12 , so that @xmath15 .
let @xmath7 be the real affine variety of all @xmath16 whose restriction to @xmath14 is @xmath13 .
then @xmath7 is an @xmath17-space , and we show ( theorem [ measurethm ] ) that @xmath2 is contained in the left regular representation if and only if @xmath1 acts freely on some @xmath18 ( and hence on a zariski open subset of @xmath19 . ) combining this result with the methods of @xcite , it follows that if @xmath0 is nonunimodular , then the preceding condition is both necessary and sufficient in order that @xmath2 have admissible vectors ( corollary [ admissible ] ) .
if @xmath0 is unimodular , then the situation for admissibility is still murky .
let @xmath0 be a connected , simply connected exponential solvable lie group with lie algebra @xmath20 .
given @xmath21 , @xmath22 , and @xmath23 , we denote both the adjoint and coadjoint actions multiplicatively : ad@xmath24 and @xmath25 . given @xmath16 , let @xmath26 be the stabilizer of @xmath27 in @xmath0 ; then @xmath0 is connected and its lie algebra is @xmath28 = 0 \text { holds for all } z \in \g\}$ ] . for the remainder of this paper , we fix a closed connected subgroup @xmath1 of @xmath0 with lie algebra @xmath14 , a unitary character @xmath12 of @xmath1 , and a _ monomial _ representation @xmath29
. let @xmath30 satisfy @xmath15 so that @xmath31 \subset \ker f$ ] ; we also use the notation @xmath32 . in our notation , we will identify representations that are unitarily equivalent .
let @xmath33 denote the borel space of equivalence classes of irreducible representations of @xmath0 .
let @xmath2 be a monomial representation .
since @xmath0 is type i , we have a unique measure class on @xmath33 such that @xmath34 in particular , the plancherel measure class @xmath35 is the measure class determined by the regular representation @xmath36 . since the multiplicity @xmath37 @xmath35-a.e . , then @xmath2 is a subrepresentation of @xmath36 if and only if @xmath38 is absolutely continuous with respect to @xmath39 .
the measure class @xmath38 is described on @xmath40 as follows : for @xmath32 , we put @xmath41 .
let @xmath42 be the canonical lebesgue measure class on @xmath7 extended to @xmath43 : for any borel subset @xmath44 of @xmath43 , @xmath45 .
by @xcite , @xmath38 is the pushforward of @xmath42 to @xmath40 . though @xmath42 is singular with respect to the lebesgue measure class on @xmath43 ( unless @xmath1 is the trivial subgroup ) , its pushforward @xmath38 may be absolutely continuous with respect to the plancherel measure . in the next section
, we will determine when this is the case .
consider the action of @xmath1 on @xmath43 by the restriction of the coadjoint action .
given @xmath16 , the lie algebra of the stabilizer of @xmath27 in @xmath1 is @xmath46 . if @xmath47 , then one observes that @xmath48 . if @xmath49 , then the preceding observation shows that each @xmath50 is @xmath0-invariant .
put @xmath51 and @xmath52 .
we claim that @xmath53 is a zariski open subset of @xmath7 .
indeed , fix a basis @xmath54 for @xmath14 and a basis @xmath55 for @xmath20 . for @xmath16
let @xmath56 be the @xmath57 matrix @xmath58 & \ell [ y_{1 } , z_2 ] & \cdots & \ell [ y_{1 } , z_n ] \\ \ell [ y_{2 } , z_1 ] & \ell [ y_{2 } , z_2 ] & \cdots & \ell [ y_{2 } , z_n ] \\ \vdots & \vdots & \cdots & \vdots \\ \ell [ y_{m } , z_1 ] & \ell [ y_{m } , z_2 ] & \cdots & \ell [ y_{m } , z_n ] \end{matrix}\right].\ ] ] * for each @xmath68 , set @xmath69
. if @xmath70 is not an ideal in @xmath20 , then @xmath71 and @xmath72 are ideals . * if @xmath73 is not an ideal in @xmath20 , then the module @xmath74 is not @xmath75-split .
then @xmath76 is a global diffeomorphism of @xmath77 onto @xmath0 .
recall the basis @xmath78 of @xmath14 and put @xmath79 .
choose @xmath80 so that @xmath81 is an ordered basis of @xmath20 .
let @xmath82 be the natural global chart for @xmath43 determined by the ordered dual basis @xmath83 .
now define @xmath84 by @xmath85 note that for @xmath86 , @xmath87 defines a global diffeomorphism from @xmath88 onto @xmath7 ( here @xmath89 is the identity in @xmath1 ) .
moreover @xmath90 is a global chart for @xmath91 and @xmath92 is a coordinatization of the map @xmath93 . for simplicity of notation we set @xmath94
; observe that the coordinate functions for the map @xmath92 are given by @xmath95 and @xmath96 0.5 cm now for each @xmath105 , we have @xmath106 + \frac{u^2}{2 ! } [ [ y_j , z_k],z_k ] + \cdots \right ) = \ell[y_j , z_k]\ ] ] holds for each @xmath107 , while @xmath108 since @xmath109 is constant . on the other hand , @xmath110
hence the differential @xmath111 is given by the matrix @xmath112\ ] ] where @xmath113 denotes the @xmath114 zero matrix and @xmath115 denotes the @xmath116 identity matrix .
now by lemma [ hdim ] , we have rank of @xmath117 , and the result follows . [ measurethm ]
let @xmath1 be a closed connected subgroup of @xmath0 with lie algebra @xmath14 and let @xmath32 .
if @xmath1 acts freely on some point of @xmath7 then @xmath38 is absolutely continuous with respect to @xmath35 .
otherwise , @xmath38 is singular with respect to @xmath35 .
suppose that there is @xmath18 such that @xmath118 . by lemmas [ hdim ] and [ rank ] , @xmath119 is a non - empty zariski open subset of @xmath7 .
for each @xmath120 there exists a rectangular open neighbourhood @xmath121 of @xmath122 such that the restriction of @xmath93 to @xmath121 is a submersion , and hence that @xmath123 is open .
now @xmath124 is open and satisfies @xmath125 .
hence @xmath126 is open in @xmath127 .
next we invoke results concerning the stratification and parametrization of coadjoint orbits @xcite : there is a @xmath0 invariant zariski - open subset @xmath128 of @xmath43 , such that @xmath129 has the structure of a smooth manifold ( with underlying quotient topology ) and such that the quotient mapping @xmath130 is real analytic . now since @xmath128 is dense in @xmath43 , then @xmath131 .
since @xmath132 and @xmath128 is @xmath0-invariant , then @xmath133 is a non - empty zariski - open subset of @xmath7 , and the lebesgue measure @xmath42 on @xmath7 is supported on @xmath133 . since @xmath134 is included in the @xmath0-invariant set @xmath135 , then @xmath136 is disjoint from the set @xmath137 , and hence @xmath38 is supported on @xmath138 .
put @xmath139 ; since @xmath140 is real analytic , then so is @xmath141 .
moreover , @xmath142 is open in @xmath127 and @xmath143 .
since @xmath141 is real analytic on @xmath7 , its set of singular points in @xmath7 has @xmath42-measure zero , and @xmath141 is a submersion on the set of regular points in @xmath7 . since the pushforward of lebesgue measure by a submersion is absolutely continuous with respect to lebesgue measure , then @xmath38 is absolutely continuous with respect to the lebesgue measure class on @xmath144 .
since plancherel measure @xmath35 on @xmath129 belongs to the lebesgue measure class on @xmath129 @xcite and @xmath144 is an open subset of @xmath129 , then @xmath38 is absolutely continuous with respect to @xmath39 .
now suppose that for all @xmath18 , @xmath145 is non - trivial .
then for all points @xmath18 , the rank of @xmath93 at @xmath122 is less than @xmath146 .
it follows that the lebesgue measure of @xmath147 is zero , and hence @xmath148 .
but since @xmath53 is a zariski - open subset of @xmath7 , then the measure @xmath38 is supported on @xmath149 , and hence @xmath38 is singular with respect to @xmath35 .
we now turn to the question of admissibility .
let @xmath150 be any representation of @xmath0 acting in @xmath151 . for @xmath152 define @xmath153 by @xmath154 .
the vector @xmath155 is said to be admissible ( or a continuous wavelet ) if @xmath156 is an isometry of @xmath157 into @xmath11 . in this case , @xmath158 intertwines the representation @xmath150 with the left regular representation @xmath36 of @xmath0 , so that @xmath159 is a closed left invariant subspace of @xmath11 and @xmath150 is equivalent with @xmath36 acting in @xmath160 .
let @xmath161 be a closed left invariant subspace of @xmath11 , and let @xmath162 be the orthogonal projection onto @xmath161 .
then there is a unique ( up to @xmath35-a.e .
equality ) measurable field @xmath163 of orthogonal projections where @xmath164 is defined on @xmath165 , and so that @xmath166 holds for @xmath35-a.e .
@xmath167 . set @xmath168 .
we recall ( * ? ? ?
* theorem 4.22 ) . [ admissible ]
let @xmath161 be a closed left invariant subspace of @xmath11 .
if @xmath0 is nonunimodular , then @xmath161 has an admissible vector .
if @xmath0 is unimodular , then @xmath160 has an admissible vector if and only if @xmath169 is integrable over @xmath33 with respect to the plancherel measure @xmath35 .
suppose that @xmath0 is unimodular .
though it is clear that the condition that @xmath1 acts freely on points of @xmath7 is still necessary for admissibility , examples indicate that the multiplicity function is never integrable , and hence that @xmath2 never has admissible vectors in this case .
thus we make the following . | intersection harmonic analysis wavelet theory notion admissibility unitary representation @xmath2 compact topological group @xmath0 vector @xmath8 admissible if mapping @xmath9 isometry @xmath10 into @xmath11 which representations admissible vectors ? answered contexts classes representations @xmath0 type i @xmath2 irreducible @xcite @xmath2 left regular representation @xmath0 @xcite monograph @xcite describes relation plancherel theory class representations @xmath0 exponential solvable lie group @xmath2 unitary representation @xmath0 induced from character @xmath1 description irreducible decomposition @xmath2 coadjoint orbit picture @xcite explicit plancherel formula for @xmath0 in @xcite coadjoint orbit parameters condition @xmath2 subrepresentation of left regular representation @xmath0 @xmath2 induced from character @xmath12 @xmath1 @xmath13 linear functional on @xmath14 @xmath12 @xmath7 real affine variety of @xmath16 restriction @xmath14 @xmath13@xmath7 @xmath17-space @xmath2 in left regular representation if @xmath1 acts on @xmath18 open subset @xmath19 if @xmath0 nonunimodular preceding condition necessary @xmath2 admissible vectors if @xmath0 unimodular situation admissibility murky @xmath0 connected exponential solvable lie group with lie algebra @xmath20 @xmath21 @xmath22 @xmath23 denote adjoint coadjoint actions@xmath24 @xmath25 @xmath26 stabilizer of @xmath27 in @xmath0 @xmath0 connected lie algebra @xmath28 = 0 fix closed connected subgroup @xmath1 of @xmath0 with lie algebra @xmath14 unitary character @xmath12 of @xmath1 monomial representation @xmath29 @xmath30 satisfy @xmath15 @xmath31 \subset use notation @xmath32 identify representations unitarily equivalent @xmath33 denote borel space of equivalence classes of irreducible representations of @xmath0 @xmath2 monomial representation@xmath0 type i unique measure class on @xmath33 @xmath34 plancherel measure class @xmath35 determined by representation @xmath36 multiplicity @xmath37 @xmath35-a.e @xmath2 subrepresentation @xmath36 if @xmath38 continuous @xmath39 measure class @xmath38 described on @xmath40 @xmath32 @xmath41 @xmath42 lebesgue measure class @xmath7 extended to @xmath43 subset @xmath44 @xmath43 @xmath45 @xmath38 pushforward of @xmath42 to @xmath40 @xmath42 singular measure class @xmath43 pushforward @xmath38 continuous plancherel measure determine action of @xmath1 on @xmath43 restriction coadjoint action @xmath16 algebra stabilizer @xmath27 @xmath1 is @xmath46 @xmath47 @xmath48 @xmath49 @xmath50 @xmath0-invariant @xmath51 and @xmath52 @xmath53 subset of @xmath7 basis @xmath54 for @xmath14 basis @xmath55 for @xmath20@xmath16 @xmath56 @xmath57 matrix @xmath58 y_{1 } z_2 ]\cdots y_{1 z_n y_{2 } z_1 y_{2 z_2 y_{2 z_n y_{m } z_1 ] y_{m } z_2 ]_ z_n{matrix @xmath68 set @xmath69 @xmath70 not ideal @xmath20 @xmath71 @xmath72 ideals @xmath73 not ideal @xmath20 @xmath74 not @xmath75-split @xmath76 global diffeomorphism @xmath77 @xmath0 recall basis @xmath78 @xmath14 @xmath79 choose @xmath80 @xmath81 ordered basis @xmath20 @xmath82 natural global chart @xmath43 dual basis @xmath83 define @xmath84 @xmath85 @xmath86 @xmath87 global diffeomorphism @xmath88 @xmath7 @xmath89 identity @xmath1 @xmath90 global chart @xmath91 @xmath92 coordinatization map @xmath93simplicity set @xmath94 coordinate functions map @xmath92 @xmath95 @xmath96 0.5 cm each @xmath105 @xmath106 + \frac{u^2}{2 ! } y_j z_k\cdots = [y_j z_k] @xmath107 @xmath108 @xmath109 constant @xmath110 differential @xmath111 given matrix @xmath112 @xmath113 @xmath114 zero matrix @xmath115 @xmath116 identity matrix rank @xmath117 result follows @xmath1 closed subgroup @xmath0 algebra @xmath14 @xmath32 if @xmath1 acts freely @xmath7 @xmath38 continuous @xmath35 otherwise @xmath38 singular @xmath35 @xmath18 @xmath118 @xmath119 non - empty open subset @xmath7 each @xmath120 rectangular open neighbourhood @xmath121 @xmath122 restriction @xmath93 to @xmath121 submersion @xmath123 open @xmath124 open satisfies @xmath125 @xmath126 open in @xmath127invoke results stratification parametrization coadjoint orbits @xmath0 invariant open subset @xmath128 @xmath43 @xmath129 smooth manifold quotient topology quotient mapping @xmath130 analytic @xmath128 dense @xmath43 @xmath131 @xmath132 @xmath128 @xmath0-invariant @xmath133 non - empty - open subset @xmath7 lebesgue measure @xmath42 @xmath7 supported @xmath133 @xmath134 @xmath0-invariant set @xmath135 @xmath136 disjoint @xmath137 @xmath38 supported @xmath138 @xmath140 analytic @xmath141 @xmath142 open @xmath127 @xmath143 @xmath141 analytic @xmath7 singular points @xmath7 @xmath42-measure zero @xmath141 submersion regular points @xmath7 pushforward lebesgue measure submersion continuous lebesgue @xmath38 lebesgue measure class @xmath144 plancherel measure @xmath35 @xmath129 lebesgue measure class @xmath144 open subset @xmath129 @xmath38 continuous @xmath39 @xmath18 @xmath145 non - trivialpoints @xmath18 rank @xmath93 @xmath122 less than @xmath146 lebesgue measure @xmath147 zero @xmath148 @xmath53 zariski - open subset @xmath7 measure @xmath38 supported @xmath149 @xmath38 singular @xmath35 admissibility @xmath150 representation @xmath0 @xmath151 @xmath152 define @xmath153 @xmath154 vector @xmath155 admissible @xmath156 isometry @xmath157 @xmath11 @xmath158 intertwines @xmath150 @xmath36 @xmath0 @xmath159 closed invariant subspace @xmath11 @xmath150 equivalent @xmath36 @xmath160 @xmath161 closed left invariant subspace @xmath11 @xmath162 orthogonal projection @xmath161 unique measurable field @xmath163 orthogonal projections @xmath164 defined @xmath165 @xmath166 holds for @xmath35-a @xmath167 set @xmath168 theorem 4.22 @xmath161 closed left invariant subspace @xmath11 @xmath0 nonunimodular @xmath161 admissible vector@xmath0 unimodular @xmath160 admissible vector if @xmath169 integrable over @xmath33 plancherel measure @xmath35 @xmath0 unimodular @xmath1 acts on points @xmath7 necessary for admissibility multiplicity function never integrable @xmath2 never admissible vectors make following | 0.491938 | 3,093 | 1,797 | , Saving $0.1 in GPT-4. |
67,557 | we present updated predictions for the total cross section of top - quark pair production at tevatron and lhc . for the lhc we also provide results at @xmath0 = 10 tev , in view of the anticipated run in 2008 and quote numbers for the production of new heavy - quark pairs with mass in the range 0.5 2 tev .
our two - loop results incorporate all logarithmically enhanced terms near threshold including coulomb corrections as well as the exact dependence on the renormalization and factorization scale through next - to - next - to - leading order in qcd . | research on top - quark physics at hadron colliders has received great interest in the past years in view of the steadily improving measurements at tevatron and the upcoming lhc ( see ref .
@xcite for a recent review ) . in this respect ,
the total cross section for top - quark pair production is a quantity of great importance for experimental analyses and even allows for measurements of the top - quark mass .
moreover , on the theory side , the total cross section has been subject to numerous studies the motivation being improved predictions beyond the long - known next - to - leading order ( nlo ) corrections in qcd @xcite .
recent work in this direction has aimed at completing the next - to - next - to - leading order ( nnlo ) qcd predictions @xcite , at resumming large sudakov logarithms to next - to - next - to - leading logarithmic accuracy @xcite and , at estimating bound state effects @xcite . also our knowledge on the parton distribution functions ( pdfs ) and the precision of the top - quark mass determination has continuously improved over the last years . in order to study the impact of the various improvements on tevatron and lhc predictions we build on the recent results of ref .
these approximate nnlo results for the total cross section are based on the complete logarithmic dependence on the heavy quark velocity @xmath1 near threshold @xmath2 .
moreover , they include the complete two - loop coulomb corrections as well as the exact dependence on the renormalization and factorization scale at nnlo @xcite .
recently , similar studies have appeared in refs .
@xcite . while ref .
@xcite largely follows our approach @xcite to describe the total top - quark pair cross section at nnlo , ref .
@xcite has limited itself to updating older predictions based on threshold resummation to next - to - leading logarithmic accuracy only .
thus , ref .
@xcite necessarily arrives at larger theoretical uncertainties .
the interesting study of ref .
@xcite on the other hand applied consistently predictions to nlo accuracy in qcd . in doing so ,
it has investigated correlations of rates for top - quark pair production with many other cross sections at lhc to quantify a potential sensitivity to the gluon luminosity .
the total hadronic cross section for top - quark pair production depends on the hadronic center - of - mass energy squared @xmath3 and the top - quark mass @xmath4 .
it is given by @xmath5 where @xmath6 are the pdfs of the proton .
the partonic cross section is given by @xmath7 and @xmath8 denotes the standard convolution ( see e.g. ref .
@xcite ) .
the generally adopted procedure to estimate the theoretical uncertainty for @xmath9 in eq . ( [ eq : totalcrs ] ) exploits the residual dependence on the renormalization and factorization scale , @xmath10 and @xmath11 , which are identified throughout this article ( i.e. @xmath12 ) .
the nlo qcd corrections for the parton cross section @xmath7 and the pdfs @xmath6 provide the first instance where a meaningful error can be determined in this way .
we define the range as @xmath13 where @xmath14 is computed from the variation of the cross section with respect to the parameters of the global fit ( see e.g. refs .
@xcite ) . in this contribution
we employ the approximate nnlo result @xcite to predict cross sections ( [ eq : totalcrs ] ) and the associated uncertainty ranges ( [ eq : range ] ) at tevatron and lhc .
let us therefore briefly comment on the anticipated accuracy .
our cross section takes along all logarithmically enhanced terms @xmath15 , @xmath16 as well as the complete coulomb corrections ( @xmath17 ) at two loops for the dominant parton channels @xmath18 and @xmath19 and adds them on top of the exact nlo predictions . in this way , our predictions rely on exact expressions in the region of phase space @xmath2 , where perturbative corrections receive the largest weight from the convolution with the parton luminosities , cf .
( [ eq : totalcrs ] ) .
the effect of new parton channels opening at nnlo ( @xmath20 and @xmath21 ) is expected to be small , cf .
the @xmath22 and @xmath23 channels at nlo .
the region of large energies @xmath24 on the other hand is inaccessible within our approach @xcite .
however , it is expected to give only small contributions in a full nnlo calculation in line with the observed small corrections for top - quark pair production together with an additional jet at nlo @xcite which are part of the full nnlo correction for top - quark pair production .
moreover , we have also included the exact @xmath10 and @xmath11 scale dependence at nnlo @xcite which can be constructed using renormalization group methods .
for the time being , we have chosen a common value @xmath25 for the scales , and we will address the independent variation of @xmath10 and @xmath11 in a future publication .
however , based on preliminary studies we do not expect large modifications here . in summary
, we have accounted for all numerically dominant contributions and are confident that this provides a very good approximation to the unknown full nnlo result as experience from other reactions , e.g. higgs - production in gluon fusion @xcite shows .
let us next present our results for tevatron and lhc .
in figs . [
fig : tev ] and [ fig : lhc ] we plot the uncertainty range ( [ eq : range ] ) comparing nlo and nnlo accuracy . the and qcd prediction for the @xmath26 total cross section at tevatron for @xmath27 tev .
the bands denote the total uncertainty from pdf and scale variations for the mrst06nnlo set @xcite according to eq .
( [ eq : range ] ) .
, width=264 ] same as fig .
[ fig : tev ] for lhc with @xmath28 tev .
, width=264 ] at tevatron ( fig .
[ fig : tev ] ) the central value at nnlo increases typically by 8% with respect to nlo .
the residual scale dependence of is 3% , which corresponds to a reduction by a factor of two compared to nlo . the overall uncertainty according to eq .
( [ eq : range ] ) is at about 8% for the cteq6.6 and 6% for the mrst06nnlo pdf set . at lhc ( fig .
[ fig : lhc ] ) our leads only to a small shift of a few percent in the central value and the band is about 6% for cteq6.6 and about 4% for mrst06nnlo , which exhibits again a drastic reduction of the scale uncertainty as compared to the prediction based on nlo qcd . for phenomenological applications , the results of eqs .
( [ eq : totalcrs ] ) , ( [ eq : range ] ) are best presented by means of simple formulae for the mass dependence of the total cross section . to that end
we make the ansatz following ref .
@xcite @xmath29 where @xmath30 .
the parameters @xmath31 are fitted to reproduce @xmath9 in the mass range @xmath32 with a typical accuracy of better than @xmath33 per mille . for tevatron and lhc the respective results for various pdf sets
are given in tabs .
[ tab : mass - fit - tev][tab : mass - fit - lhc ] .
note , that eq.([eq : mass - dep ] ) uses a polynomial of degree four and also determines the parameter @xmath34 for the central value @xmath35 from the fit .
@xmath36&e[pb ] @xmath37 + @xmath38&e[pb ] @xmath36 + @xmath38&e[pb ] @xmath36 + finally , we briefly quote some rates for the pair - production of new heavy quarks in the fundamental representation of the color @xmath39 gauge group at lhc with @xmath28 tev ( see also ref .
such particles with a mass @xmath40 appear in certain extensions of the standard model and we focus on a production model which is entirely dominated by qcd effects . thus , our cross section @xmath41 provides a meaningful and accurate prediction because its numerical values arises largely from the threshold region where the logarithms @xmath42 dominate . in tabs .
[ tab : new - hvq - mrst06 ] , [ tab : new - hvq - cteq65 ] we quote the corresponding numbers in the mass range @xmath43 ( see ref .
@xcite for results to nlo accuracy ) .
we observe that the scale dependence at nnlo accuracy is rather small , showing the expected good stability of the perturbative prediction .
the relative variation of @xmath9 with respect to the pdfs , though , is dominating by far .
note there is the usual factor of two between the pdf uncertainty quoted by mrst06nnlo @xcite and the cteq6.5 @xcite pdf sets due to the definition of the tolerance criteria in the respective fits .
the reason for the large observed pdf uncertainty is the gluon pdf being poorly constrained in the relevant region of large momentum fraction @xmath44 .
this is a fact well - known to influence many searches for high - mass particles in gluon fusion channels ( see e.g. ref .
@xcite for the correlation of top - quark pair production rate with the high mass higgs cross section ) . [ cols="^,<,<,^,<,<,^,<,<,^ " , ]
we have presented updated predictions for cross sections of top - quark pair production based on the ( approximate ) nnlo results of ref .
these represent the best present estimates for hadro - production of top - quark pairs , both at tevatron and lhc .
we have argued that the neglected contributions ( i.e. power suppressed terms away from threshold and new parton channels ) are numerically small .
we have found good convergence properties of the higher order corrections and greatly improved stability of the total cross section with respect to scale variations by our result . for applications , we have presented simple formulae ( [ eq : mass - dep ] ) with 0.1 per mille accuracy for the mass dependence of the total cross section in the range @xmath32 .
finally , we have applied our results to estimate the pair - production rates of new quarks heavier than the top - quark in the range up to @xmath45 .
is supported by the helmholtz gemeinschaft under contract vh - ng-105 and and p.u .
is a heisenberg fellow of deutsche forschungsgemeinschaft ( dfg ) .
this work is also partly supported by dfg in sfb / tr 9 .
w. bernreuther , j. phys .
g35 ( 2008 ) 083001 , arxiv:0805.1333 [ hep - ph ] p. nason , s. dawson and r.k .
ellis , nucl .
b303 ( 1988 ) 607 w. beenakker et al .
d40 ( 1989 ) 54 w. bernreuther et al .
b690 ( 2004 ) 81 , hep - ph/0403035 s. dittmaier , p. uwer and s. weinzierl , phys .
98 ( 2007 ) 262002 , hep - ph/0703120 m. czakon , a. mitov and s. moch , phys . lett .
b651 ( 2007 ) 147 , arxiv:0705.1975 [ hep - ph ] m. czakon , a. mitov and s. moch , nucl .
b798 ( 2008 ) 210 , arxiv:0707.4139 [ hep - ph ] j.g .
krner , z. merebashvili and m. rogal , ( 2008 ) , arxiv:0802.0106 [ hep - ph ] m. czakon , phys .
b664 ( 2008 ) 307 , arxiv:0803.1400 [ hep - ph ] s. moch and p. uwer , phys .
rev . d ( 2008 ) in press , arxiv:0804.1476 [ hep - ph ] k. hagiwara , y. sumino and h. yokoya , ( 2008 ) , arxiv:0804.1014 [ hep - ph ] n. kidonakis et al . , phys .
d64 ( 2001 ) 114001 , hep - ph/0105041 n. kidonakis and r. vogt , ( 2008 ) , arxiv:0805.3844 [ hep - ph ] m. cacciari et al . ,
( 2008 ) , arxiv:0804.2800 [ hep - ph ] p.m. nadolsky et al . , ( 2008 ) , arxiv:0802.0007 [ hep - ph ] a.d .
martin et al .
b652 ( 2007 ) 292 , arxiv:0706.0459 [ hep - ph ] w.k .
tung et al .
, jhep 02 ( 2007 ) 053 , hep - ph/0611254 s. moch and a. vogt , phys .
b631 ( 2005 ) 48 , hep - ph/0508265 | research on top - quark physics at hadron colliders interest improving measurements at tevatron upcoming lhc ref @xcite review total cross section for top - quark pair production for experimental analyses measurements top - quark mass total cross section subject to studies improved predictions beyond next - to - leading order corrections in qcd @xcite recent work completing qcd predictions resumming large sudakov logarithms to logarithmic accuracy estimating bound state effects knowledge on parton distribution functions precision of top - quark mass determination improved study impact of improvements on tevatron lhc predictions build on recent results ref approximate nnlo results for total cross section based on logarithmic dependence on heavy quark velocity threshold include two - loop coulomb corrections exact dependence on renormalization factorization scale at nnlo @xcite similar studies appeared in refs @xcite follows approach top - quark pair cross section updating older predictions based on threshold resummation to next - to - leading logarithmic accuracy arrives at larger theoretical uncertainties study of ref@xcite applied predictions nlo accuracy in qcd investigated correlations top - quark pair production with cross sections at lhc potential sensitivity to gluon luminosity total hadronic cross section for - production depends on hadronic center - of - mass energy @xmath3 top - quark mass @xmath4 given by @xmath5 @xmath6 are pdfs of proton partonic cross section by @xmath7 @xmath8 standard convolution procedure uncertainty for @xmath9 exploits residual dependence on renormalization factorization scale @xmath10 @xmath11 nlo qcd corrections for parton cross section @xmath7 pdfs @xmath6 provide first instance meaningful error range as @xmath13 @xmath14 computed from variation cross section global fit approximate nnlo result @xcite to predict cross sections uncertainty ranges at tevatron lhc on anticipated accuracy cross section takes logarithmically enhanced terms @xmath15 @xmath16 complete coulomb corrections ( @xmath17 ) at for dominant parton channels @xmath18 @xmath19 adds nlo predictionspredictions rely on expressions in region phase space @xmath2 perturbative corrections receive largest weight from convolution with parton luminosities effect of new parton channels opening at nnlo ( @xmath20 and @xmath21 ) expected small channels region of large energies @xmath24 inaccessible approach @xcite expected small contributions in full nnlo calculation small corrections for top - quark pair production additional jet at nlo @xcite included exact @xmath10 and @xmath11 scale dependence at nnlo @xcite constructed using renormalization group methods chosen common value @xmath25 for scales address independent variation of @xmath10 and @xmath11 in future publication expect large modifications accounted for all numerically dominant contributions provides good approximation to unknown full nnlo result results for tevatron and lhc in figs plot uncertainty range comparing nlo and nnlo accuracy qcd prediction for @xmath26 total cross section at tevatron for @xmath27 tev bands denote total uncertainty from pdf and scale variations for mrst06nnlo set @xciteeq : range ] width=264 same as fig . tev lhc with @xmath28 tev at tevatron central value at nnlo increases 8% nlo residual scale dependence 3% reduction factor two compared to nlo overall uncertainty eq . 8% for cteq6.6 6% mrst06nnlo pdf set lhc small shift few percent central value band 6% for cteq6.6 4% mrst06nnlo drastic reduction scale uncertainty prediction nlo qcd phenomenological applications results of eqs . totalcrs range presented simple formulae for mass dependence of total cross section ansatz ref @xcite @xmath29 @xmath30 parameters @xmath31 reproduce @xmath9 mass range @xmath32 accuracy better than @xmath33 per mille tevatron and lhc results pdf sets in tabs . mass - fit - mass - fit - lhc ] eq.( mass - dep ] ) uses polynomial of degree four determines parameter @xmath34 for central value @xmath35 from fit@xmath36&e[pb @xmath37 + @xmath38&e[pb quote rates pair - production new heavy quarks color @xmath39 gauge group lhc with @xmath28 tev particles with mass @xmath40 appear in extensions standard model focus production model dominated by qcd effects cross section @xmath41 provides accurate prediction numerical values from threshold region logarithms @xmath42 dominate tabs mrst06 cteq65 quote numbers in mass range @xmath43 results nlo accuracy scale dependence at nnlo accuracy small good stability perturbative prediction relative variation of @xmath9 pdfs dominating factor of two between pdf uncertainty by mrst06nnlo @xcite and cteq6.5 @xcite pdf sets due to tolerance criteria large pdf uncertainty gluon pdf poorly constrained in region large momentum fraction @xmath44 searches for high - mass particles in gluon fusion channels @xcite correlation top - quark pair production rate with high mass higgs cross sectionpresented updated predictions top - quark pair production nnlo results best estimates hadro - production tevatron lhc argued neglected contributions power threshold channels small found good convergence properties higher order corrections improved stability total cross section scale variations presented formulae 0.1 per mille accuracy mass dependence total cross section range @xmath32 applied results estimate pair - production rates new quarks heavier top - up to @xmath45 supported by helmholtz gemeinschaft - ng-105 heisenberg fellow deutsche forschungsgemeinschaft dfg partly supported by dfg sfb tr bernreuther . g35 ( 2008 ) 083001 arxiv:0805.1333 hep nason dawson ellis b303 ( 1988 ) 607 beenakker d40 ( 1989 ) 54 bernreuther b690 ( 2004 ) 81 hep - ph/0403035 dittmaier uwer weinzierl phys . 98 ( 2007 ) 262002 hep - ph/0703120 czakon mitov moch phys. lett b651 ( 2007 ) 147 arxiv:0705.1975 hep - ph m czakon mitov moch b798 2008 210 arxiv:0707.4139 krner merebashvili rogal 2008 arxiv:0802.0106 czakon b664 2008 307 arxiv:0803.1400 ph s. moch p. uwer phys 2008 arxiv:0804.1476 k hagiwara sumino h yokoya 2008 arxiv:0804.1014 - ph n kidonakis phys d64 2001 ) 114001 hep - ph/0105041 kidonakis r vogt 2008 arxiv:0805.3844 hep - ph m. cacciari arxiv:0804.2800 ph p nadolsky arxiv:0802.0007 ph a b652 2007 ) 292 arxiv:0706.0459 - ph tung 02 2007 053 hep - ph/0611254 s. moch a vogt phys b631 2005 ) 48 hep - ph/0508265 | 0.436678 | 3,029 | 1,564 | , Saving $0.1 in GPT-4. |
67,558 | abstract : : following wolszczan s landmark discovery of planets in orbit around pulsar psr b1257 + 12 in 1991 , over 300 planets in more than 200 planetary systems have been found . therefore , the meaning of wolszczan s discovery can not be overestimated . in this paper
we aim to convince the reader that the objects accompanying pulsar psr b1257 + 12 are more exotic than thought so far
. they might not be ordinary planets but dwarf strange quark stars , whereas the pulsar might be a quark star with standard mass , not a neutron star . if this was the case
, it would indicate that strange quark matter is the ground state of matter . | the millisecond pulsar psr b1257 + 12 with the estimated age of @xmath0 millions years , is located at a distance of 300 parsecs from the earth .
the tiny pulsar is being orbited by three bodies with masses @xmath1 , @xmath2 , and @xmath3 of the earth mass , typical of ordinary planets , and the respective radii of @xmath4 , @xmath5 and @xmath6 @xmath7 @xcite , @xcite .
for that reason , the bodies are considered to be planets .
it was clear from the start , that the planetary system around psr b1257 + 12 was unusual . before wolszczan
s discovery , the existence of planets around neutron stars had not been considered seriously .
if the planets were to be remnants of a planetary system existing around the progenitor star of psr b1257 + 12 , then it would be highly improbable for these planets to have survived the supernova explosion accompanying the neutron star s birth , since then compact bodies acquire velocities high enough to escape from the binding potential of the gravitational well of the surrounding matter .
these planets might also had been formed already after the supernova explosion and the neutron star s formation .
in particular , models of planetary system formation around pulsar accreting matter from its companion have been suggested @xcite . however , this scenario seems also improbable because of strong radiation in the pulsar s vicinity @xcite .
therefore , another hypothesis seems worth considering .
the objects in orbit around psr b1257 + 12 may be just miniature quark stars .
a quark star is a hypothetical object consisting of strange quark matter ( sqm ) .
sqm is a mixture of quasi - free quarks @xmath8 , @xmath9 , @xmath10 and electrons , all of which form relativistic fermi gas .
if sqm is the ground state of strong interactions , then one could expect strange stars to form in space . for sqm to be the ground state
, one needs the minimum of energy per baryon to be not greater than the energy per baryon for the strongest bound nucleus in nature the nucleus of iron-@xmath11 . hence , @xmath12 at zero pressure , where @xmath13 is the energy density and @xmath14 the baryon density .
the possibility of existence of such objects was put forward first in @xcite .
the easiest way to obtain a phenomenological equation of state for sqm , is to consider the following simplified model @xcite .
suppose that quarks form ultra - relativistic and electrically neutral quantum gas of non - interacting fermions in the @xmath15-equilibrium at temperature @xmath16 $ ] . in this case , the pressure @xmath17 and the total energy density @xmath13 are @xmath18 where @xmath19 is the baryon density and @xmath20 is the bag constant , @xmath21 $ ] .
these equations are valid when the fermi energy is much higher than the rest energy of the heaviest quark ( in our numerical calculations we included also the strange quark mass , but the full form of the resulting equations is not important here ) .
the maximal mass and the corresponding size of a quark star are close to those for a neutron star . with the assumed example values
@xmath22 $ ] and the strange quark mass @xmath23 $ ] , sqm would be the ground state for strong interactions .
then , the maximal mass of a quark star made of sqm would be @xmath24 with the corresponding areal radius of about @xmath25 $ ] times greater than its areal radius .
this distinction is usually not made in the literature and by a star s radius is always understood its areal radius . ] , whereas a neutron star with the uv14+tni equation of state @xcite attains @xmath26 at @xmath27 $ ] , see figure [ fig:1 ] . [ cols="^,^ " , ] what could have caused the mini - quark stars to find themselves in the pulsar vicinity ?
the first possibility is detachment of quark matter from the central pulsar during its formation phase , when the `` boiling '' strange matter core could eject some fragments , so that they start to orbit around the strange star .
however , this mechanism does not explain why the orbits of the three planets are nearly circular and almost coplanar ( at least , for two of them ) @xcite .
these features make the planetary disk origin of the pulsar planets very likely . the planetary disk stability and the planet formation present a lot of unknowns and require ad hoc assumptions @xcite .
here we propose another scenario .
the present rotation period of pulsar b1257 + 12 is of about @xmath28 ms and it decelerates with the rate @xmath29 @xcite . the pulsar has the feature of a pulsar spun up in a binary system it is a millisecond pulsar and , at the same time , it has comparably small deceleration rate and weak magnetic field .
it seems therefore probable the pulsar might have had a companion from which it had been accreting matter . in this case
it is difficult to estimate the age and the initial rotation velocity of a compact object .
however , during the accretion phase , the compact object , by assumption being a quark star , could have spun up to the velocities necessary to detach matter from its surface in the equatorial region .
if only the accretion were sufficient to accelerate the star to the keplerian frequency , which is about 1 khz for strange star @xcite , then several bubbles of quark matter could start to orbit the central star . in a natural way their trajectories
would be circular and coplanar .
being pieces of strange matter they would be completely resistant to the unfriendly pulsar environment heated by pulsar wind or accretion which may evaporate the normal planet .
in this paper we put forward the hypothesis that the planets orbiting the pulsar psr b1257 + 12 may be quark stars made of strange quark matter .
this is alternative to the currently accepted hypothesis that the system is composed of `` ordinary '' planets
. 99 a. wolszczan & d.a .
frail 1992 , nature vol 355 145 a. wolszczan 1994 , nature vol 264 538 m. banit , m.a .
ruderman , j. shaham , j.h .
applegate 1993 , apj 415 779 m. coleman miller & douglas p. hamilton 2001 , apj 550 863 e. witten , phys .
d * 30 * ( 1984 ) 272 .
chodos , a. et al . , 1974 , phys .
d , 9 , 3471 r. b. wiringa , v. fiks and a. fabrocini , phys . rev .
c * 38 * ( 1988 ) 1010 .
p. haensel , j. l. zdunik and f. douchin , a&a , * 385 * , 301 ( 2002 ) p.haensel acta phis .
pol b18 1987 x .- d .
li , i. bombaci , et.al .
* 83 * , 3776 ( 1999 )
, k. and yue , y. l. and xu , r. x. mnras.381l ... 1l r. n. manchester , g. b. hobbs , a. teoh and m. hobbs , astron . j. * 129 * ( 2005 ) 1993 http://www.atnf.csiro.au/research/pulsar/psrcat p. haensel , j. l. zdunik , m. bejger and j. m. lattimer , m. konacki and a. wolszczan , astrophys .
j. * 591 * ( 2003 ) l147 e. s. phinney , b. m. s. hansen _ planets around pulsars _ , asp conference series vol.36 p. 371 - 390 . | millisecond pulsar psr b1257 + 12 estimated age @xmath0 millions years distance 300 parsecs from earth orbited by three bodies masses @xmath1 @xmath2 @xmath3 earth mass radii @xmath4 @xmath5 @xmath6 @xmath7 bodies considered be planets planetary system around psr b1257 + 12 unusual before wolszczan s discovery planets around neutron stars not considered if planets remnants of planetary system around progenitor star psr b1257 + 12 improbable planets survived supernova explosion neutron star compact bodies velocities escape gravitational planets might formed after supernova explosion neutron star formation models planetary system formation around pulsar accreting matter companion suggested @xcite improbable strong radiation pulsar s vicinity another hypothesis objects around psr b1257 + 12 may be miniature quark stars quark star object strange quark matter ( sqm ) mixture quasi - free quarks electrons form relativistic fermi gas if sqm ground state strong interactions expect strange stars form spacesqm ground state needs minimum energy per baryon not greater than strongest nucleus iron-@xmath11 @xmath12 at zero pressure @xmath13 energy density @xmath14 baryon density possibility existence objects in @xcite phenomenological equation state for sqm simplified model @xcite quarks form ultra - relativistic electrically neutral quantum gas non - interacting fermions @xmath15-equilibrium at temperature @xmath16 $ pressure @xmath17 total energy density @xmath13 @xmath18 @xmath19 baryon density @xmath20 bag constant equations valid when fermi energy higher than rest energy heaviest quark included strange quark mass maximal mass size quark star close to neutron star assumed example values @xmath22 $ strange quark mass @xmath23 $ sqm ground state for strong interactions maximal mass quark star sqm @xmath24 areal radius @xmath25 $ times greater than areal radius distinction star s radius understood areal radius neutron star uv14+tni equation state attains @xmath26 at @xmath27 $ ] see figurecols="^ what caused mini - quark stars in pulsar vicinity ? first possibility is detachment of quark matter from central pulsar during formation phase strange matter core could eject fragments around strange star mechanism explain orbits of three planets nearly circular coplanar for two features make planetary disk origin of pulsar planets likely planetary disk stability planet formation present unknowns require ad hoc assumptions propose another scenario rotation period of pulsar b1257 + 12 about @xmath28 ms decelerates @xmath29 @xcite pulsar millisecond pulsar small deceleration rate weak magnetic field probable pulsar might have had companion accreting matter difficult to estimate age initial rotation velocity of compact object during accretion phase compact object quark star could have spun up to velocities to detach matter from surface if accretion star to keplerian frequency about 1 khz bubbles of quark matter could orbit central star natural trajectories be circular coplanar resistant to unfriendly pulsar environment pulsar wind accretion hypothesis planets orbiting pulsar psr b1257 + 12 may be quark stars of strange quark matter alternative hypothesis system ordinary planets wolszczan frail 1992 nature vol 355 145 wolszczan 1994 nature vol 264 538 banit ruderman shaham applegate 1993 apj 415 779 coleman miller douglas hamilton 2001 apj 550 863 witten . 30 1984 ) 272 chodos al 1974 . 9 3471 wiringa fiks fabrocini . rev 38 * 1988 ) 1010 haensel zdunik f douchin 385 301 ( 2002 ).haensel acta phis . b18 1987 bombaci 83 * 3776 1999 ) yue xu mnras.381l manchester hobbs teoh hobbs astron . j 129 * 2005 ) 1993.csiro.au/research/pulsar/psrcat haensel zdunik bejger lattimer konacki wolszczan astrophys . j 591 * 2003 ) l147 phinney hansen planets around pulsars asp conference series vol.36 p. 371 - 390 | 0.270086 | 1,821 | 927 | , Saving $0.1 in GPT-4. |
67,563 | in this paper an approach is outlined . with this approach
some explicit algorithms can be applied to solve the initial value problem of @xmath0dimensional damped oscillators .
this approach is based upon following structure : for any non - conservative classical mechanical system and arbitrary initial conditions , there exists a conservative system ; both systems share one and only one common phase curve ; and , the value of the hamiltonian of the conservative system is , up to an additive constant , equal to the total energy of the non - conservative system on the aforementioned phase curve , the constant depending on the initial conditions .
a key way applying explicit symplectic algorithms to damping oscillators is that by the newton - laplace principle the nonconservative force can be reasonably assumed to be equal to a function of a component of generalized coordinates @xmath1 along a phase curve , such that the damping force can be represented as a function analogous to an elastic restoring force numerically in advance .
two numerical examples are given to demonstrate the good characteristics of the algorithms .
hamiltonian , dissipation , non - conservative system , damping , explicit symplectic algorithm | feng@xcite , marsden@xcite , neri@xcite and yoshida@xcitehad developed a series of symplectic algorithms for hamiltonian systems .
these algorithms possess some advantages .
but it is difficult to apply these algorithms to damping dynamical systems , because it has been stated in most classical textbooks that the hamiltonian formalism focuses on solving conservative problems .
damping phenomena is very important in the modeling of dynamical systems , and can not be avoided .
our aim is to apply some explicit canonical algorithms to nonlinear damping dynamical systems , which is generated generally by fe - method .
these canonical algorithms reported in this paper can be readily utilized for computing large - scale nonlinear damped dynamical systems .
betch@xcite@xcite@xcite attempted to apply directly some implicit algorithms to damping systems .
the implicit symplectic algorithms utilized by betch@xcite possess a few good characteristic , e.g. energy - conservation , momentum - consistence , etc ... in terms of energy - conservation , implicit symplectic algorithms might be better than explicit symplectic ones .
but explicit symplectic schemes might be more suitable for nonlinear problems .
if one needs to apply symplectic algorithms to a dissipative system , one must convert the dissipative system into a hamiltonian system or find some relationship between the dissipative system and a conservative one . in the literature@xcite
, we have stated a proposition describing a relation among a damping dynamical system and conservative ones : for any non - conservative classical mechanical system and arbitrary initial condition , there exists a conservative system ; both systems sharing one and only one common phase curve ; and the value of the hamiltonian of the conservative system is equal to the sum of the total energy of the non - conservative system on the aforementioned phase curve and a constant depending on the initial condition .
[ pro:1 ] in other words , a dissipative ordinary equation and a conservative equation may possess a common particular solution . in the next section ,
an analytical examples are given to explain this proposition .
readers can find the detailed proof of proposition [ pro:1 ] in the reference@xcite in the literature @xcite a basic explicit canonical integrator is proposed .
based on this basic scheme , neri@xcite constructed 4-order explicit canonical integrator , and then yoshida @xcite proposed a general method to construct higher order explicit symplectic integrator . utilizing the proposition [ pro:1 ] , we apply this class of explicit canonical integrators to damping dynamical systems
. this point will be in detail stated in sec .
[ sec:3 ] .
consider a special one - dimensional simple mechanical system : @xmath2 where @xmath3 is a constant .
the exact solution of the equation above is @xmath4 where @xmath5 are constants .
differentiation gives the velocity : @xmath6 from the initial condition @xmath7 , we find @xmath8 .
inverting eq .
( [ eq : sol_1d ] ) yields @xmath9 and by substituting into eq .
( [ eq : sol_1dv ] ) , such we have @xmath10 the dissipative force @xmath11 in the dissipative system ( [ eq : simp_1d ] ) is @xmath12 substituting eq .
( [ eq : dx ] ) into eq .
( [ eq : f ] ) , the conservative force @xmath13 is expressed as @xmath14 clearly , the conservative force @xmath13 depends on the initial condition of the dissipative system ( [ eq : simp_1d ] ) , in other words , an initial condition determines a conservative force .
consequently , a new conservative system yields @xmath15 the stiffness coefficient in this equation must be negative .
one can readily verify that the particular solution ( [ eq : sol_1d ] ) of the dissipative system can satisfy the conservative one ( [ eq:1d_eq_consys ] ) .
this point agrees with proposition ( [ pro:1 ] ) .
the potential of the conservative system ( [ eq:1d_eq_consys ] ) is @xmath16{\mathrm{d}}x = -\frac{c^2}{2}x^2+c^2a_1x\ ] ] therefore the hamiltonian is @xmath17 where @xmath18 .
furthermore , proposition ( [ pro:1 ] ) can be depicted by fig .
[ fig : relation ] . the phase flow of conservative system ( [ eq : sol_1d ] ) transforms the red area in phase space to the purple area ; the phase flow of conservative system ( [ eq:1d_eq_consys ] ) transforms the red area to the green area . the blue curve in fig .
[ fig : relation ] illustrates the common phase curve .
if one draws more common phase curves and phase flows , the picture will like a flower , the phase flow of the nonconservative system likes a pistil and phase flows conservative systems like petals .
[ fig : relation ]
in the paper@xcite@xcite@xcite a symplectic algorithm based second kind generation function was stated @xmath19 where the superscript @xmath20 denotes the @xmath20-th time node , @xmath21 denotes coordinates and @xmath22 denotes canonical momenta , and @xmath23 denotes hamiltonian quantity , @xmath24 . if the hamiltonian is seperable , i.e. @xmath25 , then the symplectic scheme([eq : base_scheme ] ) above becomes an explicit symplectic scheme : @xmath26 for some nonlinear vibration mechanical system , @xmath27 .
let us consider an @xmath0dimensional nonlinear oscillator : @xmath28 where @xmath29 denotes a non - linear damping coefficient matrix which depends on @xmath21 , and @xmath30 denotes a non - linear stiffness matrix which depends on @xmath21 and consists of two parts @xmath31(@xmath32 is a diagonal matrix ) .
in accordance with proposition [ pro:1 ] , a conservative mechanical system was found associated with the dissipative system ( [ eq : osc ] ) in addition to its initial conditions .
subject to these initial conditions , the dissipative system ( [ eq : osc ] ) possesses a common phase curve @xmath33 with the conservative system . as in eq .
( [ eq : mf ] ) , we can consider that the components of the damping force @xmath34 determine the components of a conservative force on the phase curve @xmath33 @xmath35 for convenience , this conservative force is assumed to be an elastic restoring force : @xmath36 in a similar manner , the components of the non - conservative force @xmath37 are equal to the components of a conservative force on the phase curve @xmath33 @xmath38 the conservative force can likewise be assumed to an elastic restoring force : @xmath39 by an appropriate transformation , an equivalent stiffness matrix @xmath40 that is diagonal in form can be obtained @xmath41 consequently , an @xmath42-dimensional conservative system is obtained @xmath43 which shares the common phase curve @xmath33 with the @xmath42-dimensional damping system described by ( [ eq : osc ] ) . in this paper , the conservative system
is called the substitute conservative system .
the lagrangian of eqs.([eq : ex2 - 6 ] ) is @xmath44 with the hamiltonian @xmath45 where @xmath46 is the zero vector , and @xmath47 . here
@xmath48 in eq . (
[ eq : ex2 - 7 ] ) is the mechanical energy of the conservative system ( [ eq : ex2 - 6 ] ) , because @xmath49 is a potential function such that @xmath48 is independent of the path taken in phase space .
subject to a certain initial condition , one need merely to solve the conservative system([eq : ex2 - 6 ] ) .
but one must in advance obtain the numerical approximation of the matrix @xmath40 for a time step , such that one can utilize the algorithm ( [ eq : basee_exp_scheme ] ) to integrate the conservative system ( [ eq : ex2 - 6 ] ) for a time step .
one can repeat this process above up to the end . in this way one obtains the numerical particular solution of the conservative system ( [ eq : ex2 - 6 ] ) , which is exactly the numerical particular solution of the damping one .
the he numerical approximation of the matrix @xmath32 can be assumed as : @xmath50 \label{eq : eqv_stiffness}\\ { \tilde{k}_j}({q_j}^i ) = { c_{jl}}\dot q_l^i/{q_j}^i+\hat{k}_{jl}q_l^i/{q_j}^i \nonumber \end{aligned}\ ] ] hence the explicit canonical scheme ( [ eq : basee_exp_scheme ] ) can be modified into @xmath51{q^i } ) \\ { q_j}^{i + 1 } = { q^i } + \tau { p_j}^{i + 1 } \\
\end{array } \label{eq:1st_exp_schemeeq}\ ] ] the scheme above is a one order scheme .
furthermore one can construct higher order explicit canonical schemes utilizing the method reported in the literatures@xcite@xcite . now consider a map from @xmath52 to @xmath53 : @xmath54 where @xmath55,\bm{z}'=\left[\begin{array}{l}\bm{p}',\\ \bm{q}'\end{array}\right],\\ { { \mathsfsl{e } } } = \left [ { \begin{array}{*{20}{c } } 0 & 0 \\ 1 & 0 \\ \end{array } } \right ] { { \mathsfsl{f } } } = \left [ { \begin{array}{*{20}{c } } 0 & { - ( { { \mathsfsl{k } } } + { { \mathsfsl{\tilde{k } } } } ) } \\ 0 & 0 \\
\end{array } } \right].\end{aligned}\ ] ] in fact eq.([eq : exponent_form ] ) is the succession of the following mappings , @xmath56 in reality the difference between the equations above and eq.([eq:1st_exp_schemeeq ] ) is that the coefficients @xmath57 before the time step @xmath58 . in the literature @xcite a generalized method to determine @xmath57
were given .
therefore , the higher order explicit canonical scheme can be represented as : @xmath59 { \tilde{k } _
\alpha } ( q_\alpha ^j ) = \sum\limits_{l = 1}^n { { c_{\alpha l}}\dot q_l^j / q_\alpha ^j}+\hat{k}_{\alpha l}q_l^i / q_\alpha^i \\ { { \mathsfsl{e } } } = \left [ { \begin{array}{*{20}{c } } 0 & 0 \\ 1 & 0 \\
\end{array } } \right]\;\;\;\;{{\mathsfsl{f } } } = \left [ { \begin{array}{*{20}{c } } 0 & { - ( { { \mathsfsl{k } } } + { { \mathsfsl{\tilde{k } } } } ) } \\ 0 & 0 \\ \end{array } } \right ] \\ { { \bm{z}}^{j + 1 } } = ( \prod\limits_{i = 1}^h { { e^{{s_i}\tau{{\mathsfsl{f}}}}}{e^{{r_i}\tau{{\mathsfsl{e } } } } } } ) { { \bm{z}}^j } \\
\end{array}\ \label{eq : h_exp_scheme}\ ] ]
two examples will be given to shown this numerical method[eq : h_exp_scheme ] .
to begin , we consider a van der pol s oscillator @xmath60 where @xmath61 .
the initial conditions are given by @xmath62 .
we employee the @xmath63order explicit symplectic method ( [ eq : h_exp_scheme ] ) with coefficients @xmath64{2}+1/\sqrt[3]{2})]/6,\ \ s_2=s_3=[1-(\sqrt[3]{2}+1/\sqrt[3]{2})]/6 , \\ \ \
r_1=r_3=[2+(\sqrt[3]{2}+1/\sqrt[3]{2})]/3,\ \ r_2=-[2+(\sqrt[3]{2}+1/\sqrt[3]{2})]/3,\
\ r_4=0,\end{aligned}\ ] ] and classical explicit @xmath63order runge - kutta method to compute the resonance of the van der pol s oscillator ( [ eq : ex_duffing ] ) respectively , then employ a same method to integrate the results to the total energy , which is the sum of the mechanical energy and the work done by damping forces in the system ( [ eq : vdp ] ) .
the both methods are run with the same step size @xmath65 . the resonance is shown in fig .
[ fig : x_vdp ] , and the total energy is shown in fig .
[ fig : energy_vdp ] .
it is aparent from fig .
[ fig : energy_vdp ] that the explicit symplectic method ( [ eq : h_exp_scheme ] ) has qualitatively different behavior to the runge - kutta method .
the energy divergence between the explicit symplectic method and the exact solution is smaller than that between runge - kutta method and the exact solution .
the energy divergence between the explicit symplectic method and runge - kutta method increases with the time evolution . due to the increasement of the energy , the phase difference between both the results in fig .
[ fig : x_vdp ] increases also with the time evolution . in the second example
, we consider a @xmath66dimensional damped nonlinear duffing oscillator @xmath67 with the initial conditions @xmath68 .
the program of the both methods with the step size @xmath65 are carried out to simulate eq .
( [ eq : ex_duffing ] ) . the resonance is shown in fig .
[ fig : x_vdp ] , the numerical solution of the total energy is shown in fig.[fig : te ] .
there is only tiny difference between resonance results of the two methods , correspondingly , the difference among the total energy obtained by the numerical methods and anlytical methods is very tiny . as numerical examples in the other literatures@xcite , that explicit runge - kutta method must cause numerical pseudo dissipation which might be positive or negative .
the difference between our numerical examples and the examples in the literature@xcite is the total energy in our examples and the mechanical energy in their examples .
we have introduced a class of explicit symplectic algorithms to dissipative mechanical systems successfully , by changing these algorithms into the scheme.([eq : h_exp_scheme ] ) . because the algorithms ( [ eq : h_exp_scheme ] ) are explicit and possess good energy preserving characteristics , the explicit symplectic algorithms ( [ eq : h_exp_scheme ] ) is quite suitable for long term integration of arbitrary dimensional nonlinear dissipative mechanical systems .
13 natexlab#1#1[2]#2 , , in : , , , , pp . .
, , , , ( ) . , , , , ( ) .
, , , , ( ) . . , , , department of physics , university of maryland , .
, , , ( ) . , , , , ( ) . . , , ( ) . . , , , ( ) . , , , in : , , , , pp . . , , , , , ( ) | feng marsden neri yoshida developed symplectic algorithms for hamiltonian systems algorithms advantages difficult to apply to damping dynamical systems hamiltonian formalism focuses on solving conservative problems damping important in modeling can not be avoided aim is to apply explicit canonical algorithms to nonlinear damping dynamical systems generated by fe - method algorithms can for large - scale nonlinear damped dynamical systems betch attempted implicit algorithms to damping systems implicit algorithms possess good energy - conservation momentum - consistence implicit algorithms might better than explicit symplectic explicit symplectic more suitable for nonlinear problems apply symplectic algorithms to dissipative system must convert into hamiltonian system or find relationship between conservative proposition relation among damping dynamical system conservative ones : for non - conservative classical mechanical system exists conservative system common phase curve value of hamiltonian of conservative system equal to total energy of non - conservative system constant depending on initial condition dissipative equation and conservative equation may possess common solutionnext section analytical examples explain proposition readers find detailed proof proposition pro:1 ] reference@xcite literature basic explicit canonical integrator proposed neri@xcite constructed 4-order explicit canonical integrator yoshida proposed method construct higher order explicit symplectic integrator utilizing proposition pro:1 ] apply explicit canonical integrators damping dynamical systems stated sec sec:3 ] consider special one - mechanical system @xmath2 @xmath3 constant solution equation @xmath4 @xmath5 constants differentiation gives velocity @xmath6 initial condition @xmath7 find @xmath8 inverting eq. sol_1d yields @xmath9 substituting into eq sol_1dv @xmath10 dissipative force @xmath11 dissipative system @xmath12 substituting eq into eq conservative force @xmath13 expressed as @xmath14 depends on initial condition dissipative system simp_1d initial condition determines conservative force new conservative system yields @xmath15 stiffness coefficient negative solution eq : sol_1d ] dissipative system satisfy conservative one eq:1d_eq_consys agrees with proposition pro:1 ] )potential conservative system eq:1d_eq_consys is @xmath16{{d}}x = -\frac{c^2}{2}x^2+c^2a_1x\ hamiltonian is @xmath17 @xmath18 proposition pro:1 depicted by fig . phase flow conservative system transforms red area to purple area transforms red green area blue curve illustrates common phase curve common phase curves flows picture like flower phase flow nonconservative system pistil conservative systems like petals symplectic algorithm second kind generation function stated @xmath19 superscript @xmath20 denotes-th time node @xmath21 coordinates @xmath22 canonical momenta @xmath23 hamiltonian quantity @xmath24 if hamiltonian seperable @xmath25 symplectic scheme base_scheme becomes explicit symplectic scheme : @xmath26 nonlinear vibration mechanical system, @xmath27 .consider @xmath0dimensional nonlinear oscillator @xmath28 @xmath29 non - linear damping coefficient matrix depends on @xmath21 @xmath30 non - linear stiffness matrix depends on @xmath21 two parts @xmath31(@xmath32 diagonal matrix proposition pro:1 conservative mechanical system associated with dissipative system osc initial conditions dissipative system common phase curve @xmath33 with conservative system components damping force @xmath34 determine components conservative force phase curve @xmath33 @xmath35 conservative force elastic restoring force components non - conservative force @xmath37 equal to conservative force phase curve @xmath33 @xmath38 conservative force elastic restoring force equivalent stiffness matrix @xmath40 diagonal obtained @xmath41 @xmath42-dimensional conservative system obtained @xmath43 shares common phase curve @xmath33 with @xmath42-dimensional damping system conservative system substitute conservative system lagrangian of eqs - 6 is @xmath44 hamiltonian @xmath45 @xmath46 zero vector @xmath47 @xmath48 eq .eq : ex2 - 7 ] ) mechanical energy conservative system ex2 - 6 ] ) @xmath49 potential function @xmath48 independent of path phase space solve conservative system ex2 - 6 ] obtain numerical approximation matrix @xmath40 for time step algorithm basee_exp_scheme integrate conservative system ex2 - 6 ] repeat process end obtains numerical solution conservative system ex2 - 6 ] solution damping one numerical approximation matrix @xmath32 assumed as @xmath50 {eq : eqv_stiffness{k}_j}({q_j}^i ) = { c_{jl}} q_l^i/{q_j}^i+{k}_{jl}q_l^i/{q_j}^i canonical scheme basee_exp_scheme modified into @xmath51{q^i } { q_j}^{i + 1 } = { q^i } + \tau { p_j}^{i + 1 }{array{eq:1st_exp_schemeeq scheme one order scheme construct higher order schemes method literatures@xciteconsider map @xmath52 to @xmath53 @xmath54 @xmath55{z}=\left[\begin{array\bm{p}{q\end{array}\right] \mathsfsl{e } } = \left \begin{array{c 0 & 0 1 0 \end{array \right\mathsfsl{f } } = \left\begin{array{c 0 - \mathsfsl{k +\mathsfsl{\tilde{k 0 & 0 \end{array } \right].\end{aligned} eq.( exponent_form succession mappings @xmath56 difference between equations eq. coefficients @xmath57 before time step @xmath58 literature generalized method determine @xmath57higher order canonical scheme represented as @xmath59 \tilde{k \alpha q_\alpha ^j ) = \sum\limits_{l = 1}^n\alpha l}} q_l^j / q_\alpha ^j{k{\alpha l}q_l^i / q_\alpha^i \mathsfsl{e } \left\begin{array}{c 0 & 0 1 & 0 \end{array \right\mathsfsl{f } \left\begin{array{c 0 - \mathsfsl{k\mathsfsl\tilde{k 0 & 0 \end{array \right \bm{z}}^{j + 1 } \prod\limits_{i = 1}^h\mathsfsl{f_i\mathsfsl{e \bm{z}}^j \end{array}{eq h_exp_scheme} two examples numerical method consider van der pol s oscillator @xmath60 @xmath61 initial conditions given by @xmath62employee @xmath63order explicit symplectic method h_exp_scheme coefficients @xmath64{2}+1/\sqrt[3]{2})]/6, s_2=s_3=[1-(\sqrt[3]{2}+1/\sqrt[3]{2})]/6 r_1=r_3=[2+(\sqrt[3]{2}+1/\sqrt[3]{2})]/3, r_2=-[2+(\sqrt[3]{2}+1/\sqrt[3]{2})]/3, r_4=0,\end{aligned}\ ] ] classical explicit @xmath63order runge - kutta method compute resonance van der pol s oscillator employ same method integrate results total energy sum mechanical energy damping forces system both methods run same step size @xmath65 resonance shown in fig . x_vdp ] total energy fig . [ fig energy_vdp ] explicit symplectic method_exp_scheme different behavior runge - kutta method energy divergence symplectic method exact solution smaller runge - kutta method divergence increases time evolution phase difference results increases timesecond example consider @xmath66dimensional damped nonlinear duffing oscillator @xmath67 initial conditions @xmath68 program both methods step size @xmath65 simulate eq . eq : ex_duffing ] resonance shown in fig . fig x_vdp ] numerical solution total energy shown in fig.[fig : te ] tiny difference between resonance results two methods difference total energy numerical methods anlytical methods tiny numerical examples@xcite explicit runge - kutta method cause numerical pseudo dissipation positive or negative difference between numerical examples@xcite is total energy mechanical energy introduced class explicit symplectic algorithms to dissipative mechanical systems changing algorithms into scheme.([eq : h_exp_scheme ] ) algorithms h_exp_scheme explicit possess good energy preserving characteristics suitable for long term integration of arbitrary dimensional nonlinear dissipative mechanical systems. 13 natexlab#1#1[2]#2 , in : , pp department of physics university of maryland , , , in : , pp . ) | 0.609934 | 3,487 | 1,859 | , Saving $0.1 in GPT-4. |
67,571 | we experimentally demonstrate a broadband and an ultra - broadband spectral bandwidth polarization rotators .
both polarization rotators have modular design , that is , they are comprised of an array of half - wave plates rotated to a given angle .
we show that the broadband and ultra - broadband performance of the polarization rotators is due to the adiabatic nature of the light polarization evolution . in this paper
we experimentally investigate the performance of broadband and ultra - broadband polarization rotators comprising of ten multi - order half - wave plates or ten commercial achromatic half - wave plates , respectively .
the half - wave plates in the arrays are rotated gradually with respect to each other starting from an initial alignment between the fast polarization axis of the first one and the incoming linearly polarized light , to the desired polarization rotation angle .
_ keywords _ : broadband polarization rotator , adiabatic evolution , piecewise adiabatic passage , modular design . | the analogy between the poincar space , which is used for the description of light polarization , and the bloch space , which is used for the description of quantum - state dynamics of two - level systems , is well - known @xcite .
this has allowed a transfer of concepts from the field of quantum optics @xcite and nuclear magnetic resonance to the field of polarization optics @xcite .
for instance , the concept of composite pulses was successfully used to create broadband half- and quarter - wave plates @xcite as well as tunable polarization rotators @xcite .
another example is the application of adiabatic techniques developed for quantum systems in polarization optics @xcite to design various robust and achromatic polarization retarders and optical isolators . in this paper
we report an experimental demonstration of a broadband and an ultra - broadband polarization rotators , designed by an analogy to piecewise adiabatic passage technique @xcite .
the broadband and ultra - broadband polarization rotators are assembled according to the recent theoretical work of shore et al .
@xcite as a sequence of ten multi - order half - wave plates and ten commercial achromatic half - wave plates , respectively . for both designs , the half - wave plates are rotated gradually from an alignment between the fast - polarization axis with the initial linearly polarized light to an alignment of the fast - polarization axis with the desired final linear polarization orientation .
we first consider the propagation of a plane electromagnetic wave along the z - axis through a lossless medium . in this case polarization evolution
is given with the torque equation for the stokes vector kuratsuji1998,kuratsuji2007,sala , seto : @xmath0where @xmath1 is the distance along the propagation direction , @xmath2 $ ] is the stokes vector , and the driving torque @xmath3 $ ] is the birefringence vector of the medium . when the medium does not possess optical activity and is uniaxial with the slow and fast axes in the @xmath4 plane , then the components of the birefringence vector @xmath5 are given explicitly as @xmath6here @xmath7 and @xmath8 are the refractive indices along the fast and slow axes , @xmath9 is the light wavelength , @xmath10 is the rotary power , and @xmath11 is the angle of rotation between the fast ( slow ) axis and the @xmath12 ( @xmath13 ) cartesian axis .
now if the stokes vector @xmath14 is initially parallel to the birefringence vector @xmath15 , such that @xmath16 , and we change slowly ( adiabatically ) the angle @xmath11 , then the stokes vector will also evolve adiabatically with it .
for instance , if linearly polarized light is initially in the horizontal plane , i.e. , @xmath17 , and @xmath18 , which corresponds to aligning the @xmath12 and @xmath13 axes of the cartesian coordinate system to coincide with the fast and slow optical axes , then the birefringence vector @xmath15 is initially aligned with the stokes vector . following an adiabatic evolution ,
the final angle of linear polarization will be @xmath19 .
this adiabatic approach has the advantages of being robust with respect to large intervals of values for @xmath20 and is therefore , broadband . in this work
we propose an alternative scheme for the adiabatic evolution for the stokes vector by means of discrete modular design of the polarization rotators @xcite .
that is , we consider that the rotation of the angle @xmath11 takes place in a sequence of discrete steps , rather than continuously .
experimentally , we realize this by a sequence of @xmath21 birefringent crystals , each rotated at an angle @xmath22 with respect to the chosen cartesian coordinate system ( illustrated schematically in figure [ fig : fig1 ] ) .
for such a scheme the overall adiabatic evolution is achieved if the individual rotation angles change gradually and if there are sufficient number of steps ( @xmath23 ) .
it is necessary that each of the birefringent crystals drives not more than a " rabi cycle , in other words , that each module is not more than a half - wave plate .
we performed several experiments to verify our theoretical predictions for the performance of both broadband and ultra - broadband adiabatic polarization rotators .
the linear polarization of white light beam was rotated at 45@xmath24 , 60@xmath24 , 75@xmath25 and 90@xmath24 degrees passing through a set of ten half - wave plates . , irises @xmath26 , lens @xmath27 , lens @xmath28 and polarizer @xmath29 form a collimated beam of white polarized light . analyzer @xmath30 and lens @xmath31 focus the beam of output light onto the entrance @xmath32 of an optical fibre connected to a spectrometer .
the stack of ten wave plates rotates the linear polarization of the light in the broadband and ultra - broadband spectra . ]
a 10 w halogen - bellaphot ( osram ) lamp with dc power supply was applied as a light source with continuous spectrum in the range from 450 nm to 1100 nm ( figure [ fig : fig2 ] ) .
a sequence of an iris , two lenses and a polarizer were used for production of collimated light beam .
the iris @xmath26 was placed in the focus of plano - convex lens @xmath27 ( @xmath33=35 mm ) imitating a point light source .
the received beam was additionally collimated by a second lens @xmath28 with @xmath33=150 mm .
polarizer @xmath29 ( glan - tayler , 210 - 1100 nm , borrowed from a lambda-950 spectrometer ) linearly polarized the white light in the horizontal plane .
we used two types of half - wave plates for our experiments . in the first experiment a set of ordinary multi - order quarter - wave plates ( wpmq10m-780 , thorlabs ) which perform as half - wave plates at @xmath34=763 nm were used while in the second experiment achromatic half - wave plates ( wrm053-mica , 700 - 1100 nm , aperture 20 mm ) were exercised . each wave plate ( aperture of @xmath35 ) was assembled onto a separate rsp1 rotation mount which realizes a @xmath36 rotation .
the scale marked at @xmath37 increments allows for precise , repeatable positioning and fine angular adjustment .
we analyzed the characteristics of the transmitted light through the ten wave plates by a polarizer @xmath30 ( the same type as @xmath29 ) and a grating monochromator ( model avaspec-3648 fiber optic spectrometer with controlling software avasoft 7.5 ) .
we used a plano - convex lens @xmath31 ( @xmath33=20 mm ) and a two - axis micro - positioner to focus the light beam onto the optical fibre entrance @xmath32 of the monochromator .
( a ) and @xmath38 ( b ) as follow : red dashed line @xmath39 , green short dash line @xmath40 , blue dashed dotted line @xmath41 and black solid line @xmath42 . ]
our aim was to demonstrate broad and ultra - broad spectral bandwidth linear polarization rotators composed by two sets of ten identical wave plates .
each set was assembled by multi - order wave plates or by achromatic wave plates , respectively , as described above in subsection [ optical setup ] .
the wave plates were slightly tilted to reduce unwanted reflections @xcite . for the analysis of the experimental data we used a single beam spectrometer . to account for noise and losses due to light transmission , reflection and absorption in different media , we measured the light and dark spectra for all experiments
. furthermore , we used the dark spectrum , which is measured with the light path completely blocked , for correction of hardware offsets .
the reference spectrum is usually taken with the light source on and a blank sample instead of the sample of interest measured . in our case , however , we measured the transmission spectrum of the already assembled polarization rotator , but the axes of the polarizer @xmath29 , the analyzer @xmath30 and the fast axis of each single wave plates were all set parallel .
we used the measured light spectrum as a reference for the subsequent measurements . a linearly polarized light beam passed by the sequences of @xmath43 wave plates .
each of them was adjusted at the estimated angle by the formula : @xmath44 , where @xmath45 is the sequential number of the wave plate and @xmath21 in the present experiment .
the rotation effect was investigated for each of the arbitrary chosen angles @xmath46 .
the data were taken with analyzer @xmath30 revolved at angle @xmath47 and @xmath38 .
( a ) and @xmath38 ( b ) as follow : red dashed line @xmath39 , green short dash line @xmath40 , blue dashed dotted line @xmath41 and black solid line @xmath42 .
the black dashed dot dot curve shows the spectrum of single achromatic wave plate . ] we experimentally tested and proved development of a broad spectral bandwidth polarization rotator settled by ten multi - order half - wave plates .
the transmittance spectra were taken for four arbitrary chosen angles @xmath47 .
the obtained results are presented on the figure [ fig : fig3 ] a. the experimental curves show flat maximum in wide spectral diapason what is kept at different angles @xmath47 .
the experimental curves at angles @xmath38 of the analyzer @xmath30 were also presented on figure [ fig : fig3 ] b. the blow out the ray at @xmath38 proves that the polarization of the light is orthogonal to the one at angle @xmath47 .
even broader effect was reached by using ten achromatic half - wave plates .
we created ulta - broad bandwidth linear polarization rotator following the above explained algorithm .
we used the same rotation angles @xmath47 .
each of the wave plate in the sequences was the same type and with axis rotated at the angle @xmath48 .
the experimental results are shown on the figure [ fig : fig4 ] a. again for proving that the developed set of wave plates was working as a linear polarization rotator , the transmittance spectra at angles @xmath38 are presented ( see figure [ fig : fig4 ] b ) .
the shown experimental curves not exceed 1000 nm because of the lack of the signal after this spectral region .
in this paper we demonstrated experimental application of piecewise adiabatic passage , concept introduced by shapiro et al . in quantum optics @xcite , for an optical broadband and ultra - broadband polarization rotators .
the broadband and ultra - broadband polarization rotators are modular designed as stacks of multi order half - wave plates or commercial achromatic half - wave plates , which are rotated smoothly from alignment between the fast - polarization axes with initial linear polarized light to the alignment of the fast - polarization axes with desired linearly polarization orientation .
the experimental results show visible broadening effects which are kept the same at different angles of the chosen sequence of wave plates .
e. d acknowledges financial support by bulgarian national science fund grant : drila 1/04 .
a.r and e.k acknowledge financial support by sutd start - up grant no .
srg - epd-2012 - 029 , sutd - mit international design centre ( idc ) grant no . | analogy between poincar space light polarization and bloch space quantum - state dynamics known @xcite allowed transfer of concepts from quantum optics nuclear magnetic resonance to polarization optics concept of composite pulses used broadband half- quarter - wave plates tunable polarization rotators example application of adiabatic techniques for quantum systems in polarization optics design robust achromatic polarization retarders optical isolators paper experimental demonstration of broadband ultra - broadband polarization rotators designed piecewise adiabatic passage technique - rotators assembled shore et al as ten multi - order half - wave plates and ten commercial achromatic half - wave plates half - wave plates rotated gradually from alignment to final consider propagation of plane electromagnetic wave along z - axis through lossless medium polarization evolution given with torque equation for stokes vector @xmath0where @xmath1 is distance along propagation direction stokes vector driving torque @xmath3 is birefringence vector of mediummedium optical activity uniaxial with slow fast axes in @xmath4 plane components birefringence vector @xmath5 as @xmath6here @xmath7 @xmath8 refractive indices along fast slow axes @xmath9 light wavelength @xmath10 rotary power @xmath11 angle of rotation between fast axis @xmath12 axis stokes vector @xmath14 parallel to birefringence vector @xmath15 change angle @xmath11 stokes vector adiabatically linearly polarized light in horizontal plane @xmath17 @xmath18 aligning @xmath12 @xmath13 axes with fast slow optical axes birefringence vector @xmath15 aligned with stokes vector following adiabatic evolution final angle of linear polarization @xmath19 adiabatic approach robust large intervals for @xmath20 broadband alternative scheme for adiabatic evolution for stokes vector discrete modular design of polarization rotators @xcite rotation of angle @xmath11 in discrete stepsrealize by sequence @xmath21 crystals each rotated angle @xmath22 cartesian coordinate system illustrated in figure adiabatic evolution achieved if rotation angles change gradually sufficient steps @xmath23 not than cycle module half - wave plate performed experiments broadband ultra - broadband adiabatic polarization rotators linear polarization white light beam rotated at 45 60 75 90 degrees ten half - wave plates irises @xmath26 lens polarizer @xmath29 form collimated beam white polarized light analyzer @xmath30 lens @xmath31 focus output light entrance @xmath32 optical fibre spectrometer ten wave plates rotates linear polarization light in broadband ultra - broadband spectra 10 w halogen - bellaphot lamp with dc power supply applied light source continuous spectrum 450 nm to 1100 nm figure iris two lenses polarizer used for collimated light beam iris @xmath26 in focus of plano - convex lens @xmath27 @xmath33=35 mm point light sourcebeam collimated by second lens @xmath28 with @xmath33=150 mm polarizer @xmath29 210 - 1100 nm from lambda-950 spectrometer polarized white light horizontal used two half - wave plates experiments first experiment ordinary multi - order quarter - wave plates @xmath34=763 nm second experiment achromatic half - wave plates wrm053-mica 700 - 1100 nm aperture 20 mm each wave plate @xmath35 assembled rsp1 rotation mount @xmath36 rotation scale @xmath37 increments allows precise repeatable positioning fine angular adjustment analyzed transmitted light through ten wave plates by polarizer @xmath30 grating monochromator used plano - convex lens @xmath31 ( @xmath33=20 mm ) two - axis micro - positioner focus light beam optical fibre entrance @xmath32 monochromator red green blue black @xmath42 aim demonstrate broad ultra - broad spectral bandwidth linear polarization rotators ten identical wave platesset assembled by multi - order or achromatic wave plates wave plates tilted to reduce reflections for analysis used single beam spectrometer losses light transmission reflection absorption measured light dark spectra for experiments used dark spectrum measured with light path blocked for correction of hardware offsets reference spectrum with light source on blank sample measured transmission spectrum of assembled polarization rotator axes of polarizer @xmath29 analyzer @xmath30 fast axis of single wave plates set parallel used measured light spectrum as reference for subsequent measurements linearly polarized light beam passed by @xmath43 wave plates adjusted at estimated angle by formula : @xmath44 sequential number of wave plate rotation effect investigated for angles data taken with analyzer @xmath30 revolved at angle @xmath47 @xmath38 red line green line blue line black line @xmath42 black dashed dot curve shows spectrum of single achromatic wave plate tested broad spectral bandwidth polarization rotator settled by ten multi - order half - wave platestransmittance spectra taken for four angles @xmath47 results presented on figure fig3 experimental curves show flat maximum in wide spectral diapason at different angles @xmath47 curves at angles @xmath38 @xmath30 presented blow out ray at @xmath38 proves polarization light orthogonal to at angle @xmath47 broader effect using ten achromatic half - wave plates created ulta - broad bandwidth linear polarization rotator algorithm used same rotation angles @xmath47 wave plate same type axis rotated at angle @xmath48 experimental results shown on figure fig fig4 wave plates linear polarization rotator transmittance spectra at angles @xmath38 presented figure experimental curves not exceed 1000 nm lack of signal after spectral region demonstrated application of piecewise adiabatic passage shapiro quantum optics for optical broadband ultra - broadband polarization rotators modular as stacks of multi order half - wave plates rotated smoothly from alignment fast - polarization axes to results show visible broadening effects same at different angles wave platesacknowledges support bulgarian national science fund grant a.r e.k sutd start - up grant epd-2012 - 029 international design centre grant | 0.431995 | 2,492 | 1,240 | , Saving $0.1 in GPT-4. |
67,573 | in this paper we discuss the demographics of the radio blazar population : ( i ) what are their parent ( ` unbeamed ' ) sources and ( ii ) what magnitude and/or type of evolution have they undergone ?
the discussion is based on models of radio source evolution and beaming based on a ` dual population ' unification paradigm .
these models , developed from radio blazar properties in bright samples , predict blazar demographic trends at the lower flux - density levels ; samples from deep mjy - level surveys ( _ e.g. _ nvss and first ) may now provide direct tests of these predictions .
# 1 1.25 in .125 in .25 in | all galaxies are radio galaxies in a sense .
the scale of activity ranges from ` normal ' galaxies ( like our own ) whose radio emission emanates from supernova remnants . at higher radio powers
are the ` starburst ' galaxies , greatly - enhanced stellar activity resulting in stronger radio emission . at similar radio powers
are the seyfert galaxies with their small - scale radio core - jet structures and with clear indication of agn activity . at the top end of the scale
are the powerful radio galaxies , together with bllacs and quasars , the latter two being collectively termed blazars .
radio galaxies tend to have steep radio spectra and therefore dominate low radio - frequency ( @xmath0 0.5 ghz ) surveys .
the structures are predominantly double - lobed ; fanaroff & riley ( 1974 ) compared the distances between central maxima of the radio structures to overall size and discovered a dichotomy .
sources with regions of maximum brightness separated by more than 0.5 times the overall source size are edge - brightened ( frii ) , while when the regions of maximum brightness lies within this limit the source has diffuse outer lobes ( fri ) ( centrally concentrated ) ; and the highest radio power objects are _ predominantly _ friis whilst those of lower radio power are _ usually _ fris .
the overlap between the classes at middling radio powers is seen in the local radio luminosity function ( figure 1 ) , and the division is known to be a function of optical luminosity as well as radio ( owen and ledlow 1993 ) .
evidence is accumulating that the powerful radio sources , double - lobed radio galaxies on the one hand and blazars on the other , are ` unified ' by projection effects , with a central torus together with relativistic beaming responsible for our classification of objects as radio galaxies or blazars ( scheuer & readhead 1979 ; orr & browne 1982 ; scheuer 1987 , barthel 1989 , urry & padovani 1995 and references therein ) .
the ` dual - population ' unified scheme ( wall & jackson 1997 ) , summarised in table 1 , posits that frii radio galaxies are the parent population of _ all _ quasars and _ some _ bllac type objects .
the frii - quasar sources are galaxies with class ` a ' spectra ( hine & longair , 1979 ) , having strong , high - excitation emission lines in their optical / uv spectra .
the frii - bllac sources are galaxies with class ` b ' spectra ( hine & longair , 1979 ) , having only weak , if any , high - excitation lines in their spectra .
the second part of the scheme posits that fri radio galaxies are the parent population of the remainder of the bllac type objects .
lccc & uv / optical & radio + & emission & spectrum & source + population & type & at 5 ghz & type + + frii , class a & narrow & steep & rg + & broad & steep & rg + & broad & ` flat ' & quasar + + frii , class b & weak / none & steep & rg + & none & ` flat ' & bl lac + + fri & weak / none & steep & rg + & none & ` flat ' + +
our analysis interprets radio source counts and and identification / redshift data for samples at high flux densities , adopting a dual - population unified scheme paradigm . using a simple parametric model to describe the evolution of the underlying parent populations we first fit low - frequency radio data ( @xmath0 0.5 ghz ) , where radio samples are unbiased by the effects of doppler beaming .
then , using these evolution models for the parent populations of fri and frii objects , we fit the 5 ghz radio source count using a small number of parameters to describe the doppler beaming , which gives rise to blazars ( _ i.e. _ quasars and bl lacs with @xmath1 , where @xmath2 ) the fundamental dataset to our analysis is the differential radio source count , the count at a single observing frequency of source density on the sky as a function of flux density .
compiling data from wide - area and deep pencil - beam surveys yields source counts spanning 2 decades in frequency and 6 decades in flux density as shown in figure 2 .
in fact radio source counts contain a wealth of cosmological information , in particular providing potent evidence against a steady - state universe ( ryle & clarke 1961 ) and indicating that differential evolution must take place , with the most luminous sources showing the strongest evolution .
in addition the _ shape _ of the source count changes with observing frequency , suggesting that more complex behaviour than just number density evolution is taking place - and our analysis shows just what this behaviour is .
identification programmes and further studies of radio source samples have revealed distinct regions within the counts .
these regions are indicated along the abscissa in figure 2 and discussed briefly here : * region 1 * at the very highest flux densities the source count is near - euclidean , although typically less than 20 sources contribute to this region .
this region is comprised of a mixture of nearby sources and cosmologically - distant powerful sources . the flat count arises due to the dilution of the powerful evolving sources by the local lower - radio - power sources which are far more abundant at low redshifts . * region 2 * in this region the counts rise more rapidly with decreasing flux density than the euclidean ( -3/2 power law ) prediction , and then reach a euclidean plateau .
this region of the count is made up of the most powerful radio sources at cosmological distances .
the low - frequency counts ( @xmath3 0.5 ghz ) comprise almost entirely of steep - spectrum sources whose radio emission is greatest at these frequencies . the width of the plateau increases with increasing frequency , due to an increasing blazar ( flat - spectrum ) contribution .
* region 3 * below the euclidean plateau the counts at all wavelengths drop away from the euclidean prediction .
this region extends up to two orders of magnitude in flux density .
( limited ) identification data shows that the sources at such flux densities are lower - power objects at intermediate redshifts and not , as might be supposed , powerful radio sources at ever - increasing redshifts .
very little work has been done on the blazar population at these flux density levels .
* region 4 * at low flux densities the counts again flatten to near - euclidean . in this region
the dominant populations change dramatically to ` blue ' starburst galaxies and ` red ' fri - type galaxies , both relatively local and of low radio powers ( _ e.g. _ windhorst et al .
1993 , benn et al . 1993 ) . using the spectroscopically - complete 3crr sample at 178 mhz ( laing , riley & longair 1983 and more recently published data collated by r. laing , private communication )
we first determine an appropriate form for the evolution of the fri and frii populations using the statistic .
figure 3 shows @xmath4 values for 137 steep - spectrum frii radio galaxies and 26 fri radio galaxies .
both populations show increasing with radio power . however , fris exibit negative evolution with @xmath5 at @xmath6 and the friis exhibit strong positive evolution with @xmath7 at @xmath8 .
interestingly , there is a trend for to continuously increase with @xmath9 across the populations , suggesting luminosity - dependent evolution which may be population - independent . for all friis
+ for all fris adopting exponential luminosity - dependent density evolution ( ldde ) we fit the source count at 151 mhz comprising the 3crr and 6c survey ( hales , baldwin , & warner 1988 ) , using a minimisation routine .
the model fit requires strong evolution of the most powerful frii sources , with little or no evolution of the fri population ( table 2 ) .
thus the fit reproduces the behaviour for the radio galaxy population as a whole , in that only friis with @xmath10 undergo any evolution , the lower - power friis and all fris have a constant space density to their cut - off redshift , @xmath11 .
lcrr & exponential ldde & + population & parameter values & @xmath12 & @xmath13 + + fri & @xmath14 , @xmath11 = 5.0 + & @xmath15 & @xmath16 not used given @xmath17 = 0.0 + + frii , class a & b & @xmath18 , @xmath11 = 5.62 + & @xmath19 , @xmath20 + + & best - fit & 30.73 & 33 + @xmath21 degrees of freedom
we fit the 5 ghz source count , which is well - defined over a wide flux - density range , starting from the evolution model fit of table 2 .
however at 5 ghz we incorporate the beamed products ( blazars ) by randomly aligning the sources with respect to our line - of - sight , allowing both different lorentz factors ( @xmath22 ) and different core - to - extended flux ratios for the two ( fri , frii ) populations .
the best - fit beaming parameters are shown in table 3 .
lcrr & & + population & parameter values & @xmath12 & @xmath13 + + fri & @xmath22 = 15.0 , @xmath23 + + frii , class a & b & @xmath22 = 8.5 , @xmath24 = 0.01 + & @xmath25 + + & best - fit & 32.98 & 25 + the contribution from each population to the 5-ghz source count is shown in figure 4 .
the contribution from the quasar population peaks at a higher flux density limit than its parent population , and a similar behaviour is seen for the the bl lac objects .
we can analyse the relative importance of each class of radio source as a function of flux density .
blazars are dominated by quasars ( high - excitation frii parents ) at @xmath26 0.3 jy , but dominated by bl lac - type objects at lower flux densities .
the magnitude of the rapidly changing quasar fraction predicted by our model is well - matched by observations from the 0.25 jy - sample of flat - spectrum sources of shaver et al .
( 1996 ) - see figure 15 of jackson & wall ( 1999 ) . in terms of population mix , it is also instructive to consider the integral population count , shown in figure 5 in which the model fit has been transposed to 1.4 ghz .
this figure reveals that complete samples are predicted to comprise @xmath2720% bl lac objects for flux density limits extending down to @xmath28 0.1 mjy .
optical spectroscopy of objects selected from the first survey ( becker , white & helfand 1995 ) could provide substantial tests of this prediction to @xmath28 1 mjy .
transposing our model ( evolution plus beaming ) to 2.7 ghz we compare our predictions with the limited observational data currently available .
figure 6 shows the mean redshift , @xmath29 , across a wide flux density limit range for all source types as well as three sub - classes . for complete samples
the model predicts that the value of peaks at @xmath2720 mjy then drops rapidly . in detail
, we find that _
( i ) _ the contribution from steep - spectrum sources dominates and therefore traces the overall @xmath29 , _ ( ii ) _ for the quasar population , rises continuously as @xmath30 decreases .
however the _ number _ of quasars is negligible below @xmath30 = 20 mjy so this has little effect on the overall value and _ ( iii ) _ the peak of the bl lac sources occurs at a higher flux density than that of the quasars , although the value of is much lower . comparing the model values to the 0.25 jy flat - spectrum quasar sample of shaver et al .
( 1996 and in preparation ) we find reasonable agreement .
the data points for the 0.25 jy sample shown on figure 6 are _ lower _ limits as this sample selects quasars with ( @xmath31 ) .
our model , however , defines blazars as having ( @xmath32 ) . beamed objects with @xmath31 have parent sources of lower radio luminosity , which in turn are predicted to have a lower distribution than the less - beamed @xmath33 , higher - luminosity objects which are missing from the 0.25 jy sample .
figure 7 splits the bllac distribution of figure 6 into its contributing parent populations .
we see that low - excitation - frii bllac sources have a distribution which is very similar to the quasar distribution . however , the fri bllacs have a very different distribution , with gradually increasing with decreasing flux density .
morphological studies of a sizeable sample of bllac objects from _
e.g. _ the first survey would provide direct , powerful tests of this prediction . | galaxies are radio galaxies scale activity ranges from normal galaxies radio emission from supernova remnants higher radio powers starburst galaxies enhanced stellar activity stronger radio emission similar radio powers seyfert galaxies small - scale radio core - jet structures agn activity top end powerful radio galaxies bllacs quasars termed blazars radio galaxies steep radio spectra dominate low radio - frequency surveys structures predominantly double - lobed fanaroff riley 1974 compared distances between central maxima to size dichotomy sources with regions maximum brightness separated by more than 0.5 times size edge - brightened frii within limit diffuse outer lobes highest radio power objects predominantly friis lower radio power usually fris overlap between classes middling radio powers seen in local radio luminosity function figure 1 division function of optical luminosity radio ledlow 1993 evidence powerful radio sources double - lobed radio galaxies blazars unified by projection effects central torus relativistic beaming for classification as radio galaxies or blazars scheuer 1979 1982 1987 1989 padovani 1995dual - population unified scheme wall jackson 1997 summarised table 1 posits frii radio galaxies parent population quasars bllac objects frii - quasar sources galaxies class ` a ' spectra strong high - excitation emission lines optical uv spectra frii - bllac sources class ` b ' spectra weak high - excitation lines posits fri radio galaxies parent population remainder bllac type objects lccc uv optical radio emission spectrum source population type 5 ghz + frii class a narrow steep rg + broad flat ' quasar + frii class b weak none steep rg + none ` flat ' bl lac + fri weak none steep rg + flat ' analysis interprets radio source counts identification / redshift data for high flux densities dual - population unified scheme paradigm simple parametric model evolution fit low - frequency radio data 0.5 ghz unbiased by doppler beaming models fri frii fit 5 ghz radio source count small parameters doppler beaming rise blazarsquasars bl lacs @xmath1 @xmath2 fundamental dataset analysis differential radio source count single frequency source density function flux density data from wide - area deep pencil - beam surveys yields source counts 2 decades frequency 6 decades flux density figure 2 radio source counts contain cosmological information evidence against steady - state universe differential evolution luminous sources strongest evolution shape source count changes with frequency complex behaviour number density evolution analysis shows identification programmes studies radio source samples revealed distinct regions within counts regions indicated abscissa figure 2 region 1 * highest flux densities source count near - euclidean less than 20 sources contribute mixture nearby cosmologically - distant powerful sources flat count dilution powerful sources by local lower - radio - power sources abundant low redshifts region 2 * counts rise rapidly decreasing flux density reach euclidean plateau region most powerful radio sources cosmological distances low - frequency counts @xmath3 0.5 ghz ) steep - spectrum sources radio emission greatest width plateau increases increasing frequency due increasing blazar flat - spectrum ) contributionregion 3 below euclidean plateau counts wavelengths drop from euclidean prediction extends two orders magnitude in flux density limited data shows sources lower - power objects at intermediate redshifts not powerful radio sources at redshifts little work on blazar population region 4 low flux densities counts flatten to near - euclidean dominant populations change to blue starburst galaxies red fri - type galaxies local low radio powers windhorst 3crr sample at 178 mhz data determine appropriate form for evolution of fri frii populations figure 3 shows @xmath4 values for 137 steep - spectrum frii galaxies 26 fri radio galaxies populations with radio power fris negative evolution with @xmath5 @xmath6 friis positive evolution with @xmath7 trend increase with @xmath9 suggesting luminosity - dependent evolution all friis fris adopting exponential luminosity - dependent density evolution fit source count at 151 mhz 3crr 6c survey hales 1988 minimisation routine model fit requires strong evolution of powerful frii sources little or no evolution of fri population table 2fit reproduces behaviour radio galaxy population only friis with @xmath10 undergo evolution lower - power friis fris have constant space density redshift @xmath11 population parameter values @xmath12 + fri = 5.0 + @xmath17 = 0.0 + frii class a b @xmath18 = 5.62 + best - fit 30.73 33 + @xmath21 degrees freedom fit 5 ghz source count flux - density range evolution model at 5 ghz beamed products randomly aligning sources different lorentz factors core - to - extended flux ratios for populations best - fit beaming parameters shown in table 3 population parameter values @xmath12 + + fri @xmath22 = 15.0 + + frii class a b = 8.5 @xmath24 = 0.01 + @xmath25 + best - fit 32.98 & 25 + contribution each population to 5-ghz source count in figure 4 contribution quasar population higher flux density limit similar behaviour bl lac objects relative importance of each class radio source flux densityblazars dominated by quasars at @xmath26 0.3 jy bl lac objects at lower flux densities changing quasar fraction model matched by observations 0.25 jy - sample flat - spectrum sources shaver 1996 jackson wall 1999 integral population count in figure 5 model transposed to 1.4 ghz complete samples @xmath2720% bl lac objects for flux density limits to @xmath28 0.1 mjy optical spectroscopy of first survey tests prediction to @xmath28 1 mjy transposing model to 2.7 ghz predictions with observational data figure 6 shows mean redshift @xmath29 across flux density limit range for all source types three sub - classes complete samples predicts value of peaks at @xmath2720 mjy drops rapidly contribution from steep - spectrum sources dominates @xmath29 quasar population rises as @xmath30 decreases quasars negligible below @xmath30 = 20 mjy effect on value peak of bl lac sources at higher flux density than quasars value lower model values to 0.25 jy flat - spectrum quasar sample of shaver . reasonable agreementdata points 0.25 jy sample figure 6 lower limits selects quasars @xmath31 ) model defines blazars @xmath32 ) beamed objects with @xmath31 lower radio luminosity predicted lower distribution less - beamed @xmath33 higher - luminosity objects missing 0.25 jy sample figure 7 splits bllac distribution parent populations low - excitation - frii bllac sources similar quasar distribution fri bllacs different distribution increasing decreasing flux density studies sample bllac objects first survey tests prediction | 0.440853 | 2,880 | 1,453 | , Saving $0.1 in GPT-4. |
67,575 | i present a technique for measuring the broken phase sphaleron rate nonperturbatively .
there are three parts to the calculation : determination of the probability distribution of chern - simons number @xmath0 ; measurement of @xmath1 , the mean rate of change of @xmath0 at the barrier ; and measurement of the `` dynamical prefactor , '' the fraction of barrier crossings which result in a permanent integer change in @xmath0 . | the `` sphaleron rate , '' the spacetime diffusion constant for chern - simons number @xmath0 , is a key quantity in the study of electroweak baryogenesis , setting the efficiency of baryon number violation .
it is also a very appropriate topic at this conference , since it is nothing but the topological susceptibility of the electroweak plasma , only in _
minkowski _ , time ; @xmath2 @xmath3 can be determined reliably by perturbation theory in any regime where @xmath4 , where @xmath5 is the higgs condensate . however
, this is precisely the regime where it is so small as to be of no physical consequence .
we really want to know it in the symmetric phase at @xmath6 the phase transition temperature , to determine how many baryons can be produced ; and in the broken phase at and immediately below @xmath6 to determine whether they are subsequently erased or are preserved until the present epoch .
recently there has been significant progress in the symmetric phase rate .
it has been shown analytically @xcite and verified numerically @xcite that `` hard thermal loop '' radiative corrections are important , and the rate is known with reasonable accuracy , though it has recently been demonstrated that there are logarithmic corrections @xcite which are not well under control .
but the real time method can not measure the broken phase rate ; the rate is so low that no events will occur in any reasonable amount of time .
since we can not extend the perturbative calculation beyond 1 loop , we will instead go about a nonperturbative measurement .
we can use the dimensional reduction approximation and we will relentlessly ; we also take the dynamics to be classical , regulated to give a physically reasonable cutoff to the hard thermal loops . for
a much more detailed discussion of the work presented here , see @xcite .
here is the cartoon picture of why baryon number violation is slow in the broken phase . because the higgs condensate makes being half way between winding numbers energetically expensive
, there is a free energy barrier to being at @xmath7 .
our method of determining the rate is to find the probability distribution of @xmath0 , and to multiply the chance to be in a narrow range about @xmath7 by the mean crossing time of that range , set by @xmath8 .
then we need to determine what fraction of crossings are permanent . in the continuum ,
the meaning of @xmath0 , modulo 1 , is the integral of @xmath9 , @xmath10^a \right ) \ , .\ ] ] here @xmath11 parameterizes a path through the space of 3-d configurations .
the constant of integration is set by defining @xmath12 in vacuum .
this is consistent since the above integral around any closed loop is the second chern class of the loop , which is an integer .
however , to make a nonperturbative study we need to find a lattice implementation for the above , which is touchy since no lattice definition of @xmath9 is a total derivative .
the best thing to do is to use the above definition for a unique specific path , the cooling path , defined by @xmath13 with @xmath14 the yang - mills part of the 3-d hamiltonian .
this definition of @xmath0 is unique because the cooling path is ; it is also minimally exposed to lattice problems since the cooling process very quickly eliminates uv fluctuations , which are responsible for any poor behavior in @xmath9 .
there are three problems we must take care of , though ; a uv noise problem , a numerical cost problem , and the problem of turning a probability distribution for @xmath0 into a rate .
the first problem is that @xmath0 actually is not a very good measurable to be using .
the problem is that it contains a lot of uv noise not related to topology .
our discussion will follow that of ambjrn and krasnitz @xcite . in the uv
the 3-d classical thermal theory we are studying behaves at leading order like 3 copies of the abelian theory , in which @xmath15 the mean square value at leading order in perturbation theory is then @xmath16 which is linearly uv divergent .
the size of the divergence is extensive in volume and will be cut off by the lattice spacing , leading to a @xmath17 behavior .
the solution is not to measure @xmath0 itself , but @xmath0 _ after _ some preliminary cooling has been performed ; @xmath18^a
\right ) \ , .\ ] ] to illustrate that this does the trick , we show @xmath0 and @xmath19 for a period of hamiltonian evolution in the broken phase , and @xmath19 for a sphaleron transition : this graphically illustrates that using @xmath19 cuts out irrelevant uv noise , while still capturing interesting topological information .
there is still a problem with using @xmath19 , the numerical cost involved in performing the cooling , which we must do thousands of times to get an accurate monte - carlo study of the probability distribution .
we finesse this problem by observing that , since cooling eliminates the uv excitations , it quickly renders the lattice fields smooth enough that we lose no information by blocking by a factor of 2 .
when we use a classically improved hamiltonian and definition of @xmath9 after the blocking , there is almost no difference between the blocked and unblocked measurement of @xmath19 , provided we cool by at least @xmath20 before blocking . for lattices larger than about @xmath21
we can double block .
we can now find the ( @xmath22 dependent ) probability distribution of @xmath19 by conventional multicanonical means . to turn this into a rate , we first measure @xmath23 ( sampled over @xmath24 configurations ) by making a canonically distributed sample of @xmath24 configurations , generating momenta from the canonical ensemble , and evolving for a very short hamiltonian time , measuring @xmath19 before and after .
the mean rate of @xmath24 crossings per 4-volume is then @xmath25 to measure the dynamical prefactor , which is the fraction of crossings which lead to permanent integer change in @xmath0 , we use the same @xmath24 sample but continue the hamiltonian time evolution , both forward and backward in time , until the system settles about a vacuum ; then we count how many crossings and how many permanent @xmath19 changes occur .
the fraction of crossings leading to permanent @xmath19 change is @xmath26 for very large debye mass squared @xmath27 there are multiple crossings , but for realistic @xmath27 the prefactor is about 0.4 .
the final results for @xmath3 are compared to perturbation theory ( two loop effective potential but tree kinetic terms ) in figure [ lastfig ] .
baryon number erasure after the phase transition is prevented if @xmath28 , which occurs if @xmath29 .
99 p. arnold , d. son , and l. yaffe , phys . rev . *
d 55 * , 6264 ( 1997 ) .
g. d. moore , c. hu , and b. mller , phys . rev . * .
d 58 * , 045001 ( 1998 ) .
d. bdeker , phys . lett . * b 426 * , 351 ( 1998 ) .
g. d. moore , hep - ph/9805264 .
j. ambjrn and a. krasnitz , phys . lett .
* b 362 * , 97 ( 1995 ) . | sphaleron rate spacetime diffusion constant for chern - simons number @xmath0 key in electroweak baryogenesis setting efficiency of baryon number violation appropriate topic at conference topological susceptibility of electroweak plasma in time @xmath2 @xmath3 determined by perturbation theory in regime where @xmath4 @xmath5 is higgs condensate regime small no physical consequence want to know in symmetric phase at @xmath6 phase transition temperature determine baryons produced broken phase at below @xmath6 determine erased or preserved until present epoch significant progress in symmetric phase rate shown hard thermal loop radiative corrections important rate known with accuracy logarithmic corrections not under control real time method can measure broken phase rate rate low no events occur in reasonable time can extend perturbative calculation beyond 1 loop nonperturbative measurement use dimensional reduction approximation dynamics classical physically reasonable cutoff to hard thermal loops detailed discussion see @xcite picture baryon number violation slow in broken phase higgs condensate makes half way between winding numbers energetically expensive free energy barrier to at @xmath7 determining rate find probability distribution of @xmath0 multiply chance range @xmath7 by mean crossing time set by @xmath8 determine fraction of crossings permanent continuum @xmath0 modulo 1 is integral of @xmath9 @xmath10 @xmath11 parameterizes path through 3-d configurations constant of integration set by defining @xmath12 in vacuum integral around closed loop is second chern class of loop integer nonperturbative study lattice implementation for no lattice definition of @xmath9 total derivative use definition for unique path cooling path defined by @xmath13 with @xmath14 yang - mills 3-d hamiltonian definition of @xmath0 unique because cooling path minimally exposed to lattice problems eliminates uv fluctuations for poor behavior in @xmath9 three problems uv noise numerical cost turning probability distribution for @xmath0 into rate first problem @xmath0 not good measurable contains uv noise not related to topology discussion follow ambjrn and krasnitz @xcite uv 3-d classical thermal theory behaves like 3 copies abelian theory @xmath15 mean square value is @xmath16 linearly uv divergent divergence extensive in volume cut off by lattice spacing to @xmath17 behavior solution not measure @xmath0 but @xmath0 after preliminary cooling @xmath18 show @xmath0 @xmath19 for hamiltonian evolution broken phase @xmath19 sphaleron transition @xmath19 cuts out irrelevant uv noise capturing topological information problem with @xmath19 numerical cost cooling thousands times accurate study probability distribution cooling eliminates uv excitations renders lattice fields smooth lose no information by blocking by factor of 2 improved hamiltonian definition of @xmath9 after blocking almost no difference between blocked and unblocked measurement of @xmath19 cool by @xmath20 before blocking lattices larger than @xmath21 can double block find @xmath22 dependent probability distribution of @xmath19 by conventional meansturn into rate measure @xmath23 sampled over configurations distributed sample generating momenta evolving short hamiltonian time measuring @xmath19 before after mean rate @xmath24 crossings per 4-volume @xmath25 measure dynamical prefactor fraction crossings permanent integer change in @xmath0 use same @xmath24 sample continue hamiltonian time evolution until system settles vacuum count crossings permanent @xmath19 changes fraction of crossings leading to permanent @xmath19 change is @xmath26 large mass @xmath27 multiple crossings realistic @xmath27 prefactor 0.4 final results for @xmath3 compared to perturbation theory figure baryon number erasure after phase transition prevented if @xmath28 occurs if @xmath29 p. arnold d. son l. yaffe phys . rev 6264 ( 1997 ) moore hu b mller 045001 ( 1998 ) bdeker phys 426 351 ( 1998 ) j. ambjrn a. krasnitz 362 97 ( 1995 ) | 0.268967 | 1,656 | 861 | , Saving $0.0 in GPT-4. |
67,580 | it has recently been shown that strong interactions underlying electroweak symmetry breaking will induce four - fermion amplitudes proportional to @xmath0 , which in turn will influence a variety of flavor changing processes .
we argue that the size of these effects are likely to be far below the current experimental bounds .
( 0,0)(0,0 ) ( 310,205)utpt-97 - 14 ( 310,190)hep - ph/9706333 the corrections induced by a new strong sector underlying electroweak symmetry breaking are conveniently encoded in an effective chiral lagrangian .
recently attention has focused on a particular term in this effective lagrangian which induces corrections proportional to @xmath0 in charged current interactions @xcite . integrating out
the @xmath1 quark then yields interesting effects in the down - type quark sector , inducing corrections to @xmath2 and @xmath3@xmath4 mixing @xcite , and various rare @xmath5 and @xmath6 decays @xcite .
all these effects are correlated since they are related to one parameter in the effective lagrangian @xcite . in this note
we will provide an estimate of the size of the new parameter , expressed in terms of the number of new fermion doublets in some underlying theory of electroweak symmetry breaking .
such an estimate in the case of the @xmath7 parameter proved useful to constrain technicolor theories .
the @xmath7 parameter is related to a term in an effective lagrangian with coefficient @xmath8 which is completely analogous to the @xmath9 term appearing at order @xmath10 in the low energy qcd chiral lagrangian @xcite . since @xmath9 is a measured quantity , an estimate for @xmath7
is thereby obtained @xcite for qcd - like technicolor theories .
the situation is somewhat different for the parameter of interest here , which is the coefficient of another term appearing at order @xmath10 . in the qcd case
this coefficient is not a measurable quantity since the term can be removed by the equations of motion .
fortunately in the qcd case there are quark models which model the chiral symmetry breaking and which quite successfully reproduce the values of all ten measured parameters , @xmath11@xmath12 .
such models can then be expected to provide a reasonable estimate of the new parameter , which we will refer to as @xmath13 ( corresponding to @xmath14 in @xcite and @xmath15 in @xcite ) .
a naive quark model may be based on the nonlinear sigma model where effects of a single quark loop are considered .
this does a fair job of reproducing those @xmath16 s which happen to correspond to convergent loop integrals @xcite .
a more sophisticated quark - based approach leads to the extraction of all ten parameters @xcite .
most convenient for our purposes is the gauged nonlocal constituent ( gnc ) quark model @xcite , which incorporates the momentum dependence expected for dynamically generated fermion masses .
in qcd the mass function is known to fall as the square of the momentum ( up to logarithms ) for large momentum .
the gnc model can incorporate such momentum dependence in an manner which preserves the local chiral symmetries of the underlying theory .
the mass function then naturally regulates the loop integrals , and successful values for @xmath11@xmath12 are obtained .
here we are interested in @xmath13 and its sensitivity to the choice of mass function .
we will focus on the following two terms in the chiral lagrangian , @xmath17 with @xmath18{,\nonumber}\\&&{l_{\mu \nu } } = { { \partial}_{\mu } } { l_{\nu } } - { { \partial}_{\nu } } { l_{\mu } } - [ { l_{\mu } } , { l_{\nu } } ] { , \nonumber}\\&&{l_{\mu } } = ( { v_{\mu } ^a } - { a_{\mu } ^a})\sigma ^{a},{\ \ \ } { r_{\mu } } = ( { v_{\mu } ^a } + { a_{\mu } ^a } ) \sigma ^{a}{.\nonumber}\end{aligned}\ ] ] by identifying @xmath19 with weak gauge fields @xmath20 and using the equation of motion for @xmath21 the @xmath13 term can be transformed into a sum of 4-quark operators , which include the following charged current interaction @xcite .
@xmath22 @xmath23 are the ckm matrix elements
. the chiral symmetry breaking physics in the underlying theory will produce finite contributions to the @xmath12 and @xmath13 coefficients .
these coefficients become running couplings within the effective theory , and the finite values we are referring to correspond to the values renormalized at the chiral symmetry breaking scale .
these values may be determined by choosing some convenient amplitude and using that to match the effective theory onto the underlying theory ( or at least a model of the underlying theory ) .
we shall consider the two - point function @xmath24 where the only contributions from the order @xmath10 chiral lagrangian are from @xmath12 and @xmath13 .
@xmath25 the naive quark loop yields @xmath26 where @xmath27 is the number of fermion doublets contributing in the loop .
the couplings of the gauge fields to the fermions in the gnc model are the following , where @xmath28 is the dynamical fermion mass function .
@xmath29{\nonumber}\\&&{\ \ \ \ } - i[({q_{1 } } , \mu , a)^{{}}{\ \leftrightarrow\ } ( { q_{2 } } , \nu , b)^{{}}]\\&&{\gamma _ { { \it aa}}^{\mu \nu
ab}}=0\\&&{\it g}({p_{2 } } , { p_{1}})=\frac { \sigma ( - { p_{2}^2 } ) - \sigma ( - { p_{1}^2})}{{p_{2}^2 } - { p_{1}^2}}\end{aligned}\ ] ] thus as a consequence of gauge invariance one diagram will have the two vector fields attached at the same point on the loop . to study the effect of the momentum dependence in @xmath28 we will consider the following one parameter family of mass functions satisfying @xmath30 .
@xmath31 for @xmath32 , the naive quark loop result is obtained , whereas qcd is better modeled by a value of @xmath33 closer to unity .
we plot @xmath34 and @xmath35 in fig .
( 1 ) as a function of @xmath33 .
when @xmath36 we see that @xmath37 is more than twice as large as the naive quark loop result , which brings it into line with the measured qcd value @xcite .
@xmath38 on the other hand shows surprisingly little sensitivity to @xmath33 , and its value for all @xmath33 is close to @xmath39 .
term in those references is defined differently so that it appears in ( [ b ] ) with the opposite sign .
] we also checked other functional forms for @xmath28 and found that @xmath38 is generally quite insensitive to the momentum dependence of the fermion mass function . in @xcite and
@xcite it was argued that the experimental constraints from @xmath2 and @xmath3@xmath4 mixing imply that @xmath40 .
but we now see that this limit is far above what we could reasonably expect .
our results imply that the limit is saturated for @xmath41 , which is clearly impossible given the constraint on @xmath27 from @xmath7 . to put it another way ,
if we assume that a heavy fourth family ( @xmath42 ) is still allowed by the constraint on @xmath7 , then the contribution to @xmath13 is only @xmath43 .
the shifts in the various rare @xmath5 and @xmath6 decay modes from standard model values as described @xcite would then be less than 5% . while our models of strong interactions are admittedly simple - minded , they strongly suggest that the effects of @xmath13 will be very difficult to observe given the constraints on @xmath7 .
we conclude that if flavor changing effects beyond the standard model are seen , they will likely have more to do with the physics responsible for quark and lepton masses rather than the physics responsible for electroweak symmetry breaking . | this research was supported in part by the natural sciences and engineering research council of canada .
99 d. comelli , j. bernabeu , a. pich , a. santamaria , phys .
lett . * 78 * , 2902 ( 1997 ) .
g. burdman , wisconsin preprint , madph-97 - 996 , hep - ph/9705400 . j. gasser and h. leutwyler , nucl . phys . *
b250 * , 465 ( 1985 ) .
b. holdom and j. terning , phys .
b247 * , 88 ( 1990 ) .
e.g. j. balog , phys . lett .
* 149b * , 197 ( 1984 ) .
d. espriu , e. de rafael , and j. taron , nucl . phys . *
b345 * , 22 ( 1990 ) , and references therein .
b. holdom , phys . rev . *
d45 * , 2534 ( 1992 ) ; j. terning , phys . rev . * d44 * , 887 ( 1991 ) ; b. holdom , j. terning , and k. verbeek , phys . lett . * b245 * , 612 ( 1990 ) ; * b273 * , 549e ( 1991 ) . and
@xmath38 are coefficients of terms in the chiral lagrangian ( [ a ] ) as determined by the gauged nonlocal constituent quark model .
the naive quark model gives @xmath44 .
the parameter @xmath33 appears in the fermion mass function in ( [ d ] ) . ] | research supported by natural sciences engineering research council canada comelli bernabeu pich santamaria phys . lett 78 2902 ( 1997 ) burdman wisconsin preprint madph-97 - 996 hep - ph/9705400 gasser leutwyler phys b250 465 ( 1985 ) holdom terning phys b247 88 ( 1990 ) balog. 149b 197 ( 1984 ) espriu de rafael j taron phys. b345 22 ( 1990 ) holdom phys d45 2534 ( 1992 ) j terning d44 887 ( 1991 ) holdom terning k verbeek phys b245 612 ( 1990 ) b273 549e ( 1991 ) @xmath38 coefficients chiral lagrangian determined nonlocal constituent quark model naive quark model gives @xmath44 parameter @xmath33 fermion mass function | 0.07318 | 383 | 206 | , Saving $0.0 in GPT-4. |
67,584 | methods and results for numerical simulations of one and two interacting rf - systems suitable for quantum gates are presented .
these are based on high accuracy numerical solutions to the static and time dependent schroedinger equation for the full in one and two variables . among the points examined in the static analysis
is the range of validity of the effective two - state or `` spin 1/2 '' picture .
a range of parameters is determined where the picture holds to good accuracy as the energy levels undergo gate manipulations .
some general points are presented concerning the relations between device parameters and `` good '' quantum mechanical state spaces .
the time dependent simulations allow the examination of suitable conditions for behavior , and permits the introduction of a random noise to simulate the effects of .
a formula is derived and tested relating the random noise to the rate .
sensitivity to device and operating parameters for the logical gates not and cnot are examined , with particular attention to values of the tunnel parameter @xmath0 slightly above one .
it appears that with values of @xmath0 close to one , a quantum cnot gate is possible even with rather short times .
many of the methods and results will apply to coupled double - potential well systems in general . | in previous work we have described quantum logic gates based on the rf .
the basic operation involved is an adiabatic inversion , where the reverses flux states under the sweep of an external field @xmath1 .
this is equivalent to the logical not @xcite .
when a second is added whose flux can add or subtract from , parameter ranges were found for the two @xmath2 so that the two - undergoes a reshuffling of levels equivalent to the logical operation cnot @xcite,@xcite . in this paper
we present a study of these systems by numerical simulations which enable us to examine these processes in more quantitative detail . among the points we can examine
is the validity of the `` spin 1/2 picture '' . in the previous work we often found good agreement with a simplified
spin 1/2 analogy " where the two lowest states of each are treated as an effective two - state system .
this picture is very useful in understanding and predicting the behavior of the s and here we examine its validity by simulations for the full many - state system
. a further question which can be studied in detail is .
this is the operating principle for our quantum gates and it is necessary to know under which conditions it holds . finally , we can use our programs to study the effects of on the quantum gates .
we will examine how to introduce as a noise signal and study its effects on our operations .
it should be stressed that our results would have been difficult if not impossible to obtain without the extensive of numerical programs . due to the sensitivity to the various parameters and the subtleties of the tunneling problem
, analytic methods would be difficult and uncertain . with the high accuracy programs
, a short run can replaces otherwise long , complicated , and often approximate formulas .
we begin by shortly reviewing our approach @xcite , @xcite as applied to the one variable , one rf , at first without . this will allow us to fix the notation and parameters , and to establish the connection between the full and the `` spin 1/2 picture '' .
the hamiltonian in terms of the capacitance @xmath3 and inductance @xmath4 of the junction is @xmath5 with @xmath6+\beta cos\phi\}$ ] , where @xmath7 , @xmath8 being the critical current for the junction @xcite .
the variable @xmath9 is the flux @xmath10 in the loop in flux quantum units @xmath11 , and furthermore shifted by @xmath12 so that @xmath13 corresponds to the maximum at the center of the double potential well @xmath14 .
similarly @xmath1 is an external field , where due to the shift , @xmath15 corresponds to a non - zero applied field of size @xmath16 . by factoring out an overall energy scale @xmath17 so that @xmath18 we obtain the dimensionless hamiltonian @xmath19+\beta\ , cos\phi\}={-1\over 2\xi } { \partial ^2\over\partial \phi^2 } + \xi \{{{\textstyle\frac{1}{2}}}[(\phi-\phi^{ext})^2 ] + \beta\ , cos\phi\}\;.\ ] ] with this choice of the factor @xmath20 the `` mass '' @xmath21 and the `` potential '' @xmath22 are equal , hence the second form in terms of a common parameter @xmath23 , whose value is @xmath24 . in effect
the capacitance c and inductance l have been exchanged for an energy scale @xmath20 and a dimensionless number @xmath23 .
we shall discuss the physical meaning of @xmath23 below . in the following we shall endeavor to express all energies in terms of the general energy unit @xmath20 and all times in the time unit @xmath25 . to converted these to dimensional units one multiplies by @xmath26 and for the time by @xmath27 @xcite .
thus results with the dimensionless , eq [ hama1 ] , which we will use in our computer simulations , are to be converted to physical energies by multiplying by @xmath20 .
typical values @xmath28 and @xmath29 for example , yield @xmath30 and @xmath31 , while the time unit is @xmath32 .
we will use these values for `` typical examples '' or when absolute times or energies are needed .
we shall usually work with @xmath33 , which for @xmath28 corresponds to a critical current of about @xmath34 . to carry out calculations with eq [ hama1 ] for the eigenvalues and eigenfunctions , the numerical method employed a large basis of harmonic oscillator and expanded the cosine as a low order polynomial . since in the harmonic oscillator basis
polynomials are simple , sparse , matrices , the problem is reduced to algebraic manipulations and a matrix diagonalization .
it was usually found that an expansion up to 8th order for the cosine and a basis of 256 oscillator states gave numerical stability .
a typical run with a few hundred basis vectors on a pentium 4 machine lasts less than a minute and resolves our smallest energy splitting to better than two significant figures .
eq [ hama1 ] contains three parameters , , @xmath0 and @xmath23 .
the applied flux controls the asymmetry of the potential and for = 0 the potential is symmetric .
fig.1 , left , shows an example with the potential in this symmetric configuration . in this `` level crossing ''
situation the energy splitting of the lowest pair of levels is just the tunneling energy @xmath35 and so is quite small .
here it is 0.0044 and not clearly resolved on the plot . [ cols="^,^ " , ]
we have presented a description of one or two interacting rf s in a form suitable for numerical simulation and a of programs for carrying out these simulations .
in addition to static properties , the behavior under time dependent conditions , including flux sweeps for gates , and the effects of noise or , were studied . among the points investigated was the validity of the `` spin 1/2 picture '' where the behavior governed by the full is approximated by the components of an effective `` spin 1/2 '' using the lowest levels of the double - potential well .
the identification between the parameters of the spin and the full was established and the regime of validity of the approximation were investigated numerically .
it is found , using various criteria , including `` hilbert space completeness '' , that the `` spin 1/2 '' picture is approximately valid up to variations of the external bias which moves the basic pair of levels a substantial fraction of the principal level splitting .
for the two , we have found it is capable of performing the logical operation cnot , and have been able to determine regions of the plane with definite configurations of the suitable for cnot sweeps .
the conditions for behavior under such sweeps were examined and times found for typical parameters .
the splitting @xmath35 of the basic pair enters into the condition and is very sensitive to the barrier parameter @xmath0 .
it is found that the most interesting regions for this parameter are values slightly greater than one , and detailed studies were presented for @xmath33 .
noise or were simulated as a random flux noise and a formula was derived relating the parameters of this random noise to the parameter d. numerical simulations verify the validity of the formula .
this random noise is then applied to the logic gates and possible regimes of operation for the gates are found .
one of the interesting conclusions is that the sweep time can be substantially longer than the time 1/d before effects become very large .
this is traced to the fact that , and also non- , are mostly effective at `` level crossing '' , which is only a small part of the sweep . indeed for the two - cnot gate it is found that noise applied to the control bit , which does not undergo a `` level crossing '' , has almost no effect .
one of our general conclusions therefore , for any devices of this general type , is that `` level crossings '' should be held to a minimum , and when they occur a large tunneling energy or splitting at `` crossing '' is beneficial both for reducing both and non - effects .
it thus appears that an interesting region for the engineering of gates involves rather small values of @xmath0 , where the splitting is large .
however the are then nearly delocalized .
we have examined the case @xmath36 , where the reduced tunnel barrier and thus large @xmath35 permit fast operation of the gates .
the resulting short time for the to act leads to operations which are successful with presently discussed values of the time , some nanoseconds .
however , a careful choice of parameters is necessary . generally speaking
, we may say that our studies verify that it is in principal possible to have a regime where one can create and manipulate quantum linear combinations of quasi - classical objects , as pictured in fig.9 .
while the parameters and examples we have used are specific to the rf , it will be recognized that our methods and results will apply , after some adjustments , to almost all interacting double - potential well systems .
p. silvestrini and l. stodolsky , _ physics letters _ * a280 * 17 - 22 ( 2001 ) .
see also p. silvestrini and l. stodolsky in _ macroscopic quantum coherence and quantum computing _
271 , eds .
d. averin , b. ruggiero and p. silvestrini , kluwer academic / plenum , new york ; cond - mat/0004472 .
we express frequencies as angular frequency , radians / second , since this what naturally arises with @xmath37 units .
to get herz , divide by @xmath38 .
we take this opportunity to correct a misunderstanding in refs @xcite where frequencies were expressed as hz : the units should have been radians / second .
other quantities , not cited as hz , need not be changed .
harris and l. stodolsky , _ phys .
lett . _ * b116 * 464 ( 1982 ) . for a general introduction and review of these concepts
see l. stodolsky , quantum damping and its paradoxes " in _ quantum coherence _ ,
j. s. anandan ed .
world scientific , singapore ( 1990 ) or l. stodolsky .
`` coherence and the clock '' in _ time and matter _
, i. i. bigi and m. faessler , eds .
world scientific , ( 2006 ) , pg.117 ; quant - ph/0303024 . increased robustness with respect to with increased tunnel splitting might , at first sight , appear to be unclear in view of the approach reviewed in section iv b of ref@xcite , where there seem to be two rates involved with the , with one of them apparently proportional to the tunnel splitting .
however , for small @xmath39 , using @xmath40 one verifies that the two rates eqs 4.11 and 4.12 of ref@xcite are really the same , controlled by the same factor @xmath41 .
that is , both terms are induced by the same physical processes and there is really only one parameter .
this is in fact represented by our @xmath42 , and at small x our simulations automatically include `` both kinds '' of .
the assumption of small x ( see discussion following eq[id ] ) is approximately correct in view of the small tunneling splittings , but for very low temperature where `` spontaneous emission '' becomes important , the point might have to be reconsidered .
this might be handled by modifying our eq[pdota ] so that with constant @xmath43 the density matrix relaxes not to @xmath44 at infinite time but to the more strictly correct limit given by a boltzmann factor .
of course , the argument that an increased tunnel splitting allows a faster sweep and so less time for to act remains true , independently of this issue . | previous work described quantum logic gates based on rf basic operation adiabatic inversion reverses flux states under sweep external field @xmath1 equivalent to logical not @xcite second added flux add or subtract parameter ranges found for two @xmath2 two - undergoes reshuffling levels equivalent to logical operation @xcite paper study systems by numerical simulations examine processes quantitative detail examine validity of `` spin 1/2 picture '' previous work agreement with simplified spin 1/2 analogy two lowest states treated as effective two - state system picture useful understanding predicting behavior s examine validity by simulations for full many - state system operating principle for quantum gates necessary to know under conditions holds use programs to study effects quantum gates examine introduce noise signal study effects on operations results difficult obtain without numerical programs to sensitivity to parameters tunneling problem analytic methods difficult uncertain high accuracy programs short run replaces long complicated formulas begin reviewing approach @xcite applied to one variable one rf fix notation parameters establish connection between full `` spin 1/2 picture ''hamiltonian capacitance @xmath3 inductance @xmath4 junction @xmath5 @xmath6+\beta cos\phi\}$ @xmath7 @xmath8 critical current junction @xcite variable @xmath9 flux @xmath10 flux quantum units @xmath11 shifted @xmath12 @xmath13 maximum center double potential well @xmath14 @xmath1 external field shift @xmath15 non - zero field @xmath16 factoring energy scale @xmath17 @xmath18 dimensionless hamiltonian @xmath19+\beta\ cos\phi\}={-1\over 2\xi }\partial \phi^2\phi-\phi^2 + \beta\ cos\phi\ factor @xmath20 mass @xmath21 potential @xmath22 equal second form common parameter @xmath23 value @xmath24 capacitance c inductance l exchanged energy scale @xmath20 dimensionless number @xmath23 discuss physical meaning @xmath23 express energies general energy unit @xmath20 times time unit @xmath25 dimensional units multiplies @xmath26 time @xmath27results dimensionless eq [ hama1 ] computer simulations converted to physical energies multiplying by @xmath20 values @xmath28 @xmath29 yield @xmath30 @xmath31 time unit @xmath32 use values for typical examples absolute times energies needed work with @xmath33 corresponds critical current @xmath34 calculations eq [ hama1 eigenvalues eigenfunctions method employed large basis harmonic oscillator expanded cosine as low order polynomial harmonic oscillator polynomials simple matrices problem reduced to algebraic manipulations matrix diagonalization expansion to 8th order cosine basis 256 oscillator states numerical stability typical run hundred basis vectors on pentium 4 machine lasts less than minute resolves smallest energy splitting two figures eq [ hama1 ] contains three parameters @xmath0 @xmath23 applied flux controls asymmetry potential = 0 symmetric fig.1 shows example potential symmetric configuration level crossing situation energy splitting lowest pair levels tunneling energy @xmath35 small 0.0044 not resolved plot description interacting rf s numerical simulation programsstatic properties behavior under time dependent conditions flux sweeps for gates effects of noise studied investigated validity of `` spin 1/2 picture '' behavior governed by full approximated by components effective spin 1/2 using lowest levels double - potential well identification between parameters spin full established regime validity approximation investigated numerically hilbert space `` spin 1/2 '' picture valid up to variations external bias basic pair of levels principal level splitting two logical operation cnot regions suitable for cnot sweeps conditions for behavior under sweeps examined times found for typical parameters splitting @xmath35 of basic pair sensitive to barrier parameter @xmath0 interesting regions for parameter values slightly greater than one detailed studies for @xmath33 noise simulated as random flux noise formula derived relating parameters noise to parameter d numerical simulations verify validity formula random noise applied to logic gates possible regimes of operation for found sweep time longer than time 1/d before effects large , non- effective at `` level crossing small part sweep two - cnot gate noise applied to control bit level crossing almost no effectgeneral conclusions for level crossings minimum large tunneling energy splitting at crossing beneficial for reducing non - effects interesting region for engineering gates small values of @xmath0 splitting large nearly delocalized examined case @xmath36 reduced tunnel barrier large @xmath35 permit fast operation gates short time for act leads to operations successful with values time nanoseconds careful choice of parameters necessary studies verify possible create manipulate quantum linear combinations of quasi - classical objects in.9 parameters examples specific to rf methods results apply to all interacting double - potential well systems p. silvestrini l. stodolsky physics letters _ * a280 * 17 - 22 ( 2001 ) macroscopic quantum coherence quantum computing _ 271 averin ruggiero silvestrini kluwer academic / plenum new york - mat/0004472 express frequencies as angular frequency radians / second @xmath37 units herz divide by @xmath38 misunderstanding in @xcite frequencies expressed as hz units radians / second other quantities not hz need not changed harris l.stodolsky phys . lett. * b116 * 464 ( 1982 ) introduction review concepts see stodolsky quantum damping paradoxes quantum coherence j. s. anandan ed world scientific singapore ( 1990 ) l stodolsky coherence clock time matter i i. bigi m. faessler eds world scientific 2006 ) pg.117 ; quant - ph/0303024 increased robustness increased tunnel splitting unclear approach section iv b ref@xcite two rates one proportional to tunnel splitting small @xmath39 @xmath40 verifies two rates 4.11 4.12 ref@xcite same controlled by same factor @xmath41 both terms induced same physical processes one parameter represented by @xmath42 small x simulations include both kinds assumption small x discussion correct small tunneling splittings very low temperature spontaneous emission important point reconsidered modifying eq[pdota ] constant @xmath43 density matrix relaxes not to @xmath44 correct limit boltzmann factor argument increased tunnel splitting allows faster sweep less time act remains true | 0.360635 | 2,566 | 1,307 | , Saving $0.1 in GPT-4. |
67,588 | the scalar wave equation is considered in the background of a charged vaidya metric in double null coordinates @xmath0 describing a non - stationary charged black hole with varying mass @xmath1 and charge @xmath2 .
the resulting time - dependent quasinormal modes are presented and analyzed .
we show , in particular , that it is possible to identify some signatures in the quasinormal frequencies from the creation of a naked singularity . | even though macroscopically charged black holes are not expected to be produced in realistic astrophysical scenarios , we can obtain fascinating information on the structure and behavior of einstein equations and general relativity with the study of these objects . here , we used the charged vaidya metric to model time - dependent black holes with an electric charge , and study the behavior of the quasinormal modes in a non - stationary background .
the charged vaidya metric@xcite is an asymptotic flat and spherically symmetric solution of einstein
maxwell equations in the presence of a radial flow of a null - type fluid . for our purposes , it is more convenient to consider the charged vaidya metric in double - null coordinates , as proposed by waugh and lake in @xcite . in double - null coordinates @xmath3 , the charged vaidya metric reads @xmath4 where @xmath5 stands for the metric on the unity sphere @xmath6 .
the einstein - maxwell equations imply that the functions @xmath7 and @xmath8 and @xmath9 obey @xmath10 and @xmath11 where @xmath12 , @xmath1 and @xmath2 are arbitrary integration functions . with the choice of @xmath13 constant , one can interpret @xmath1 and @xmath2 , respectively , as the total mass and the total charge of the solution . both functions are only required to be monotonic , but a further condition must be imposed due to the weak energy condition in the case where both @xmath14 and @xmath15 are non null @xcite . the solutions of eq .
( [ r ] ) corresponds to the congruence of ingoing and outgoing null geodesics of the metric ( [ metric ] ) .
we adopt here the semi - analytical method proposed in @xcite to construct such geodesics . in fig .
[ fig1 ] we present as an example the null geodesics obtained for a case with @xmath16 , that is , a reissner - nordstrm ( rn ) black hole , in this parametrization .
[ fig3 ] presents the same solutions for a time dependent background . for an interesting analysis of the time dependent horizons in the vaidya metric ,
see @xcite . , but this time for a time dependent charged black hole , with the choices @xmath17 , @xmath18 and @xmath19 + q_1 $ ] , with @xmath20 , @xmath21 , @xmath22 and @xmath23 . ] , but this time for a time dependent charged black hole , with the choices @xmath17 , @xmath18 and @xmath19 + q_1 $ ] , with @xmath20 , @xmath21 , @xmath22 and @xmath23 . ]
the evolution of a massless scalar field in this background is given by the klein - gordon equation : @xmath24 decomposing the field @xmath25 in spherical harmonics @xmath26 and using eqs .
( [ f ] ) and ( [ r ] ) , we find @xmath27 with the potential @xmath28 given by @xmath29 for more details on the numerical algorithm , initial conditions and wave extraction process for eq .
( [ onda ] ) we refer the reader to refs .
we first tested our code for static schwarzschild and reissner - nordstrm configurations , with very good agreement with values found in the literature @xcite . in fig .
( [ wrxwi_01 ] ) we present the time variation in the @xmath30 plane of the real and imaginary parts of the qnm of the scalar perturbation with @xmath31 for a rn black hole with constant mass and a time dependent charge function @xmath2 .
the frequencies change in a similar way with the time variation of the mass or the charge , but are typically more sensitive to the charge . the specific case depicted in fig .
( [ wrxwi_01 ] ) corresponds to a charge variation where @xmath32 goes from 0.7 to 0.8 , where we see how the variation goes from one stationary state to another through a non - adiabatic trajectory .
the relevant timescale here for characterizing mass or charge changes which are adiabatic or not is essentially the same one of the @xmath33 case studied in @xcite : qnm modes will relax and follow an adiabatic trajectory in the @xmath30 plane whenever @xmath34
. plane , showing the transition between stationary states for the @xmath31 scalar perturbation of a black hole with constant mass @xmath35 and @xmath19 + q1 $ ] , with @xmath36 , @xmath37 , @xmath38 and different values of @xmath39 .
right plot : the minima of @xmath40 obtained in the transition shown in the left plot for different values of @xmath39 , to quantify the non - adiabatic behavior.,title="fig : " ] plane , showing the transition between stationary states for the @xmath31 scalar perturbation of a black hole with constant mass @xmath35 and @xmath19 + q1 $ ] , with @xmath36 , @xmath37 , @xmath38 and different values of @xmath39 .
right plot : the minima of @xmath40 obtained in the transition shown in the left plot for different values of @xmath39 , to quantify the non - adiabatic behavior.,title="fig : " ] we would like also to use our formalism and numerical setup to probe the formation of a naked singularity in the spacetime . in order to do so and avoid numerical problems as @xmath41 , we choose @xmath42 and @xmath43 + q_1 $ ] , taking @xmath44 .
this implies that the rn black hole will be ( rigorously ) extremal only for @xmath45 , but allows us to probe configurations very near this end state at finite times . by varying the parameter @xmath39
, we can control the speed with which the black approaches the extremal state .
the results are shown in fig .
( [ wrxwi_04 ] ) . in the left plot
we show how the results depend on the manner in which @xmath46 .
we remind here that an extremal rn black hole is not a naked singularity , and its qnms are well defined as the appropriated limit of the non - extremal case @xcite . the results in the right plot of fig .
[ wrxwi_04 ] show what happens when we try to artificially create a naked singularity .
we chose configurations with constant charge @xmath47 and time dependent mass functions with final mass @xmath48 approaching and smaller than @xmath47 .
the results for @xmath49 are shown in the right plot of fig .
[ wrxwi_04 ] and are truncated when @xmath50 since the simulation crashes at this point , as expected . ) , but this time with @xmath51 so that the rn black hole becomes extreme as @xmath45 .
right plot : @xmath52 of a scalar perturbation with @xmath31 obtained for `` artificial '' background configurations with @xmath48 approaching and smaller than @xmath47 ; @xmath53 presents a similar behavior.,title="fig : " ] ) , but this time with @xmath51 so that the rn black hole becomes extreme as @xmath45 .
right plot : @xmath52 of a scalar perturbation with @xmath31 obtained for `` artificial '' background configurations with @xmath48 approaching and smaller than @xmath47 ; @xmath53 presents a similar behavior.,title="fig : " ]
our results confirm , for the charged case , the same non - stationary behavior corresponding to some sort of inertia for qnm frequencies identified in @xcite . in particular , for situations with @xmath54 , the qnm frequencies will not follow in the @xmath30 plane a trajectory corresponding to the instantaneous frequency associated to a rn black hole of a given @xmath32 ratio , and a certain inertial behavior for the qnm can be identified .
moreover , we see that the faster the change in @xmath1 or @xmath2 is , the bigger the deviation from the stationary regime will be ( see fig .
these points are now under investigation , and a more complete analysis will appear soon @xcite . as for the @xmath33 case considered in @xcite , an interesting extension of this work would be analysis of the highly damped qnm , the so - called overtones . since for such overtones
the ratio @xmath55 is always larger than for the fundamental ( @xmath56 ) qnm considered here , including , for sufficient large @xmath57 , cases for which @xmath58 , it would be interesting to check if the non - stationary inertial behavior could somehow be attenuated for @xmath59 .
we notice that the numerical analysis presented here can not be extended directly to the @xmath59 case since one can not identify the @xmath59 frequencies with sufficient accuracy .
we believe this could be attained , in principle , by means of the wkb approximation .
this work was supported by cnpq , fapesp and the max planck society .
it is a pleasure to thank rodrigo panosso macedo and luciano rezzolla for useful discussions .
99 berti e 2004 `` black hole quasinormal modes : hints of quantum gravity ? '' _ proc . of int .
workshop on dynamics and thermodynamics of black holes and naked singularities _
( milan ) p 145 ( _ preprint _
gr - qc/0411025v1 ) | macroscopically charged black holes not expected astrophysical scenarios obtain information on structure behavior einstein equations general relativity used charged vaidya metric model time - dependent black holes electric charge behavior quasinormal modes in non - stationary background charged vaidya metric@xcite asymptotic flat spherically symmetric solution einstein maxwell equations radial flow null - type fluid convenient consider metric in double - null coordinates proposed waugh lake @xcite double - null coordinates @xmath3 metric reads @xmath4 @xmath5 metric unity sphere @xmath6 einstein - maxwell equations imply functions @xmath7 @xmath8 @xmath9 obey @xmath10 @xmath11 arbitrary integration functions @xmath13 constant @xmath1 as total mass charge of solution functions required monotonic further condition weak energy condition @xmath14 @xmath15 non null @xcite solutions of ingoing outgoing null geodesics of metric adopt semi - analytical method in @xcite construct geodesics fig .fig1 null geodesics case @xmath16 reissner - nordstrm rn ) black hole parametrization fig3 same solutions time dependent background analysis time dependent horizons vaidya metric see @xcite time dependent charged black hole choices @xmath17 @xmath18 @xmath19 + q_1 $ @xmath20 @xmath23 time dependent charged black hole choices @xmath17 @xmath19 + q_1 $ evolution of massless scalar field klein - gordon equation @xmath24 decomposing field @xmath25 in spherical harmonics @xmath26 using eqs find @xmath27 potential @xmath28 @xmath29 details numerical algorithm initial conditions wave extraction process eq refer reader to refs tested code for static schwarzschild reissner - nordstrm configurations agreement with values literature @xcite fig . time variation in @xmath30 plane real imaginary parts qnm scalar perturbation with @xmath31 rn black hole constant mass time dependent charge function @xmath2 frequencies change with time variation mass charge more sensitive to charge specific case fig .wrxwi_01 corresponds charge variation @xmath32 0.7 to 0.8 variation stationary state non - adiabatic trajectory relevant timescale mass charge changes adiabatic same @xmath33 case @xcite qnm modes relax follow adiabatic trajectory @xmath30 plane @xmath34 plane transition between stationary states @xmath31 scalar perturbation black hole constant mass @xmath35 @xmath19 + q1 $ ] @xmath36 values @xmath39 plot minima @xmath40 transition values @xmath39 quantify non - adiabatic behavior plane transition between stationary states @xmath31 perturbation black hole constant mass @xmath35 @xmath19 + q1 $ @xmath36 values @xmath39 right plot minima @xmath40 transition values @xmath39 quantify non - adiabatic behavior formalism numerical setup probe formation naked singularity spacetime avoid numerical problems choose @xmath42 @xmath43 + q_1 $ ] taking @xmath44 implies rn black hole extremal for @xmath45 allows probe configurations near end state finite times varying parameter @xmath39 control speed black approaches extremal state results shown in fig .wrxwi_04 ] left plot results depend on @xmath46 extremal rn black hole not naked singularity qnms defined appropriated limit non - extremal case @xcite results right plot show artificially create naked singularity chose configurations constant charge @xmath47 time dependent mass functions final mass @xmath48 approaching smaller than @xmath47 results for @xmath49 right plot truncated when @xmath50 crashes with @xmath51 rn black hole extreme as @xmath45 right plot @xmath52 scalar perturbation with @xmath31 artificial background configurations @xmath48 approaching smaller @xmath47 @xmath53 similar behavior @xmath51 rn black hole extreme @xmath45 right plot @xmath52 scalar perturbation with @xmath31 @xmath48 approaching smaller than @xmath47 @xmath53 similar behavior results confirm charged case non - stationary behavior inertia for qnm frequencies in @xcite situations with @xmath54 qnm frequencies not follow @xmath30 plane trajectory instantaneous frequency rn black hole @xmath32 ratio inertial behavior for qnm identifiedfaster change in @xmath1 or @xmath2 bigger deviation from stationary regime see fig . points under investigation complete analysis soon @xcite @xmath33 case extension analysis of highly damped qnm overtones ratio @xmath55 larger than fundamental @xmath56 ) qnm large @xmath57 cases check if non - stationary inertial behavior attenuated for @xmath59 numerical analysis not extended to @xmath59 case identify frequencies accuracy wkb approximation work supported by cnpq fapesp max planck society thank rodrigo panosso macedo luciano rezzolla for discussions 99 berti e 2004 black hole quasinormal modes hints of quantum gravity proc int . workshop on dynamics thermodynamics of black holes naked singularities milan ) p 145 preprint gr - qc/0411025v1 ) | 0.318257 | 2,095 | 1,111 | , Saving $0.1 in GPT-4. |
67,591 | a hermite type formula is introduced and used to study the zeta function over the real and complex @xmath0-projective space .
this approach allows to compute the residua at the poles and the value at the origin as well as the value of the derivative at the origin , that gives the regularized determinant of the associated laplacian operator . | zeta functions on the sphere ( and in general on a closed riemannian manifold ) were first introduced by minakshisundaram and peijel as extensions of the classical riemann zeta function @xcite .
an analytic definition , by a mellin transform of the trace of the heat operator associated to the laplacian in the standard metric , shows how the residua at the poles are given by the coefficients in the asymptotic expansion of the trace of the heat operator @xcite .
such an expansion can be obtained in a large number of cases using global analysis @xcite @xcite @xcite @xcite , and in particular the first coefficients in the case of the laplacian on a riemannian manifold can be computed using local invariants associated to the curvature tensor @xcite .
the constant term , that corresponds to the value of the zeta function at the origin , is important in physics @xcite and in conformal theory @xcite , being associated to the conformal anomaly . on the other hand ,
the derivative of the zeta function at the origin gives the atiyah regularized determinant of the laplacian @xcite .
early approaches to calculate these quantities give explicit results for the 2 sphere @xcite @xcite , while more recently an explicit formula for the residua has been obtained in @xcite for the @xmath0-sphere , using a result that allows to write a dirichlet series as a sum of classical hurwitz zeta functions @xcite .
this method fails to compute the derivative , but an alternative one is provided by @xcite , where a factorization theorem for zeta regularized products is introduced and explicit results for the low dimensional spheres are given .
the aim of this paper is to show how these zeta functions can be treated very easily by classical methods using a hermite type formula , in exactly the same way as for the one dimensional case of the riemann ( hurwitz ) zeta function .
this approach allows us to deal not only with the zeta function on the spheres , but also with the zeta function on the real and complex projective spaces .
for all these cases , we localize the poles and give explicit formulas for the residua .
we show that the origin @xmath1 is a regular point , and compute the value of the zeta function together with the value of its first derivative at @xmath1 .
the main feature of this approach is that it can be used more generally to deal with the zeta function of any operator whose eigenvalues are explicitly known with their multiplicity .
there is work in progress where further cases are under consideration .
consider the function @xmath2 of the complex variable @xmath3 for @xmath4 , where @xmath5 is a polynomial of degree @xmath6 , and @xmath7 real constants with @xmath8 , @xmath9 .
introduce the function of the complex variable @xmath10 : @xmath11 then we have the following : for @xmath4 , @xmath12 where @xmath13 is an integral function of @xmath3 . [ l1 ] the infinite sum , is actually a finite sum of infinite sums of the following type , that can be treated by using the plana theorem when @xmath4 , @xmath14 @xmath15 @xmath16^{-s/2}[(b+c)^2+y^2]^{-s/2}\times\ ] ] @xmath17 \frac{dy}{e^{2\pi y}-1},\ ] ] indeed , the last integral converges for all values of @xmath3 .
furthermore , the last integral converges uniformly by standard estimates on @xmath18 , so the given formula defines an analytic function of @xmath3 .
@xmath19 from proposition [ l1 ] , it is clear that all the poles and relative residua come from the first two terms .
this expression can also be used to compute the value of @xmath10 and its derivative at @xmath1 .
spheres and real projective spaces can be treated together as follows , complex projective spaces will be considered afterwards .
the polynomial @xmath5 , with @xmath20 for the sphere @xmath21 and the projective space @xmath22 respectively ( @xmath23 ) , is : @xmath24 it is now convenient to distinguish odd and even cases .
we introduce the numbers @xmath25 , by for @xmath26 , the numbers @xmath25 are defined by the equations @xmath27^l,\ ] ] @xmath28^l.\ ] ] [ d1 ] in particular : @xmath29 , @xmath30 , @xmath31 .
.2 in this allows us to write the zeta function as a sum of some standard ones .
let @xmath32^s},\ ] ] be the zeta function in dimension @xmath33 ; then @xmath34 @xmath35 now , consider the two functions @xmath36^s},\ ] ] @xmath37^s},\ ] ] then : for @xmath38 , @xmath39 @xmath40 [ l3 ] to write simpler formulae , let us also introduce the function @xmath41^{(l - s)/2 } [ ( b+c)^2+y^2]^{-s/2}\times\ ] ] @xmath42\frac{dy}{e^{2\pi y}-1}.\ ] ] by expanding the trigonometric functions and recalling the standard integral representation of the bernoulli numbers @xmath43 , we get . ]
@xmath44 @xmath45 @xmath46 where @xmath47 denotes the integer part of the rational number @xmath48 .
we can now state the main properties of the functions @xmath49 .
the function @xmath50 has a simple pole at @xmath51 with residuum @xmath52 , while for @xmath53 , @xmath54(c+a)^m(c+b)^m+\ ] ] @xmath55 [ l4 ] proceeding as in proposition [ l1 ] , we get @xmath56 @xmath57^{1/2-s/2}[(b+c)^2+y^2]^{-s/2}\times\ ] ] @xmath58 \frac{dy}{e^{2\pi y}-1}+\ ] ] @xmath59^{-s/2}[(b+c)^2+y^2]^{-s/2}\times\ ] ] @xmath60 \frac{dy}{e^{2\pi y}-1};\ ] ] then the poles are given by the second term , while the values at the non positive integers can be computed by using the formulae introduced above .
@xmath19 the function @xmath61 has simple poles at @xmath62 , @xmath63 , with residua @xmath64 for @xmath53 , @xmath65 @xmath66 [ l5 ] for the odd case , difficulties arise when dealing with the first integral .
if we restrict ourselves to the interesting case of @xmath67 , we get @xmath68 @xmath69^{-s/2}[(b+c)^2+y^2]^{-s/2 } \sin\left[s\left(\arctan\frac{y}{c}+\arctan\frac{y}{b+c}\right)\right ] \frac{dy}{e^{2\pi y}-1},\ ] ] and we have the following two possible ways of treating it : @xmath70 when @xmath71 and where @xmath72 is the hyper geometric function , and @xmath73 when @xmath74 , where the last integral is actually convergent if @xmath75 , and can be expressed in term of a hyper geometric function : @xmath76 in particular , we use the first representation to get the analytical continuation in the negative half plane and to compute the residua at the poles , and the second one to compute the value at @xmath77 , @xmath53 , and the derivative at @xmath1 ( see section [ s4 ] ) ( both expressions are good to compute the residua ) .
@xmath19 from the previous lemmas we immediately get the following : the function @xmath78 has simple poles at @xmath79 , for @xmath80 , with residua : @xmath81 the function @xmath82 has simple poles at @xmath83 , @xmath53 , with residua @xmath84 @xmath85 [ p2 ] these should be compared with theorem 2 of @xcite , where we note that a factor @xmath86 is missing . for what concerns the value of @xmath87 ,
this can be easily computed ( using any mathematical software ) from the formulae in lemmas [ l3 ] , [ l4 ] and [ l5 ] .
explicit results for @xmath88 and @xmath89 are given in section [ s4 ] .
next , consider the complex projective spaces @xmath90 , @xmath23 .
the polynomial @xmath5 is @xmath91 ^ 2 } \prod_{i=1}^{k-1 } ( x+i)^2,\ ] ] and @xmath92 if we introduce the numbers @xmath93 by for @xmath26 , the numbers @xmath93 are defined by the equations @xmath94 [ d2 ] ( in particular : @xmath95 , @xmath96 ^ 2 $ ] ) ; then we can write [ l6 ] for @xmath97 , @xmath98 ^ 2 } \sum_{l=0}^{k-1 } a_{k , l } z_{even}(s - l,0,k,1).\ ] ] from this and lemma [ l4 ] , we immediately get [ p3 ] the function @xmath99 has simple poles at @xmath79 , for @xmath100 , with residua @xmath101 ^ 2 } a_{k , n-1}.\ ] ]
in this last section we show how to use proposition [ l1 ] to compute the regularized determinant by considering the case of @xmath102 .
next , we give explicit results for dimensions 2,3 and 4 . even if this is beyond the purposes of these notes ,
we observe that such results show that a unique formula for the determinant for a general @xmath33 is likely to exist , and could be determined using the approach outlined here .
we have @xmath102 , @xmath103 , @xmath67 and @xmath104 .
writing @xmath105 as in proposition [ l1 ] , we decompose the second term as in lemma [ l3 ] , but the third one by developing the numerator .
we get @xmath106^s}=\ ] ] @xmath107^s } + \int_1^\infty [ cx(cx+2)]^{-s}dx+ \int_1^\infty [ cx(cx+2)]^{1-s}dx+\ ] ] @xmath108^{-s/2}\sin[-s(\theta+\phi ) ] \frac{dy}{e^{2\pi y}-1}+\ ] ] @xmath109^{-s/2}\sin[(1-s)\theta - s\phi ] \frac{dy}{e^{2\pi y}-1}+\ ] ] @xmath110^{-s/2}\sin[(2-s)\theta - s\phi ] \frac{dy}{e^{2\pi y}-1},\ ] ] where @xmath111 , @xmath112 .
the derivative at @xmath1 of the first term is immediate , @xmath113^s}\right|_{s=0}=-(c+1)\log\sqrt{c(c+1)},\ ] ] while that of the second one can be computed with little effort by using the representation introduced in the proof of lemma [ l5 ] .
we get @xmath114 for the last three integrals we get @xmath115 @xmath116\}\sin\theta \frac{dy}{e^{2\pi y}-1}+\ ] ] @xmath117 @xmath118\}\sin 2\theta \frac{dy}{e^{2\pi y}-1 } + \ ] ] @xmath119 that can be expressed in terms of derivatives of the hurwitz zeta function , @xmath120 @xmath121 @xmath122 collecting and simplifying , we get the results shown below , together with the other low dimensional cases .
@xmath123 @xmath124 @xmath125 @xmath126 @xmath127 @xmath128 @xmath129 @xmath130 @xmath131 @xmath132 @xmath133 @xmath134 @xmath135 @xmath136 the values for @xmath137 agree with the ones provided by @xcite , section 4 ; the value for @xmath138 confirms the one originally given by @xcite , against that of @xcite(see also section [ s5 ] for further remarks ) ; the values for @xmath139 agrees with the one computed from the formula in the corollary of section 3 of @xcite ( see also section 3 of @xcite ) .
numerical computations were done by using maple .
to conclude some remarks are in order .
the first one concerns the one dimensional case , where we get the following relations @xmath140 @xmath141 @xmath142 also notice that , while the value of the zeta function at @xmath1 seems to depend only on the topology , the value of its derivative , and hence the regularized determinant , does not . for what concerns the first statement ,
just recall that , for any closed riemannian manifold @xmath144 of dimension @xmath145 , @xmath146 where @xmath147 is the laplacian operator in the standard metric , and @xmath148 is the coefficient of the constant term in the asymptotic expansion of the heat operator @xmath149 ( see for example @xcite ) .
in particular , compare with the value computed above for @xmath138 , where @xmath150 . our final remark , concerns the possibility of using the approach introduced to treat the case of the laplacian coupled with a constant potential .
this is an important tool to face the problem of a generic potential ( see @xcite for the one dimensional case ) . for the sake of simplicity , we restrict ourselves to the case of the 2-sphere .
the eigenvalues are then @xmath151 , and the zeta function is @xmath152^s}.\ ] ] we can compute @xmath154 @xmath155 where @xmath156 , and @xmath157 is the integral function of order 2 and zeros @xmath158 , of multiplicity @xmath159 , defined by the canonical product of genus 2 : @xmath160^{2n+1 } { { \rm e}}^{-\frac{2n+1}{n(n+1 ) } q^2}.\ ] ] the main difference with respect to the one dimensional case is due to the slower rate of increasing of the eigenvalues ; this reflects in the appearance of a further contribution coming from the singularity of @xmath161 at @xmath51 , thorough the term @xmath162 . in the following figures
are plotted the functions @xmath157 and @xmath163 , where the infinite product is approximated by the product on the first 1000 terms . | zeta functions on sphere closed riemannian manifold introduced by minakshisundaram peijel riemann zeta function @xcite analytic definition mellin transform heat operator laplacian standard metric shows residua at poles by coefficients asymptotic expansion of heat operator @xcite expansion obtained using global analysis first coefficients laplacian riemannian manifold computed using local invariants curvature tensor constant term value zeta function at origin important in physics conformal theory associated to conformal anomaly derivative of zeta function at origin gives regularized determinant of laplacian @xcite early approaches give explicit results for 2 sphere @xcite explicit formula for residua obtained in @xcite for @xmath0-sphere dirichlet series sum of classical hurwitz zeta functions method fails to compute derivative alternative by @xcite factorization theorem for zeta regularized products introduced explicit results for low dimensional spheres aim paper show zeta functions treated by classical methods using hermite type formula one dimensional case riemann hurwitz zeta functionapproach allows zeta function spheres real complex projective spaces localize poles give formulas for residua show origin @xmath1 regular point compute value zeta function first derivative at @xmath1 approach used deal zeta function operator eigenvalues known multiplicity work in progress further cases consideration consider function @xmath2 complex variable @xmath3 for @xmath4 @xmath5 polynomial of degree @xmath6 @xmath7 real constants with @xmath8 @xmath9 introduce function complex variable @xmath10 @xmath11 @xmath4 @xmath12 @xmath13 integral function of @xmath3 infinite sum finite sum of infinite sums treated plana theorem @xmath4 @xmath14 @xmath15 @xmath16^{-s/2}+c^2+y^2]^-s/2 @xmath17 last integral converges for all values @xmath3 last integral converges uniformly by standard estimates on @xmath18 formula defines analytic function of @xmath3 @xmath19 proposition poles relative residua from first two terms expression compute value @xmath10 derivative at @xmath1spheres projective spaces treated complex spaces considered afterwards polynomial @xmath5 @xmath20 sphere @xmath21 projective space @xmath22 @xmath23 is @xmath24 distinguish odd even cases introduce numbers @xmath25 defined by equations @xmath27^l @xmath28^l @xmath29 @xmath30 @xmath31 write zeta function sum standard ones @xmath32^s zeta function dimension @xmath33 @xmath34 @xmath35 consider two functions @xmath36^s} @xmath37^s} @xmath38 @xmath39 @xmath40 simpler formulae introduce function @xmath41^{(l - s)/2+c)^2+y^2]{-s/2} @xmath42 expanding trigonometric functions recalling standard integral representation bernoulli numbers @xmath43 @xmath44 @xmath45 @xmath46 @xmath47 denotes integer part rational number @xmath48 state main properties functions @xmath49 .function @xmath50 pole @xmath51 residuum @xmath52 @xmath53 @xmath54(c+a)+b)^m+ @xmath55 @xmath56 @xmath57{1/2-s/2}[(b+c)^2+y^2]} @xmath58{e^{2\pi y}-1} @xmath59^{-s/2}[(b+c)^2+y^2]{-s/2} @xmath60{dy}{e^{2\pi y}-1} poles second term values non positive integers computed formulae function @xmath61 poles @xmath62 @xmath63 residua @xmath64 @xmath53 @xmath65 @xmath66 odd case difficulties first integral@xmath67 @xmath68 @xmath69^{-s/2}[(b+c)^2+y^2]^{-s/2 } \sin\left[s\left(\arctan\frac{y}{c}+\frac{y}{b+c}\right) ] \frac{dy}{e^{2\pi y}-1} two ways @xmath70 @xmath71 @xmath72 hyper geometric function @xmath73 @xmath74 last integral convergent @xmath75 hyper geometric function @xmath76 first representation analytical continuation negative half plane compute residua poles second compute value @xmath77 @xmath53 derivative @xmath1 section s4 ] compute residua @xmath19 function @xmath78 poles @xmath79 @xmath80 residua @xmath81 @xmath82 poles @xmath83 @xmath53 residua @xmath84 @xmath85 compared with theorem 2 @xcite factor @xmath86 missing value @xmath87 computed from formulae lemmas [ l3 ] [ l4 ] [ l5 ] results @xmath88 @xmath89 section [ s4 ] consider complex projective spaces @xmath90 @xmath23 polynomial @xmath5 @xmath91 ^ 2 \prod_{i=1}^{k-1 } x+i)^2 @xmath92 introduce numbers @xmath93 @xmath26 defined equations @xmath94 [ d2 @xmath95 @xmath96 ^ 2 write [ l6 ] @xmath97 @xmath98 ^ 2 \sum_{l=0}^{k-1 }_ l z_{even}(s - l,0,k,1) [ p3 ] function @xmath99 poles @xmath79 @xmath100 residua @xmath101 ^ 2 a_{k , n-1} section proposition [ l1 ] compute regularized determinant case @xmath102 results dimensions 2,3 4 unique formula determinant @xmath33 determined approach @xmath102 @xmath103 @xmath67 @xmath104 @xmath105 proposition [ l1 ] decompose second term lemma [ l3 ] third one numerator .@xmath106^s @xmath107 cx-s}dx{1-s}dx+ @xmath108-s/2}\sin-s\theta+\phi\frac{dy}{e^{2\pi y}-1}+ @xmath109-s/2}-s)\theta - s\phi{2\pi y}-1}+ @xmath110-s/2}\sin-s)\theta - s\phi{dy{e^{2\pi y}-1} @xmath111 @xmath112 derivative @xmath1 first term immediate @xmath113{s=0}=-(c+1)\log\sqrt{c(c+1)} second one computed representation lemma @xmath114 last three integrals @xmath115 @xmath116\sin\theta \frac{dy}{e^{2\pi y}-1}+ @xmath117 @xmath118 2\theta \frac{dy}{e^{2\pi y}-1 + @xmath119 derivatives hurwitz zeta function @xmath120 @xmath121 @xmath122 results low dimensional cases@xmath123 @xmath130 @xmath133 @xmath134 @xmath136 values @xmath137 agree with @xcite section 4 value @xmath138 confirms @xcite(see section [ s5 ] values @xmath139 agrees with from formula section 3 of @xcite numerical computations maple conclude remarks order first one concerns one dimensional case relations @xmath140 @xmath141 @xmath142 value zeta function at @xmath1 depend on topology value derivative regularized determinant does not first statement recall closed riemannian manifold @xmath144 dimension @xmath145 @xmath146 @xmath147 is laplacian operator in standard metric @xmath148 coefficient of constant term in asymptotic expansion of heat operator @xmath149 @xcite compare with value for @xmath138 @xmath150 final remark possibility approach to treat case of laplacian with constant potential important tool problem generic potential see @xcite for one dimensional case restrict to case of 2-sphere eigenvalues are @xmath151 zeta function is @xmath152^s}.compute @xmath154 @xmath156 @xmath157 integral function order 2 zeros @xmath158 multiplicity @xmath159 defined canonical product genus 2 @xmath160^{2n+1 }^{-\frac{2n+1}{n(n+1 ) } q^2 main difference one dimensional case slower rate increasing eigenvalues contribution singularity @xmath161 @xmath51 term @xmath162 figures plotted functions @xmath157 @xmath163 infinite product approximated product first 1000 terms | 0.557158 | 3,352 | 1,880 | , Saving $0.1 in GPT-4. |
67,610 | i describe the ir and x - ray campaign we have undertaken to determine the nature of the faint discrete x - ray source population discovered by _
chandra _ in the galactic center .
these results will provide the input to the flamingos-2 galactic center survey ( f2gcs ) . with flamingos-2 s multi - object ir spectrograph we will obtain 1000s of ir spectra of candidate x - ray source counterparts , allowing us to efficiently identify the nature of these sources , and thus dramatically increase the number of known x - ray binaries and cvs in the milky way . | the unprecedented sensitivity and angular resolution of _ chandra _ has been utilized by wang _
et al . _ ( 2002 ; hereafter w02 ) and muno _ et al . _ ( 2003 ; hereafter m03 ) to investigate the x - ray source population of the galactic center ( gc ) .
the w02 acis - i survey of the central 0.8@xmath0@xmath12@xmath0of the gc revealed a large population of previously undiscovered discrete weak sources with x - ray luminosities of @xmath2@xmath3erg@xmath3s@xmath4 .
m03 imaged the central 40 pc@xmath5 ( at 8 kpc ) around sgr a * ( fig .
1a ) , finding an additional @xmath62300 discrete point sources down to a limiting flux of @xmath7 erg / s .
some individual sources have been identified as x - ray transients , high - mass stars , lmxbs , and cvs . however , the nature of the majority of these newly detected sources is as yet unknown .
we have undertaken an ir and x - ray observational campaign to determine the nature of the faint discrete x - ray source population discovered by _
chandra _ in the gc .
data obtained to date includes a deep _
chandra _ survey of the nuclear bulge ; deep , high resolution ir imaging from vlt / isaac , ctio / ispi , and the ukidss galactic plane survey ; and ir spectroscopy from vlt / isaac and irtf / spex . by cross - correlating the x - ray imaging with our ir surveys ,
we initially identify candidate counterparts to the x - ray sources via astrometry . using a detailed ir extinction map
, we are deriving magnitudes and colours for all the candidates .
having thus established a target list , we will use the multi - object ir spectrograph flamingos-2 on gemini south to carry out a spectroscopic survey of the candidates , to search for the emission lines which are a hallmark of accreting binaries . by determining the nature of these sources
, this flamingos-2 galactic center survey will have a dramatic impact on our knowledge of the galactic accreting binary population .
we have cross - correlated the source catalog derived from the w02 _ chandra _ survey with @xmath8images of 26 selected 2.5@xmath5 regions obtained with isaac on the vlt to identify candidate ir counterparts to the x - ray sources @xcite .
ir spectroscopy to search for accretion signatures will be required for definitive identifications . using this technique
, we were able to conclusively identify the ir counterpart to `` edd-1 '' , one of these newly discovered low - luminosity x - ray sources @xcite .
cross - correlation of our ctio ispi @xmath8image ( fig .
1b ) of the central 10 pc@xmath5 of the gc with the deep sgr a * image of m03 is underway ; this will produce a large number of ir candidate counterparts to the x - ray sources .
due to the extremely high stellar density in the nuclear bulge , many of these astrometric `` matches '' are likely to be chance superpositions . with thousands of candidate counterparts , traditional long - slit single - target spectroscopy
would be a prohibitively inefficient method by which to identify true counterparts .
thus we will need to follow - up with multi - object ir spectroscopy to find the true matches : this is the work which will be performed with the flamingos-2 galactic center survey ( f2gcs ) .
[ fig : sub : a ] [ fig : sub : b ]
we find that the ir extinction in the gc can vary on scales as small as 5 ( 0.2 - 0.6 pc at 8 kpc ; @xcite ) . some areas show little evidence of this `` granularity '' , while others are highly structured .
to obtain reddening - corrected stellar photometry , a local value for the @xmath8extinction ( on scales @xmath9 ) must be measured and applied .
the relationship of extinction to wavelength in the ir is a power law with slope @xmath10 @xcite .
the `` canonical '' value for @xmath10 is @xmath62 ( _ e.g . _
@xcite ) . in contrast , for the gc we find a mean value of @xmath11 ; and furthermore , along any given line of sight to the gc @xmath10 varies substantially , ranging from @xmath61.83.6 @xcite .
thus we find that the `` universal '' ir extinction law is _ not _ universal in the gc !
[ fig : sub : a ] [ fig : sub : b ] the united kingdom infrared deep sky survey ( ukidss ) is an imaging survey covering 7500@xmath0@xmath5 of the northern sky using a wide - field ir camera ( wfcam ) on ukirt . the galactic plane survey ( gps ) component of ukidss covers the nuclear bulge ( -2@xmath0@xmath1215@xmath0 , @xmath132@xmath0 ) to a depth of @xmath8@xmath618 with a resolution of @xmath60.8 @xcite .
we have obtained an additional 600 ksec of _ chandra _ observations of the central degree of the gc ; cross - correlation of the resultant x - ray source catalog , combined with the earlier w02 and m03 datasets , with the ukidss gps imaging of the gc ( fig .
2a ) will provide additional targets for the f2gcs . with the ukidss gps data we can also map the 1 - 2 @xmath14 m `` granularity '' in the gc ( fig .
2b ) and compare it to the spitzer map of the same region @xcite , to search for correlations in the dust and gas structure .
bandyopadhyay , r.m . , _
et al . _ , 2005 , mnras 364 , 1195 gosling , a.j . ,
blundell , k.m . , & bandyopadhyay , r.m . , 2006 ,
apj 640l , 171 gosling , a.j . ,
bandyopadhyay , r.m . , & blundell , k.m . , 2008 , mnras ,
_ in prep _
lucas , p.w . , _
et al . _ , 2007 , mnras , _ submitted _ martin , p.g . &
whittet , d.c.b . , 1990 , apj 357 , 113 mikles , v.m .
_ , 2006 , apj 651 , 408 nishiyama , s. , _ et al .
_ , 2006 , apj 638 , 839 muno , m.p . , _
et al . _ , 2003 ,
apj 589 , 225 rieke , g.h . & lebofsky , m.j . , 1985 ,
apj 288 , 618 stolovy , s. , _ et al .
_ , 2006 , jphcs 54 , 176 wang , q.d . , gotthelf , e.v . , & lang , c.c . , 2002 ,
nature 415 , 148 | unprecedented sensitivity angular resolution of chandra utilized by wang muno investigate x - ray source population galactic center w02 survey central 0.8 revealed population weak sources with x - ray luminosities @xmath2 m03 imaged central 40 pc@xmath5 around sgr fig additional @xmath62300 discrete sources flux @xmath7 erg s sources identified as x - ray transients high - mass stars lmxbs cvs nature majority unknown undertaken ir x - ray observational campaign to determine nature of discrete x - ray source population by chandra in gc data includes deep _ chandra _ survey nuclear bulge high resolution ir imaging from vlt / isaac ctio / ispi ukidss galactic plane survey ir spectroscopy from vlt / isaac irtf / spex cross - correlating x - ray imaging with ir surveys candidate counterparts via astrometry detailed ir extinction map magnitudes colours for candidates target multi - object ir spectrograph flamingos-2 on gemini south spectroscopic survey candidates for emission lines accreting binariesdetermining nature sources flamingos-2 galactic center survey galactic accreting binary population cross - correlated source catalog w02 _ chandra _ survey with @xmath8images of 26 2.5@xmath5 regions isaac identify ir counterparts to x - ray sources @xcite ir spectroscopy accretion signatures required for definitive identifications ir counterpart to edd-1 '' discovered low - luminosity x - ray @xcite cross - correlation ctio ispi @xmath8image central 10 pc@xmath5 with deep sgr a * image m03 underway large number ir candidate counterparts to x - ray sources high stellar density nuclear bulge astrometric matches likely chance superpositions traditional single - target spectroscopy inefficient identify counterparts need follow - up with multi - object ir spectroscopy find true matches work performed with flamingos-2 galactic center survey ir extinction in gc on scales 0.2 - 0.6 pc at 8 kpc some areas little others highly structured obtain reddening - corrected stellar photometry local value for @xmath8extinction must measured appliedrelationship extinction to wavelength in ir is power law with slope @xmath10 @xcite canonical value for @xmath10 is @xmath62 for gc mean value @xmath11 @xmath10 varies from @xmath61.83.6 @xcite universal ir extinction law not universal in gc united kingdom infrared deep sky survey imaging survey covering 7500@xmath0@xmath5 northern sky wide - field ir camera on galactic plane survey covers nuclear bulge ( -2@xmath0@xmath1215 @xmath132 to depth @xmath8@xmath618 with resolution @xmath60.8 @xcite obtained additional 600 ksec of chandra observations of central degree gc cross - correlation of x - ray source catalog with w02 and m03 datasets ukidss gps imaging additional targets for f2gcs gps data map 1 - 2 @xmath14 m granularity in gc compare to spitzer map region @xcite correlations in dust and gas structure bandyopadhyay , r.m. et al . 2005 mnras 364 , 1195 gosling blundell2006 apj 640l 171 gosling bandyopadhyay r.m blundell k.m 2008 mnras prep lucas p.w et al 2007 mnras p.g whittet d.c.b 1990 apj 357 113 mikles v.m 2006 apj 651 408 nishiyama al 2006 apj 638 839 muno m.p 2003 apj 589 , 225 rieke g.h lebofsky m.j 1985 apj 288 618 stolovy al 2006 jphcs 54 176 wang gotthelf e.v lang c.c 2002 nature 415 , 148 | 0.267012 | 1,618 | 846 | , Saving $0.0 in GPT-4. |
67,613 | we consider secular perturbations of nearly keplerian two - body motion under a perturbing potential that can be approximated to sufficient accuracy by expanding it to second order in the coordinates . after averaging over time to obtain the secular hamiltonian
, we use angular momentum and eccentricity vectors as elements .
the method of variation of constants then leads to a set of equations of motion that are simple and regular , thus allowing efficient numerical integration .
some possible applications are briefly described .
[ firstpage ] celestial mechanics binaries : general comets : general | if a binary star , or a planet revolving around a star , is perturbed by one or more distant relatively slowly moving masses the perturbation can be approximated by a tidal field .
fast computation of secular perturbations in such a potential is thus desirable .
potential applications include the kozai resonance , motion of a well isolated binary in a star cluster and cometary motion under the galactic tidal field .
several authors have considered the computation of cometary motion perturbations due to the galactic tidal field e.g. @xcite , @xcite , @xcite .
more recently @xcite suggested the use of the time averaged secular perturbing function and the lagrangian equations for computing the evolution of orbital elements of a comet . while @xcite compared various ways to compute the evolution of orbital elements ( using keplerian or delaunay elements ) .
@xcite used the so called vectorial elements , i.e. the areal velocity vector and the eccentricity vector .
however , the perturbing potential considered in all these studies was a simplified one , in most cases just the term causing the force perpendicular to the galactic plane was included .
in this paper we consider a general quadratic potential . for more mathematical generality
we include the possibility of a constant force ( although such a term does not appear in a tidal field ) and also a slow rotation of the quadratic perturbing field is allowed .
we form the simple regular equations of motion using the vectorial elements .
finally it is necessary to stress that the main purpose of this paper is to introduce the simple secular equations of motion when a two - body system is perturbed by a ( at most ) quadratic perturbing potential .
consider a system in which the keplerian motion of a particle around a central mass @xmath0 is perturbed by a perturbing potential @xmath1 .
in addition , a slow rotation ( with angular velocity @xmath2 ) is assumed .
the hamiltonian may be written @xmath3 where @xmath0 is the mass of the central body , @xmath4 is the position vector with components @xmath5 , @xmath6 is the velocity vector with components @xmath7 i.e. @xmath8 ( which also can be identified as the momentum vector ) and @xmath9 is the semi - major axis of the kepler motion . the function @xmath10 is the perturbing function , assumed at most quadratic in the coordinates .
if the time dependence of @xmath10 is due to rotation of the potential with a constant angular velocity , then one can consider the system in the rotating coordinate system in which the perturbing function is a constant .
thus , using the suffix notation , one may write in place of @xmath10 @xmath11 where there is no time dependence , @xmath12 are components of a ( possible ) constant acceleration vector @xmath13 and @xmath14 is the second derivative matrix : @xmath15 the last term in ( [ perturbingfunction ] ) is the scalar product of the angular velocity vector @xmath16 of the perturbing potential and the angular momentum vector of the moving particle , here written using the levi - civita symbol @xmath17 .
since the matrix @xmath14 is symmetric it can be diagonalized by selecting a suitable coordinate system .
however , there may be situations in which this is not convenient .
therefore we consider the general non - diagonal case . in the theory presented here that does not add any substantial complications .
the auxiliary quantities needed in averaging the hamiltonian are : 1 . the angular momentum vector per unit of mass ( actually areal speed ) of the particle @xmath18 2 .
the eccentricity vector ( also known as the runge - lenz- or laplace vector ) @xmath19 or , for our purposes , the vector @xmath20 is more convenient .
3 . the mean values of the position vector components @xmath21 4 .
the time averaged coordinate products can be written @xmath22 where @xmath23 is the kronecker @xmath24 .
this result was easy to derive for @xmath25 .
the generalization was then formed by educated guess and use of computer algebra for confirmation .
the secular perturbing function @xmath26 takes the form @xmath27 if one introduces the matrixes @xmath28 and @xmath29 with the elements @xmath30 and the vector @xmath31 @xmath32 then the averaged perturbing function can be written in the simple form @xmath33 where the vectors @xmath31 , @xmath2 , and matrices @xmath34 , @xmath35 are constants in any particular orbit .
the reason for the choice of the elements @xmath36 and @xmath37 was that one obtains for their components the simple poisson bracket relations @xmath38 with these formulae one gets the equations of motion ( breiter and ratajczak(2005 ) and references therein ) @xmath39 and by ( [ quadratich ] ) the partial derivatives of @xmath40 with respect to the vectorial elements are @xmath41 where we have used the vector - matrix notation . in deriving the above equations one should note that @xmath36 and @xmath37 are constants of motion under the two - body hamiltonian @xmath42 .
thus @xmath9 can be considered a ( numerical ) constant since @xmath43 .
the equations of motion have the integrals @xmath44 here we must remark that the form of the perturbing function is not unique .
this is because of the larger - than - necessary number of variables and because one may e.g add terms proportional to @xmath45 and/or to @xmath46 which would not affect the final expressions for the derivatives of the elements in ( [ jequation ] ) and ( [ eequation ] ) .
note that when the rotation term @xmath47 is included , the result for the evolution of the vectorial elements will be obtained in the rotating coordinate system .
possible applications of the vectorial element equations include : 1 . constant force , or rotating constant force .
this may be restricted to toy applications , but was included in our treatment for the sake of generality and because it is mathematically simple .
isolated binary in a star cluster . assuming the perturbing stars are distant and
move relatively slowly , one may approximate the perturbing potential by @xmath48 where @xmath49 are the positions of the perturbers relative to the centre - of - mass of the binary .
writing @xmath50 we get for the components of the matrix @xmath51 @xmath52 where the matrix elements may be time dependent , but this does not invalidate the equations presented above .
3 . kozai resonance .
this term is often used in connection of triple systems in which averaging over the outer orbit gives the perturbing potential of the form @xmath53 where @xmath54 is the mass of the third body , @xmath55 is the semi - minor axis of the outer ellipse and @xmath56 is the unit normal vector of the orbital plane of the distant body . in this case
we have @xmath57 where @xmath58 are the components of the vector @xmath56 .
cometary motion under the galactic tide . in this case
the tidal potential is usually written @xmath59 which is possible by choosing the coordinate system such that @xmath60 axis is perpendicular to the galactic plane and the @xmath61-axis points to the galactic centre .
5 . the perturbing potential due to a small disk of mass
can also be expressed easily in terms of the vectorial elements , although not in quadratic form . the secular perturbing function for this case
is often written in terms of the delaunay elements in the form @xmath62 where @xmath63 is a ( small ) constant and @xmath64 are the delaunay elements . here
@xmath65 is a constant , @xmath66 and @xmath67 , where @xmath56 is the unit direction vector of the symmetry axis of the potential .
thus @xmath68 consequently it is easy to include the effect of any such potential into the equations of motion for the vectorial elements .
note that this kind of a term could be used to approximate the effect of the planets ( averaged over orbits ) to comets that come close , but not too close , to the solar system .
we ran some numerical test computations and compared results against direct numerical integration of the equations of motion for the coordinates .
the tests used the tidal field of the galaxy in the form @xmath69 ( parameters from @xcite ) and a value for the semi - major - axis of an oort cloud comet of @xmath70au .
the experiments suggest : if one needs results for a long time and infrequent output is allowed , then one of the best numerical integrators available is the burlirsh - stoer method ( actually the code known as @xmath71 , originally written by burlirsh ) . however , this is efficient only when run with optimally long steps , which may be too long for some applications .
if one needs frequent output , then the method of choice is the implicit midpoint method .
it is very simple to implement and gives quite high precision , especially it conserves the quadratic integrals of motion ( @xcite , @xcite ) . in this case one
finds that the stepsize equal to one period ( i.e. @xmath72 years ) gives a clearly satisfactory accuracy : the plots of the vectorial elements from both computations ( secular and coordinate integration ) agree well , except for a small phase error .
however , for @xmath73au the results differ and the reliability of a secular theory becomes questionable .
we have generalized the equations of @xcite who considered the simple case of perturbing function of the form @xmath74 .
breiter , s. , ratajczak , r. 2005 .
monthly notices of the royal astronomical society 364 , 1222 - 1228 .
fouchard , m. 2004 .
monthly notices of the royal astronomical society 349 , 347 - 356 .
fouchard , m. , froeschl , c. , matese , j. j. , valsecchi , g. 2005 .
celestial mechanics and dynamical astronomy 93 , 229 - 262 .
heisler , j. , tremaine , s. 1986 .
icarus 65 , 13 - 26 .
huang , w. and leimkuhler , b. 1997 .
siam j. sci .
, vol . , 18 , # 1 . | binary star or planet perturbed by distant slowly moving masses approximated by tidal field fast computation of secular perturbations desirable applications include kozai resonance motion of isolated binary in star cluster cometary motion under galactic tidal field authors considered computation of cometary motion perturbations tidal field @xcite @xcite suggested time averaged secular perturbing function lagrangian equations for evolution of orbital elements of comet @xcite compared ways evolution elements elements @xcite used vectorial elements areal velocity vector eccentricity vector perturbing potential considered simplified force perpendicular to galactic plane included paper consider general quadratic potential constant force tidal field slow rotation of quadratic perturbing field form simple regular equations of motion using vectorial elements main purpose introduce simple secular equations of motion when two - body system perturbed by quadratic perturbing potential consider system keplerian motion of particle around central mass @xmath0 perturbed by perturbing potential @xmath1 slow rotation with angular velocity @xmath2 assumedhamiltonian written @xmath3 @xmath0 mass central body @xmath4 position vector components @xmath5 @xmath6 velocity vector components @xmath7 @xmath8 momentum vector @xmath9 semi - major axis kepler motion function @xmath10 perturbing function quadratic coordinates time dependence @xmath10 due to rotation potential constant angular velocity system rotating coordinate system perturbing function constant suffix notation write place @xmath10 @xmath11 no time dependence @xmath12 components constant acceleration vector @xmath13 @xmath14 second derivative matrix @xmath15 last term perturbingfunction scalar product angular velocity vector @xmath16 perturbing potential angular momentum vector moving particle written levi - civita symbol @xmath17 matrix @xmath14 symmetric diagonalized suitable coordinate system situations not convenient consider general non - diagonal case substantial complications auxiliary quantities averaging hamiltonian angular momentum vector per unit mass areal speed particle @xmath18 2 eccentricity vector runge - lenz- laplace vector @xmath19 @xmath20 3 mean values position vector components @xmath21 4time averaged coordinate products written @xmath22 @xmath23 kronecker @xmath24 result easy for @xmath25 generalization formed by educated guess computer algebra secular perturbing function @xmath26 form @xmath27 matrixes @xmath28 @xmath29 elements @xmath30 vector @xmath31 @xmath32 averaged perturbing function simple form @xmath33 vectors @xmath31 @xmath2 matrices @xmath34 @xmath35 constants in orbit choice elements @xmath36 @xmath37 obtains simple poisson bracket relations @xmath38 equations of motion @xmath39 partial derivatives of @xmath40 vectorial elements @xmath41 used vector - matrix notation @xmath36 @xmath37 constants of motion under two - body hamiltonian @xmath42 @xmath9 numerical constant since @xmath43 equations of motion integrals @xmath44 form perturbing function not unique larger - - necessary number of variables add terms proportional to @xmath45 @xmath46 affect final expressions derivatives elements rotation term @xmath47 included result evolution of vectorial elements obtained in rotating coordinate system possible applications vectorial element equations includeconstant force rotating force restricted to toy applications included treatment generality mathematically simple isolated binary in star cluster perturbing stars distant move slowly perturbing potential by @xmath48 @xmath49 positions perturbers relative to centre mass binary @xmath50 components of matrix @xmath51 @xmath52 matrix elements time dependent invalidate equations kozai resonance used triple systems averaging over outer orbit perturbing potential @xmath53 @xmath54 mass third body @xmath55 semi - minor axis outer ellipse @xmath56 unit normal vector of orbital plane distant body @xmath57 @xmath58 components of vector @xmath56 cometary motion under galactic tide tidal potential written @xmath59 coordinate system @xmath60 axis perpendicular to galactic plane @xmath61-axis galactic centre perturbing potential due small disk of mass expressed vectorial elements not quadratic form secular perturbing function written delaunay elements @xmath62 @xmath63 constant @xmath64 delaunay elements @xmath65 constant @xmath66 @xmath67 @xmath56 unit direction vector of symmetry axis potential@xmath68 easy include effect potential into equations motion for vectorial elements term effect of planets to comets close to solar system ran numerical test computations compared results against direct numerical integration equations motion for coordinates tests used tidal field galaxy form @xmath69 parameters from @xcite value semi - major - axis of oort cloud comet @xmath70au experiments suggest results long time infrequent output best numerical burlirsh - stoer method code @xmath71 efficient with long steps frequent output method implicit midpoint method simple implement high precision conserves quadratic integrals of motion @xcite stepsize equal to one period @xmath72 years ) gives satisfactory accuracy plots vectorial elements computations agree well except small phase error for @xmath73au results differ reliability secular theory questionable generalized equations of @xcite considered case of perturbing function of form @xmath74 breiter , s ratajczak , r. 2005 monthly notices royal astronomical society 364 , 1222 - 1228 fouchard , m. 2004 monthly notices society 349 , 347 - 356 fouchardfroeschl matese valsecchi 2005 celestial mechanics dynamical astronomy 93 229 - 262 heisler tremaine 1986 icarus 65 13 - 26 huang leimkuhler 1997 siam 18 # 1 | 0.352913 | 2,318 | 1,216 | , Saving $0.1 in GPT-4. |
67,623 | vertex detectors at future hadron colliders will need to cope with large particle fluences .
diamond is a particularly radiation hard material and exhibits further properties that makes it an attractive material for such detectors . within the rd42 collaboration
several chemical vapor deposition diamond samples are being studied in the form of strip and pixel detectors . while the quality of the poly - crystalline diamond samples is constantly increasing and the feasibility of producing wafers has been demonstrated , recently a single - crystal diamond pixel detector has been assembled and characterized in a 100 gev particle beam at cern .
results on performance , detection efficiency , spatial resolution and charge collection are reported here together with the latest radiation damage studies .
[ [ fn1]for the rd42 collaboration vertex detectors , radiation hard sensors , lhc upgrade , atlas pixel detector | while the large hadron collider is scheduled to start collecting data this year , the design luminosity of the planned super large hadron collider ( slhc ) of @xmath0@xmath1s@xmath2 poses a serious challenge for future vertex detectors close to the interaction region , in particular due to the harsh radiation environment . in slhc scenarios
the total expected fluence at a radius of about 5 cm will exceed @xmath3particles/@xmath4 .
a number of studies are currently being performed to find solutions for detectors that can operate in such environments .
chemical vapor deposition diamond , in either single - crystal ( sccvd ) or polycrystalline ( pcvd ) form , is one material which is being considered for such detectors . besides being radiation tolerant ,
diamond is also an attractive material , since it is an excellent thermal conductor and it exhibits no leakage current , fast collection times and a small input capacitance .
recently pcvd diamond material has become available in the form of wafers .
figure [ fig : fig1 ] shows two such diamond wafers with as - grown collection distances of 315@xmath5 m and 310@xmath5 m respectively , produced by element six @xcite .
the collection distance is a measure of the quality of material and routinely exceeds 300@xmath5 m . the growth side of two full 12 cm diameter wafers metallised with gold contacts 1 cm apart.,title="fig : " ] the growth side of two full 12 cm diameter wafers metallised with gold contacts 1 cm apart.,title="fig : " ] after characterizing four full wafers , all indications are that element six can reproducible grow and process high - quality detector - grade cvd diamond material .
advances in the production of sccvd diamonds have lead to the production of sensors with sizes larger than 1@xmath4 and thicknesses of up to 700@xmath5 m .
the production process is still in its development and the sensors often come in irregular shapes .
sccvd detector material is superior in terms of charge collection and homogeneity than poly - crystalline diamond sensors .
several diamond pixel detectors have been constructed within the rd42 collaboration , developing and characterizing detectors , in cooperation with the fraunhofer institute for reliability and microintegration ( izm ) @xcite , performing fine - pitch flip - chip bump bonding .
single - chip devices and also multi - chip module detectors have been assembled and tested in particle beams .
testbeam results for a sccvd pixel detector @xcite and a full module pcvd pixel detector @xcite are presented in this paper , together with radiation tolerance studies @xcite .
two diamond pixel detectors have been assembled with standard atlas pixel electronics , calibrated in the laboratory and characterized in high - energy particle beams . for assembly
, both sides are metallised with ti / w .
one side is segmented into a @xmath6m@xmath7 pixel pattern to match the pattern of the atlas pixel front - end chip ( fe - i3 ) .
solder ( pbsn ) bumping and flip - chipping technology at izm has been employed to mate both parts .
figure [ fig : fig2 ] shows the two assemblies , a single - chip detector , with an irregularly shaped @xmath8 m thick single - crystal diamond sensor ( left ) and a @xmath9 m thick full - size , @xmath10m@xmath7 polycrystalline module with 16 front - end chips ( right ) . a sccvd single - chip ( left ) and a full - size pcvd diamond module connected with atlas pixel front - end chips through standard flip - chip bump - bonding.,title="fig : " ] a sccvd single - chip ( left ) and a full - size pcvd diamond module connected with atlas pixel front - end chips through standard flip - chip bump - bonding.,title="fig : " ] the atlas pixel electronics provides zero suppression in the readout with a typical threshold level of 3000 - 4000 electrons and a noise of 160 electrons when used with silicon detectors . as diamond sensors are basically free of leakage currents and have lower input capacitance , the sensor was operated at thresholds of about 1700 electrons with a noise level of about 130 electrons , as determined from fitting the threshold curves of the pixels ( s - curves ) .
analog information is obtained via the in - pixel measurement of time - over - threshold ( tot ) with an approximate resolution of 8 bit .
the tot calibration is obtained for every pixel individually , by fitting the tot response to input pulses injected over a capacitance of 41.4ff .
the pcvd module has been tested in a 6 gev electron beam at desy .
the measurements were affected by multiple scattering as reported in @xcite . with a bias voltage of 800v a noise of 130 electrons at a threshold of @xmath11 electrons were measured . more than @xmath12 of the electronics channels were operational .
the sccvd single - chip module was tested in a 100 gev pion test beam at cern , using the bonn - atlas telescope @xcite as a reference for tracking and efficiency measurements .
it consists of double - sided silicon microstrip detectors with @xmath13 m pitch . only tracks with a single hit in each telescope plane
are selected .
the extrapolated spatial uncertainty at the device under test ( dut ) is @xmath14 m . unlike in silicon
the electric field inside good sccvd diamond is fairly uniform and thus the spatial response of the detector was uniform for bias voltages larger than 0.25 v/@xmath5 m . due to the faster charge collection
smaller cluster sizes are observed at increasing bias voltages .
the measured most probable charge value ( mpv ) in the sensor at 400 v is 13.1 ke compatible with the expected value using the tot charge calibration outlined in @xcite .
the measured mpv increases with the applied bias voltage and saturate above 100 v , indicating that the charge collection is complete above these values . for full charge collection
single and double hit clusters have roughly equal sharing .
the overall efficiency to find a hit near a track point extrapolated on the plane of the dut is found to be @xmath15 .
charge collection and space point reconstruction in pcvd are affected by the grain structure and the trapping of electric charges at their boundaries which may cause horizontal polarization fields . in
@xcite it was observed that the distribution of residuals across the sensor is not statistically distributed but exhibits regions with systematically positive or negative residuals on a scale smaller than @xmath16 m . to quantify this apparent clustering , the linear correlation coefficient between the residuals of all track pairs has been determined . since each selected event has exactly one reconstructed track , pairs are formed considering tracks of two different events .
the average correlation coefficients are determined in bins of the distance between the two tracks of a pair .
an empirical function , a sum of a sinusoidal modulation with the periodicity of the pixel pitch with a falling exponential that accounts for the residual shifts introduced by the grain structure , was found to provide a good description of the data . in
@xcite the data for a low - quality pcvd diamond were fit , resulting in a correlation length ( the decay factor in the exponential ) of @xmath17 m . here , we repeat this measurement for the single - chip detector . since there are no grain boundaries the device exhibits optimal charge collection and a correlation length compatible with @xmath18 m
is observed @xcite , indicating the absence of horizontal polarization fields .
this is illustrated in figure [ fig : fig3 ] .
correlation coefficient as a function of the track distance between the tracks in two events .
the lines are fits to the data in @xcite and @xcite with a functional form of an exponential term plus a damped sinusoidal , describing the pixel pitch of @xmath19 m . ]
the spatial resolution is shown in figure [ fig : fig4 ] , where the difference between the track position predicted by the telescope on the plane of the dut ( the sccvd sensor ) and the reconstructed hit of the dut device is shown for both directions of the pixel
. spatial resolution of the sccvd device measured with respect to the reference telescope , using digital information only ( light shaded histograms ) and analog information via the @xmath20-algorithm ( dark histograms ) .
bias voltage was 200v .
plotted is the difference between the telescope space point and the measured space point .
( a ) resolution in the 50@xmath5 m direction , ( b ) resolution in the 400@xmath5 m direction . ] for the hit reconstruction first only the digital information is used ( light gray distribution ) , i.e. the pixel with the largest charge signal above threshold collected in a cone around the extrapolated track position is taken as the hit pixel and its center is assumed to be the reconstructed position .
the expected digital resolutions of pixel pitch divided by @xmath21 in both directions , i.e. 14.4 @xmath5 m and 115.5 @xmath5 m , respectively , folded by the resolution of the track extrapolation of about 5 @xmath5 m , is observed .
position reconstruction which exploits the analog information available via the tot readout is done by using the @xmath20-algorithm @xcite .
the corresponding residual distribution is shown by the dark histograms in figure [ fig : fig4 ] .
it shows a pronounced peak with good resolution for clusters with two pixel hits , for which the @xmath20-algorithm improves the resolution substantially . in the short pixel direction where the fraction of two hit clusters is about 40% the intrinsic resolution of the device , after quadratically subtracting
the telescope extrapolation error of 5@xmath5 m , becomes ( @xmath22)@xmath5 m for a bias voltage of 200v .
the error quoted is obtained from the fit to the residual distribution . repeating the analysis on several data samples indicate an additional systematic uncertainty in the order of 0.5@xmath5 m . in the direction of the long pixel dimension , the resolution for events with two - hit clusters ( @xmath2310% )
is an order of magnitude better than for events with single hit clusters .
the space resolution depends on the bias voltage .
the charge collection efficiency and hence the total collected charge are field dependent until full charge collection is obtained .
also the sharing of charge between pixels due to diffusion depends on the field inside the sensor .
the measured resolution shows an optimum between both effects at bias voltages of 200v @xcite for this device .
the rd42 irradiation program consists of testing each sample prepared as a detector in a cern test beam before and after irradiation .
previously , diamond samples were irradiated with reactor neutrons and charged pions .
most recently we irradiated the highest quality polycrystalline cvd diamond with 24 gev protons from cern ps to a fluence of @xmath24 p/@xmath4 .
summary of proton irradiation results for pcvd ( blue points ) and sccvd ( red points ) material at an electric field of 1v/@xmath5 m and 2v/@xmath5 m ( green square ) .
the black curve is a standard damage curve 1/ccd = 1/ccd@xmath25 + k@xmath26 . with a shift of the sccvd doses of @xmath27
p/@xmath4 the sccvd and pcvd data fall on a single curve indicating the damage due to irradiation is common to both . ]
figure [ fig : fig6 ] shows the signals observed in various pcvd and sccvd diamond sensors after a range of irradiations .
the sccvd diamond is expected to be representative of the next generation high quality polycrystalline material .
a single damage curve given by the equation ccd@xmath2 = ccd@xmath28 , with ccd@xmath25 the initial collection distance and @xmath29 the damage constant can accommodate for the different measurements shown , with one universal constant @xmath29 . in particular , also the measurements on the sccvd fall on the same curve if shifted by @xmath27p/@xmath4 .
the interpretation is that , owing to the intrinsic crystallite structure , the unirradiated pcvd material behaves as effectively having the same number of trapping centers as the sccvd material after a dose of @xmath27p/@xmath4 .
after ten years of slhc running typical fluences of @xmath30p/@xmath4 are anticipated for vertex detectors at 4 cm . from figure
[ fig : fig6 ] we see that diamond sensors still retain an average signal of @xmath232500 electrons after such fluences @xcite .
00 element six ltd .
, king s ride park , ascot , berkshire , sl5 8bp , united kingdom .
fraunhofer institut fr mikrointegration , gustav - meyer - allee 25 , berlin , germany .
m. mathes et al . , 2008 _ jinst _ * 3 * p12002 .
m. mathes , ph.d thesis , urn : nbn : de : hbz:5n-16430 .
m. barbero et al .
( rd42 collaboration ) , cern / lhcc 2008 - 005 , lhcc - rd-016 . j. treis et al.,nucl .
instr . and meth .
a 490 ( 2002 ) 112 .
t. lari et al . , nucl .
instr . and meth .
a 537 ( 2005 ) 581 .
e. belau et al.,nucl
. instr . and meth .
a 214 ( 1983 ) 253 .
w. trischuk for the rd42 collaboration , proc . of the 34@xmath31 ichep ,
to appear . | large hadron collider collecting data this year design luminosity @xmath0@xmath1s@xmath2 challenge for future vertex detectors due harsh radiation environment in slhc scenarios fluence at radius 5 cm exceed @xmath3particles/@xmath4 studies solutions for detectors in such environments chemical vapor deposition diamond in single - crystal or polycrystalline form considered for detectors radiation tolerant attractive excellent thermal conductor no leakage current fast collection times small input capacitance pcvd diamond material available in wafers figure shows two diamond wafers with collection distances of 315@xmath5 m and 310@xmath5 m produced by element six @xcite collection distance quality exceeds 300@xmath5 m growth side two 12 cm diameter wafers gold element six can grow process high - quality detector - grade cvd diamond material advances in production sccvd diamonds lead to sensors with sizes larger than 1@xmath4 thicknesses up to 700@xmath5 m production process development sensors in irregular shapes sccvd detector material superior in charge collection homogeneity than poly - crystalline diamond sensorsdiamond pixel detectors constructed rd42 collaboration with fraunhofer institute reliability microintegration @xcite fine - pitch flip - chip bump bonding single - chip multi - chip detectors assembled tested in particle beams testbeam results sccvd full module pcvd detector presented radiation tolerance studies two diamond pixel detectors assembled with atlas pixel electronics calibrated characterized high - energy particle beams sides metallised with ti / w one side segmented @xmath6m@xmath7 pixel pattern atlas pixel front - end chip solder bumping flip - chipping technology izm parts figure shows two assemblies single - chip detector shaped @xmath8 m thick single - crystal diamond sensor @xmath9 m thick size @xmath10m polycrystalline module 16 front - end chips sccvd single - chip full - size pcvd diamond module connected with atlas pixel front - chips flip - chip bump - bonding atlas pixel electronics provides zero suppression readout threshold 3000 - 4000 electrons noise 160 electrons silicon detectorsdiamond sensors free leakage lower input capacitance operated at thresholds 1700 electrons noise 130 electrons determined fitting threshold curves analog information in - pixel measurement time - over - threshold tot ) resolution 8 bit tot calibration pixel fitting response input pulses capacitance 41.4ff pcvd module tested 6 gev electron beam at desy measurements affected by multiple scattering bias voltage 800v noise 130 electrons threshold @xmath11 electrons @xmath12 electronics channels operational sccvd single - chip module tested 100 gev pion test beam at cern bonn - atlas telescope @xcite double - sided silicon microstrip detectors @xmath13 m pitch tracks single hit selected extrapolated spatial uncertainty @xmath14 m electric field sccvd diamond uniform spatial response uniform for bias voltages larger than 0.25 v/@xmath5 m faster charge collection smaller cluster sizes at increasing bias voltages measured probable charge value at 400 v 13.1 ke compatible expected tot charge calibration mpv increases with bias voltage above 100 v charge collection completefull charge collection single double hit clusters equal sharing efficiency to find hit near track point @xmath15 charge collection space point reconstruction in pcvd affected by grain structure trapping electric charges horizontal polarization fields @xcite distribution residuals across sensor not statistically distributed regions positive or negative residuals smaller than @xmath16 m linear correlation coefficient between residuals track pairs determined each selected event one reconstructed track pairs formed two events average correlation coefficients determined distance between two tracks empirical function sum sinusoidal modulation periodicity pixel pitch falling exponential residual shifts grain structure good description data @xcite data low - quality pcvd diamond fit correlation length @xmath17 m repeat measurement for single - chip detector no grain boundaries device optimal charge collection correlation length compatible with @xmath18 m absence of horizontal polarization fields illustrated in figure correlation coefficient function of track distance between tracks two events lines to data @xcite functional form exponential term damped sinusoidal pixel pitch @xmath19 mspatial resolution shown in figure difference between track position predicted by telescope reconstructed hit for both directions spatial resolution sccvd device measured reference telescope digital information light shaded histograms analog information @xmath20-algorithm dark histograms bias voltage 200v difference between telescope space point measured space point resolution in 50@xmath5 m direction 400@xmath5 m direction hit reconstruction digital information used light gray distribution pixel with largest charge signal above threshold hit pixel center reconstructed position expected digital resolutions of pixel pitch divided by @xmath21 directions 14.4 @xmath5 m and 115.5 @xmath5 m folded by resolution track extrapolation 5 @xmath5 m position reconstruction analog information @xmath20-algorithm residual distribution shown by dark histograms in figure shows peak good resolution for clusters with two pixel hits @xmath20-algorithm improves resolution short pixel direction fraction of two hit clusters 40% intrinsic resolution device telescope extrapolation error 5@xmath5 m becomes ( @xmath22)@xmath5 m for bias voltage 200v error obtained from fit residual distributionanalysis data samples systematic uncertainty 0.5@xmath5 m resolution two - hit clusters @xmath2310% better than single hit clusters space resolution depends bias voltage charge collection efficiency total collected charge field dependent until full charge collection sharing charge between pixels depends field inside sensor resolution optimum at voltages 200v @xcite rd42 irradiation program sample cern test beam before after irradiation diamond samples irradiated with reactor neutrons charged pions irradiated polycrystalline cvd diamond with 24 gev protons @xmath24 p/@xmath4 proton irradiation results for pcvd sccvd at electric field 1v/@xmath5 m 2v/@xmath5 m black curve standard damage curve 1/ccd = 1/ccd@xmath25 + k@xmath26 shift sccvd doses @xmath27 p/@xmath4 data single curve damage irradiation common figure shows signals in pcvd sccvd diamond sensors after irradiations sccvd diamond representative of next generation high quality polycrystalline materialsingle damage curve equation ccd@xmath2 = ccd@xmath28 ccd@xmath25 initial collection distance @xmath29 damage constant different measurements universal constant @xmath29 measurements sccvd same curve shifted by @xmath27p/@xmath4 crystallite structure unirradiated pcvd material same trapping centers sccvd after dose @xmath27p/@xmath4 ten years slhc running fluences @xmath30p/@xmath4 vertex detectors at 4 cm diamond sensors retain average signal @xmath232500 electrons after fluences element six ltd king s ride park ascot berkshire sl5 8bp united kingdom fraunhofer institut mikrointegration gustav - meyer - allee 25 berlin germany mathes al 2008 p12002 ph.d thesis:5n-16430 barbero et al . rd42 collaboration lhcc 2008 - 005 - rd-016 treis et al. 490 ( 2002 ) 112 lari 537 ( 2005 ) 581 belau. 214 ( 1983 ) 253 trischuk rd42 collaboration34@xmath31 ichep appear | 0.501003 | 3,028 | 1,541 | , Saving $0.1 in GPT-4. |
67,624 | the notion of death line of rotating pulsars is applied to model of oscillating neutron stars .
it is shown that the magnetosphere of typical non - rotating oscillating stars may not contain secondary plasma to support the generation of radio emission in the region of open field lines of plasma magnetosphere . | the model describing formation of the magnetosphere around highly magnetized rotating neutron star was first proposed by @xcite . they argued that rotating magnetized neutron star can not be surrounded by vacuum .
electric field , generated on the surface of the star due to rotation , pulls particles out of the surface and accelerates them along open magnetic field lines .
accelerated particles emit gamma rays due to curvature of magnetic field lines .
gamma rays then produce electron - positron pairs , which in turn accelerate and produce photons again .
there arises a chain reaction of multiplication of electrons , positrons and gamma - ray quanta near the surface of polar cap which leads to the formation of the plasma magnetosphere .
there are , however , certain conditions limiting the values of the pulsars rotation period @xmath0 and the strength of magnetic field @xmath1 under which formation of the magnetosphere can be realized .
death line is the @xmath2 ( or @xmath3 , or @xmath4 ) diagram which indicates the region where pulsar can support radio emission from magnetosphere ( here @xmath5 is the potential of accelerating field , @xmath6 means the time derivative of a period ) .
the essence of death line is that beyond certain conditions on the period and surface magnetic field strength of the pulsar sufficient amount of secondary plasma which is responsible for radio emission can not be generated . in this case
either potential drop can not accelerate the primary particles along magnetic field lines enough to produce high energy gamma rays , or the component of magnetic field perpendicular to gamma ray propagation is not high enough to produce the pair .
the mechanism of pair creation and pulsar radio emission is thoroughly described in the papers of @xcite , @xcite , @xcite , @xcite . since its appearance in 1975 in the work of
@xcite the notion of death line has been reviewed and revised by many authors ( see e.g. @xcite , @xcite , @xcite ) . numerical approach to the calculation of death line for the case of dipolar magnetic field of rotating neutron star is presented by @xcite .
up to now the notion of death line is satisfied well enough and all known radio pulsars are located in the region , which is predicted by theory to be `` radio - loud '' , i. e. capable to produce radio emission ( see @xcite , @xcite ) .
investigation of death line may help in better understanding the mechanism of radio emission from neutron stars ; it can serve as a tool for checking various models of pulsar radiation .
the works of @xcite , @xcite are devoted to exploring the model dependence of the pulsar death lines where two different types of acceleration models are investigated in detail . in our research we will study the death line s formalism for oscillating non - rotating neutron stars by using approach developed by @xcite .
our intention is to show that in the framework of present concepts of death line of pulsar one can predict that oscillating non - rotating typical neutron stars do not produce enough plasma in the region of open field lines to generate radio emission and , therefore , they are radio - quiet
. however the oscillating star may have plasma magnetosphere formed in earlier stages of stellar evolution when it was rotating . according to @xcite
the cohesive energy of the particles on surface of a pulsar is small to prevent them escape into the magnetosphere under the vacuum electric field . but
theory of pulsar activity tells that for producing radio emission there should be continuous processes of plasma formation maintained in the vicinity of the neutron star s polar cap .
this condition is not satisfied on investigated stage of neutron star s evolution .
as it was shown in a number of papers ( see @xcite , @xcite , @xcite ) the production of secondary plasma in the open field lines region of the pulsar may be realized if the potential accelerating the primary particle is large enough ( for the inverse compton scattering of thermal photons the lorentz factor of primary particles must rich @xmath7 , where @xmath8 is the particle s rest mass , @xmath9 is the temperature of the beam , @xmath10 is the speed of light ) and if the perpendicular to @xmath11-ray component of magnetic field gets the value @xmath12 here @xmath13 is the energy of photon emitted by the primary particle and @xmath14 g is the natural quantum mechanical unit for magnetic field strength ( @xmath15 is the charge of electron ) .
the conditions for pair production are better satisfied at the point @xmath16 , where @xmath17 is the distance from the stellar center ( in units of stellar radius ) to the place where the gamma ray is emitted and @xmath18 is the place where the pair is arising . below this condition the angle between magnetic field and @xmath11-ray is small , up to this condition the strength of magnetic field is low . using this assumption @xcite got the expression for the death line of radiopulsars in the following form @xmath19 where @xmath20 is the strength of pulsar magnetic field , normed on @xmath21 , @xmath22 is the radius of the star . the accelerating electric field in the region of the open field lines of the pulsar @xmath23 , while corresponding potential drop @xmath24 .
consider now non - rotating neutron star oscillating with the frequency @xmath25 .
oscillations of the star are assumed to be toroidal , what means that the components of the oscillation velocity in spherical coordinates @xmath26 have the following form @xmath27 here @xmath28 is the transversal velocity amplitude , oscillation are assumed to be small - amplitude and the magnetic field of the star has dipolar configuration .
oscillations of the star produce electric field which in analogy with the pulsar s accelerating electric field can be written as @xmath29 the effective frequency of stellar pulsations will be @xmath30 , where @xmath31 is the amplitude of oscillations .
the physical processes in the magnetosphere of oscillating neutron stars were thoroughly studied by timokhin et .
al . ( 2000 ) . in particular
, it was discussed an important question on formation of the region of open magnetic field lines in the magnetosphere of an oscillating star .
it was shown that for every mode of oscillations the angle ( on the surface of the star ) of the last open field line @xmath32 should be determined self - consistently using equation for the alfvenic surface .
it was found that for oscillation modes with @xmath33 the angle @xmath32 is small and expression for the goldreich - julian charge density in the approximation of small angles @xmath34 may be written as @xmath35 accelerating scalar potential above the surface of the pulsar , which can be found by means of poisson equation @xmath36 where @xmath37 is the gravitational lapse function and @xmath38 is the effective space charge density , is proportional to @xmath39 for rotating neutron star . in the case of oscillating magnetized neutron star
one can find by introducing dimensionless angular variable @xmath40 and taking into account angular dependence of @xmath41 that scalar potential @xmath42 .
consequently , for oscillation modes with @xmath33 accelerating scalar potential will contain additional small parameter .
taking into account that oscillation amplitude @xmath31 is about @xmath43 times smaller than radius of star @xmath22 and assuming that @xmath44 one can find that potential drop accelerating the primary particles for modes with @xmath33 will be @xmath45 times smaller with compare to that in the case of rotating neutron stars .
the result is that this potential will not be large enough to provide the primary particle with sufficient lorentz factor to emit gamma ray which would be possible to generate a pair .
figure [ dl ] illustrates the situation .
dashed curve is the death line for the rotating radiopulsar in frames of the space - charge - limited flow model , adopted from the paper @xcite .
solid line presents the death line for non - rotating oscillating neutron star .
it is plotted for real ( not effective ) values of pulsation periods , which are @xmath46 times shorter than @xmath47 . using the graph
one can see that radioemission from non - rotating stars may be observed only if the frequency of oscillations is rather large .
typical oscillating non - rotating stars with parameters @xmath48 and surface magnetic field @xmath49 will lie beyond the radio - loud region .
it means that plasma may not be produced in the region of open field lines of oscillating non - rotating neutron star s magnetosphere with typical parameters .
it means that vacuum electrodynamic model of oscillating stars ( @xcite , @xcite , @xcite ) is more preferable rather than plasma magnetosphere model ( @xcite ) .
observations of radio emission from a star s environment are possible only in the case if the magnetosphere of the star is filled with enough amount of secondary plasma . primary particles , extracted from the surface of the star , generate secondary plasma oscillations resulting in radio photons emission .
as it is seen from our results , typical non - rotating oscillating neutron star has not enough conditions to produce secondary plasma and probably may not produce plasma magnetosphere
. however magnetars ( @xcite ) with the high surface magnetic fields of order @xmath50 are above solid line from figure [ dl ] . furthermore , there is now observational evidence for stellar oscillations coming from the observation of quasi - periodic oscillations ( qpos ) following giant flares of soft gamma - ray repeaters ( sgrs ) ( @xcite , @xcite , @xcite , @xcite , @xcite ) .
the analysis of the x - ray data from sgrs has in fact revealed that the decaying part of spectrum exhibits a number of oscillations with frequencies in the range of a few tenths of @xmath51 to a few hundred @xmath51 that agree reasonably well with the expected toroidal modes of the magnetar crust ( @xcite ) . due to this
one may expect formation of plasma magnetosphere around oscillating relativistic stars with stronger magnetic fields as magnetars .
this research is supported in part by the uzffr ( projects 5 - 08 and 29 - 08 ) and projects fa - f1 - 079 , fa - f2-f061 and a13 - 226 of the uzas , by the ictp through the oea - prj-29 and the regular associateship grants .
arons , j. , & scharlemann , e. t. 1979 , , 231 , 854 chen , k. , & ruderman , m. a. 1993 , , 402 , 264 erber , t. 1966 , rev .
phys . , 38 , 626 goldreich , p. , & julian , w. h. 1969 , , 157 , 869 jones , p. b. 1986 , , 218 , 477 kantor , e. m. , & tsygan , a. i. 2004 , astronomy reports , v. 48 , no .
12 , 1029 manchester , r. n. , & taylor , j. h. : pulsars .
freeman , san francisco , 1977 medin , z. , & lai , d. 2007 , , 382 , 1833 muslimov , a. g. , & harding , a. k. 2002 , , 568 , 862 qiao , g. j. , xue , y. q. , zhang , b. , xu , r. x. , ye , f. f. , & wang , h. g. 2003 : high energy processes and phenomena in astrophysics iau symposium , vol .
214 , 171 rudak , b. , & ritter , h. 1994 , , 267 , 513 ruderman , m. a. , & sutherland , p. g. 1975 , , 196 , 51 zhang , b. , harding , a. k. , & muslimov , a. g. 2000 , , 531 , l135 zhang , b. , harding , a. k. , & muslimov , a. g. 1999 , aas , 19513303z | model describing formation magnetosphere around rotating neutron star proposed by @xcite argued be surrounded by vacuum electric field surface due to rotation pulls particles accelerates along magnetic field lines accelerated particles emit gamma rays due to curvature field lines gamma rays produce electron - positron pairs accelerate produce photons chain reaction of multiplication of electrons positrons gamma - ray quanta near surface polar cap leads to formation plasma magnetosphere conditions pulsars rotation period strength of magnetic field @xmath1 formation magnetosphere death line is @xmath2 diagram indicates region where pulsar radio emission from magnetosphere potential field time derivative of period death line beyond certain conditions period surface magnetic field strength pulsar secondary plasma responsible for radio emission be generated potential drop accelerate primary particles high energy gamma rays or magnetic field perpendicular to gamma ray propagation not high enough pair mechanism of pair creation pulsar radio emission described in papers of @xcite since notion of death line reviewed and revised by many authorsnumerical approach to calculation death line for dipolar magnetic field of rotating neutron star presented by @xcite notion death line satisfied radio pulsars in region predicted radio - loud radio emission @xcite investigation death line radio emission from neutron stars tool for checking models pulsar radiation works of @xcite model dependence pulsar death lines acceleration models our research study death line formalism for oscillating non - rotating neutron stars approach by @xcite intention to show oscillating non - rotating neutron stars produce enough plasma radio emission radio - quiet oscillating star may have plasma magnetosphere formed in earlier stages stellar evolution @xcite cohesive energy of particles on surface pulsar small to prevent escape into magnetosphere theory pulsar activity for radio emission continuous processes plasma formation in vicinity neutron star s polar cap condition not satisfied on investigated stage neutron star s evolutionshown in papers @xcite production of secondary plasma in open field lines region pulsar realized if potential accelerating primary particle large for inverse compton scattering thermal photons lorentz factor of primary particles rich @xmath7 particle s rest mass temperature beam speed of light perpendicular to @xmath11-ray component magnetic field gets value @xmath12 energy of photon primary particle natural quantum unit for magnetic field strength charge of electron conditions for pair production satisfied at point @xmath16 distance from stellar center to gamma ray emitted @xmath18 place pair arising below condition angle between magnetic field and @xmath11-ray small strength magnetic field low @xcite expression for death line of radiopulsars @xmath19 strength of pulsar magnetic field radius of star accelerating electric field in open field lines pulsar @xmath23 potential drop @xmath24 non - rotating neutron star oscillating with frequency @xmath25 oscillations toroidal components oscillation velocity spherical coordinates @xmath26 transversal velocity amplitude oscillation small - amplitude magnetic field star has dipolar configurationoscillations star produce electric field pulsar s written as @xmath29 effective frequency pulsations @xmath30 @xmath31 amplitude physical processes magnetosphere oscillating neutron stars studied by timokhin et 2000 discussed formation open magnetic field lines angle last open field line @xmath32 determined equation alfvenic surface oscillation modes @xmath33 angle @xmath32 small goldreich - julian charge density @xmath34 written as @xmath35 accelerating scalar potential above surface pulsar poisson equation @xmath36 @xmath37 gravitational lapse function @xmath38 effective space charge density proportional to @xmath39 for rotating neutron star oscillating magnetized neutron star introducing dimensionless angular variable @xmath40 angular dependence @xmath41 scalar potential @xmath42 oscillation modes with @xmath33 accelerating scalar potential additional small parameter oscillation amplitude @xmath31 @xmath43 times smaller than radius star @xmath22 potential drop accelerating primary particles modes @xmath33 @xmath45 times smaller rotating neutron starspotential not large enough primary particle sufficient lorentz factor emit gamma ray figure dl ] illustrates dashed curve death line for rotating radiopulsar space - charge - limited flow model from paper @xcite solid line death line for non - rotating neutron star plotted for pulsation periods @xmath46 times shorter than @xmath47 radioemission from non - rotating stars observed if frequency oscillations large typical oscillating stars with parameters @xmath48 @xmath49 beyond radio - loud region plasma may not produced in open field lines neutron star s magnetosphere vacuum electrodynamic model of oscillating stars preferable than plasma magnetosphere model radio emission star possible if magnetosphere filled with secondary plasma primary particles generate secondary plasma oscillations radio photons emission typical non - rotating neutron star not enough conditions produce secondary plasma may not produce plasma magnetosphere magnetars ( @xcite ) with high surface magnetic fields order @xmath50 above solid line figure observational evidence for stellar oscillations from quasi - periodic oscillations following flares soft gamma - ray repeaters ( @xciteanalysis x - ray data sgrs revealed decaying spectrum exhibits oscillations frequencies few tenths @xmath51 to hundred with expected toroidal modes magnetar crust expect formation plasma magnetosphere around oscillating relativistic stars stronger magnetic fields research supported by uzffr projects 5 - 08 29 - 08 ) projects fa - f1 - 079 fa - f2-f061 a13 - 226 uzas ictp oea - prj-29 regular associateship grants arons scharlemann 1979 , 231 , 854 chen ruderman 1993 402 , 264 erber 1966 rev. . 38 , 626 goldreich julian 1969 157 , 869 jones 1986 218 , 477 kantor tsygan 2004 astronomy reports v. 48 no . 12 , 1029 manchester taylor pulsars freeman san francisco 1977 medin lai 2007 , 382 , 1833 muslimov harding 2002 568 , 862 qiao xue zhang xu ye wang2003 high energy processes phenomena astrophysics symposium vol 214 , 171 rudak ritter 1994 267 , 513 ruderman m a. sutherland 1975 196 , 51 zhang b. harding k muslimov 2000 531 , l135 zhang harding muslimov 1999 19513303z | 0.433731 | 2,631 | 1,313 | , Saving $0.1 in GPT-4. |
67,626 | given a video and its incomplete textural description with missing words , the * video - fill - in - the - blank ( vifitb ) * task is to automatically find the missing word .
the contextual information of the sentences are important to infer the missing words ; the visual cues are even more crucial to get a more accurate inference . in this paper
, we presents a new method which intuitively takes advantage of the structure of the sentences and employs merging lstms ( to merge two lstms ) to tackle the problem with embedded textural and visual cues . in the experiments
, we have demonstrated the superior performance of the proposed method on the challenging `` movie fill - in - the - blank '' dataset @xcite . | video - fill - in - the - blank ( vifitb ) is a new computer vision problem .
( figure [ fig_example ] shows an example ) .
it is related to the standard video - question - and - answer ( viqaa ) problem , but have significant differences .
a major difference is that , for a standard vqaa problem , it has a complete sentence of the `` question '' , and requires the methods to output the `` answer '' . in this framework , it is easier to encode the `` question '' in the model ( e.g. neural networks ) and then use it to predict an output .
however , in the video - fill - in - the - blank ( vifitb ) problem , it is much trickier to encode the `` question '' , since the `` question '' is broken into parts and is much more challenging to encode the pieces of question to find the missing word of `` blank '' efficiently . in this paper
, we propose a new method to encode the sentence fragments before and after the blank using two lstms ( long short term memory ) , and use these two lstms outputs to find the correct answer .
this is an intuitive way to encode the semantic and contextual information from the sentence efficiently and makes the model easier to train .
similar to traditional visual - question - and - answer ( vqaa ) , we encode both textural and visual cues . from the experiments we found that both cues are crucial to improve the performance .
the visual - fill - in - the - blank ( vfitb ) problem is related to visual - captioning ( vc ) problem @xcite , for which the deep neural networks , especially recurrent neural networks ( e.g. lstm : long short term memory ) are intensely employed @xcite .
it s also highly related to the visual - question - and - answer ( vqaa ) problem @xcite .
for vqaa problem , usually there is a `` question '' with a question mark , and the answer is a word from the dictionary @xcite .
the answer can be numbers , colors , objects , etc .
some of the datasets @xcite provide multiple choices as the candidates for the answer .
strictly speaking , a visual - fill - in - the - blank ( vfitb ) problem can be considered as a vqaa problem ; however , this question is dramatically different from standard vqaa problems .
there are already some work for image - fill - in - the - blank ( imfitb ) problem @xcite , however , the blanks in the dataset are usually at the end of the sentence and it is similar to sentence prediction . also there is a dataset for video - fill - in - the - blank ( vifitb ) problem @xcite , however , the dataset was collected in a multiple choice fashion and limited number of words are available for the blank and mostly are about objects or actions . to our knowledge , `` movie fill - in - the - blank '' dataset @xcite is the first large - scale dataset for the vifitb problem in the wild and we show superior performance of the proposed method on the dataset .
the proposed method encodes the sentence with two lstms , and encodes the video by max - pooling over features coming out of a cnn .
an attention model @xcite helps us to leverage more spatial information from videos .
figure [ fig_framework ] shows an illustration of the proposed method .
each of the words from the sentence is converted into a feature vector using the word2vec approach @xcite , and then pass to lstms one by one . the left part ( before blank ) and the right part ( after blank )
are fed into two different lstms , aiming to capture the structure of the sentence .
we use @xmath0 and @xmath1 to represent the output from the left and right lstms respectively , and we combine them using by @xmath2),\ ] ] where @xmath3 is the final output , @xmath4 is a trainable parameter matrix , and `` @xmath5 '' means concatenation of left and right vectors .
the output @xmath3 from the two lstms are then input into the visual attention model . on the other hand ,
the video frames are processed by a pre - trained vgg19 network @xcite and the last pooling layer is extracted .
a max - pooling has been applied to all the frames features and one @xmath6 tensor is the outcome of this part as a visual feature . within the attention model , this visual feature and the textural feature @xmath3 are combined by @xmath7 where @xmath8 is output feature , @xmath9 and @xmath10 are trainable parameter matrices , @xmath11 is the visual feature , and `` @xmath12 '' means summation of the feature vector @xmath3 wit all @xmath13 vectors in @xmath11 .
the combined feature @xmath8 is then used to find attention model : @xmath14 where @xmath15 is the trainable parameter matrix .
the attention model @xmath16 is employed to output the weights for @xmath13 visual features .
finally , the textural and weighted visual features @xmath17 are combined to predict the word in the blank @xmath18
there are two methods to train the proposed deep network : end - to - end training , and incremental training . for end - to - end training ,
we train the whole network together , and for incremental training , we first train the sentence network ( lstms ) , then combine it with the visual attention network and train them together again .
we found the incremental training method has better performance .
we did our experiments on the `` movie fill - in - the - blank '' dataset @xcite .
we use categorical cross - entropy as our loss function and adagrad optimizer with early stoping strategy by observing the validation loss .
we show the performances of the proposed method with different setups , and compared it with several state - of - the - art methods .
table [ table_results ] shows some of the quantitative results.the end - to - end " is our method which trains all the parameters from scratch .
however , our incremental " method uses the method sentence " to initialize two lstms and the matrix @xmath4 , since they are in common between both of them .
the left sentence " method just uses the left part of the blank to answer the question . visual lstm "
method uses just cnn features coming out of last fully connected layer of vgg-19 and pass them through an lstm to get the answer with out considering the question .
.results on `` movie fill - in - the - blank '' dataset . [ cols="<,<,^",options="header " , ] [ table_results ]
we have proposed a new method for the video - fill - in - the - blank ( vifitb ) problem which takes advantage of the sentence structure before and after the blank and employed two lstms to encode the textural information efficiently .
we have demonstrated that , by incorporating the visual cues with the spatial attention model , the performance can be further improved .
we verified our ideas on the new `` movie fill - in - the - blank '' dataset of 2016 large scale movie description and understanding challenge ( lsmdc ) , and showed improved results compared with several baseline methods . in experiments ,
we have considered the output dimension of @xmath3 vector as @xmath19 .
also the batch size in training stage is @xmath20 .
each epoch takes about @xmath21 seconds in our implementation settings .
s. antol , a. agrawal , j. lu , m. mitchell , d. batra , c. lawrence zitnick , and d. parikh .
vqa : visual question answering . in _ proceedings of the ieee international conference on computer vision _ , pages 24252433 , 2015 .
t. mikolov , i. sutskever , k. chen , g. s. corrado , and j. dean . distributed representations of words and phrases and their compositionality . in _ advances in neural information processing systems _ ,
pages 31113119 , 2013 .
s. venugopalan , m. rohrbach , j. donahue , r. mooney , t. darrell , and k. saenko .
sequence to sequence - video to text . in _ proceedings of the ieee international conference on computer vision _ , pages 45344542 , 2015 .
l. yao , a. torabi , k. cho , n. ballas , c. pal , h. larochelle , and a. courville .
describing videos by exploiting temporal structure . in _ proceedings of the ieee international conference on computer vision _ ,
pages 45074515 , 2015 . | video - fill - in - the - blank ( vifitb ) new computer vision problem figure_example shows example related to standard video - question - answer viqaa ) problem significant differences major difference standard vqaa problem complete sentence of `` question '' requires methods to output answer easier to encode `` question '' model predict output video - fill - in - - blank vifitb ) problem trickier to encode `` question '' broken into parts challenging find missing word blank new method to encode sentence fragments before and after blank using two lstms ( long short term memory ) to find correct answer intuitive way to encode semantic contextual information model easier to train similar encode textural and visual cues both crucial improve performance visual - fill - blank vfitb ) problem related to visual - captioning ( vc ) problem @xcite deep neural networks recurrent neural networks memory employed related to visual - question - and - answer ( vqaa ) problem @xcite vqaa problem `` question '' with question mark answer is word from dictionary answer can be numbers colors objectsdatasets @xcite provide multiple choices answer visual - fill - in - the - blank ( ) problem vqaa problem question different from standard vqaa problems work for image - fill - in - - blank problem @xcite blanks end of sentence similar to sentence prediction dataset for video - fill - in - the - blank ( vifitb ) problem @xcite collected multiple choice limited words for blank mostly about objects or actions movie fill - in - the - blank '' dataset @xcite first large - scale dataset for vifitb problem superior performance of proposed method encodes sentence with two lstms video by max - pooling over features attention model @xcite spatial information from videos figure method words sentence converted into feature vector word2vec approach pass to lstms left and right part fed into two lstms capture structure sentence @xmath0 and @xmath1 output from left right lstms combine @xmath2) @xmath3 final output @xmath4 trainable parameter matrix @xmath5 concatenation of left and right vectors output @xmath3 from two lstms into visual attention modelvideo frames processed by pre - trained vgg19 network @xcite last pooling layer extracted max - pooling applied to features one @xmath6 tensor visual feature visual textural feature @xmath3 combined by @xmath7 output feature trainable parameter matrices visual feature summation of feature vector @xmath3 @xmath13 vectors in @xmath11 combined feature @xmath8 used attention model @xmath14 @xmath15 trainable parameter matrix attention model @xmath16 weights for @xmath13 visual features textural weighted visual features @xmath17 combined to predict word in blank @xmath18 two methods to train deep network end - to - end incremental training end - to - end training whole network incremental training sentence network visual attention network incremental training better performance experiments on movie fill - in - the - blank dataset @xcite categorical cross - entropy loss function adagrad optimizer early stoping strategy validation loss performances setups compared with state - of - the - art methods table shows quantitative results end - to - end trains parameters incremental method uses sentence initialize two lstms matrix @xmath4left sentence method uses left part blank answer question visual lstm method uses cnn features last layer vgg-19 through lstm answer considering question.results movie fill - in - the - blank dataset proposed new method video - fill - in - - blank problem advantage sentence structure before after blank two lstms encode textural information incorporating visual cues spatial attention model performance verified ideas on new movie fill - in - blank dataset 2016 large scale movie description understanding challenge showed improved results baseline methods considered output dimension @xmath3 vector @xmath19 batch size training stage @xmath20 each epoch takes @xmath21 seconds s. antol agrawal j lu mitchell batra lawrence zitnick parikh visual question answering proceedings ieee international conference computer vision pages 24252433 2015 t. mikolov sutskever chen corrado j dean distributed representations of words phrases compositionality advances neural information processing systems pages 31113119 2013 venugopalan rohrbach donahue mooney darrell saenko sequence - video to textproceedings ieee conference computer vision pages 45344542 2015 l. yao torabi k cho ballas pal h larochelle courville describing videos temporal structure proceedings conference computer vision pages 45074515 2015 | 0.316381 | 1,887 | 924 | , Saving $0.1 in GPT-4. |
67,627 | we study the possibility of gauging the standard model flavor group .
anomaly cancellation leads to the addition of fermions whose mass is inversely proportional to the known fermion masses . in this case all flavor violating effects turn out to be controlled roughly by the standard model yukawa , suppressing transitions for the light generations . due to the inverted hierarchy
the scale of new gauge flavor bosons could be as low as the electroweak scale without violating any existing bound but accessible at the tevatron and the lhc .
the mechanism of flavor protection potentially provides an alternative to minimal flavor violation , with flavor violating effects suppressed by hierarchy of scales rather than couplings . | in the standard model ( sm ) the only source of flavor violation arises from the yukawa couplings . in the limit of vanishing quark masses
the sm lagrangian acquires a large global symmetry ( known as flavor or horizontal symmetry ) mixing sm fermions of different generations .
it is quite tempting to impose such a symmetry on the physics beyond the sm in order to suppress extra flavor violation . in the extreme case where all new physics effects are flavor universal up to small corrections satisfying the full flavor symmetry and proportional to the sm yukawa ,
the scale of such new physics can be a tev .
this idea , known as minimal flavor violation ( mfv ) is quite old @xcite , but recently is gaining ever more interest ( see @xcite ) as a consequence of the persistent failure to find flavor violation beyond the standard model .
the idea of assuming horizontal symmetries to be true symmetries of nature is even older @xcite .
unfortunately such assumption itself is not enough to suppress flavor violation below the experimental bounds when the flavor symmetry is broken at low scales .
the classical arguments against low - scale flavor symmetry work as follows ( see , , @xcite for a nice review ) . in order to produce the sm yukawa couplings flavor symmetry must eventually be broken by the vacuum expectation value ( vev ) of some field ( `` flavon '' ) .
this implies the presence of massless goldstone bosons ( gb ) and bounds from hadron decays and astrophysics on such states are even stronger than those from flavor physics .
of course gbs can be easily avoided by gauging . in this case
, however , there are flavor gauge bosons that mediate dangerous flavor changing neutral currents ( fcnc ) and their masses must be well above the tev scale . even requiring that the only sources of flavor breaking are the sm yukawas is not enough to avoid large fcnc from the flavor gauge bosons . indeed if the masses of the gauge bosons are proportional to the sm yukawa couplings , they generate tree - level four - fermion operators proportional to inverse powers of the sm yukawa couplings , enhancing fcnc among the first generations .
therefore despite the fact that the only spurions breaking flavor are the sm yukawa couplings , as in mfv , inverse powers of the spurions appear in the higher dimensional operators , producing _ de facto _ `` maximal '' flavor violating operators .
this argument shows how mfv models can not arise directly from having a fundamental flavor symmetry in the underling theory but rather from accidental ones . in the classical argument against low - scale flavor gauge bosons sketched above we
can easily identify a way out .
if the fields breaking the flavor symmetry are instead proportional to the inverse of the sm yukawa couplings , the effective operators generated by integrating out the flavor gauge bosons will be roughly proportional to positive powers of the yukawa couplings , suppressing flavor violating effects for the light generations , much like in mfv models .
the spectrum of the extra flavor states , controlled by the flavon vevs , will thus present an inverted hierarchy , with states associated to the third generation much lighter than those associated to the first two .
models implementing this inverted hierarchy were first introduced in @xcite . remarkably , as we will show in this paper , the mechanism described above is automatic in the minimal extension of the sm with gauged flavor symmetries . with just the sm fermion content the full sm plus
flavor gauge group would be anomalous .
extra flavorful fermions have to be added to cancel these anomalies .
such fermions are also welcome as they can make the sm yukawa terms arise from a renormalizable lagrangian , now that the yukawa couplings have been uplifted to dynamical `` flavon '' fields .
the smallest set of fermions canceling all anomalies leads automatically to the inverted - hierarchy structure mentioned above .
the quantum numbers of these extra fermions are indeed such that the mixing with the sm fermion is flavor diagonal while their masses are proportional to the flavon vevs .
the sm fermion masses arise via a see - saw like mechanism , after integrating out the extra fermions , and are thus proportional to the inverse of the flavon vevs .
all non - sm particle masses ( fermions , vector bosons and flavon fields ) are controlled by the flavon vevs , thus they are roughly proportional to the inverse of the sm yukawa .
the resulting inverted hierarchy in the new physics sector protects the sm fermions from getting large flavor breaking effects even when the lightest new states lie at the electroweak scale .
there are a number of analogies with mfv models : new physics effects are controlled by the flavor group , we may have models where only one spurion for each sm yukawa matrix breaks the flavor symmetry , and the flavor breaking effects follow the hierarchical structure of the yukawa couplings .
however , this kind of models are not mfv . indeed , in these models there is a limit
where all yukawa couplings vanish but flavor breaking effects remain finite . contrary to the naive intuition that flavor violating effects must be larger than in mfv , it is very easy to find values of the parameters that produce extra flavor non - universal states without incurring into a flavor problem .
moreover , unlike in mfv , these could be light , even below the electroweak scale .
the tightest bounds on this kind of models do not come from flavor breaking observables but rather from electroweak precision tests ( ewpt ) and direct searches for new particles , opening the possibility for direct discoveries of flavor physics at the lhc .
the mechanism protecting from flavor violations is robust against deformations of the model , both when more flavon fields are considered and when the nature of the flavor subgroup that is actually gauged is changed .
of course the detailed structure of the flavor sector as well as the size of the flavor violations will depend on these modifications but the latter will continue to remain sufficiently small in most of the parameter space of the theory . we should point out that the possibility of having non - universal gauge bosons ( and other flavorful physics ) at the tev scale is well known in the literature . in composite higgs models and similar extra - dimensional constructions , for example , there is the possibility of having extra flavorful gauge bosons . unlike in our case
however , these vector fields are not `` the '' flavor gauge bosons ( _ viz .
_ they are not the states eating the goldstone bosons of the broken flavor symmetry ) , but rather gauge bosons in some non - trivial representation of the flavor group , getting mass splitting from the breaking of the flavor symmetry ( which generically is explicit in these models ) . in this respect
they are closer to realizing the mfv idea ( see , , @xcite ) . in the rest of the paper
we give the details on how the mechanism works , we will discuss where the strongest bounds on the model come from and possible signatures at hadron colliders . for definiteness we will focus on the quark sector , gauging the full flavor group and considering mainly the minimal set of flavon fields , although the same mechanism can easily be applied to more general situations .
in the absence of yukawas , focusing on the quark sector , the sm enjoys at the classical level the global symmetry @xmath0 where @xmath1 , @xmath2 and @xmath3 transform as fundamentals .
we assume this to be an exact symmetry of nature . in order to allow yukawa couplings
the flavor symmetry should be broken spontaneously by the vacuum .
this can be most simply realized by the vevs of two bifundamentals flavon fields transforming as @xmath4 in general the vevs of these fields , while related , should not be confused with the yukawa matrices , as functions of @xmath5 may have equal transformation properties .
indeed this will be the crucial feature of our model . to avoid problematic flavor violating gbs , the symmetry should be gauged . within the sm
the gauging of the sm flavor symmetry ( [ flavorsymmetry ] ) is anomalous due to cubic and mixed hypercharge anomalies .
the simplest option to cancel the cubic non - abelian anomalies is to add two right - handed colored fermions in the fundamental of @xmath6 , one left handed fundamental of @xmath7 and one left - handed fundamental of @xmath8 . in this way the fermions are vector - like with respect to the flavor gauge group but remain chiral with respect to the sm gauge symmetry .
the other possibility , with the two right - handed triplets in an @xmath9 doublet is an uninteresting , non - chiral model .
we are therefore led rather uniquely to the following model : [ cols="^,^,^,^,^,^,^",options="header " , ] and with @xmath10 the mass of the lightest ckm mode eigenstate .
the contributions are so suppressed that @xmath11 can be as light as 100 mev without incurring into problems with flavor .
the quantitative results above nicely fit with what is observed in the two flavor case , and the couplings of the canonically normalized ckm modes to the fermions are numerically compatible with the short - hand formula @xmath12 which is similar to the flavor preserving one for the radial modes of the diagonal flavon fields .
in this appendix we derive the one loop expression for the @xmath13 , @xmath14 and @xmath15 parameters in our model . for simplicity we work in the limit where only the mixing of the top is important .
from the couplings ( [ wcouplings ] ) , ( [ zcouplings ] ) the contribution of the third generation to @xmath14 ( obtained as the difference between the correlators in our model and their sm values corresponding to @xmath17 ) is given by , @xmath18\end{gathered}\ ] ] where we have introduced the self energies @xmath19 with two left currents . in the limit
@xmath20 one finds , @xmath21\ , .
\label{t}\ ] ] s. m. barr and a. zee , phys .
d * 17 * ( 1978 ) 1854 .
f. wilczek and a. zee , phys .
lett . * 42 * ( 1979 ) 421 . c. l. ong , phys .
d * 19 * ( 1979 ) 2738 .
j. chakrabarti , phys .
d * 20 * ( 1979 ) 2411 .
t. maehara and t. yanagida , prog .
. phys .
* 61 * ( 1979 ) 1434 .
a. davidson , m. koca and k. c. wali , phys .
* 43 * ( 1979 ) 92 .
_ ibid , _ phys .
b * 86 * ( 1979 ) 47 .
d. d. wu , high energy phys .
nucl . phys . * 4 * ( 1980 ) 455 .
t. yanagida , phys .
d * 20 * ( 1979 ) 2986 .
a. davidson and k. c. wali , phys .
d * 21 * ( 1980 ) 787 .
z. g. berezhiani and j. l. chkareuli , `` quark - leptonic families in a model with su(5 ) x su(3 ) symmetry .
( in russian ) , '' sov .
j. nucl .
* 37 * ( 1983 )
618 [ yad . fiz .
* 37 * ( 1983 ) 1043 ] .
+ z. g. berezhiani , `` the weak mixing angles in gauge models with horizontal symmetry : a new approach to quark and lepton masses , '' phys .
b * 129 * ( 1983 ) 99 .
a. a. affolder _ et al .
_ [ cdf collaboration ] , `` search for a fourth - generation quark more massive than the z0 boson in @xmath26 collisions at @xmath27 tev , '' phys . rev .
* 84 * ( 2000 ) 835 [ arxiv : hep - ex/9909027 ] .
t. aaltonen _ et al . _
[ cdf collaboration ] , `` search for new particles leading to @xmath28 jets final states in @xmath29 collisions at @xmath30 = 1.96-tev , '' phys .
d * 76 * ( 2007 ) 072006 [ arxiv:0706.3264 [ hep - ex ] ] .
e. salvioni , g. villadoro and f. zwirner , `` minimal z models : present bounds and early lhc reach , '' jhep * 0911 * ( 2009 ) 068 [ arxiv:0909.1320 [ hep - ph ] ] .
+ e. salvioni , a. strumia , g. villadoro and f. zwirner , `` non - universal minimal z models : present bounds and early lhc reach , '' jhep * 1003 * ( 2010 ) 010 [ arxiv:0911.1450 [ hep - ph ] ] .
m. e. albrecht , t. feldmann and t. mannel , `` goldstone bosons in effective theories with spontaneously broken flavour symmetry , '' arxiv:1002.4798v2 [ hep - ph ] ; two approaches towards the flavour puzzle .
michaela e. albrecht , phd thesis , tu munich , aug 2010 . | standard model flavor violation from yukawa couplings sm lagrangian acquires large global symmetry flavor horizontal symmetry mixing fermions different generations tempting to impose symmetry on physics beyond sm suppress flavor violation extreme case new physics effects flavor universal small corrections full flavor symmetry proportional to sm yukawa scale new physics can be tev idea minimal flavor violation old gaining interest failure to find flavor violation beyond standard model assuming horizontal symmetries true symmetries nature older not enough to suppress flavor violation below experimental bounds flavor symmetry broken at low scales arguments against low - scale flavor symmetry produce sm yukawa couplings flavor symmetry must broken by vacuum expectation value vev ) field implies massless goldstone bosons bounds from hadron decays astrophysics stronger than flavor physics gbs avoided by gauging flavor gauge bosons mediate dangerous flavor changing neutral currents masses must above tev scale only sources flavor breaking sm yukawas not enough to avoid large fcnc from flavor gauge bosonsmasses gauge bosons proportional to sm yukawa couplings generate tree - level four - fermion operators proportional to inverse powers enhancing fcnc first generations despite spurions breaking flavor sm yukawa couplings inverse powers appear in higher dimensional operators producing maximal flavor violating operators mfv models from fundamental flavor symmetry accidental ones argument against low - scale flavor gauge bosons way out if fields breaking flavor symmetry proportional to inverse sm yukawa couplings effective operators generated flavor gauge bosons proportional to positive powers yukawa couplings suppressing flavor violating effects for light generations extra flavor states controlled by flavon vevs present inverted hierarchy states third generation lighter than first two models introduced in @xcite mechanism automatic in minimal extension sm with gauged flavor symmetries sm fermion content full sm plus flavor gauge group anomalous extra flavorful fermions added to cancel anomalies fermions sm yukawa terms arise from renormalizable lagrangian smallest set of fermions canceling anomalies leads to inverted - hierarchy structurequantum numbers extra fermions mixing with sm fermion flavor diagonal masses proportional to flavon vevs sm fermion masses arise via see - saw mechanism integrating extra fermions proportional to inverse flavon vevs non - sm particle masses controlled by flavon vevs proportional to inverse sm yukawa inverted hierarchy new physics sector protects sm fermions from large flavor breaking effects states electroweak scale analogies with mfv models new effects controlled by flavor group one spurion each sm yukawa matrix breaks flavor symmetry flavor breaking effects follow hierarchical structure yukawa couplings models not mfv limit yukawa couplings vanish flavor breaking effects finite flavor violating effects easy to find values extra flavor non - universal states without flavor problem could be light below electroweak scale tightest bounds models from flavor breaking observables electroweak precision tests direct searches for new particles direct discoveries flavor physics at lhc mechanism protecting flavor violations robust against deformations model more flavon fields nature flavor subgroupdetailed structure flavor sector size flavor violations depend on modifications latter small in parameter space theory possibility non - universal gauge bosons flavorful physics at tev scale known in literature in composite higgs models extra - dimensional constructions extra flavorful gauge bosons vector fields not flavor gauge bosons states eating goldstone bosons of broken flavor symmetry gauge bosons in non - trivial representation flavor group mass splitting from breaking flavor symmetry explicit in models closer to realizing mfv idea rest of paper details on mechanism discuss strongest bounds possible signatures at hadron colliders focus on quark sector gauging full flavor group considering minimal set of flavon fields mechanism applied to general situations absence of yukawas quark sector sm enjoys global symmetry @xmath0 where @xmath1 @xmath2 @xmath3 transform as fundamentals exact symmetry of nature allow yukawa couplings flavor symmetry should broken by vacuum realized by vevs two bifundamentals flavon fields transforming as @xmath4 vevs fields not confused with yukawa matrices functions of @xmath5 have equal transformation propertiescrucial feature model avoid problematic flavor violating gbs symmetry gauged sm flavor symmetry anomalous due to cubic mixed hypercharge anomalies option cancel cubic non - abelian anomalies add two right - handed colored fermions in @xmath6 one left @xmath7 @xmath8 fermions vector - like flavor gauge group remain chiral sm gauge symmetry two right - handed triplets in @xmath9 doublet uninteresting non - chiral model led to model @xmath10 lightest ckm mode eigenstate contributions suppressed @xmath11 light as 100 mev without problems flavor quantitative results fit with two flavor case couplings normalized ckm modes to fermions compatible with short - hand formula @xmath12 similar to flavor preserving one for radial modes diagonal flavon fields appendix one loop expression for @xmath13 @xmath14 @xmath15 parameters work in limit mixing of top importantcouplings contribution third generation @xmath14 difference correlators values @xmath17 @xmath18 introduced self energies @xmath19 two left currents limit @xmath20 finds @xmath21 s m. barr a. zee phys . 17 1978 ) 1854 f. wilczek a. zee 42 1979 ) 421 l. ong phys 19 1979 ) 2738 . chakrabarti phys 20 1979 ) 2411 t. maehara t. yanagida phys 61 1979 ) 1434 a. davidson m. koca k. c. wali phys 43 1979 ) 92 86 1979 ) 47 wu high energy phys 4 1980 ) 455 yanagida 20 1979 ) 2986 a. davidson k. c. wali phys 21 1980 ) 787 z. g. berezhiani j. l. chkareuli quark - leptonic families model su(5 ) x su(3 ) symmetry russian nucl 37 1983 ) 618 1983 1043 z gberezhiani weak mixing angles gauge models horizontal symmetry new approach quark lepton masses phys . 129 1983 ) 99 affolder search fourth - generation quark massive z0 boson @xmath26 collisions @xmath27 phys . rev 84 2000 ) 835 hep - ex/9909027 aaltonen search new particles @xmath28 jets final states @xmath29 collisions @xmath30 1.96-tev . 76 2007 ) 072006 arxiv:0706.3264 hep - salvioni villadoro zwirner minimal z models present bounds early lhc reach jhep 0911 2009 ) 068 arxiv:0909.1320 salvioni zwirner non - universal minimal z models present bounds early lhc reach jhep 1003 2010 ) 010 arxiv:0911.1450 albrecht mannel goldstone bosons theories broken flavour symmetry arxiv:1002.4798v2 two approaches flavour puzzle michaela e albrecht phd thesis tu munich aug 2010 | 0.441995 | 2,999 | 1,495 | , Saving $0.1 in GPT-4. |
67,630 | we examine a model in which a nonequilibrium phase transition from an active to an extinct state is observed .
the order of this phase transition has been shown to be either continuous or first - order , depending on the parameter values and the dimension of the system .
using increasingly large clusters , we use the cluster approximation method to obtain estimates for the critical points in 1 + 1 dimensions . for the continuous phase transitions
only , extrapolations of these approximations show excellent agreement with simulation results .
further , the approximations suggest that , consistent with simulation results , in 1 + 1 dimensions no first - order phase transitions are observed . | the order of phase transitions in 1 + 1 dimensions has long been a topic of debate among researches in the field of nonequilibrium phase transitions .
while it has been argued that first - order phase transitions are impossible in this dimension ( see for example @xcite ) , most agree that such transitions are feasible .
for example , dickman and tom sought to find the simplest model with short - range interactions that exhibited a first - order phase transition in one spatial dimension @xcite .
they examined the pair- and triplet - creation models with the reactions @xmath0 with @xmath1 , 3 respectively . the mean field ( mf ) of such reactions yields a first - order phase transition , yet often , continuous transitions are thought to be observed in ( 1 + 1)-dimensional monte carlo ( mc ) simulations .
they varied the diffusion rate @xmath2 to see what effect this had since they expected that , with larger diffusion , the model would exhibit more mf - like behaviour due the better mixing of particles , and hence , a first - order phase transition might be observed .
they found a first - order phase transition for the triplet - creation model only .
a continuous phase transition was observed in the pair - creation model even when 95% of the attempted moves were diffusive .
for the triplet - creation model , hysteresis was observed , indicative of a first - order transition , for sufficiently high diffusion rate .
recently , fiore and de oliveira @xcite used the conservative diffusive contact process , in which the number of particles is fixed , to confirm dickman and tom s original findings ( see also @xcite ) .
cardozo and fontanari @xcite also examined the triplet - creation model and found that the critical exponents change continuously from dp values to compact directed percolation values as the diffusion rate was increased . due to the strong crossover effects ,
however , they were unable to locate the precise position of the point where the transition changed order . in this paper
, we wish to examine a slightly modified version of the pair - creation model in the ( 1 + 1)-dimensional case through both simulation and , importantly , mf techniques .
the model has previously been shown , by mf , to exhibit both continuous and first - order phase transitions whose lines in parameter space meet at a tricritical point @xcite ( see also @xcite ) . in ( 1 + 1)-dimensional mc simulations ,
however , the model exhibits a continuous phase transition across the whole phase space . to examine this model further ,
we employ a technique originally introduced by ben - avraham and khler @xcite which they called the @xmath3-site cluster approximation . in most cases , as @xmath3 increases
, the method predicts increasingly accurate behaviour of the model in question . in this paper , we collect data for @xmath4 and extrapolate our findings as @xmath5 to obtain approximations for the order of the phase transition and value of the critical point .
we have a @xmath6-dimensional square lattice of linear length @xmath7 where each site is either occupied by a single particle or is empty .
a site is chosen at random .
the particle on an occupied site dies with probability @xmath8 , leaving the site empty . if the particle does not die , a nearest neighbour site is randomly chosen .
if the neighbouring site is empty the particle moves there and produces a new individual at the site that it has just left with probability @xmath9 . if the chosen site is , however , occupied the particle reproduces with probability @xmath10 producing a new particle on another randomly selected neighbouring site , conditional on that site being empty .
a time step is defined as the number of lattice sites @xmath11 and periodic boundary conditions are used .
we have the following reactions for a particle @xmath12 for proliferation and annihilation respectively , @xmath13 assuming the particles are spaced homogeneously , the mf equation for the density of active sites @xmath14 is given by @xmath15 the first two terms consider the sexual and asexual reproduction reactions respectively and the final term death of an individual .
equation ( [ e : mean field ] ) has three stationary states : @xmath16 .
\end{aligned}\ ] ] for @xmath17 , @xmath18 continuously as @xmath19 , indicative of a continuous phase transition with critical point @xmath20 for @xmath21 , we have a jump in @xmath22 from @xmath23 to zero , this time at the critical point @xmath24 further , for @xmath21 , we have a region @xmath25 where the survival of the population is dependent on the population density .
in fact , we have extinction for @xmath26 for @xmath21 we therefore have a first - order phase transition .
the two phase transition lines meet at the point @xmath27 , defining the position of the tricritical point @xmath28 . at the mf level then , we have a phase diagram as shown in figure [ f : phase diagram ] . and
the first - order phase transition for @xmath29 . equations ( [ e : critical point 1]-[e : critical point 2 ] ) have been re - arranged in the figure to make @xmath9 the dependent variable.,width=321 ] in the region to the left of the transition lines , there exists at least one real and positive steady state . in the shaded region
only , there exist two such steady states with none existing to the right of the transition lines .
the tricritical point not only marks the intersection of the first - order and continuous phase lines , but also the line bordering the region with population density dependence .
the mf technique that we employed in the previous section assumed independence between individual sites . in reality there exist , of course , correlations between nearby sites . increasing improvements to our original mf equation
could then be made by considering pairs , triplets , quadruplets and so on , of adjacent sites .
such an approach was developed by ben - avraham and khler , called the @xmath30-cluster approximation method @xcite .
we direct the interested reader to ben - avraham and khler s paper @xcite and others ( for example @xcite and references therein ) for a detailed explanation of the approach .
briefly , the method involves examining clusters of size @xmath3 . for sites @xmath31 ,
we denote @xmath32 as the probability that the @xmath33 sites are in state @xmath34 at time @xmath35 .
the method is concerned with how @xmath36 changes in time and involves deriving a system of master equations for @xmath37 . since @xmath38,1 in our case , there are @xmath39 such equations .
it is , however , easy to see that the number of independent equations is much smaller than this ( see , for example , @xcite ) .
since we are concerned with clusters of size @xmath3 , we need a way to approximate @xmath40 from clusters of size @xmath41 . in their @xmath30-approximation ,
ben - avraham and khler consider adjacent clusters of size @xmath3 with an overlap of @xmath42 sites .
using the bayesian extension process , the approximation is then , @xmath43 for example , the @xmath44 approximation for a cluster of size six is given by @xmath45 ben - avraham and khler found that the @xmath46-approximation yields the most accurate results and is termed the @xmath3-site approximation for short .
the simple 1-site approximation is just our original mf equations , so we examine the 2-site approximation . introducing the subscripts @xmath47 and @xmath48 for occupied and empty sites respectively , we have two independent variables , chosen to be the particle density @xmath49 and the pair density @xmath50 .
it is then simple to derive the other 2-site probabilities @xmath51 to obtain the master equations , we consider the reactions which change the number of occupied sites @xmath52 and the number of pairs of sites @xmath53 .
the reactions are listed in table [ t : pair approximation ] along with their probabilities of occurring , given the particle configuration . from this table ,
the master equations for @xmath54 and @xmath55 are then derived in the usual way .
.reactions for the 2-site approximation where , reactions such as @xmath56 for which @xmath57 , have been ignored . a symmetry factor arising from the parity symmetry
has been included ( right column ) rather than writing both equations down . [ cols="^,^,^,>,>,<,>",options="header " , ]
we clearly see how increasing the size of the clusters gives more accurate approximations for the behaviour of the model when comparing to the mc simulation results . in particular , for @xmath58
, we show in the main plot of figure [ f : ca phase diagram ] b ) , the approximated values for @xmath59 against @xmath60 .
an extrapolation of the results as @xmath5 shows excellent agreement with the mc value .
unfortunately , at the first - order phase transition , when the number of real and positive steady state solutions decrease from two to zero , such an extrapolation does not lead to good agreement .
this is likely to be due to the fact that at the continuous phase transition , the correlation length is infinite and therefore considering increasingly large numbers of adjacent sites will lead to more accurate approximations . at first - order phase transitions , since the correlation length remains finite , for @xmath61 we would expect the approximations for @xmath62 to be independent of @xmath3 .
we can further plot the value of the tricritical point as a function of cluster size @xmath3 and , again , extrapolate . as the inset of figure [ f : ca phase diagram ] b ) shows , the position of @xmath63 decreases with @xmath3 , appearing to become zero for some finite @xmath3 . in other words , for sufficiently large @xmath3 , the corresponding @xmath3-site cluster approximation would predict no first - order phase transition .
this , however , would be surprising since we would expect such behaviour to be true only as @xmath5 .
since the tricritical transition is infinitesimally close to a first - order transition though , we may well expect a tailing - off of this apparent linear behaviour for large @xmath3 .
we have seen how , for this model , the analytical cluster - approximation method seems to correctly predict a continuous phase transition in 1 + 1 dimensions across the whole phase space as @xmath5 .
further , the method predicts , with a high degree of accuracy , the critical point for continuous phase transitions only .
these findings highlight the power of the method for this case and will hopefully lead on to further examination of the techniques involved .
all computer simulations were carried out on the imperial college london s hpc for which we thank matt harvey and simon burbidge .
alastair windus would also like to thank epsrc for his ph.d .
10 url # 1#1urlprefix[2][]#2 hinrichsen h 2000 _ cond - mat/0006212 _ | order phase transitions in 1 + 1 dimensions debate argued first - order phase transitions impossible most agree feasible dickman and tom sought simplest model first - order phase transition in dimension examined pair- triplet - creation models with reactions @xmath0 @xmath1 , 3 mean field mf ) reactions yields first - order phase transition continuous transitions observed in 1 + 1)-dimensional monte carlo simulations varied diffusion rate @xmath2 expected larger diffusion model more mf - like behaviour mixing first - order phase transition found first - order phase transition for triplet - creation model continuous phase transition observed in pair - creation model 95% moves diffusive triplet - creation model hysteresis observed indicative of first - order transition for high diffusion rate fiore and de oliveira used conservative diffusive contact process confirm dickman tom s findings cardozo and fontanari examined triplet - creation model critical exponents change from dp to compact percolation values as diffusion rate increased unable to locate precise position point where transition changed orderpaper examine modified pair - creation model ( 1 + 1)-dimensional case through simulation mf techniques model shown mf exhibit continuous first - order phase transitions meet at tricritical point in ( 1 + 1)-dimensional mc simulations model exhibits continuous phase transition across phase space examine employ technique introduced by ben - avraham khler @xmath3-site cluster approximation @xmath3 increases method predicts accurate behaviour model collect data for @xmath4 extrapolate findings as @xmath5 approximations for order phase transition value critical point @xmath6-dimensional square lattice linear length each site occupied by particle or empty site chosen at random particle occupied site dies leaving site empty if particle die nearest neighbour site randomly chosen if site empty particle moves produces new individual if chosen site occupied particle reproduces new particle on another randomly selected site conditional on site empty time step defined as number of lattice sites @xmath11 periodic boundary conditions usedreactions for particle @xmath12 proliferation annihilation @xmath13 mf equation for density of active sites @xmath14 by @xmath15 first two terms sexual asexual reproduction reactions final term death equation mean field three stationary states @xmath16 @xmath17 @xmath18 @xmath19 continuous phase transition with critical point @xmath20 @xmath21 jump in @xmath22 from @xmath23 to zero at critical point @xmath24 @xmath21 region @xmath25 survival population dependent on population density extinction for @xmath26 @xmath21 first - order phase transition two phase transition lines meet at point @xmath27 tricritical point @xmath28 phase diagram figure first - order phase transition for @xmath29 equations critical point 1- critical point 2 re - arranged @xmath9 dependent variable region left of transition lines one real positive steady state shaded region two steady states none right of transition lines tricritical point marks intersection of first - order continuous phase lines line bordering region with population density dependence mf technique assumed independence between sites correlations between nearby sitesimprovements to original mf equation considering pairs triplets quadruplets of adjacent sites approach developed by ben - avraham khler @xmath30-cluster approximation method @xcite direct to khler s paper @xcite others for detailed explanation method examining clusters of size @xmath3 sites @xmath31 @xmath32 as probability @xmath33 sites in state @xmath34 @xmath35 method @xmath36 changes deriving master equations for @xmath37 @xmath38,1 @xmath39 such equations number of independent equations smaller concerned with clusters size @xmath3 need approximate @xmath40 from clusters @xmath41 @xmath30-approximation ben - avraham khler consider adjacent clusters size @xmath3 with overlap of @xmath42 sites bayesian extension process approximation is @xmath43 @xmath44 approximation for cluster six by @xmath45 @xmath46-approximation accurate results @xmath3-site approximation simple 1-site approximation original mf equations examine 2-site approximationsubscripts @xmath47 and @xmath48 for occupied and empty sites two independent variables particle density @xmath49 and pair density @xmath50 simple to derive 2-site probabilities @xmath51 master equations consider reactions occupied sites @xmath52 pairs sites @xmath53 reactions listed in table [ pair approximation probabilities particle configuration master equations for @xmath54 and @xmath55 derived usual for 2-site approximation reactions @xmath56 @xmath57 ignored symmetry factor from parity symmetry included right column increasing size of clusters gives accurate approximations for behaviour model mc simulation results for @xmath58 in main plot of figure f phase diagram approximated values for @xmath59 against @xmath60 extrapolation of results as @xmath5 shows agreement with mc value at first - order phase transition real positive steady state solutions decrease from two to zero extrapolation to good agreement continuous phase transition correlation length infinite large numbers of adjacent sites more accurate approximationsfirst - order phase transitions correlation length finite @xmath61 approximations @xmath62 independent of @xmath3 plot value tricritical point function of cluster size @xmath3 extrapolate figure phase diagram position @xmath63 decreases with @xmath3 zero for finite @xmath3 large @xmath3 @xmath3-site cluster approximation no first - order phase transition surprising behaviour true only as @xmath5 tricritical transition close to first - order transition expect tailing - off linear behaviour for large @xmath3 analytical cluster - approximation method continuous phase transition in 1 + 1 dimensions across phase space @xmath5 method predicts high accuracy critical point for continuous phase transitions findings highlight power method lead further examination techniques computer simulations on imperial college london s hpc thank matt harvey simon burbidge alastair windus epsrc for ph.d . hinrichsen h 2000 cond - mat/0006212 | 0.35936 | 2,401 | 1,233 | , Saving $0.1 in GPT-4. |
67,633 | the ground state of a two - dimensional , harmonically confined mesoscopic assembly of up to thirty polar molecules is studied by computer simulations . as the strength of the confining trap
is increased , clusters evolve from superfluid , to supersolid , to insulating crystals . for strong confinement ,
the crystalline structure can be predicted based on classical energetics .
however , clusters of specific numbers of particles ( i.e. , @xmath0=12 and @xmath0=19 ) display a _
non - classical crystalline structure _ , stabilized by quantum effects , in an intermediate range of confinement strength . in these cases , coexistence of quantum and classical crystalline configurations is observed at finite temperature . | one of the most remarkable recent experimental achievements is the preparation of cold ensembles of dipolar particles .
gases of ultracold ground state magnetic atoms have been shown to display dipole - dominated dynamics in the quantum regime @xcite .
huge electric dipole moments , of up to thousands of debye , are present in highly excited rydberg states of alkali atoms @xcite , and experiments are under way to demonstrate and control interactions in these systems @xcite . strong correlations with ground state atoms can be obtained by weakly admixing with laser light these rydberg states @xcite , at the price of finite heating and losses due to spontaneous emission @xcite .
polar molecules prepared in the electronic and rovibrational ground state @xcite combine the stability of ground state particles with dipole moments of up to a few debye , leading to large dipolar interactions .
experiments are well under way to exploit these interactions in combination with reduced trapping geometries , to add collisional stability @xcite .
this opens the door to the study of strongly correlated quantum phases with designed long range interactions @xcite , e.g. , for dipolar bose systems the density - driven superfluid - crystal quantum phase transition in 2d @xcite .
+ finite , confined dipolar assemblies are worthy of investigation for a number of reasons , the most obvious being the study of the evolution of the physical properties of the system as its size is increased , approaching the bulk limit @xcite .
for example , recent theoretical work with fermionic molecules @xcite has analyzed the effects of quantum statistics on transitions between wigner - type states in traps @xcite . in this work ,
we focus our attention on quasi - two - dimensional , harmonically confined dipolar systems of mesoscopic size , comprising a relatively small number of particles .
one of the fundamental issues that one may address , is the occurrence of superfluidity in a cluster , until now mostly explored in the context of helium @xcite and hydrogen @xcite droplets .
+ based on results of first principle quantum simulations , an intriguing deviation from bulk behaviour that occurs in a dipolar system of small size is predicted here , namely the occurrence of _ non - classical _ ground state crystalline arrangements , in an intermediate range of confinement strength .
these configurations differ from the classical arrangement by the number of particles in the inner shell of the crystal , and are stabilized by the energetic contribution of zero - point motion . to our knowledge , this is the first demonstration of such a physical effect , which can be observed experimentally in mesoscopic dipolar clusters of polar molecules .
it is observed for clusters of specific sizes , namely @xmath0=12 and @xmath1 , for @xmath2 .
this effect takes place in the mesoscopic crystalline phase , in which quantum - mechanical exchanges of indistinguishable particles are stroingly suppressed ; it is therefore expected to take place irrespective of the quantum statistics of the particles .
+ we describe the mathematical model of the system utilized here in the next section ; we then briefly review the computational methodology utilized here , and illustrate the results of the calculations and main physical conclusions in the following sections .
we consider a setup where @xmath0 dipolar particles of mass @xmath3 are confined to a 2d plane by applying a strong transverse trapping field @xcite , e.g a 1d optical lattice .
dipole moments are aligned perpendicular to the plane , with a dc induced dipole moment @xmath4 .
we assume an additional in - plane parabolic trap with frequency @xmath5 , as realized by a magnetic dipole trap , or a single site of a large spacing optical lattice .
for pure dipolar interactions , the many - body hamiltonian in dimensionless form is given by @xmath6 + \sum_{i < j}\frac{1}{|{{\mbox{\boldmath$r$}}}_i-{{\mbox{\boldmath$r$}}}_j|^3},\end{aligned}\ ] ] where the unit of length is @xmath7 , that of energy is @xmath8 , and where @xmath9 , @xmath10 being the characteristic ( dimensionless ) confining length of the parabolic trap .
@xmath11 plays the role of control parameter here , as a greater value of @xmath11 means that the trap is compressed , and the particle density correspondingly increased .
we assume for definiteness that particles obey bose statistics , and all of the numerical results presented here are obtained with this assumption ; however , the conclusions of this study pertaining to the crystalline phases are independent of quantum statistics
. + the basic ground state physics of the system described by ( [ eq : eqhamres ] ) has been characterized in previous works @xcite ; generally speaking , it mimics qualitatively that of the bulk system @xcite . for sufficiently small @xmath11 ,
the cluster is in a low - density , weakly interacting superfluid phase @xcite . in the limit of strong confinement , the potential energy dominates , and the ground state takes on the classical lowest - energy crystalline configuration , predictable by straightforward potential energy minimization ( last two terms of eq . ) .
+ for a cluster comprising a sufficiently small number of particles ( i.e. , @xmath12 ) , the finite size of the system allows for the existence of several intermediate ,
supersolid " phases @xcite displaying simultaneously a finite superfluid response and crystalline order @xcite .
these phases are not expected to survive in the thermodynamic limit ; indeed , the nature of a supersolid phase of a system of dipolar bosons in two dimensions is predicted @xcite to deviate significantly from that of mesoscopic supersolids .
+ in this manuscript we mostly focus our attention on mesoscopic crystalline phases , in which particles are localized and consequently quantum - mechanical exchanges are suppressed .
we investigated the low temperature ( @xmath13 ) properties of the system described by ( [ eq : eqhamres ] ) by means of computer simulations , based on the continuous - space worm algorithm @xcite . since this technique is by now
fairly well - established , and extensively described in the literature , we shall not review it here .
details of the simulation are standard .
+ because we are interested in the physics of the system in the @xmath13 limit , we report here results corresponding to a temperature @xmath14 sufficiently low to regard them as essentially ground state estimates . a quantitative criterion to assess whether the temperature @xmath14 of the simulation is sufficienty low ,
consists of comparing it to the computed particle mean kinetic energy @xmath15 .
for all the simulations for which we present results in figures [ f2 ] and [ f4 ] , it is @xmath16 .
+ the use of a finite temperature technique to investigate ground state physics might appear counterintuitive , considering that methods exist purposefully designed to study the ground state of a many - body system ( e.g. , diffusion monte carlo ) . in practice , however , finite - temperature techniques often prove superior , even to determine ground state properties ( naive statements by dmc practitioners notwithstanding ) .
this is mainly owing to the unbiasedness of finite temperature methods , which , unlike their @xmath14=0 counterparts , require no _ a priori _ physical input ( e.g. , a trial wave function ) . moreover , finite temperature methods allow one to assess more easily and reliably quantities other than the energy , including off - diangonal correlations .
we compute both global and local superfluid responses , using standard methodology @xcite .
all of the results presented here are extrapolated to the limit of zero imaginary time step @xcite .
for a two - dimensional system of @xmath0=12 dipolar bosons confined in a planar harmonic trap of different characteristic length @xmath17 . ] figure [ fig : fig2 ] displays radial density profiles ( computed with respect to the center of the trap ) for the ground state of a mesoscopic assembly of @xmath0=12 particles , confined in two - dimensional harmonic traps of varying strength ( i.e. , characteristic length ) . in order to facilitate the comparison
, we plot the radial density @xmath18 as a function of @xmath19 .
figure [ f2 ] shows particle density maps obtained from statistically representative configuration snapshots ( i.e. , particle world lines ) for four distinct cases , corresponding to different values of the harmonic confining length @xmath17 , namely @xmath20 and 0.00056 . by statistically " representative , we mean that every configuration generated in the simulation is physically equivalent to that shown in the figure ; in particular , for the case in which the system takes on the crystalline arrangements shown in panels ( c ) and ( d ) , every configuration in the monte carlo random walk only differs from that shown by a mere rotation .
[ cols="^,^ " , ] + most interestingly , we find that in the range of values of @xmath17 in which the non - classical crystal ( e.g. , for @xmath0=12 , @xmath21 ) is the ground state , the classical one ( @xmath22 ) remains a low - temperature metastable conguration .
specifically , in a finite range of temperature the system is observed in the monte carlo simulation to switch between @xmath21 and @xmath22 .
this is shown qualitatively in figure [ f3 ] , displaying typical configuration snaphots yielded by a simulation carried out at a temperature @xmath23 ; this is approximately thirty times higher than that yielding essentially ground state estimates , for which the configuration is that shown in the right panel of figure [ f3 ] .
the kinetic energy per particle at this high temperature is around @xmath24 .
melting of the cluster into a featureless fluid takes place at a higher temperature , in fact at a value close to that of the bulk crystal @xcite .
a similar coexistence is observed for the @xmath0=19 cluster as well .
the physical behavior described above , namely the coexistence of two phases " of the cluster at finite temperature , with the classical one emerging as the temperature is raised , is analogous to what predicted for some small parahydrogen clusters ( between 20 and 30 molecules ) @xcite .
this intriguing type of phase coexistence is allowed by the finite size of the system , with the ensuing interplay between bulk and surface energy .
an important difference is that , in the case of @xmath25-h@xmath26 clusters , coexistence is between superfluid and crystalline phases , the superfluid being the ground state , underlain by particle exchanges . at higher temperature ,
as the thermal wavelength of the molecules becomes shorter , exchanges are suppressed and the system finds it energetically advantageous to take on solid - like . in the case of the dipolar cluster discussed here ,
coexistence is between two crystalline phases .
exchanges are infrequent in these phases , quantum - mechanical effects consisting almost exclusively of zero - point motion .
an important consequence of this fact , is that a transition between non - classical and classical ground states can be expected to occur in mesoscopic dipolar systems of either bose or fermi particles .
this is because any physical difference between bose and fermi systems can emerge only in the presence of quantum - mechanical exchanges of indistinguishable particles , which are strongly suppressed in these mesoscopic crystals .
+ in conclusion , a numerical studies of mesoscopic systems of dipolar particles confined to two dimensions yields evidence of non - classical crystalline ground states for clusters of specific numbers , underlain by quantum zero - point motion .
the mesoscopic phases described in this work can be realized with polar molecules of current experimental interest .
for example ,
for a moderate in - plane confinement @xmath27 1khz and fully polarized rbcs ( @xmath28 debye ) , lics ( @xmath29 debye ) , and sro molecules ( @xmath30 debye ) , @xmath17 reaches values as small as @xmath31 and 0.002 , respectively . since the in - situ interparticle distances can be of the order of several hundreds of nm , it may be possible to directly address single particles _ in situ _ , and thus image the _ spatial structure _ of the crystalline phases above using , e.g. , tightly focused beams .
alternatively , we propose the following method , which amounts to a version of a _ magnifying lens_. at a given time @xmath32 the ( dc or ac ) fields inducing the dipole - dipole interactions are switched off , and the in - plane harmonic confinement is inverted in sign . because of this inverted potential , each particle experiences a radial acceleration which depends on its _ spatial position _ at time @xmath32 . after a certain time - of - flight , the particles can be , e.g , ionized and their positions recorded on a ion plate with unit efficiency , providing a magnified picture of the in - situ spatial configuration .
+ obviously , in a strict thermodynamic sense one can not speak of `` phase '' of a finite system . henceforth ,
therefore , this word should be interpreted loosely , in reference to qualitative , average properties of the system in its most probable configurations .
obviously , any finite cluster can rotate without dissipation at a temperature lower than @xmath33 , @xmath34 being the classical moment of inertia of the cluster .
such a free rigid rotation constitutes a conceptually different scenario from the one referred to here as superfluid " . | remarkable preparation of cold dipolar particles ultracold ground magnetic atoms display dipole - dominated dynamics in quantum regime @xcite huge electric dipole moments to thousands present in rydberg states alkali atoms experiments to control interactions correlations with ground state atoms by admixing with laser light rydberg states finite heating losses emission polar molecules in electronic rovibrational ground state combine stability with dipole moments to large dipolar interactions experiments to exploit interactions with reduced trapping geometries add collisional stability opens door to study of correlated quantum phases with long range interactions density - driven superfluid - crystal quantum phase transition finite confined dipolar assemblies worthy investigation evolution of physical properties as size bulk limit work with fermionic analyzed effects quantum statistics on transitions between wigner - type states in traps attention on quasi - two - dimensional harmonically confined dipolar systems of mesoscopic size small number particlesfundamental superfluidity in cluster explored in helium hydrogen @xcite droplets quantum simulations deviation from bulk behaviour dipolar system small size predicted non - classical ground state crystalline arrangements intermediate range confinement strength configurations differ classical by number particles inner shell crystal stabilized by energetic contribution zero - point motion first demonstration of physical effect observed in mesoscopic dipolar clusters observed for specific sizes @xmath0=12 @xmath1 effect in mesoscopic crystalline phase quantum - mechanical exchanges of suppressed expected irrespective of quantum statistics particles describe mathematical model system review computational methodology illustrate results calculations physical conclusions consider setup @xmath0 dipolar particles mass @xmath3 confined to 2d plane strong transverse trapping field @xcite 1d optical lattice dipole moments aligned perpendicular to plane with dc induced dipole moment @xmath4 assume additional in - plane parabolic trap with frequency @xmath5 realized by magnetic dipole trap large spacing optical latticedipolar interactions many - body hamiltonian dimensionless form given by @xmath6 + \sum_{i < j}\frac{1}{|{{\mbox_i-{{_j|^3},\end{aligned} unit of length @xmath7 energy @xmath8 @xmath9 @xmath10 characteristic confining length of parabolic trap @xmath11 control parameter greater value trap compressed particle density increased assume particles obey bose statistics numerical results obtained assumption conclusions crystalline phases independent of quantum statistics basic ground state physics system described by eq eqhamres characterized previous works @xcite mimics bulk system @xcite small @xmath11 cluster in low - density weakly interacting superfluid phase strong confinement potential energy dominates ground state classical lowest - energy crystalline configuration predictable by potential energy minimization cluster small number particles @xmath12 ) finite size system allows several intermediate supersolid phases @xcite finite superfluid response crystalline orderphases not expected survive thermodynamic limit nature supersolid phase of system dipolar bosons predicted to deviate significantly from mesoscopic supersolids in manuscript focus attention on mesoscopic crystalline phases particles localized quantum - mechanical exchanges suppressed investigated low temperature ( @xmath13 ) properties system described by eq eqhamres computer simulations based on continuous - space worm algorithm technique well - established described not review here details simulation standard interested in physics system in @xmath13 limit report results corresponding to temperature @xmath14 low ground state estimates quantitative criterion assess temperature @xmath14 simulation low comparing to computed particle mean kinetic energy @xmath15 simulations present results in figures [ f2 ] and [ f4 ] it is @xmath16 use of finite temperature technique to investigate ground state physics might counterintuitive methods designed to study ground state system practice finite - temperature techniques often prove superior ground state properties owing to unbiasedness of finite temperature methods require no physical inputfinite temperature methods assess quantities other than energy including off - diangonal correlations compute global local superfluid responses using standard methodology @xcite results extrapolated to zero imaginary time step @xcite two - dimensional system of @xmath0=12 dipolar bosons in planar harmonic trap different characteristic length @xmath17 figure displays radial density profiles center trap for ground state of mesoscopic assembly of @xmath0=12 particles in two - dimensional harmonic traps varying strength length plot radial density @xmath18 as function of @xmath19 figure f2 shows particle density maps from statistically representative configuration snapshots for four cases different values of harmonic confining length @xmath17 @xmath20 and 0.00056 statistically representative every configuration simulation physically equivalent to figure system crystalline arrangements configuration random walk differs by rotation in range of values of @xmath17 non - classical crystal is ground state classical @xmath22 ) remains low - temperature metastable congurationfinite range temperature system simulation switch between @xmath21 and @xmath22 shown in figure [ f3 ] typical configuration snaphots simulation at temperature @xmath23 thirty times higher than ground state estimates configuration shown in right panel kinetic energy per particle at high temperature around @xmath24 melting cluster into featureless fluid at higher temperature close to bulk crystal @xcite similar coexistence for @xmath0=19 cluster physical behavior coexistence of two phases cluster at finite temperature classical one emerging as temperature raised analogous to small parahydrogen clusters between 20 30 molecules ) @xcite phase coexistence allowed by finite size system interplay between bulk and surface energy @xmath25-h@xmath26 clusters coexistence between superfluid and crystalline phases superfluid ground state particle exchanges higher temperature thermal wavelength shorter exchanges suppressed system advantageous solid - like dipolar cluster coexistence between two crystalline phases exchanges infrequent quantum mechanical effects zero - point motion transition between non - classical and classical ground states expected in mesoscopic dipolar systemsphysical difference between bose fermi systems quantum - mechanical exchanges of indistinguishable particles suppressed in mesoscopic crystals numerical studies of mesoscopic systems of dipolar particles two dimensions non - classical crystalline ground states for specific numbers underlain by quantum zero - point motion mesoscopic phases realized with polar molecules experimental interest moderate in - plane confinement @xmath27 1khz fully polarized rbcs lics sro molecules @xmath17 reaches values small as @xmath31 and 0.002 in - situ interparticle distances several hundreds of nm to address single particles image spatial structure of crystalline phases using tightly focused beams method magnifying lens dc or ac fields dipole - interactions switched off in - plane harmonic confinement inverted each particle experiences radial acceleration depends on spatial position after time flight particles be ionized positions recorded on ion plate efficiency magnified picture of in - situ spatial configuration thermodynamic sense can not speak of `` phase '' of finite systemword interpreted loosely reference to qualitative average properties system in probable configurations finite cluster can rotate without dissipation at temperature lower than @xmath33, @xmath34 classical moment inertia free rigid rotation different scenario from superfluid | 0.508948 | 2,862 | 1,399 | , Saving $0.1 in GPT-4. |
67,637 | we have studied time variability in the flux from the flat spectrum synchrotron radiation of the blandford & knigl ( 1979 ) model for relativistic , conical jets .
the resulting model has been applied to the flux variation of the flat spectrum of grs 1915 + 105 observed by fender & pooley ( 2000 ) .
this comparison has highlighted a fundamental problem of the blandford & knigl ( 1979 ) model in that it requires unphysically large electron densities to explain the flux levels observed from the flat spectrum of grs 1915 + 105 .
2@xmath0 1s@xmath1 | to explain the infrared and millimetre emission oscillations from grs 1915 + 105 , we have adopted the blandford & knigl ( 1979 , hereafter bk79)@xcite model of a relativistic jet , containing power - law electron distributions .
adiabatic expansion gives the jet a conical shape with an electron density and magnetic field strength that decreases with radius .
the jet emits partially self - absorbed synchrotron emission over a radial range covering several orders of magnitude , and hence from a wide range of optical depths .
therefore , the total jet emission consists of the summation self - absorbed synchrotron spectra from regions of the jet of decreasing optical depth ( or increasing radius ) , and hence electrons at each radius , from the base to the top of the jet , emit a spectrum that is progressively shifted towards lower frequencies .
this summation results in a ` flat ' region to the spectrum , which has a slope of zero if energy losses due to adiabatic decompression are ignored ( as in the bk79 model ) .
however , if some energy losses do occur then this region has a positive , inverted slope , the magnitude of which depends upon the magnitude of the energy loss .
the extent of the flat spectral region , @xmath2 , is solely determined by the emission region size , which in our model is fixed by the jet velocity and time - scale of variability . at the lower frequency end of the flat spectrum there is a smooth transition to an optically thick synchrotron spectrum with @xmath3 , and the high frequency end smoothly terminates into an optically thin , @xmath4 , synchrotron spectrum . full detail of the derivation of our model may be found in collins , kaiser & cox ( 2002)@xcite . the emission spectrum predicted by this model is given in terms of the optical depth function , @xmath5 .
@xmath6 \,\mathrm{d}r \mbox { mjy},\ ] ] where , @xmath7 @xmath8 is the distance to the jet , @xmath9 defines the extent of the emission region , and all parameters have si units unless stated otherwise .
the model has two parameters .
the first , which we have named @xmath10 , controls the optical depth as a function of radius , @xmath5 , and thus shifts the flat region of the spectrum to higher or lower frequencies .
the second model parameter , which we have named @xmath11 , only effects the flux normalisation .
the @xmath10 parameter also has an effect upon the flux normalisation , and therefore we first fit the @xmath10 parameter to the spectral shape , before fitting @xmath11 to the flux normalisation .
the model parameter , @xmath10 , is related to the physical parameters of the system by @xmath12 , where @xmath13 is the normalisation value of the electron density distribution , and @xmath14 is the magnetic field strength . the @xmath11 parameter is defined as @xmath15 .
the half - width of the jet @xmath16 is fixed by observational constraints , and therefore from @xmath10 and @xmath11 , we may determine @xmath17 and @xmath18 . to implement time dependence in this model as a variation in the flux level of the flat spectrum it would be simplest to make @xmath11 a function of time with @xmath10 fixed .
however , physically this would require the electron density to vary by a process that exactly compensates variations in the magnetic field strength .
therefore we choose to make @xmath10 a function of time , and fix @xmath11 , which may be simply interpreted , physically , as a variation to the electron density injected into the jet .
time dependence is then included into equation [ efn ] by defining @xmath19 , where @xmath20 is the bulk velocity of the jet material , and by increasing the injected value of @xmath21 with time according to a gaussian function .
from this model we can predict three observable parameters to which our model parameters may be fit ; the time lag between the peak fluxes at each frequency of observation , the flux ratio of these peak fluxes , and the flux normalisation value .
the value of the @xmath10 model parameter affects all three of these observables .
the flux ratio is determined by the position of the flat spectral region with respect to the two frequencies of observation , which is defined by @xmath10 .
the time lags are also determined by the position of the flat spectral region , because the optically thin end of the flat spectrum is formed at small radii , whereas the optically thick end is formed at much larger radii .
therefore the closer that both frequencies of observation are to the optically thin end , the smaller the time lag .
fender & pooley ( 2000)@xcite observed large amplitude oscillations in the emission from grs 1915 + 105 at an infrared wavelength ( 2.2 @xmath22 m ) and a millimetre wavelength ( 1.3 mm ) .
the time lag between the peak fluxes from each wavelength for the first oscillation was approximately 25 s , with the dereddened infrared flux reaching 525 mjy and the millimetre flux reaching 340 mjy , giving a peak flux ratio of 1.5 .
however , the extinction correction applied to the infrared flux is uncertain , such that the peak flux ratio is @xmath23 .
such a small flux ratio over a frequency range covering three orders of magnitude is suggestive of the ` flat ' spectrum produced by relativistic jets . for the fixed parameters of the model we assume that the distance to grs 1915 + 105 is 11 kpc , the opening half - angle of its jet is @xmath24 , the base of the emission region is at @xmath25 m , and the jet material has a bulk velocity of @xmath26 .
the value of the @xmath10 parameter was fit to the observed time lag , and , for this value of @xmath10 , @xmath11 was fit to the flux normalisation of the peak millimetre flux . from these model parameters
the corresponding physical parameters of jet were determined to be @xmath27 t and @xmath28 m@xmath29 .
the model also predicts a flux ratio of 1.1 which agrees with the observations .
we have found that it is impossible to obtain flux levels of the order of @xmath30 mjy without unphysically high electron densities , for time lags of less than 300 s , as demonstrated by figure 1 .
this is not just a problem with this time dependent model , but is a fundamental problem of the bk79 model , as can be seen by performing the same fitting process to the observed flux ratio . to agree with the observed flux levels and the time lag between peaks with realistic electron densities ,
the model must either predict a flux @xmath31 times larger with the fitted value of @xmath21 , or predict the observed time lag with a value of @xmath10 that is @xmath32 times larger . since relativistic effects are negligible in the grs 1915 + 105 jet ( see collins et al .
2002 ) , the later case can be immediately ruled out , as it would require the model s fixed parameters to be adjusted beyond physical limits .
it is also difficult to justify the higher flux required by former solution , as doppler beaming and geometrical effects are not sufficient . | infrared millimetre emission oscillations from grs 1915 + 105 adopted blandford & knigl 1979@xcite model relativistic jet power - law electron distributions adiabatic expansion gives jet conical shape electron density magnetic field strength decreases with radius jet emits self - absorbed synchrotron emission radial range orders magnitude wide range optical depths total jet emission self - absorbed synchrotron spectra from regions decreasing optical depth increasing radius electrons each radius emit spectrum shifted towards lower frequencies results in flat region slope zero if energy losses ignored energy losses region positive inverted slope magnitude depends energy loss extent flat spectral region determined by emission region size fixed jet velocity time - scale variability lower frequency end smooth transition to optically thick synchrotron spectrum @xmath3 high frequency end terminates into optically thin @xmath4 synchrotron spectrum detail derivation in collins , kaiser cox ( 2002)@xcite emission spectrum predicted optical depth function @xmath5@xmath6 \mathrm{d}r \mbox { mjy} @xmath7 @xmath8 distance to jet @xmath9 defines emission region all parameters have si units unless stated model has two parameters first @xmath10 controls optical depth radius shifts flat region spectrum to higher or lower frequencies second parameter @xmath11 effects flux normalisation flux normalisation fit @xmath10 to spectral shape before @xmath11 flux normalisation @xmath10 related to physical parameters by @xmath12 @xmath13 normalisation value electron density distribution @xmath14 magnetic field strength @xmath11 parameter defined as @xmath15 half - width of jet @xmath16 fixed by observational constraints from @xmath10 and @xmath11 determine @xmath17 and @xmath18 implement time dependence as variation in flux level spectrum make @xmath11 function of time with @xmath10 fixed electron density make @xmath10 function of time fix @xmath11 as variation to electron density injected into jettime dependence included into equation defining @xmath19 @xmath20 bulk velocity jet material increasing injected value @xmath21 with time gaussian function model three observable parameters time lag between peak fluxes flux ratio flux normalisation value @xmath10 model parameter affects observables flux ratio determined by position flat spectral region frequencies defined by @xmath10 time lags determined by position spectral region optically thin end formed at small radii thick end larger radii closer frequencies to optically thin end smaller time lag fender & pooley observed large amplitude oscillations in emission from grs 1915 + 105 at infrared wavelength 2.2 m millimetre wavelength ( 1.3 mm time lag between peak fluxes approximately 25 s infrared flux 525 mjy millimetre flux 340 mjy peak flux ratio 1.5 extinction correction infrared flux uncertain peak flux ratio is @xmath23 small flux ratio over frequency range suggestive of flat spectrum by relativistic jetsfixed parameters model assume distance to grs 1915 + 105 11 kpc opening half - angle jet @xmath24 emission region @xmath25 m jet material bulk velocity @xmath26 @xmath10 parameter fit observed time lag @xmath11 flux normalisation peak millimetre flux physical parameters jet @xmath27 t @xmath28 m@xmath29 model predicts flux ratio 1.1 agrees with observations impossible obtain flux levels @xmath30 mjy without high electron densities time lags less than 300 s demonstrated figure 1 problem fundamental problem bk79 model fitting process observed flux ratio agree flux levels time lag peaks electron densities model must predict flux @xmath31 times larger with value @xmath21 or time lag @xmath10 @xmath32 times larger relativistic effects negligible in grs 1915 + 105 jet 2002 later case ruled out model s fixed parameters beyond physical limits difficult to justify higher flux required former solution doppler beaming geometrical effects not sufficient | 0.269757 | 1,599 | 832 | , Saving $0.0 in GPT-4. |
67,645 | motivated by the recent demonstration of its use as a tool for the detection and characterization of phase - shape correlations in multivariate time series , we show that eigenvalue decomposition can also be applied to a matrix of indices of bivariate phase synchronization strength .
the resulting method is able to identify clusters of synchronized oscillators , and to quantify their strength as well as the degree of involvement of an oscillator in a cluster .
since for the case of a single cluster the method gives similar results as our previous approach , it can be seen as a generalized synchronization cluster analysis , extending its field of application to more complex situations .
the performance of the method is tested by applying it to simulation data . | the eigenvalue decomposition of the equal - time correlation matrix of a set of signals is one of the standard tools of multivariate data analysis ( _ cf . _
recently , @xcite demonstrated the usefulness of the eigenvalue decomposition of the correlation matrix specifically as a tool for the detection of phase - shape correlations in multivariate data sets .
they showed that changes in the degree of synchronization in all or a subset of signals are reflected in coordinated changes in the highest and lowest eigenvalues , and that information about the channels involved and the type of their interaction can be obtained from the corresponding eigenvectors .
while the correlation of time series may indicate the synchronization of the oscillators they are obtained from , the physical concept of synchronization refers specifically to the adjustment of the rhythms of oscillators , i.e. to the relative dynamics of their phases rather than their amplitudes @xcite .
moreover , there is a regime in the dynamics of coupled chaotic oscillators in which the phase difference is bounded while the amplitudes remain uncorrelated @xcite , called phase synchronization . in this paper , we show that in order to focus the analysis on synchronization relations , it is possible to replace the matrix of correlation coefficients with a matrix of indices of bivariate phase synchronization strength . combined with an additional step of sorting signals into groups , eigenvalue decomposition can operate as a synchronization cluster analysis , generalizing the previous approach of @xcite .
the correlation matrix @xmath0 of a set of data channels @xmath1 , @xmath2 , consists of the correlation coefficients @xmath3 $ ] between channels .
its eigenvalues @xmath4 and eigenvectors @xmath5 are defined by the equation @xmath6 which in general has @xmath7 different solutions , @xmath8 . in the following
we assume that the eigenvectors are normalized , @xmath9 , and the solutions have been sorted according to the eigenvalues , @xmath10 .
the eigenvectors and -values of @xmath0 are real - valued , and the eigenvalues are non - negative . because a matrix becomes the diagonal matrix of its eigenvalues by being transformed into the basis of its eigenvectors and the trace of a matrix is invariant under such a transform , for every correlation matrix holds @xmath11 . in the uncorrelated case
this is trivially fulfilled by @xmath12 for all @xmath13 . with a deviation from this
, each increase of an eigenvalue above @xmath14 has to be compensated for by at least one other eigenvalue becoming smaller than @xmath14 , such that this value gives a natural distinction between `` large '' and `` small '' eigenvalues .
the quantification of phase synchronization is based on the instantaneous phase @xmath15 of each oscillator @xmath2 . how these phases are determined in the special case
is not important here ; if the given data are time series , the standard approach is the hilbert transform for narrowband data , or the morlet wavelet transform for broadband signals ( for a discussion , see @xcite or @xcite , sec .
the statistical strength of phase synchronization of two oscillators @xmath16 and @xmath17 can then be defined as the `` peakedness '' of the distribution of the phase difference @xmath18 ; here we use the measure @xmath19 where @xmath20 enumerates the realizations in the given sample . for the continuum from no to perfect phase synchronization , this measure takes on values from @xmath21 to @xmath14 .
@xmath22 can be seen as the modulus of the complex correlation coefficient of signals @xmath23 , and its decomposition shares the properties given above for @xmath0 . since @xmath22 is a nonlinear measure , in this case the eigenvalue decomposition
can no longer be interpreted with regard to a linear transform of data channels into source channels .
but still the result of the decomposition can be used as a means to analyze the structure of synchronization relations .
we will demonstrate this with two basic , artificially constructed examples of synchronization matrices .
describe the cluster structure . ] , but only in different superpositions . ]
figure [ attenuated ] shows the synchronization matrix for a system consisting of three clusters of synchronized oscillators ( with no synchronization between clusters and a different degree of involvement of each oscillator in its cluster ) along with the result of the eigenvalue decomposition .
there are three eigenvalues larger than one , and the corresponding eigenvectors describe clearly the extent of the clusters as well as the degree of involvement of the individual oscillators .
there is also a correspondence between the size and strength of internal synchronization of the clusters and the three eigenvalues .
in contrast , the remaining eigenvectors seem not to contribute to the description of the synchronization clusters .
the interpretation of the eigenvalue decomposition of @xmath22 therefore has these aspects : 1 ) synchronization clusters are identified by eigenvalues @xmath24 .
the eigenvalues themselves quantify the _ strength _ of the clusters .
2 ) for each cluster , the corresponding eigenvector describes its internal structure . because @xmath25 , the index @xmath26 quantifies the relative involvement of channel ( oscillator ) @xmath16 in cluster @xmath13 .
3 ) combining both , the `` absolute '' involvement of channel @xmath16 in cluster @xmath13 can be quantified by the _ participation index _ @xmath27 .
figure [ synchronized ] gives the result for a matrix consisting of the same three clusters , but with additional inter - cluster synchronization . because of the coupling between them , the three clusters no longer appear in separate components of its eigenvalue decomposition .
there are still three @xmath24 , but the eigenvectors consist of superpositions of the clusters . to account for this , the oscillators belonging to the three clusters have to be identified explicitely . for the synchronization matrix of fig .
[ synchronized ] .
oscillators are attributed to that cluster for which its participation index is maximal : blue , green , or red . center : synchronization matrix with removed inter - cluster synchronization .
right : result of the eigenvalue decomposition of the trimmed matrix ; shown are the participation indices and cluster strengths . ]
this can be done by means of the participation indices ; the procedure is illustrated in fig .
[ synchronized : ca ] .
each oscillator is attributed to that cluster for which its participation is maximal . in this way
, the three clusters consisting of oscillators # 15 ( blue ) , 69 ( green ) , and 1012 ( red ) are correctly identified . in a second step , all of the indices for inter - cluster synchronizations
are set to zero , and the eigenvalue decomposition is repeated on the trimmed matrix . for the result of this decomposition ,
the interpretation given above is valid again .
the procedure described in the last section is an approach to synchronization cluster analysis .
the method identifies clusters of synchronized oscillators and quantifies the strength of the clusters as well as the degree of involvement of each oscillator in its cluster . in a previous paper
, @xcite introduced another approach that was limited to a single cluster of synchronization ; it assumed that all of the oscillators belong to the same cluster , and focused on the quantification of the degree of oscillator participation .
the algorithm derived from the observation that under relatively general conditions the synchronization indices within a cluster can be written as the product of factors @xmath28 , @xmath29 which can be interpreted as the synchronization strength between oscillator @xmath16 and the cluster itself , its `` to - cluster synchronization strength '' . ) and the participation indices given by eigenvalue decomposition ( vertical scale : @xmath30 ) in 20 simulation runs . ] we investigated the relationship between the results of the two methods by applying them to simulation data that conforms to the presupposition of a single factorizable synchronization cluster . for @xmath31 oscillators
, the @xmath28 were drawn randomly from the uniform distribution over @xmath32 $ ] ; then the @xmath33 were calculated based on @xmath34 samples of the phase difference from a wrapped normal distribution ( _ cf .
_ @xcite ) .
figure [ scacluster ] shows the result for the relation between the output of the earlier method and the participation indices for the strongest cluster .
though the mathematical background of both methods is clearly different , the plot shows an almost functional dependency , which can be roughly described by @xmath35 .
the region of zero participation indices for @xmath36 below about @xmath37 comes from the attribution of weakly synchronized oscillators to another cluster by the new method ; in this range the observed value can no longer be reliably distinguished from no synchronization for the given sample size .
since there is a clear relationship between the results of both methods in the case where the assumptions of the earlier method hold , the cluster analysis based on the eigenvalue decomposition of the synchronization matrix can be seen as a generalization of our previous approach to synchronization cluster analysis .
to check the performance of the new method , we applied it to data obtained from the numerical simulation of a system that is known to exhibit clusters of phase synchronization , previously investigated by @xcite ( see eq . 1 & sec . iv a ) .
the system consists of a chain of @xmath38 phase - coherent rssler oscillators with a diffusive coupling in the @xmath39-component of strength @xmath40 .
the natural frequencies of the oscillators increase linearly along the chain , in the range @xmath41 .
phases were defined as the angle in the @xmath42-plane . to also test for the influence of small sample size and noise
, we additionally applied the algorithm to the synchronization matrix obtained from data reduced to @xmath34 approximately independent samples with @xmath43 white noise added to the time series data .
samples , plus @xmath43 measurement noise . ]
the result is shown in fig .
[ rchain ] .
the analysis identifies two clusters of synchronized oscillators ( # 116 & 1720 ) .
there is an almost constant high participation of oscillators in their cluster , resulting in cluster strengths close to the number of involved oscillators .
decreased participation occurs near the border between the clusters , where the coupling along the chain pulls oscillators away from the common dynamics of their group .
the result for reduced sample size and measurement noise is very similar to the original .
though participation indices and cluster strengths are slightly smaller due to the noise , the same two clusters are clearly identified .
this result indicates that the cluster analysis based on eigenvalue decomposition is relatively robust against small sample size and noise .
we introduced a new approach to synchronization cluster analysis based on eigenvalue decomposition .
the new method can be seen as a generalization of our previous approach , extending its field of application to situations including multiple clusters .
the algorithm was tested on simulation data and shown to be robust against small sample size and noise .
future work will be on the use of the method to investigate synchronization patterns in eeg data
. a matlab implementation of the algorithm can be obtained from the corresponding author.this work has been supported by deutsche forschungsgemeinschaft . | eigenvalue decomposition of equal - time correlation matrix of signals is standard multivariate data analysis @xcite demonstrated usefulness for detection of phase - shape correlations in data sets showed changes in synchronization in signals reflected in changes highest lowest eigenvalues information about channels interaction obtained from eigenvectors correlation of time series synchronization of oscillators physical concept of synchronization refers to adjustment of rhythms of oscillators relative dynamics of phases amplitudes regime in dynamics of coupled chaotic oscillators phase difference bounded amplitudes uncorrelated called phase synchronization focus on synchronization to replace matrix of correlation coefficients with matrix of indices of bivariate phase synchronization strength with sorting signals into groups eigenvalue decomposition can as synchronization cluster analysis previous approach correlation matrix @xmath0 of data channels @xmath1 consists of correlation coefficients @xmath3 between channels eigenvalues @xmath4 and eigenvectors @xmath5 defined by equation @xmath6 different solutions assume eigenvectors are normalized solutions sorted according to eigenvalueseigenvectors -values of @xmath0 real - valued non - negative matrix becomes diagonal matrix of eigenvalues transformed into basis trace invariant under every correlation matrix holds @xmath11 uncorrelated case fulfilled by @xmath12 for @xmath13 each increase eigenvalue above @xmath14 compensated by one other eigenvalue smaller than distinction between large small eigenvalues quantification of phase synchronization based on instantaneous phase @xmath15 of each oscillator @xmath2 phases data time series standard approach hilbert transform for narrowband or morlet wavelet transform for broadband statistical strength of phase synchronization of oscillators @xmath16 and @xmath17 defined as peakedness of distribution phase difference @xmath18 measure @xmath19 where @xmath20 enumerates realizations sample continuum no to perfect phase synchronization takes on values from @xmath21 to @xmath14 @xmath22 modulus of complex correlation coefficient of signals @xmath23 decomposition shares properties for @xmath0 @xmath22 nonlinear measure eigenvalue decomposition interpreted linear transform data into result decomposition analyze synchronization relationsdemonstrate with two examples synchronization matrices describe cluster structure different superpositions figure attenuated shows synchronization matrix for system three clusters synchronized oscillators no synchronization between clusters different degree involvement each result of eigenvalue decomposition three eigenvalues larger than one describe extent clusters degree involvement oscillators correspondence between size strength internal synchronization clusters three eigenvalues remaining eigenvectors not contribute to description clusters interpretation of eigenvalue decomposition @xmath22 synchronization clusters identified by eigenvalues @xmath24 quantify strength clusters each cluster eigenvector describes internal structure index @xmath26 quantifies relative involvement of channel @xmath16 in cluster @xmath13 absolute involvement of channel @xmath16 in cluster @xmath13 quantified by participation index @xmath27 figure synchronized result for matrix same three clusters additional inter - cluster synchronization coupling three clusters no longer appear in separate components eigenvalue decomposition still three @xmath24 eigenvectors consist of superpositions of clusters oscillators belonging to three clusters identified explicitely synchronization matrix fig synchronizedoscillators attributed to cluster participation index maximal : blue green red center synchronization matrix with removed inter - cluster synchronization right : result eigenvalue decomposition trimmed matrix participation indices cluster strengths participation indices procedure illustrated in fig . synchronized ca each oscillator attributed to cluster participation maximal three clusters oscillators # 15 blue 69 green 1012 red identified indices for inter - cluster synchronizations set to zero eigenvalue decomposition repeated on trimmed matrix result decomposition interpretation valid procedure last approach synchronization cluster analysis identifies clusters synchronized oscillators quantifies strength degree involvement each oscillator previous paper @xcite introduced approach limited to single cluster synchronization assumed oscillators same cluster focused on quantification degree oscillator participation algorithm derived observation synchronization indices cluster product of factors @xmath28 , @xmath29 interpreted synchronization strength between oscillator @xmath16 cluster to - cluster synchronization strength participation indices given by eigenvalue decomposition vertical scale : @xmath30 ) in 20 simulation runs investigated relationship between results methods simulation data single factorizable synchronization cluster@xmath31 oscillators @xmath28 drawn randomly from uniform distribution over @xmath32 @xmath33 calculated based on @xmath34 samples phase difference from normal distribution figure scacluster shows relation between output earlier method participation indices strongest cluster mathematical background different plot shows functional dependency described by @xmath35 region zero participation indices for @xmath36 below @xmath37 from attribution weakly synchronized oscillators to another cluster new method observed value distinguished from no synchronization sample size clear relationship between results methods cluster analysis based eigenvalue decomposition synchronization matrix generalization of previous approach synchronization cluster analysis check performance new method applied to data numerical simulation system phase synchronization investigated @xcite system chain @xmath38 phase - coherent rssler oscillators diffusive coupling in @xmath39-component strength @xmath40 natural frequencies increase linearly along chain range @xmath41 phases defined as angle in @xmath42-plane influence small sample size noise applied algorithm to synchronization matrix from data reduced to @xmath34 independent samples @xmath43 white noise added time series data result shown in fig .analysis identifies two clusters synchronized oscillators ( # 116 & 1720 ) almost constant high participation cluster strengths close to number involved decreased participation near border between clusters coupling pulls oscillators away from dynamics result reduced sample size measurement noise similar to original participation indices cluster strengths smaller due to noise same two clusters identified cluster analysis eigenvalue decomposition robust against small sample size noise introduced new approach to synchronization cluster analysis generalization of previous approach extending application to multiple clusters algorithm tested on simulation data robust against small sample size noise future work synchronization patterns in eeg data matlab implementation algorithm from author supported by deutsche forschungsgemeinschaft | 0.361114 | 2,335 | 1,152 | , Saving $0.1 in GPT-4. |
67,648 | strong repulsive interactions within a one dimensional fermi system in a two - probe configuration normally lead to a reduced off - resonance conductance .
we show that if the repulsive interaction extends to the contact regions , a strong increase of the conductance may occur , even for systems where one would expect to find a reduced conductance .
an essential ingredient in our calculations is a momentum - space representation of the leads , which allows a high energy resolution .
further , we demonstrate that these results are independent of the high - energy cutoff and that the relevant scale is set by the fermi velocity . | constructing a transport theory for strongly correlated systems is one of the major challenges of condensed matter physics .
even though many interesting ideas have been proposed during recent years , no consensus has yet emerged as to the general validity and applicability of the various schemes . with this state of affairs , it is of high importance to establish reliable benchmarks for simple model systems , which then can be used to validate new approaches .
recently we presented a novel method for calculating linear response conductance@xcite using the density matrix renormalization group ( dmrg ) method@xcite . a major challenge in this work consisted in minimizing finite size effects , which was achieved via modified boundary conditions .
in this paper we circumvent these technical problems by reformulating the leads in momentum space .
this approach enables us to ( i ) reach a much higher energy resolution ( @xmath0 ) and ( ii ) allows for a greater flexibility in the choice of discretization schemes . in two recent papers mehta and andrei @xcite presented nonequilibrium bethe ansatz results for the interacting resonant - level model ( irlm ) , where a single spinless level is coupled to a left and a right lead both via a tunneling and a density - density interaction term
however , their work currently excludes the regime of resonant tunneling i.e . , the regime where the conductance is close to unity.@xcite in this work we study the linear conductance of the irlm on a lattice to provide a benchmark for the universal properties of the model .
in addition , we present results for an extended model , where the central region consists of three sites , with a similar interaction as in the irlm model . as we will show
, this model displays the same qualitative behavior as the irlm .
it should be noted that despite its simplicity , the irlm captures much of the physics of transport through an arbitrary interacting nanostructure provided that only a single level is close to the fermi energy of the leads , with all other levels well separated and outside the voltage window within which the transport takes place . for perfect coupling
the irlm model corresponds to the one - dimensional model studied by vasseur _
et al._,@xcite and molina _ et al._,@xcite obtained by restricting their nanostructure to a single site . using the embedding method they showed that smoothing the ramp of interaction for perfect contacts can compensate for the decrease of transmission due to interaction on the nanostructure . here
we go far beyond the energy resolution attained in previous work and show that interaction on the contact links can lead to strong renormalization effects , enhancing transport beyond the noninteracting system .
we use the dmrg method to evaluate the linear response conductance of the interacting nanostructure . in previous work@xcite
the leads were modeled in _ real - space _ by nearest - neighbor hopping chains . while simple to implement there are several drawbacks of this method , most prominently the need for `` damped boundary conditions '' and the resulting problem of trapping of fermions on the wilson chain ( the damped region).@xcite in the present work we introduce a setup where the leads are described in _
momentum space_. specifically , a short part of the lead close to the nanostructure is represented in real - space , accounting for local ( i.e. , high energy ) physics , while further away from the nanostructure the lead is represented in momentum space ; see fig .
[ fig : sitelayout ] . since the low - energy modes of the momentum leads are now directly coupled to the extended structure ( the nanostructure plus additional real - space sites ) , as illustrated in fig .
[ fig : sitelayout ] , the trapping of fermions on the low energy sites@xcite is avoided and no scaling sweeps are needed .
this enables much higher energy resolution , and in the current work we resolve resonances of widths @xmath1 . by virtue of the momentum representation of the leads
the discretization scheme can be chosen arbitrarily to suit the problem at hand . in the present work we use a logarithmic discretization to cover a large energy range , while switching to a linear discretization for the lowest - energy states in order to describe fermi - surface physics accurately .
the linear discretization on the low - energy scale allows for a better representation of the low - energy physics relevant for transport properties
excitations created by @xmath2 .
the models considered in this work are the irlm and the natural extension of this model to resonant linear chains , defined by the hamiltonians @xmath3 where @xmath4 and @xmath5 are the ( spinless )
fermionic creation and annihilation operators at site @xmath6 , @xmath7 .
@xmath8 , @xmath9 , and @xmath10 denote real - space , momentumspace , and tunneling between real- and momentum - space hamiltonians , respectively . the symbols @xmath11 and @xmath12 denote the nanostructure and the extended nanostructure ( the full real - space chain ) , respectively .
the indices @xmath13 and @xmath14 denote the first and last site in @xmath12 . the general setup and the specific values of the hopping matrix elements @xmath15 and
the interactions @xmath16 are indicated in fig .
[ fig : sitelayout ] , and note specifically the interactions on the contact links , @xmath17 .
the coupling @xmath18 of the extended real - space structure to the momentum leads is chosen in such a way that in the case of a cosine band it corresponds to a nearest - neighbor hopping chain in real - space with a hopping parameter of @xmath19 . in the following
we measure all energies in units of @xmath20 . for a single - site nanostructure and
@xmath21 this model reduces to the irlm .
the properties of the leads are defined by the band structure @xmath22 , which can take any form . in this work
we use either the cosine band , @xmath23 , or the linear band , @xmath24 .
@xmath25 is a cutoff parameter such that the fermi velocity @xmath26 is kept constant in all work presented here , and the band ranges between energies @xmath27 and @xmath25 . throughout this work
we use the notion of `` contact interaction '' for interaction on the link between the nanostructure and the leads .
the aim of this work is to study the effect of contact interaction .
it is known from previous work @xcite that strong repulsive interactions within the nanostructure lead to suppression of the transport off resonance due to the formation of a density - wave - like state on the dot . of the resonant level and @xmath28 of the real - space sites attached to the level vs the link interaction in the irlm for t=0.01 and two different gate voltages .
] in fig .
[ fig : all g-1 ] we show results for the conductance versus gate potential for different couplings to the leads and different contact interactions for the irlm ( @xmath21 ) .
the calculations have been performed with typically 130 sites in total , @xmath29 real - space sites and 120 momentum - space sites . due to the symmetry of the band
, we use a discretization that is symmetric around @xmath30 , and further use identical discretization of the two leads . to represent the `` large '' energy span in the band we use
20 logarithmically scaled sites , and thereafter use 10 linearly spaced sites to represent the low - energy scale correctly . in the dmrg calculations presented we used at least 1300 states per block and 10 finite lattice sweeps . to each set of dmrg results in fig .
[ fig : all g-1 ] is added a lorentzian of half width @xmath31w as a guide to the eye .
as the interaction is turned up the width of the resonance is increased far beyond the noninteracting result , up to an order of magnitude larger ; e.g. , for @xmath32 and @xmath33 the resonance width is increased by a factor of 10 . however , for a larger interaction @xmath34 , transport is suppressed , and for very large interactions the width even becomes smaller than the noninteracting resonance .
a similar nonmonotonic behavior is observed by borda _ _ et al.__@xcite using a perturbative calculation and is opposite to the one originally reported by mehta and andrei@xcite which , however , has been corrected in an erratum.@xcite where preceding work@xcite failed to reach the unitary limit , we demonstrate that indeed the resonant value remains unitary . further by changing the bandwidth @xmath25 for linear bands we have verified that the relevant energy scale is the _ fermi velocity _
@xmath35 of the leads , while the bandwidth @xmath25 _ does not _ influence the conductance , as long as @xmath36 ; compare fig .
[ fig : g - nph ] .
borda _ _ et al.__@xcite conclude in their work that `` in the case of repulsive interaction the site next to the occupied @xmath37 level is empty and thus that electron can easily jump to the conduction band '' , while for attractive interaction fermions accumulate close to the impurity . from that reasoning we would expect an asymmetric conductance curve depending on whether the impurity is filled or depleted .
however , this would violate particle - hole symmetry of the model . in fig .
[ fig : siteoccupation ] we plot the site occupation @xmath38 of the resonant level and the averaged site occupation @xmath28 of the left and right real - space sites attached to the level .
the occupations are plotted versus the contact link interaction for the interacting resonant - level model , and for two different gate voltages .
the site occupation of the resonant level and the neighboring sites are both enhanced by the repulsive interaction as long as interaction is in the range that enhances the conductance . for stronger interaction the site occupancy of the resonant level
is indeed reduced ; however , this is the regime where the conductance is reduced .
we would like to remark that in the noninteracting case and for a weak contact , @xmath39 , the site occupations of the real - space sites in the leads changes only slightly with gate voltage and are all very close to half filling .
thus it seems that the densities of the hybridizing lead levels are _ not _ the determining quantity for the interaction - induced changes of transport properties .
w has been added as a guide to the eye .
the leads are described by a cosine band between @xmath40 such that @xmath41 .
the interdot interaction suppresses the transport while the contact interaction is seen to enhance the transport . ] and a contact hopping of @xmath42 for a linear band with cutoff parameter @xmath43 , and @xmath44 and constant fermi velocity , @xmath41 .
the conductance is independent of the cutoff .
the solid line is a fit with a lorentzian of half width w@xmath45 . ]
the strong renormalization of the resonance width and the non - monotonic behavior is , however , not specific to the irlm . in fig .
[ fig : g3 ] we show results for the center peak of a three - site nanostructure . without a contact interaction
we find that the intradot interaction @xmath46 leads to a _ suppression _ of the transport in agreement with previous results.@xcite as in the single - level case already a small contact interaction increases again the width of the resonance at zero gate potential .
the enhancement of the conductance by a contact interaction is stronger than the corresponding suppression by the intradot interaction .
therefore we conjecture that the enhancement of conductance due to the contact interaction is a universal feature , which should also be present in other systems .
these findings may also be relevant for disordered structures , where repulsive interaction was found to enhance transport in the case of strong disorder.@xcite finally we have considered a non - particle - hole - symmetric irlm to address the question of parameter renormalization versus bandwidth cutoff .
the non - particle - hole - symmetric model is defined by replacing the @xmath47 terms in @xmath8 by @xmath48 .
the results are shown in fig .
[ fig : g - nph ] .
it is clearly seen from the calculation that varying the cutoff over an order of magnitude does not change the resonance , providing the interaction is _ not _ cut off by the band .
neither the position nor the width of the resonance peak is influenced by the change of the cutoff @xmath25 , which is in contrast to the the renormalization group flow that follows from the nonequilibrium bethe ansatz.@xcite there , all transport quantities depend on the cutoff @xmath25 and the conductance changes with the cutoff . while it is often difficult to compare a field theoretical model , like the irlm of mehta and andrei , with a lattice model
, we can at least conclude that the rg flow found in their work is absent in our model with regularized ( tight binding ) leads and that the relevant energy scale is the fermi velocity .
a normal paradigm in transport calculations is to make a principal division between transport region , the nanostructure or `` molecule '' , and leads , where all correlation effects are excluded from the leads . in this work
we have investigated the influence of an interaction on the contact between a nanostructure and the leads in a simple tight binding model . using the nonperturbative dmrg method to evaluate the linear conductance we have demonstrated that a contact interaction significantly influences the transport properties .
a repulsive interaction smaller or comparable to the fermi velocity in the leads enhances the conductance , while a large interaction leads to a suppression of the conductance .
our work shows that even a slight spread of the interaction on the contacts influences the transport strongly .
this demonstrates that particular care should be taken in treating the contacts correctly , especially regarding the interaction .
acknowledges support from the hpc - europa under project no .
rii3-ct-2003 - 506079 , supported by the european commission .
this work also profited from project 710 of the landesstiftung baden - wrttemberg and partial support through project b2.10 of the dfg center for functional nanostructures .
parts of the computations were performed on the xc1 and xc2 at the ssc karlsruhe .
the discussion of the the site occupation is attributed to an anonymous referee . | constructing transport theory for correlated systems condensed matter physics many ideas proposed no consensus emerged validity applicability of schemes importance to establish benchmarks for simple model systems validate new approaches presented novel method for calculating linear response conductance@xcite using density matrix renormalization group dmrg method challenge in minimizing finite size effects achieved via modified boundary conditions paper circumvent problems by reformulating leads in momentum space approach enables higher energy resolution allows greater flexibility in discretization schemes papers mehta and andrei presented nonequilibrium ansatz results for interacting resonant - level model irlm ) single spinless level coupled to left right lead via tunneling density - density interaction work excludes regime resonant tunneling conductance close to unity work study linear conductance of irlm on lattice benchmark for universal properties model present results for extended model central region three sites similar interaction as irlm model model displays same qualitative behavior as irlm irlm captures physics of transport through arbitrary interacting nanostructure single level close to fermi energy of leads other levels separated outside voltage window transportperfect coupling irlm model corresponds to one - dimensional model by vasseur@xcite molina obtained restricting nanostructure to single site embedding method showed smoothing ramp of interaction for perfect contacts decrease transmission due interaction nanostructure beyond energy resolution interaction on contact links to renormalization effects enhancing transport beyond noninteracting system dmrg method evaluate linear response conductance of interacting nanostructure previous work leads modeled in real - space by nearest - neighbor hopping chains drawbacks need for damped boundary conditions problem trapping fermions on wilson chain present work setup leads described in momentum space short part of lead close to nanostructure represented in real - space local high energy physics further away from nanostructure represented in momentum space low - energy modes of momentum leads coupled to extended structure nanostructure real - space sites trapping of fermions on low energy sites avoided no scaling sweeps needed enables higher energy resolution current work resolve resonances of widths momentum representation leads discretization scheme chosen arbitrarily suit problempresent work logarithmic discretization large energy range switching linear lowest - energy states fermi - surface physics linear discretization low - energy scale better representation physics transport properties excitations @xmath2 models considered irlm natural extension resonant linear chains defined hamiltonians @xmath3 @xmath4 @xmath5 fermionic creation annihilation operators site @xmath6 denote real - space momentumspace tunneling hamiltonians symbols @xmath11 @xmath12 denote nanostructure extended nanostructure full real - space chain indices @xmath13 denote first last site @xmath12 general setup specific values hopping matrix elements @xmath15 interactions @xmath16 indicated in fig . interactions contact links @xmath17 coupling @xmath18 extended real - space structure to momentum leads cosine band corresponds nearest - neighbor hopping chain real - space hopping parameter @xmath19 measure energies in units @xmath20 single - site nanostructure @xmath21 model reduces to irlm properties leads defined by band structure @xmath22 form use cosine band @xmath23 linear band @xmath24@xmath25 cutoff parameter fermi velocity @xmath26 constant work band ranges between energies @xmath27 and @xmath25 notion contact interaction link between nanostructure leads aim study effect contact interaction strong repulsive interactions nanostructure lead to suppression transport off resonance density - wave - like state dot resonant level @xmath28 real - space sites vs link interaction irlm for t=0.01 two different gate voltages fig . results conductance versus gate potential for different couplings leads contact interactions for irlm ( @xmath21 ) calculations performed with 130 sites @xmath29 real - space sites 120 momentum - space sites discretization symmetric around @xmath30 identical discretization of two leads large energy span use 20 logarithmically scaled sites 10 linearly spaced sites low - energy scale dmrg calculations used 1300 states per block 10 finite lattice sweeps each set dmrg results added lorentzian of half width @xmath31w guide eye interaction turned up width resonance increased noninteracting result @xmath32 @xmath33 resonance width increased by factor of 10 .larger interaction @xmath34 transport suppressed large interactions width smaller than noninteracting resonance similar nonmonotonic behavior observed by borda et al.__@xcite calculation opposite to mehta and andrei@xcite corrected in erratum.@xcite preceding work failed reach unitary limit resonant value remains unitary changing bandwidth @xmath25 for linear bands relevant energy scale is fermi velocity @xmath35 of leads bandwidth @xmath25 influence conductance @xmath36 compare fig . g - nph ] borda et al.__@xcite conclude repulsive interaction site next to occupied @xmath37 level empty electron jump to conduction band attractive interaction fermions accumulate close to impurity asymmetric conductance curve impurity filled or depleted particle - hole symmetry model fig . siteoccupation site occupation @xmath38 of resonant level averaged site occupation @xmath28 of left and right real - space sites attached occupations plotted versus contact link interaction two different gate voltages site occupation neighboring sites enhanced by repulsive interaction interaction in range conductance stronger interaction site occupancy reduced regime conductance reducedremark noninteracting case weak contact @xmath39 site occupations real - space sites leads slightly with gate voltage close to half filling densities hybridizing lead levels not determining interaction - induced changes transport properties w added guide eye leads described cosine band between @xmath40 @xmath41 interdot interaction suppresses transport contact interaction transport contact hopping @xmath42 linear band cutoff parameter @xmath43 @xmath44 constant fermi velocity @xmath41 conductance independent of cutoff solid line fit half width@xmath45 strong renormalization resonance width non - monotonic behavior not specific to irlm fig . g3 results center peak three - site nanostructure without contact interaction intradot interaction @xmath46 leads suppression transport previous results single - level case small contact interaction increases width resonance at zero gate potential enhancement conductance contact interaction stronger than suppression intradot interaction enhancement conductance contact interaction universal feature present other systems relevant for disordered structures repulsive interaction transport strong disorder considered non - particle - hole - symmetric irlm parameter renormalization versus bandwidth cutoffnon - particle - hole - symmetric model defined replacing @xmath47 terms in @xmath8 by @xmath48 results shown in fig . varying cutoff over change resonance interaction not cut off by band neither position nor width of resonance peak influenced by change cutoff @xmath25 contrast to renormalization group flow from nonequilibrium transport quantities depend on cutoff @xmath25 conductance changes difficult to compare field theoretical model with lattice model rg flow in absent in our model with regularized tight binding leads relevant energy scale is fermi velocity normal paradigm in transport calculations division between transport region nanostructure molecule leads correlation effects excluded from leads investigated influence of interaction on contact between nanostructure leads in tight binding model nonperturbative dmrg method contact interaction influences transport properties repulsive interaction smaller or comparable to fermi velocity enhances conductance large interaction conductance shows slight spread of interaction on contacts influences transport care in treating contacts interaction support from hpc - europa under project no . rii3-ct-2003 - 506079 supported by european commission work profited project 710 landesstiftung baden - wrttemberg project b2.10 dfg center functional nanostructures computations performed xc1 xc2 ssc karlsruhe site occupation anonymous referee | 0.507569 | 3,091 | 1,545 | , Saving $0.1 in GPT-4. |
67,652 | we determine non - perturbatively a fixed - point ( fp ) action for fermions in the two - dimensional u(1 ) gauge ( schwinger ) model .
our parameterization for the fermionic action has terms within a @xmath0 square on the lattice , using compact link variables . with the wilson fermion action as starting point
we determine the fp - action by iterating a block spin transformation ( bst ) with a blocking factor of 2 in the background of non - compact gauge field configurations sampled according to the ( perfect ) gaussian measure .
we simulate the model at various values of @xmath1 and find excellent improvement for the studied observables . | we are interested in a lattice representation of the continuum action for the 2d u(1 ) gauge theory with massless fermions @xcite ( the schwinger model ) .
the lattice action should respect the basic symmetries and should have the correct ( naive ) classical limit . also one has to take care for the fermion doubling problem . beyond these requirements
the form of the lattice action is largely arbitrary .
the traditional lattice actions are ultra - local but have corrections @xmath2 ( of the lattice spacing constant ) for bosons and @xmath3 for fermions .
improved actions have smaller corrections in powers @xmath4 but introduce more terms .
as long as the contributions are exponentially damped with regard to their extension in real space one calls the action local .
optimally , an improved action has no such corrections and thus no corrections to the leading critical behaviour .
such actions have been called `` perfect '' .
more terms complicate the simulation and one has to find a compromise between efficiency and perfectness .
various approaches to improvements are reviewed in @xcite .
we follow the path inspired by scale transformations leading to fp - actions @xcite .
we determine an optimal fermion action in the background of gauge field configurations sampled according to their ( optimal ) gaussian measure .
the resulting fp - action in this approximation defines a classically perfect action and for large @xmath1 we expect that it is a good approximation for the renormalized trajectory . in this limit the gauge field acts like a background field for the fermionic sector @xcite , and the fermionic action stays quadratic in the fermionic field variables .
further details may be found in @xcite .
fp - actions for that model were also studied with the method of small fields @xcite and recently in a perturbative expansion @xcite .
we denote the lattice action by @xmath5 where @xmath6 denotes the gauge field part , @xmath7 the lattice dirac operator matrix , and @xmath8 is the gauge field coupling .
we block from a so - called fine square lattice with sites @xmath9 to a coarse lattice organizing the fine lattice in @xmath10 blocks which constitute the points @xmath11 of the coarse lattice .
the grassmann fields are @xmath12 ( respectively @xmath13 on the coarse lattice ) ; the real , non - compact gauge fields @xmath14 live on the links . for the fermions we use anti - periodic boundary conditions and for the gauge field periodic ones .
the bst is defined as @xmath15 we fix the gauge within each block to the so - called _ fine _ gauge , where the plaquette field strength is distributed equally among the four link variables @xcite .
we integrate only over gauge field configurations in the fine gauge , defining the path integral measure @xmath16 .
the kernel of the bst for the fermions was taken from @xcite with parameters suitable chosen in order to have maximum locality in the situation of free fermions .
for the kernel of the gauge field we define an average over the four 2-link connections between corresponding sites in adjacent blocks .
more details are discussed in @xcite .
the resulting action on the coarse lattice is gauge invariant , hermitian invariant , invariant under the charge conjugation and respects the lattice symmetry .
it does violate chiral symmetry .
there exists a unique minimizing configuration of @xmath17 which we denote by @xmath18 . in our case
it can be computed straightforwardly by solving a set of linear equations . for @xmath19
this saddle point @xmath18 dominates the path integral .
the grassmann integration with subsequent identification of all bilinear terms in the fermionic variables defines the fermionic block action and the fp equations ( in the limit @xmath20 ) . in order to use the action also at moderate @xmath1 values
, we have to calculate the blocked action for strongly fluctuating configurations @xmath21 , too . * gauge field fp - action : * + for our bst the ultra - local standard ( non - compact ) plaquette action is a fp , up to the wave function renormalization @xcite . in @xmath22
this action is @xmath23 .
* fermion fp - action : * + we parameterize @xmath24 with a finite number of coupling constants , @xmath25 here @xmath26 is the parameterized fermion matrix , @xmath27 denotes a closed loop through @xmath28 or a path from the lattice site @xmath28 to @xmath29 ( distance vector @xmath30 ) and @xmath31 is the parallel transporter along this path .
the @xmath32-matrices denote the pauli matrices for @xmath33 and the unit matrix for @xmath34 .
the conditions @xmath35 and @xmath36 guarantee , that @xmath37 is normalized and @xmath38 reproduces the action of the massless model in the naive continuum limit ( @xmath39 denotes the component in the 1-direction ) .
we first considered terms connecting the central site @xmath28 with any other site @xmath29 in a @xmath40 lattice .
invariances of the action under certain symmetries provide further reductions . in the iteration process we found that one may omit several of the original terms .
altogether we finally considered 33 different geometric shapes corresponding to 123 independent coupling constants @xcite .
the fp - action is determined in an iteration procedure , starting from the wilson action for the massless schwinger model ( @xmath41 ) .
* we generate 50 gauge field configurations @xmath21 on the coarse @xmath0 lattice according to their probability distribution @xmath42 . for each of these
we then find the minimizing configuration @xmath18 . * with @xmath18
we construct the fermion matrix on the fine lattice and perform the bst ( grassmann integral ) giving the @xmath43 fermion matrix @xmath44 on the coarse lattice .
this is done for all 50 gauge field configurations .
* the resulting fermion matrices are compared with the fermion matrices @xmath45 for the coarse lattice .
a new set of parameters according ( [ eq : fermfit ] ) is determined by minimizing @xmath46 for the matrix norm @xmath47 .
these steps are iterated until ( typically after 10 iterations ) the couplings remain stable within statistical fluctuations .
we worked at @xmath48 .
part of the observed ( small ) fluctuations in the couplings may be due to cancellations of certain terms in the fermionic action ( redundancies ) . in fig .
[ fig : locality ] we demonstrate the locality of our fp - action .
our values are comparable or smaller than those obtained for the free fermion perfect action @xcite .
the complete set of couplings may be retrieved from @xcite .
for our check we rely on direct simulations with the fp - action on lattices of size @xmath49 ( for details cf .
@xcite ) . at each value of @xmath1
considered we generated 10000 gauge field configurations with appropriate measure and performed the grassmann integrals explicitly , i.e. by computing the corresponding determinant and inverse fermion matrix .
thus we obtain results for both situations , the 1-flavour and the 2-flavour model . for the error
estimates we repeated the procedure several times .
we studied propagators of various `` mesonic '' operators bilinear in the fermion fields . comparing the propagators with analogous simulations for wilson - fermions we find significantly improved rotational invariance , demonstrated both in real space and for the energy - momentum dispersion relations . in the 2-flavour model one
expects one massive mode and a massless flavour - triplet ( fig . [ fig_scaling ] ) .
we find non - vanishing @xmath50-masses only at small @xmath1 , indicating deviation of our fp - action from the renormalized trajectory , i.e. a small signal of `` imperfectness '' to be expected .
however , the overall scaling behaviour predicted from theory ( for the 2-flavour model ) , @xmath51 , is nicely recovered for moderately large values of @xmath52 . | interested in lattice representation continuum action for 2d u(1 ) gauge theory with massless fermions @xcite schwinger model lattice action respect symmetries correct classical limit fermion doubling problem form lattice action arbitrary traditional lattice actions ultra - local corrections @xmath2 for bosons @xmath3 for fermions improved actions have smaller corrections introduce more terms contributions damped extension real space action local improved action no corrections no critical behaviour actions called `` perfect '' more terms complicate simulation compromise between efficiency perfectness approaches improvements reviewed in @xcite follow path scale transformations fp - actions determine optimal fermion action gauge field configurations gaussian measure resulting fp - action defines classically perfect action good approximation for renormalized trajectory gauge field acts background field for fermionic sector fermionic action stays quadratic in field variables details in @xcite fp - actions studied with small fields perturbative expansion @xcite denote lattice action by @xmath5 @xmath6 gauge field part dirac operator matrix gauge field couplingblock fine square lattice @xmath9 to coarse lattice @xmath10 blocks points @xmath11 coarse lattice grassmann fields are @xmath12 @xmath13 coarse lattice real non - compact gauge fields @xmath14 live on links fermions anti - periodic boundary conditions gauge field periodic bst defined @xmath15 fix gauge to fine _ gauge plaquette field strength distributed four link variables integrate gauge field configurations fine gauge path integral measure @xmath16 kernel bst fermions from @xcite parameters maximum locality free fermions gauge field average over four 2-link connections sites adjacent blocks details @xcite action coarse lattice gauge invariant hermitian invariant invariant charge conjugation respects lattice symmetry chiral symmetry unique minimizing configuration @xmath17 @xmath18 computed linear equations @xmath19 saddle point @xmath18 dominates path integral grassmann integration identification bilinear terms fermionic variables defines fermionic block action fp equations limit @xmath20 action moderate @xmath1 values calculate blocked action for fluctuating configurations @xmath21gauge field fp - action ultra - local standard plaquette action fp wave function renormalization @xcite @xmath22 action @xmath23 fermion fp - action parameterize @xmath24 finite coupling constants parameterized fermion matrix closed loop @xmath28 lattice site to @xmath31 parallel transporter @xmath32-matrices pauli matrices @xmath33 unit matrix @xmath34 conditions @xmath35 @xmath36 guarantee @xmath37 normalized @xmath38 reproduces action massless model naive continuum limit @xmath39 component 1-direction terms connecting central site @xmath28 site @xmath29 @xmath40 lattice invariances symmetries provide reductions iteration process omit original terms considered 33 geometric shapes 123 coupling constants @xcite fp - action determined iteration procedure wilson action massless schwinger model @xmath41 generate 50 gauge field configurations @xmath21 coarse @xmath0 lattice probability distribution @xmath42 find minimizing configuration @xmath18 construct fermion matrix fine lattice perform bst @xmath43 fermion matrix @xmath44 coarse lattice all 50 gauge field configurationsresulting fermion matrices compared with @xmath45 for coarse lattice new parameters determined by minimizing @xmath46 for matrix norm @xmath47 steps iterated until 10 iterations couplings stable within statistical fluctuations worked at @xmath48 small fluctuations in due to cancellations terms in fermionic action redundancies fig . locality locality of fp - action values comparable or smaller than free fermion perfect action @xcite complete couplings retrieved from @xcite rely on direct simulations with fp - action on lattices size @xmath49 @xmath1 generated 10000 gauge field configurations performed grassmann integrals determinant inverse fermion matrix results for both situations 1-flavour and 2-flavour model error estimates repeated procedure several times studied propagators of mesonic operators in fermion fields improved rotational invariance energy - momentum dispersion 2-flavour model expects one massive mode massless flavour - triplet non - vanishing @xmath50-masses at small @xmath1 deviation of fp - action from trajectory small signal of imperfectnessoverall scaling behaviour theory 2-flavour model @xmath51 recovered for large values @xmath52 | 0.313189 | 1,743 | 886 | , Saving $0.1 in GPT-4. |
67,658 | supergiant fast x - ray transients ( sfxts ) are a class of high - mass x - ray binaries with possible counterparts in the high energy gamma rays . the _ swift _ sfxt project has conducted a systematic investigation of the properties of sftxs on timescales ranging from minutes to years and in several intensity states ( from bright flares , to intermediate intensity states , and down to almost quiescence ) .
we also performed broad - band spectroscopy of outbursts , and intensity - selected spectroscopy outside of outbursts .
we demonstrated that while the brightest phase of the outburst only lasts a few hours , further activity is observed at lower fluxes for a remarkably longer time , up to weeks .
furthermore , we assessed the fraction of the time these sources spend in each phase , and their duty cycle of inactivity .
we present the most recent results from our investigation .
the spectroscopic and , most importantly , timing properties of sfxts we have uncovered with _
swift _ will serve as a guide in search for the high energy emission from these enigmatic objects .
address = istituto di astrofisica spaziale e fisica cosmica , via u. la malfa 153 , i-90146 palermo , italy address = istituto di astrofisica spaziale e fisica cosmica , via u. la malfa 153 , i-90146 palermo , italy address = institut fr astronomie und astrophysik , universitt tbingen , sand 1 , d-72076 tbingen , germany address = istituto di astrofisica spaziale e fisica cosmica , via e. bassini 15 , i-20133 milano , italy address = dipartimento di fisica , universit di ferrara , via saragat 1 , 44122 , ferrara , italy address = dipartimento di fisica , universit di ferrara , via saragat 1 , 44122 , ferrara , italy address = istituto di astrofisica spaziale e fisica cosmica , via u. la malfa 153 , i-90146 palermo , italy address = department of astronomy & astrophysics , pennsylvania state university , university park , pa 16802 , usa address = department of astronomy & astrophysics , pennsylvania state university , university park , pa 16802 , usa address = nasa / goddard space flight center , greenbelt , md 20771 , usa address = nasa / goddard space flight center , greenbelt , md 20771 , usa | supergiant fast x - ray transients ( sfxt ) are detected as fast hard x - ray transients whose hour - long outbursts reach @xmath0@xmath1 erg s@xmath2 and whose spectra resemble those of accreting neutron stars ( ns ) .
indeed , some sfxt are shown to have pulsations ranging from a few seconds to @xmath3s .
this flaring gives sfxts quite a large dynamical range , which ranges between 3 and 5 orders of magnitude in flux .
optical identification of the counterparts has led to classify sfxts as a peculiar class of high mass x
ray binaries ( hmxb ) , associated to a blue ( ob ) supergiant companion .
we generally distinguish between confirmed and candidate sfxts based on the availability of an optical classification of the companion , and we now have 10 confirmed and as many candidate sfxts .
the details of the physical origin of the outbursts are still unclear , but the ingredients that need to be taken into account include the properties of the wind from the supergiant companion @xcite and the presence of a centrifugal or magnetic barrier @xcite .
since 2007 the sfxt project has performed an investigation of the properties of sfxts with _ swift _ , whose fast - slewing capability and broad - band energy coverage , combined with its flexible observing scheduling , make it an ideal facility to study both the bright outburst and the out - of - outburst behavior .
our strategy pairs monitoring programs with outburst follow - ups ( see , e.g. @xcite ) and has allowed us to catch several outbursts from most confirmed and several candidate sfxts ( see , e.g. @xcite and references therein ) , to monitor them during their evolution , and to study the long term properties of this class of objects with a highly sensitive soft x ray telescope .
our long - term monitoring of 4 sfxts ( igr j16479@xmath44514 , xte j1739@xmath4302 , igr j17544@xmath42619 , ax j1841.0@xmath40536 ) has allowed us to determine that they spend 35% of the time in bright outbursts ; that their most probable flux level is f(210kev)@xmath5@xmath6 erg @xmath7 s@xmath2 ( unabsorbed , @xmath810@xmath9 erg s@xmath2 ) that their duty cycle of inactivity @xcite is in the range 1955% , so that these sources accrete matter for most of the time @xcite .
[ fig1 ] and [ fig2 ] illustrate our long - term monitoring for the two sxft prototypes , xte j1739@xmath4302 and igr j17544@xmath42619 ( up to @xmath10 mjd 55150 ) and how we keep observing bright outbursts also after the end of the campaign .
2619 and subsequent outbursts .
the downward - pointing arrows are 3@xmath11 upper limits .
the upward - pointing arrows mark flares that triggered the bat transient monitor on mjd 54414 and 54906 . adapted from @xcite . ] further monitoring programs aimed at sfxts with known orbital period .
the first campaign was on igr j18483@xmath40311 @xcite ( @xmath12d ) which we followed for a whole orbital period , thus allowing us to constrain the different mechanisms proposed to explain the sfxt nature .
in particular , we applied the clumpy wind model for blue supergiants @xcite to the observed x - ray light curve . by assuming an eccentricity of @xmath13
, we could explain the x - ray emission in terms of the accretion from a spherically symmetric clumpy wind , composed of clumps with different masses , ranging from @xmath14 to @xmath15 g .
we also monitored igr j16418@xmath44532 ( @xmath16d ) for three orbital cycles @xcite and could account for the observed emission with accretion from a spherically symmetric clumpy wind ( see , also @xcite ) . among the most recent outbursts
are those of xte j1739@xmath4302 and igr j17544@xmath42619 that triggered the _ swift_/bat on 2011 february 22 and march 24 , respectively ( fig .
[ fig1 ] , last peaks at @xmath10 mjd 55600 , @xcite ) and those of igr j08408@xmath44503 @xcite . in @xcite
, we find that the spectra of both sources can be well fitted either with a two - blackbody model , or with a single unsaturated comptonization model . in the latter case , the model can be either a classical static comptonization model , such as , or the recently developed model , which includes thermal and bulk comptonization for cylindrical accretion onto a magnetized neutron star .
we refer to @xcite for a detailed description of the algorithm . here
we briefly remind the reader that is based on the solution of the radiative transfer equation for the case of cylindrical accretion onto the polar cap of a magnetized ns .
the velocity field of the accreting matter can be increasing towards the ns surface , or it may be described by an approximate decelerating profile . in the former case ,
the free parameters are the terminal velocity at the ns surface , @xmath17 , and the index of the law @xmath18 , while in the second case the law is given by @xmath19 .
the other free parameters of the model are the temperature of the black - body seed photons , @xmath20 , the electron temperature and vertical optical depth of the comptonization plasma @xmath21 and @xmath22 , respectively , and the radius of the accretion column @xmath23 , in units of the ns radius .
the different combinations of these parameters determine the steepness of the spectrum at high energies and the rollover energy position .
[ fig3 ] shows the unfolded ef(e ) models and spectra for the model . for xte j1739@xmath4302 .
_ right panel : _ absorption - corrected unfolded ef(e ) models and spectra , and residuals between the data and the model in units of @xmath11 for igr j17544@xmath42619 . adapted from @xcite .
, title="fig : " ] for xte j1739@xmath4302 .
_ right panel : _ absorption - corrected unfolded ef(e ) models and spectra , and residuals between the data and the model in units of @xmath11 for igr j17544@xmath42619 . adapted from @xcite .
, title="fig : " ] p. romano , s. vercellone , v. la parola , et al .
`` the swift supergiant fast x - ray transients project : recent results , '' in _ proceedings of the 25th texas symposium on relativistic astrophysics , online at http://pos.sissa.it/cgi-bin/reader/conf.cgi?confid=123_ , 2010 , p. 117 , . | supergiant fast x - ray transients detected as fast x - outbursts reach @xmath0@xmath1@xmath2 spectra resemble accreting neutron stars some sfxt pulsations seconds to @xmath3s flaring gives large dynamical range between 3 and 5 orders of magnitude in flux optical identification sfxts as high mass x ray binaries associated to blue supergiant companion distinguish between confirmed and candidate sfxts optical classification 10 confirmed many candidate sfxts details physical origin outbursts unclear ingredients include properties wind from supergiant companion @xcite centrifugal magnetic barrier @xcite since 2007 sfxt project properties sfxts with swift _ fast - slewing capability broad - band energy coverage flexible observing scheduling ideal facility to study bright outburst out - of - outburst behavior strategy pairs monitoring programs with outburst follow - ups several outbursts from confirmed candidate sfxts monitor during evolution study long term properties with sensitive soft x ray telescopelong - term monitoring 4 sfxts j16479@xmath44514 j1739@xmath4302 j17544@xmath42619 j1841.0@xmath40536 spend 35% time in bright outbursts probable flux level f(210kev)@xmath5@xmath6 @xmath7@xmath2 duty cycle inactivity @xcite 1955% sources accrete matter most time @xcite fig1 fig2 illustrate long - term monitoring two sxft prototypes j1739 j17544@xmath42619 @xmath10 55150 bright outbursts after end campaign downward - arrows 3@xmath11 upper limits arrows mark flares bat transient monitor on mjd 54414 54906 monitoring programs sfxts known orbital period first campaign on igr j18483@xmath40311 @xcite @xmath12d ) followed orbital period mechanisms sfxt nature applied clumpy wind model for blue supergiants @xcite x - ray light curve assuming eccentricity of @xmath13 x - ray emission from spherically symmetric clumpy wind @xmath14 to @xmath15monitored igr j16418@xmath44532 @xmath16d ) three orbital cycles @xcite account emission accretion spherically symmetric clumpy wind recent outbursts xte j1739@xmath4302 igr j17544@xmath42619 triggered swift_/bat 2011 february 22 march 24 fig peaks @xmath10 mjd 55600 @xcite igr j08408@xmath44503 @xcite spectra sources fitted two - blackbody model or single unsaturated comptonization model model classical static comptonization model or recently developed model thermal bulk comptonization cylindrical accretion magnetized neutron star refer @xcite detailed description algorithm based solution radiative transfer equation cylindrical accretion polar cap magnetized ns velocity field accreting matter increasing towards ns surface or decelerating profile former free parameters terminal velocity ns surface @xmath17 index law @xmath18 second case law given by @xmath19 free parameters model are temperature black - body seed photons electron temperature vertical optical depth comptonization plasma @xmath21 @xmath22 radius accretion column @xmath23 ns radius combinations parameters determine steepness spectrum at high energies rollover energy position fig3 ] shows unfolded ef(e ) models spectra model xte j1739@xmath4302 absorption - corrected ef(e ) models spectra residuals data model units @xmath11 j17544@xmath42619 absorption - corrected models spectra residuals data units @xmath11 j17544@xmath42619 p. romano s. vercellone v. la parola swift supergiant fast x - ray transients project recent results proceedings 25th texas symposium relativistic astrophysics online http://pos.sissa.it/conf=123 2010 p. 117 | 0.266786 | 1,589 | 835 | , Saving $0.0 in GPT-4. |
67,661 | a first measurement of the cross section of the process @xmath0 is reported using a total integrated luminosity of 231pb@xmath1 collected with the l3 detector at centre - of - mass energies of @xmath2 and @xmath3 . by selecting hadronic events with two isolated photons
the @xmath0 cross section is measured to be @xmath4 at @xmath2 and @xmath5 at @xmath3 .
the measurements are consistent with standard model expectations
. limits on quartic gauge boson couplings @xmath6 and @xmath7 of @xmath8 and @xmath9 are derived at 95% confidence level . | the lep centre - of - mass energy ( @xmath10 ) for @xmath11 collisions has now exceeded the w pair and z pair production thresholds , allowing the study of triple gauge boson production processes such as @xmath0 and @xmath12 .
measurements of these processes give a new insight into the standard model of electroweak interactions ( sm ) @xcite .
the possibility of these triple gauge boson production processes proceeding via @xmath13-channel exchange of a fourth boson provides a probe of quartic gauge boson couplings ( qgc ) .
such measurements were recently performed for the @xmath12 process @xcite .
this letter describes the first measurement of the cross section of the process @xmath0 followed by the hadronic decay of the z. in the sm this process occurs by radiation of the photons from the incoming electron and/or positron , corresponding to a total of six diagrams , three of which are presented in figure [ fig:0 ] .
no qgc contribution is predicted at the tree level and z@xmath14 events are sensitive to anomalous qgc @xcite contributions , as shown in figure [ fig:0 ] .
the measurement uses data collected with the l3 detector @xcite at lep in 1997 and 1998 at average centre - of - mass energies of @xmath15 and @xmath16 corresponding to integrated luminosities of 55pb@xmath1 and 176pb@xmath1 , respectively .
these energies are respectively denoted as @xmath17 and @xmath18 hereafter .
the @xmath19 signal is defined by three phase space cuts : photon energies above @xmath20 , photon angles with respect to the beam axis between @xmath21 and @xmath22 and invariant mass of the primary produced quarks within a @xmath23 window around the z mass , @xmath24 being the z width .
the kk2f monte carlo ( mc ) program @xcite predicts signal cross sections of about @xmath25 at both energies .
the selection of events satisfying the signal definition given above is optimised using hadronic events generated at @xmath26 with the kk2f mc program and at @xmath27 with the pythia5.72 @xcite mc program .
events from these mc programs failing the signal definition are termed qcd@xmath14 background .
other background processes are generated both at @xmath27 and @xmath26 with the mc programs pythia ( @xmath28 and @xmath29 ) , koralz4.02 @xcite ( @xmath30 ) , phojet1.05c @xcite ( @xmath31 ) and koralw1.21 @xcite for @xmath32 production except for the @xmath33 final states which are generated with excalibur @xcite .
additional background sources are found to be negligible .
the l3 detector response is simulated using the geant 3.15 program @xcite , which takes into account the effects of energy loss , multiple scattering and showering in the detector .
time dependent detector inefficiencies , as measured in each data taking period , are reproduced in these simulations .
the selection of @xmath19 candidates from balanced hadronic events with two photons and little energy deposition at low polar angles is based on photon energies and angles together with the invariant mass of the hadronic system .
the photon energy and angle criteria follow directly from the signal definition whereas the invariant mass of the hadronic system is required to be between @xmath34 and @xmath35 .
the main background after these selection requirements is due to the radiation of two initial state photons with a hadronic system failing the z signal definition criteria .
the boost @xmath36 of the recoiling system to the photons , assuming its mass to be the nominal z mass , is on average larger for these background events .
candidate events are hence required to have @xmath37 at @xmath27 and @xmath38 at @xmath26 .
another class of background events is the so called radiative return to the z , where a photon in the initial state is emitted , bringing the effective @xmath10 to the z resonance .
a z boson is then produced decaying into a hadronic system with an electromagnetic energy deposition ( photon , misidentified electron or unresolved @xmath39 ) faking the least energetic photon of the signal selection .
these events are rejected by an upper bound on the energy @xmath40 of the most energetic photon and a lower bound on the angle @xmath41 between the least energetic photon and its closest jet .
numerically @xmath42 at @xmath27 and @xmath43 at @xmath44 with @xmath45 at both the energies . data and mc distributions of the selection variables are presented in figure [ fig:1 ] for @xmath26 , where selection criteria on all the other variables are applied . a good agreement between data and mc is observed .
the application of the selection procedure described above yields the signal efficiencies and selected data and mc events summarised in table 1 .
a clear signal structure is observed in the recoil mass spectra of the two photons , presented in figure [ fig:2 ] for the two centre - of - mass energies under study .
the @xmath19 cross section at @xmath27 and @xmath26 is then determined from the number of events selected and the efficiency and background estimates from mc . | lep centre mass energy collisions exceeded w z pair thresholds study triple gauge boson production processes @xmath0 measurements insight model electroweak interactions triple gauge boson @xmath13-channel exchange fourth boson quartic gauge boson couplings measurements performed for @xmath12 process measurement cross section process @xmath0 hadronic decay z process occurs by radiation photons from incoming electron positron six diagrams three presented in figure no qgc contribution predicted tree level z@xmath14 events sensitive to anomalous qgc @xcite contributions measurement uses data l3 detector @xcite 1997 1998 average centre - - mass energies @xmath15 @xmath16 luminosities 55pb@xmath1 176pb@xmath1 energies denoted as @xmath17 and @xmath18 @xmath19 signal defined by three phase space cuts photon energies above @xmath20 photon angles between @xmath21 @xmath22 invariant mass primary quarks @xmath23 window around z mass @xmath24 z width program predicts signal cross sections of @xmath25 at both energiesselection of events satisfying signal definition optimised using hadronic events @xmath26 kk2f @xmath27 pythia5.72 @xcite events failing signal definition termed qcd@xmath14 background other background processes generated at pythia koralz4.02 phojet1.05c koralw1.21 except @xmath33 final states generated excalibur @xcite additional background sources negligible l3 detector response simulated using geant 3.15 program energy loss multiple scattering showering time dependent detector inefficiencies reproduced simulations selection of @xmath19 candidates from balanced hadronic events two photons little energy deposition angles based on photon energies angles invariant mass hadronic system photon energy angle criteria follow from signal definition invariant mass between @xmath34 @xmath35 main background selection due to radiation of two initial state photons hadronic system failing z signal definition criteria boost @xmath36 recoiling system to photons larger for background events candidate events required @xmath37 @xmath27 @xmath38 @xmath26background events radiative return to z photon initial emitted @xmath10 to z resonance z boson into hadronic system electromagnetic energy deposition faking least energetic photon rejected by upper bound energy @xmath40 most energetic photon lower bound angle @xmath41 between least energetic photon closest jet @xmath42 at @xmath27 @xmath43 at @xmath44 @xmath45 at both energies data mc distributions selection variables in figure for @xmath26 selection criteria good agreement between data mc selection procedure yields signal efficiencies data mc events summarised in table 1 clear signal structure in recoil mass spectra two photons:2 mass energies @xmath19 cross section at @xmath27 @xmath26 determined from number events selected efficiency background estimates from mc | 0.184795 | 1,172 | 595 | , Saving $0.0 in GPT-4. |
67,665 | three - jet production has been studied in deep - inelastic positron - proton scattering .
the measurement carried out with the h1 detector at hera covers a large range of four - momentum transfer squared @xmath0 and invariant three - jet masses @xmath1 .
jets are defined by the inclusive @xmath2 algorithm in the breit frame .
the size of the three - jet cross section and the ratio of the three - jet to the dijet cross section @xmath3 are described over the whole phase space by the predictions of perturbative qcd in next - to - leading order .
the shapes of angular jet distributions deviate significantly from a uniform population of the available phase space but are well described by the qcd calculation . | multi - jet production in deep - inelastic scattering ( dis ) has been successfully used at hera to test the predictions of perturbative qcd ( pqcd ) over a large range of four - momentum transfer squared @xmath4 .
recently the h1 collaboration has determined the strong coupling @xmath5 and the gluon density in the proton @xcite from the inclusive jet and the dijet cross sections measured in the breit frame . while these cross sections are directly sensitive to qcd effects of @xmath6 , the three - jet cross section in dis is already proportional to @xmath7 in leading order ( lo ) in pqcd . the higher sensitivity to @xmath8 and the greater number of degrees of freedom of the three - jet final state thus allow the qcd predictions to be tested in more detail . here
we present differential measurements of the three - jet cross section in neutral current dis and measurements of shapes of angular jet distributions which are sensitive to dynamic effects of the interaction @xcite .
the present analysis includes the first comparison of three - jet distributions measured in hadron induced reactions with a pqcd calculation in next - to - leading order ( nlo ) i.e. @xmath9 @xcite .
in neutral current dis the lepton interacts with a parton in the proton via the exchange of a boson ( @xmath10 , @xmath11 ) . at a fixed center - of - mass energy the kinematics of the lepton inclusive reaction ( for unpolarized lepton and proton beams )
is given by two variables which are here chosen to be the four - momentum transfer squared @xmath4 and the bjorken scaling variable @xmath12 .
the subprocess @xmath13 in which three massless jets emerge from the boson - parton reaction is fully described by six further variables which can be constructed from the energies @xmath14 and the momenta @xmath15 of the jets in the three - jet center - of - mass ( cm ) frame .
it is convenient to label the three jets ( @xmath16 ) in the order of decreasing energies in the three - jet cm frame .
two of these variables are the angles @xmath17 and @xmath18 that specify the relative orientation of the jets , @xmath19 where @xmath20 denotes the direction of the proton beam .
@xmath17 is the angle of the highest energy jet with respect to the proton beam direction .
@xmath18 is the angle between the plane spanned by the highest energy jet and the proton beam and the plane containing the three jets .
the angle @xmath18 indicates whether the third jet ( i.e. the lowest energy jet ) is radiated within ( @xmath21 or @xmath22 ) or perpendicular ( @xmath23 ) to the plane containing the highest energy jet and the proton beam .
the observable @xmath3 , defined by the ratio of the inclusive three - jet cross section and the inclusive two - jet cross section , is of interest especially for quantitative studies , since in this ratio both experimental and some theoretical uncertainties cancel to a large extent .
this contribution presents measurements of the inclusive three - jet cross section in dis as a function of @xmath4 .
distributions of three - jet events are measured in the variables @xmath24 and @xmath18 .
the ratio @xmath3 is measured as a function of @xmath4 .
measurements of further observables can be found in @xcite .
the analysis is based on data taken in positron - proton collisions with the h1 detector in the years 19951997 with a positron beam energy of @xmath25 and a proton beam energy of @xmath26 , leading to a center - of - mass energy @xmath27 .
we present results from two event samples which correspond to integrated luminosities of @xmath28 ( low @xmath4 : @xmath29 ) and @xmath30 ( high @xmath4 : @xmath31 ) , respectively .
jets are selected in the breit frame with @xmath32 .
the data are corrected for effects of detector resolution and acceptance , as well as for inefficiencies of the selection and for higher order qed effects .
( 20,154 ) ( -205,76 ) ( -5,76 ) ( -160,166)(a ) ( 42,193)(b ) -10 mm the measured three - jet distributions are presented in figs . 1 and 2 .
the inner error bars represent the statistical uncertainties and the outer ones the quadratic sum of all uncertainties .
the data are compared to pqcd predictions in lo and nlo with and without hadronization corrections .
the lo and nlo calculations are carried out in the @xmath33-scheme for five massless quark flavors using the program nlojet @xcite .
renormalization and factorization scales ( @xmath34 , @xmath35 ) are set to the average transverse energy @xmath36 of the three jets in the breit frame .
hadronization corrections @xmath37 are determined using the monte carlo event generator lepto @xcite as the relative change of an observable before and after hadronization .
[ fig : xsectq2 ] ( a ) shows that over the whole range of @xmath4 the nlo prediction ( corrected for hadronization effects ) gives a good description of the data not only at high @xmath4 , where nlo corrections are small , but also at low @xmath4 , where the nlo prediction is a factor of two above the lo prediction . the theoretical prediction is subject to several uncertainties , the dominant sources being the value of the strong coupling , the parton density functions of the proton ( especially the gluon density ) and the renormalization scale dependence of the nlo calculation ( see lower part of fig . [
fig : xsectq2 ] ( a ) ) . while for @xmath38 the variation of @xmath8 gives the largest effect ,
the renormalization scale dependence is the dominant source of uncertainty at lower values of @xmath4 , i.e. in the region where nlo corrections are also large . over the whole @xmath4 range the change of the cross section , induced by the variation of the gluon density is approximately half as large as the change induced by the @xmath8 variation . while the lo calculation predicts a stronger @xmath4 dependence of the ratio @xmath3 than observed in the data ( fig .
[ fig : xsectq2 ] ( b ) ) , the nlo calculation gives a good description of the data over the whole @xmath4 range .
uncertainties in the theoretical prediction for @xmath39 are investigated in the same way as for the three - jet cross section .
the nlo corrections for the three - jet and the dijet cross sections are of similar size . at low @xmath4
this leads to a smaller nlo correction and renormalization scale dependence for the ratio @xmath3 than for the cross section .
furthermore , when measured in the same region of @xmath40 and @xmath4 , with the same cut on the invariant multi - jet mass , the three - jet and the dijet cross sections probe the parton density functions of the proton in the same range of proton momentum fractions @xmath41 . since both jet cross sections are dominated by gluon induced processes , @xmath3 is almost insensitive to variations of the gluon density in the proton . for the central value of @xmath42 used in the calculations , the theoretical predictions are consistent with the data for the three - jet cross section and the ratio @xmath3 .
( 0,100 ) ( -155 , 77 ) ( 0 , 77 ) ( -112,110)(a ) ( 48,110)(b ) the normalized distributions of the angular variables @xmath43 and @xmath18 are shown in fig .
[ fig : th3psi3lo ] .
the @xmath43 distributions in both @xmath4 regions ( here only the low @xmath4 region is shown ) are peaked at the cut value of @xmath44 , corresponding to angles close to the proton and the photon direction .
the phase space prediction shows the opposite behavior , falling towards @xmath45 .
the qcd calculation in nlo gives a good description of the effects seen in the data .
this is also the case for the measured @xmath18 distributions ( here only the high @xmath4 region is shown ) which are relatively flat , while the underlying phase space is peaked at @xmath46 .
although at low @xmath4 the nlo prediction is almost a factor of two higher than the lo prediction , the shapes of the angular jet distributions are almost unaffected by the nlo correction .
qcd calculations , in contrast to the phase space predictions , indicate that , due to the bremsstrahlung nature of the process , configurations are preferred in which the plane containing the two lowest energy jets coincides with the plane spanned by the proton beam and the highest energy jet .
this corresponds to values of @xmath47 and @xmath22 .
the effects of cuts reduce this preference ( fig .
[ fig : th3psi3lo ] ( b ) ) .
these results are in qualitative agreement with measurements in @xmath48 collisions @xcite and in @xmath49 collisions @xcite .
the inclusive three - jet cross section and the ratio @xmath3 of the inclusive three - jet and the inclusive dijet cross section have been measured as a function of @xmath4 .
the predictions of perturbative qcd in next - to - leading order give a good description of the three - jet cross section and the ratio @xmath3 over the whole range of @xmath4 for values of the strong coupling close to the current world average of @xmath50 @xcite .
angular jet distributions have been measured in the three - jet center - of - mass frame .
the angular orientation of the three - jet system follows the radiation pattern expected from perturbative qcd . while the angular distributions are not consistent with a uniform population of the available phase space , they are well described by the qcd predictions .
this is the first successful test of @xmath51 calculations at hera .
g. ingelman , a. edin and j. rathsman , comput .
commun .
* 101 * ( 1997 ) 108 [ hep - ph/9605286 ] ( version used : 6.5 ) .
g. arnison _ et al . _
[ ua1 collaboration ] , phys .
b * 158 * ( 1985 ) 494 ; + j. a. appel _ et al . _
[ ua2 collaboration ] , z. phys .
c * 30 * ( 1986 ) 341 . | multi - jet production deep - inelastic scattering used perturbative qcd four - momentum transfer h1 collaboration determined coupling @xmath5 gluon density proton from inclusive jet dijet cross sections breit frame cross sections sensitive to qcd effects @xmath6 three - jet cross section dis proportional to @xmath7 pqcd higher @xmath8 degrees freedom three - jet final state allow qcd predictions present differential measurements three - jet cross section neutral current dis angular jet distributions sensitive dynamic effects interaction analysis includes comparison three - jet distributions hadron induced reactions pqcd calculation next - - leading order @xmath9 neutral current dis lepton interacts with parton proton exchange boson fixed center - of - mass energy kinematics lepton reaction given by two variables four - momentum transfer squared @xmath4 bjorken scaling variable @xmath12 subprocess @xmath13 three massless jets emerge from boson - parton reaction described by six variables energies @xmath14 momenta three - jet center - of - mass framelabel three jets @xmath16 decreasing energies three - jet cm frame variables angles @xmath17 specify orientation jets @xmath19 direction proton beam @xmath17 angle highest energy jet proton beam @xmath18 angle between plane highest energy jet proton beam plane three jets angle @xmath18 indicates third jet lowest energy radiated within or perpendicular plane highest energy jet proton beam observable @xmath3 defined ratio inclusive three - jet cross section two - jet cross section interest for quantitative studies experimental theoretical uncertainties cancel contribution presents measurements inclusive three - jet cross section function @xmath4 distributions of three - jet events measured in variables @xmath24 @xmath18 ratio @xmath3 measured function of @xmath4 measurements observables in @xcite analysis based on data positron - proton collisions h1 detector 19951997 positron beam energy @xmath25 proton energy @xmath26 center - of - mass energy @xmath27 results from two event samples integrated luminosities @xmath28 jets selected frame with @xmath32data corrected for detector resolution acceptance inefficiencies selection higher order qed effects ( 20,154 ) -205,76 ) ( -5,76 ) -160 42,193) -10 mm measured three - jet distributions presented in figs 1 2 inner error bars represent statistical uncertainties outer quadratic sum uncertainties data compared to pqcd predictions in lo nlo with without hadronization corrections lo nlo calculations in @xmath33-scheme for five massless quark flavors program nlojet @xcite renormalization factorization scales set to average transverse energy @xmath36 of three jets hadronization corrections @xmath37 determined using monte carlo event generator lepto @xcite relative change observable before after hadronization nlo prediction corrected for hadronization effects good description data high @xmath4 low @xmath4 theoretical prediction subject to uncertainties dominant sources strong coupling parton density functions renormalization scale dependence nlo calculation @xmath38 variation of @xmath8 largest effect renormalization scale dependence dominant source uncertainty at lower values @xmath4region nlo corrections large @xmath4 range change cross section variation gluon density half as large as @xmath8 variation lo calculation predicts stronger @xmath4 dependence of ratio @xmath3 nlo calculation good description data over @xmath4 range uncertainties in theoretical prediction for @xmath39 investigated three - jet cross section nlo corrections dijet cross sections similar size low @xmath4 leads smaller nlo correction renormalization scale dependence for ratio @xmath3 cross section measured same region @xmath40 @xmath4 - jet mass three - jet dijet cross sections probe parton density functions proton same range proton momentum fractions @xmath41 jet cross sections dominated by gluon induced processes @xmath3 almost insensitive to variations gluon density proton central value @xmath42 theoretical predictions consistent with data for three - jet cross section ratio @xmath3 normalized distributions of angular variables @xmath43 and @xmath18 shown in fig .@xmath43 distributions in @xmath4 regions low peaked at cut value @xmath44 angles close to proton and photon direction phase space prediction shows opposite behavior towards @xmath45 qcd calculation in nlo effects data for @xmath18 distributions high region relatively flat phase space peaked at @xmath46 at low @xmath4 nlo prediction two higher than angular jet distributions unaffected by nlo correction qcd calculations indicate configurations preferred plane two lowest energy jets coincides with plane by proton beam highest energy jet corresponds to values @xmath47 and @xmath22 effects of cuts reduce preference results with measurements in @xmath48 collisions @xmath49 collisions inclusive three - jet cross section ratio @xmath3 function of @xmath4 predictions of perturbative qcd in next - leading order of three - jet cross section ratio @xmath3 over range @xmath4 for values strong coupling close to average of @xmath50 @xcite angular jet distributions measured in three - jet center - of - mass frame angular orientation three - jet system follows radiation pattern perturbative qcd angular distributions not consistent uniform population phase space described qcd predictions first successful test @xmath51 calculations hera ingelman edin rathsman comput 101 1997 ) 108 hep - ph/9605286 version 6.5 ) arnison al . ua1 collaboration . 158 ( 1985 ) 494 appel ua2 collaboration phys . 30 1986 ) 341 | 0.352238 | 2,329 | 1,155 | , Saving $0.1 in GPT-4. |
67,668 | we present the alice results on open heavy flavour , focusing on the exclusive reconstruction of charmed mesons via displaced decay topologies .
these measurements benefit from the large pb - pb statistics collected in 2011 .
the results on the nuclear modification factor @xmath0 for @xmath1 mesons indicate a suppression of their yield in central collisions relative to binary - scaled pp collisions in a large momentum range .
the comparison to the @xmath0 of non - prompt j/@xmath2 ( measured with cms ) indicates a difference in the suppression of charm and beauty , as expected according to the predicted mass hierarchy in energy loss models .
the measurement of the azimuthal anisotropy of charmed mesons is also discussed .
the observed positive second fourier harmonic @xmath3 for transverse momentum @xmath4 gev/@xmath5 in semi - peripheral events is a hint for collective motion of charm quarks .
the results discussed above are also compared to theoretical models . | a quantitative picture of the medium created in high - energy pb - pb collisions can be obtained by measuring its properties with observables that give access to the energy - loss mechanism for different parton species and to the transport properties of the medium .
theoretical models of energy loss predict a hierarchical dependence on the colour charge and mass of the parton propagating through the medium .
a larger energy loss is expected for gluons , together with a suppression of gluon radiation at small angles for partons with larger mass ( dead - cone " effect @xcite ) .
it is therefore interesting to compare medium effects ( i.e. the path - length , colour - charge and mass dependence of the energy loss , as well as the collective motion ) on heavy quarks versus light quarks and gluons .
such challenging goals require precise heavy - flavour measurements in pb - pb collisions and a clear understanding of the reference systems provided by pp and p - pb collisions .
the latter provides the control experiment needed to disentangle effects from the cold matter in absence of a hot and dense medium .
the alice @xcite detectors used in these analyses are the inner tracking system ( its ) , the time projection chamber ( tpc ) and the time of flight ( tof ) detector .
they provide tracking , primary and secondary vertex reconstruction and particle identification , like @xmath6 separation up to transverse momentum @xmath7 gev/@xmath5 .
these features are essential ingredients to perform open - heavy flavour analyses over a broad transverse momentum range .
in addition , the vzero detector is used for triggering on collisions and for centrality measurements .
charmed mesons are reconstructed in alice through their hadronic decays : @xmath8 , @xmath9 , @xmath10 and @xmath11 .
the analysis strategy exploits the particle identification and the detector capability to reconstruct secondary vertices displaced by a few hundred @xmath12 m from the primary vertex .
topological cuts are applied on the reconstructed secondary vertices and the signals are extracted via fits to the invariant mass distributions . in these proceedings
we report on the prompt @xmath1-meson nuclear modification factor @xmath0 , defined as the yield in pb - pb collisions divided by the cross - section in pp collisions scaled by the average nuclear overlap function in a given pb - pb centrality range , as a function of @xmath13 and collision centrality .
the prompt @xmath1-meson elliptic flow @xmath3 , defined as the second harmonic coefficient of the fourier azimuthal expansion , is also reported as a function of @xmath13 in semi - peripheral collisions . in the @xmath0 analysis , the prompt @xmath1-meson yield was obtained after subtracting the contribution of @xmath1 mesons from beauty hadron decays based on perturbative qcd calculations ( fonll , @xcite ) .
the @xmath0 of non - prompt @xmath1 mesons was assumed to be @xmath14(on the basis of the cms results of non - prompt @xmath15 , instead of @xmath16 as done in the previous analysis @xcite ) and the systematic uncertainty was assessed by varying its value between 1 and 3 times the prompt-@xmath1 r@xmath17 . due to the lack of statistics of the pp data at @xmath18 tev , the pp reference needed for the r@xmath17 analysis was determined by scaling the @xmath1-meson cross section measured at @xmath19 tev to @xmath18 tev using fonll calculations @xcite .
the scaling uncertainty varies from @xmath20 to @xmath21 from low to high @xmath13 .
the contribution of @xmath1 mesons from beauty decays was also subtracted from the @xmath3 measurement with an assumption on the @xmath0 ( as above ) and @xmath3 of non - prompt @xmath1 mesons ( @xmath22 @xcite ) .
these analyses were performed on the data from the pb - pb runs at @xmath23 tev collected in 2010 ( minimum bias trigger , @xmath24b@xmath25 ) and 2011 ( @xmath26 and @xmath27b@xmath25 for 0 - 7.5% and 10 - 50% respectively ) .
the @xmath0 of prompt @xmath1 mesons was measured with alice in the @xmath28 centrality class using the data sample collected in 2011 , which extends the measurements to a wider transverse momentum range ( @xmath29 gev/@xmath5 ) compared to the published results from the 2010 data @xcite . as shown in fig .
[ fig : raa ] ( left ) , the @xmath0 values for @xmath30 , @xmath31 and @xmath32 agree within the uncertainties and show a strong suppression ( factor of @xmath33 for @xmath34 gev/@xmath5 ) of the @xmath1-meson yields in pb - pb collisions relative to pp collisions . the first measurement of the @xmath35 meson in pb - pb collisions is also shown .
a suppression of the @xmath35 is observed for @xmath36 gev/@xmath5 , in agreement within the uncertainties with the non - strange charmed mesons @xmath0 in this @xmath13 range .
the @xmath35-meson yield could be less suppressed at lower @xmath13 because of the predicted c - quark recombination with the enhanced strange quarks in the medium @xcite , however a more precise measurement is needed to draw a conclusion .
the @xmath0 as a function of the number of participant nucleons weighted by the number of binary collisions is shown in fig .
[ fig : raa ] ( right ) for @xmath30 mesons with @xmath37 gev/@xmath5 and compared to pions measured in the same centrality and @xmath13 intervals .
the main correlated systematic uncertainties for the @xmath1-meson yields come from the pp cross section and beauty feed down .
the uncorrelated systematic uncertainties are mainly due to the @xmath1-meson yield extraction and cut variation .
the suppression of @xmath1 mesons with @xmath37 gev/@xmath5 is almost independent of centrality , while the pion @xmath0 decreases with increasing centrality .
the magnitude of the suppression is similar for @xmath1 mesons and pions .
a hint of a difference in central collisions is suggested but a larger dataset is needed to draw conclusion on the expected difference between @xmath1 mesons and pion suppression due to the mass and colour charge dependence of the energy loss . in the higher @xmath13 range the @xmath1-meson suppression is no longer flat , rather it increases from peripheral to central collisions ( fig .
[ fig : raab ] , left ) .
the @xmath0 of non - prompt j/@xmath2 ( @xmath38 gev/@xmath5 and @xmath39 ) measured with cms is reported in fig .
[ fig : raab ] ( left ) as a function of the number of participants weighted by the number of binary collisions @xcite .
the results are shown together with the @xmath1-meson @xmath0 from alice ( @xmath40 gev/@xmath5 and @xmath41 ) .
the selected @xmath13 ranges correspond to similar kinematical ranges for the parent b and c quarks , but the measurements are performed in different rapidity ranges .
an indication of a difference in the suppression of charm and beauty can be observed in central collisions , consistent with the mass hierarchy expected from various energy - loss models , such as those reported in the figure @xcite .
while the model @xcite based on collisional and radiative energy loss in an anisotropic medium agrees with both prompt @xmath1 mesons and non - prompt j/@xmath2 , the others seem to underestimate both ( i.e. @xcite , based on collisional energy loss in an expanding medium ) or to underestimate the suppression of non - prompt j/@xmath2 ( i.e. @xcite , based on radiative energy loss and @xmath1-meson in - medium formation and dissociation ) .
the @xmath3 for prompt @xmath1 mesons , reported in fig .
[ fig : raab ] ( right ) , was measured as a function of @xmath13 in the range @xmath42 gev/@xmath5 in the 30 - 50% centrality class @xcite .
we observe a positive @xmath3 for prompt @xmath1 mesons in the transverse momentum range @xmath4 gev/@xmath5 , which is comparable to that of charged particles ( not shown ) .
this result suggests a collective motion also for heavy quarks at low @xmath13 . at higher momentum a positive @xmath3
is expected due to the path - length dependence of the in - medium energy loss , given the different paths traveled by a parton propagating through the medium .
the measurements of the prompt @xmath1-meson @xmath3 are also compared to various energy - loss models @xcite .
while several theoretical models based on in - medium parton energy loss reproduce reasonably well the @xmath0 of prompt @xmath1 mesons as a function of @xmath13 ( not shown , see @xcite ) , they are challenged by simultaneously reproducing results from prompt @xmath1-meson @xmath0 and @xmath3 .
the suppression of prompt @xmath1 mesons observed in the momentum range @xmath43 gev/@xmath5 in central pb - pb collisions at @xmath44 tev is evidence for a strong modification of charm production in pb - pb compared to pp collisions due to energy loss .
results for the @xmath35 meson also show a suppression in @xmath36 gev/@xmath5 but suffer from statistics limitations that do not allow us to draw conclusions on a possible enhancement at low @xmath13 due to a coalescence of charm quarks with strange quarks .
the measurements of prompt @xmath1-meson @xmath0 as a function of centrality indicate that the suppression tends to be constant with centrality in the lowest @xmath13 range , @xmath37 gev/@xmath5 , while it increases with centrality at intermediate / high @xmath13 .
the observed difference in the suppression of @xmath1 mesons and non - prompt j/@xmath2 from b - meson decays at high @xmath13 in central collisions suggests a mass dependence in the energy loss of heavy quarks .
the positive @xmath3 for prompt @xmath1 mesons in the momentum range @xmath4 gev/@xmath5 in semi - peripheral pb - pb collisions suggests that the interactions with the medium constituents transfer to charm quarks information on the azimuthal anisotropy of the system .
theoretical models of in - medium energy loss reproduce reasonably well the measured prompt @xmath1-meson @xmath0 .
nevertheless , a simultaneous description of both @xmath0 and @xmath3 remains a challenge for models , indicating that the energy - loss mechanisms and the participation of heavy quarks in the collective behavior of the quark - gluon plasma are not yet fully understood today .
9 k. aamodt _ et al . _
[ alice collaboration ] , jinst * 3 * ( 2008 ) s08002 .
m. cacciari _ et al .
_ , jhep * 1210 * ( 2012 ) 137 [ arxiv:1205.6344 ] . b. abelev _ et al . _
[ alice collaboration ] , jhep * 1209 * ( 2012 ) 112 [ arxiv:1203.2160 ] .
r. averbeck _ et al .
_ , arxiv:1107.3243
. b. abelev _ et al . _
[ alice collaboration ] , phys .
* 111 * ( 2013 ) 102301 [ arxiv:1305.2707 ] .
z. conesa del valle [ alice collaboration ] , nucl .
a * 904 - 905 * ( 2013 ) 178c [ arxiv:1212.0385 ] .
i. kuznetsova _ et al .
. phys .
j. c * 51 * ( 2007 ) 113 .
m. he _ et al .
_ , phys .
* 110 * ( 2013 ) 112301 .
a. andronic _ et al .
_ , phys .
b * 659 * ( 2008 ) 149 .
s. chatrchyan _ et al . _
[ cms collaboration ] , jhep * 1205 * ( 2012 ) 063 [ arxiv:1201.5069 ] , cms - pas - hin-12 - 014 .
o. fochler , j. uphoff , z. xu and c. greiner , j. phys .
g * 38 * ( 2011 ) 124152 [ arxiv:1107.0130 ] .
w. a. horowitz and m. gyulassy , j. phys .
g * 38 * ( 2011 ) 124114 [ arxiv:1107.2136 ] .
i. vitev _ et al .
_ , phys .
c * 80 * ( 2009 ) 054902 , w. m. alberico _ et al .
_ , eur .
j. c * 71 * ( 2011 ) 1666 [ arxiv:1101.6008 ] .
b. gossiaux _ et al .
_ , phys .
c * 79 * ( 2009 ) 044906 , j. phys .
g * 37 * ( 2010 ) 094019 . | quantitative picture medium in high - energy pb - pb collisions measuring properties with observables energy - loss mechanism parton species transport properties theoretical models energy loss predict dependence on colour charge mass parton propagating larger energy loss for gluons suppression of gluon radiation small angles for partons larger mass dead - effect interesting to compare medium effects path - length colour - charge mass dependence collective motion on heavy quarks versus light quarks gluons goals require precise heavy - flavour measurements in pb - pb collisions understanding of reference systems pp p - pb collisions provides control experiment to disentangle effects from cold matter hot dense medium alice @xcite detectors are inner tracking system time projection chamber time of flight detector provide tracking primary secondary vertex reconstruction particle identification essential open - heavy flavour analyses vzero detector used for triggering collisions centrality measurements charmed mesons reconstructed in alice through hadronic decays analysis strategy exploits particle identification detector capability to reconstruct secondary vertices displacedtopological cuts applied reconstructed secondary vertices signals extracted invariant mass distributions report @xmath1-meson nuclear modification factor yield pb - pb collisions divided cross - section pp collisions average nuclear overlap function centrality range function @xmath13 collision centrality @xmath1-meson elliptic flow @xmath3 second harmonic coefficient fourier azimuthal expansion function @xmath13 semi - peripheral collisions @xmath1-meson yield obtained subtracting contribution @xmath1 mesons from beauty hadron decays @xmath0 non - prompt @xmath1 mesons assumed @xmath14 systematic uncertainty assessed value 1 3 times prompt-@xmath1 r@xmath17 lack statistics pp data @xmath18 tev pp reference r@xmath17 analysis determined scaling @xmath1-meson cross section @xmath19 to @xmath18 tev scaling uncertainty varies @xmath20 to @xmath21 low to high @xmath13 contribution @xmath1 mesons from beauty decays subtracted from @xmath3 measurement @xmath0 non - prompt @xmath1 mesonsanalyses performed on data pb - pb runs @xmath23 2010 @xmath24b 2011 @xmath26 0 - 7.5% 10 - 50% @xmath0 @xmath1 mesons measured with @xmath28 centrality class data 2011 extends measurements wider transverse momentum range @xmath29/@xmath5 compared 2010 data @xmath0 values for @xmath30 agree show strong suppression @xmath33 @xmath34/@xmath5 of @xmath1-meson yields in pb - pb collisions first measurement @xmath35 meson in pb - pb collisions suppression @xmath35 for @xmath36 gev/@xmath5 non - strange mesons @xmath0 @xmath13 range @xmath35-meson yield less suppressed at lower @xmath13 c - quark recombination quarks precise measurement needed @xmath0 function of number participant nucleons weighted by binary collisions shown in fig . for @xmath30 mesons with @xmath37 gev/@xmath5 compared to pions same centrality @xmath13 intervalsuncertainties for @xmath1-meson yields from pp cross section beauty feed down due to yield extraction cut variation suppression of @xmath1 mesons with @xmath37 gev/@xmath5 independent of centrality pion @xmath0 decreases with centrality magnitude suppression similar for @xmath1 mesons pions difference in central collisions suggested larger dataset needed pion higher @xmath13 range suppression increases from peripheral to central collisions @xmath0 of non - prompt j/@xmath2 @xmath38@xmath5 ) measured with cms reported in fig function of number of participants by number binary collisions results shown with @xmath1-meson @xmath0 from alice ( @xmath40 @xmath41 selected @xmath13 ranges correspond to similar kinematical ranges for parent b c quarks in different rapidity ranges difference in suppression of charm beauty in central collisions consistent with mass hierarchy energy - loss models model collisional radiative energy loss anisotropic medium agrees with prompt @xmath1 mesons non - prompt j/@xmath2 others underestimate both@xcite collisional energy loss expanding medium underestimate suppression non - prompt@xmath2 radiative energy loss @xmath1-meson - medium formation dissociation @xmath3 @xmath1 mesons fig measured function @xmath13 @xmath42 gev/@xmath5 30 - 50% centrality class positive @xmath3 @xmath1 mesons transverse momentum range @xmath4 gev/@xmath5 comparable charged particles suggests collective motion heavy quarks low @xmath13 higher momentum positive @xmath3 path - length dependence - medium energy loss different paths parton measurements @xmath1-meson @xmath3 compared energy - loss models theoretical models parton energy loss reproduce @xmath0 @xmath1 mesons @xmath13 challenged reproducing results @xmath1-meson @xmath3 suppression prompt @xmath1 mesons momentum range @xmath43 gev/@xmath5 central pb - pb collisions @xmath44 tev strong modification charm production pb - pb collisions energy lossresults @xmath35 meson show suppression @xmath36@xmath5 statistics limitations enhancement low @xmath13 coalescence charm strange quarks measurements prompt @xmath1-meson @xmath0 indicate suppression constant lowest @xmath13 range increases intermediate high @xmath13 difference suppression @xmath1 mesons non - prompt@xmath2 b - meson decays high @xmath13 collisions suggests mass dependence energy loss heavy quarks positive @xmath3 prompt @xmath1 mesons momentum range @xmath4/@xmath5 - peripheral collisions suggests interactions medium transfer charm quarks information azimuthal anisotropy theoretical models in - medium energy loss reproduce prompt @xmath1-meson @xmath0 simultaneous description @xmath0 @xmath3 challenge energy - loss mechanisms participation heavy quarks quark - gluon plasma understood aamodt et al . alice collaboration jinst * 3 ( 2008 ) s08002 cacciari al jhep 1210 2012 ) 137 arxiv:1205.6344 abelev jhep 1209 2012 ) 112 arxiv:1203.2160 averbeck arxiv:1107.3243abelev alice collaboration phys . 111 2013 ) 102301 arxiv:1305.2707 ] conesa del valle alice collaboration 904 - 905 2013 ) 178c arxiv:1212.0385 ] kuznetsova . j c 51 2007 ) 113 he . 110 2013 ) 112301 andronic . 659 2008 ) 149 chatrchyan cms collaboration 1205 2012 ) 063 arxiv:1201.5069 ] cms - - hin-12 - 014 fochler uphoff xu greiner phys . g 38 2011 ) 124152 arxiv:1107.0130 ] horowitz gyulassy . 38 2011 ) 124114 arxiv:1107.2136 vitev phys . c * 80 * 2009 ) 054902 alberico j c 71 2011 ) 1666 arxiv:1101.6008 ] gossiaux phys . c 79 * 2009 ) 044906 phys . g 37 2010 ) 094019 | 0.43468 | 3,052 | 1,592 | , Saving $0.1 in GPT-4. |
67,679 | we consider the dynamics of two interacting lumps / solitons in a noncommutative gauge model . we show that equations of motion describing this dynamics can be reduced to ones of a two - dimensional mechanical system which is well studied and was shown to exhibit stochastic behaviour . | recent development of string theory has shown that noncommutative ( nc ) models can arise in certain limits of string theory @xcite . moreover
the string field theory @xcite , which is the theory of second quantised string , is based on noncommutative geometry too .
due to their proximity to string theory noncommutative theories have many features in common with it .
it is worthwhile to mention the ir / uv mixing @xcite , which says that in nc theories there is a correspondence analogous to string theory between contributions at high energies with those at low ones .
another common feature is a number of equivalences / dualities which relate different noncommutative models @xcite .
this list is continued by the so called noncommutative soliton which was found in @xcite , in the limit of large noncommutativity @xmath0 and further generalised to finite @xmath1 but nontrivial gauge field background @xcite .
these solitons are believed to correspond in the string theory language to _ branes .
_ noncommutative theories can alternatively be treated as theories of linear operators defined over a ( infinite - dimensional separable ) hilbert space . in the operator language
, noncommutative solitons look like projectors to finite - dimensional subspaces of the hilbert space , while the gauge fields split into parts corresponding respectively to the finite dimensional subspace and its orthogonal completion @xcite . in this note
we are going to consider gauge field `` solitons '' , i.e. solutions for gauge fields which are projectors to finite - dimensional subspaces of the hilbert space rather than ones for the scalar field . for more details the reader
is referred to ref . @xcite .
consider the noncommutative gauge model described by the following action , @xmath2 ^ 2\right).\ ] ] here fields @xmath3 , @xmath4 are time dependent hermitian operators defined on some separable infinite - dimensional hilbert space @xmath5 .
provided the gauss law constraint @xmath6=0 $ ] , the model ( [ action ] ) describes the noncommutative yang
mills(higgs ) model in the temporal gauge @xmath7 in a sense we explain below @xcite .
this appears also to be the same with the bosonic part of the hilbert space i.e. @xmath8 bfss m(atrix ) model @xcite .
equations of motion corresponding to this action look as follows , @xmath9=0.\ ] ] one may find static classical solution @xmath10 , @xcite , satisfying , @xmath11={\mathrm{i}}\theta_{ij}^{-1},\ ] ] for which we assume irreducibility , i.e. any quantity @xmath12 commuting with all @xmath13 : @xmath14=0 $ ] have to be proportional to unity operator . in particular , @xmath15 should be nondegenerate .
this property ensures that any operator can be formally expressed as an operator function of of the operators @xmath16 , by means of its weyl symbol .
expanding fluctuations around this solution , @xmath17 , and weyl ordering operators @xmath18 with respect to @xmath19 one gets precisely the @xmath20-dimensional noncommutative yang
mills model for the fields @xmath21 .
weyl symbol @xmath22 of an operator @xmath23 can be treated as an ordinary function subject to the moyal or star product , @xmath24 here @xmath22 and @xmath25 are weyl symbols of some operators @xmath23 and @xmath26 , @xmath27 is the weyl symbol of their product and @xmath28 , @xmath29 denote the derivatives with respect to @xmath19 and @xmath30 .
integration of a weyl symbol corresponds to @xmath31trace of the respective operator .
an important difference of the weyl
moyal algebra of functions defined above from the commutative one ( of course , beyond its noncommutativity ) is that the derivative is an internal automorphism of the algebra .
the partial derivative with respect to @xmath19 here corresponds to the commutator , @xmath32(x),\ ] ] getting another solution with a smaller number of independent operators say @xmath33 , @xmath34 , expanding and weyl ordering fluctuations around it , one gets yang mills higgs model in a smaller number of dimensions equal to @xmath35 , the number of independent irreducible operators . in the remaining part of this note we will actually consider the model corresponding to the expansion around a two - dimensional solution , @xmath36={\mathrm{i}}\theta^{-1},\ ] ] but one should keep in mind the above equivalence .
in this section we review a completely different class of localised solutions . in order to do this consider the noncommutative analog of complex coordinates which is given by oscillator rising / lowering operators @xmath37 and @xmath38 , @xmath39=1,\ ] ] and the oscillator basis , @xmath40 there is a solution to different from one discussed in the previous section which is given by commutative matrices . due to commutativity
they can be diagonalised by a proper gauge transformation and represented in the oscillator basis as follows , @xmath41 where @xmath42 represents @xmath43-th eigenvalue of the ( finite rank ) operator @xmath3 .
the rank @xmath44 of the solution is called the `` soliton number '' .
the simplest one - soliton solutions can be written in the form , @xmath45 _ we should warn that the term `` soliton '' applied to the above configuration is conventional . it is not a soliton in the sense of solutions to the solitonic equations .
moreover , in this case the solution has zero energy at rest , however geometrically it is `` noncommutative soliton '' in the sense of @xcite . in this particular case one
also can call them `` vacuum solitons '' or just `` vacua '' .
_ in the star - product form operator is represented by the weyl symbol @xmath46 the soliton shifted along noncommutative plane by a ( c - number ) vector @xmath47 is given by @xmath48 its weyl symbol , correspondingly , is given by @xmath49 .
the shifted soliton with constant @xmath47 is a solution again .
when @xmath47 becomes time - dependent one can perform a gauge transformation @xmath50 which shifts the soliton back to the centre , but produce a kinetic term @xmath51 .
thus a single noncommutative soliton moves like a free _ non - relativistic _ particle .
it is also stable since its energy at rest is zero . in what follows we are going to analyse the situation when there is a couple of solitons separated by a distance @xmath47 .
a single noncommutative soliton can be always rotated in the hilbert space by a gauge transformation to have the `` polarisation '' @xmath52 and `` orientation '' along @xmath53 . when there are two such solitons one can choose _ without loss of generality _ the `` orientation '' to involve nontrivially only @xmath53 and @xmath54 .
consider two solitons which are obtained from @xmath55 by shifts along the noncommutative plane by respectively @xmath56 and @xmath57 .
since the dynamics of the centre is free and can be decoupled by a time - dependent gauge transformation similar to , where @xmath58 is the coordinate of the centre .
the coordinates of the solitons reduced to the centre become @xmath59 .
thus , the configuration we consider looks like , [ 2solit ] @xmath60 where we introduced the shorthand notations , @xmath61 the hight @xmath62 can be absorbed by the rescaling of the coupling and the time , therefore we put it to @xmath63 . in what follows also @xmath64 @xmath4 will enter trivially in the equation , so in the remaining part of the paper we simply neglect them and allow the index @xmath65 to run in the range @xmath66 .
if we were considering more than two noncommutative solitons we would have to keep more fields .
the symmetry of the model allows its further simplification . in particular
, the gauss law conservation , @xmath67 provided by the unitary gauge invariance , implies that operators @xmath64 are nontrivial only on a two - dimensional subspace @xmath68 of the hilbert space spanned by @xmath69 .
therefore , one can restrict the equations of motion to @xmath68 assuming thus that @xmath64 are 2@xmath702 matrices acting on it .
this model has a number of interesting features studied in @xcite .
the motion resulting from above equations appears to be _
stochastic _ for all values of @xmath47 except @xmath74 .
when @xmath74 the motion is periodic but unstable , infinitesimal deviation from this position brings it back to stochasticity .
the solution to initial equations is expressed in terms of solution @xmath75 and @xmath76 to eq .
as follows , @xmath77 where @xmath78 are the weyl symbols of the two - dimensional pauli matrices which form the basis for hermitian operators acting on @xmath68 , qualitatively , the solution , describes lumps bouncing ( stochastically ) around `` points '' @xmath80 of the noncommutative plane . in the string theory language
the heights of the lumps can be interpreted as transversal ( to the noncommutative 2-brane ) coordinates of two 0-branes .
n. seiberg and e. witten , `` string theory and noncommutative geometry , '' _ jhep _ * 09 * ( 1999 ) 032 , http://www.arxiv.org/abs/hep-th/9908142[hep-th/9908142 ] .
e. witten , `` noncommutative geometry and string field theory , '' _ nucl .
* b268 * ( 1986 ) 253 .
s. minwalla , m. v. raamsdonk , and n. seiberg , `` noncommutative perturbative dynamics , '' _ jhep _ * 02 * ( 2000 ) 020 , http://www.arxiv.org/abs/hep-th/9912072[hep-th/9912072 ] .
c. sochichiu , `` on the equivalence of noncommutative models in various dimensions , '' http://www.arxiv.org/abs/hep-th/0007127[hep-th/0007127 ] . c. sochichiu ,
`` exercising in k - theory : brane condensation without tachyon , '' http://www.arxiv.org/abs/hep-th/0012262[hep-th/0012262 ] .
r. gopakumar , s. minwalla , and a. strominger , `` noncommutative solitons , '' _ jhep _ * 05 * ( 2000 ) 020 , http://www.arxiv.org/abs/hep-th/0003160[hep-th/0003160 ] . c. sochichiu ,
`` noncommutative tachyonic solitons : interaction with gauge field , '' _ jhep _ * 08 * ( 2000 ) 026 , http://www.arxiv.org/abs/hep-th/0007217[hep-th/0007217 ] .
c. sochichiu , `` some notes concerning the dynamics of noncommutative solitons in the m(atrix ) theory as well as in the noncommutative yang - mills model , '' http://www.arxiv.org/abs/hep-th/0104076[hep-th/0104076 ] . c. sochichiu ,
`` m(any ) vacua of iib , '' _ jhep _ * 05 * ( 2000 ) 026 , http://www.arxiv.org/abs/hep-th/0004062[hep-th/0004062 ] .
t. banks , w. fischler , s. h. shenker , and l. susskind , `` m theory as a matrix model : a conjecture , '' _ phys .
* d55 * ( 1997 ) 51125128 , http://www.arxiv.org/abs/hep-th/9610043[hep-th/9610043 ] .
g. z. baseian , s. g. matinian , and g. k. savvidy , `` nonlinear plane waves in massless yang - mills theory .
( in russian ) , '' _ pisma zh .
* 29 * ( 1979 ) 641 .
s. g. matinian , g. k. savvidy , and n. g. ter - arutunian - savvidy , `` stochasticity of yang - mills classical mechanics and its elimination by higgs mechanism , '' .
efi-512 - 55 - 81-yerevan .
b. v. medvedev , `` dynamic stochasticity and quantization , '' _ theor .
* 60 * ( 1985 ) 782797 .
i. y. arefeva , a. s. koshelev , and p. b. medvedev , `` chaos - order transition in matrix theory , '' _ mod .
lett . _ * a13 * ( 1998 ) 2481 , http://www.arxiv.org/abs/hep-th/9804021[hep-th/9804021 ] .
i. y. arefeva , p. b. medvedev , o. a. rytchkov , and i. v. volovich , `` chaos in m(atrix ) theory , '' http://www.arxiv.org/abs/hep-th/9710032[hep-th/9710032 ] . | development string theory shown noncommutative nc models arise in limits string theory string field theory second quantised string based on noncommutative geometry proximity to string theory noncommutative theories have features common ir / uv mixing correspondence between contributions high energies low equivalences / dualities noncommutative models by noncommutative soliton found in @xcite limit large noncommutativity generalised to finite @xmath1 nontrivial gauge field background solitons correspond string theory to branes noncommutative theories treated as theories linear operators over infinite - dimensional separable hilbert space noncommutative solitons look like projectors to finite - dimensional subspaces hilbert space gauge fields split into parts subspace completion gauge field solitons solutions gauge fields projectors to finite - dimensional subspaces hilbert space details to ref . @xcite consider noncommutative gauge model described by action @xmath2 fields @xmath3 @xmath4 are time dependent hermitian operators on separable infinite - dimensional hilbert space @xmath5gauss law constraint @xmath6=0 model describes noncommutative yang mills(higgs model temporal gauge @xmath7 @xcite same bosonic part hilbert space @xmath8 bfss m(atrix model @xcite equations motion @xmath9=0. classical solution @xmath10 @xmath11=\theta_{ij}^{-1} irreducibility quantity @xmath12 commuting @xmath13 @xmath14=0 proportional to unity operator @xmath15 nondegenerate ensures operator expressed operator function operators @xmath16 weyl symbol expanding fluctuations solution @xmath17 weyl ordering operators @xmath18 @xmath19 @xmath20-dimensional noncommutative yang mills model fields @xmath21 weyl symbol @xmath22 operator @xmath23 ordinary function moyal star product @xmath24 @xmath22 @xmath25 weyl symbols operators @xmath23 @xmath26 @xmath27 weyl symbol product @xmath28 @xmath29 derivatives @xmath19 @xmath30 integration weyl symbol corresponds @xmath31trace operatordifference weyl moyal algebra of functions from commutative derivative is internal automorphism algebra partial derivative @xmath19 corresponds to commutator @xmath32(x), solution with smaller independent operators @xmath33 , @xmath34 expanding ordering fluctuations yang mills higgs model smaller dimensions equal to @xmath35 consider model expansion around two - dimensional solution , @xmath36={\mathrm{i}}\theta^{-1}, equivalence review different class localised solutions consider noncommutative analog of complex coordinates by oscillator rising / lowering operators @xmath37 and @xmath38 , @xmath39=1, oscillator basis , @xmath40 solution different by commutative matrices diagonalised by gauge transformation represented in oscillator basis, @xmath41 @xmath42 represents @xmath43-th eigenvalue of finite rank operator @xmath3 rank @xmath44 solution `` soliton number '' simplest one - soliton solutions written , @xmath45 _ term `` soliton '' conventional not soliton solutions solitonic equationssolution zero energy at rest noncommutative soliton '' @xcite vacuum solitons star - product form operator represented by weyl symbol @xmath46 soliton shifted noncommutative plane by c - vector @xmath47 @xmath48 weyl symbol @xmath49 shifted soliton constant @xmath47 solution @xmath47 time - dependent gauge transformation @xmath50 shifts soliton back to centre kinetic term @xmath51 single noncommutative soliton moves like free non - relativistic particle stable energy at rest zero situation couple solitons separated by distance @xmath47 single noncommutative soliton rotated hilbert space by gauge transformation polarisation @xmath52 orientation along @xmath53 two solitons choose orientation @xmath53 and @xmath54 two solitons obtained from @xmath55 by shifts noncommutative plane @xmath56 and @xmath57 dynamics centre free decoupled by time - dependent gauge transformation @xmath58 coordinate of centre coordinates solitons reduced to centre become @xmath59configuration @xmath60 introduced shorthand notations @xmath61 hight @xmath62 absorbed by rescaling coupling time put to @xmath63 @xmath64 @xmath4 enter trivially equation remaining neglect allow index @xmath65 run in range @xmath66 more than two noncommutative solitons more fields symmetry model allows simplification gauss law conservation @xmath67 unitary gauge invariance implies operators @xmath64 nontrivial on two - dimensional subspace @xmath68 hilbert space spanned @xmath69 restrict equations motion to @xmath68 @xmath64 2@xmath702 matrices acting model features motion equations stochastic for values @xmath47 except @xmath74 motion periodic unstable infinitesimal deviation brings stochasticity solution equations @xmath75 @xmath76 to eq @xmath77 @xmath78 weyl symbols of two - dimensional pauli matrices basis for hermitian operators on @xmath68 solution describes lumps bouncing stochastically around points @xmath80 noncommutative planestring theory heights lumps transversal noncommutative 2-brane coordinates two 0-branes n. seiberg e. witten string theory noncommutative geometry 09 1999 ) 032 .arxiv/hep-th/9908142 e. witten noncommutative geometry string field theory b268 1986 ) 253 s. minwalla. raamsdonk n. seiberg noncommutative perturbative dynamics 02 * ( 2000 ) 020 /hep-th/9912072 c. sochichiu equivalence noncommutative models dimensions/hep-th/0007127 sochichiu exercising k - theory brane condensation without tachyon/hep-th/0012262 r. gopakumar s. minwalla a. strominger noncommutative solitons 05 * ( 2000 ) 020 /hep-th/0003160 c. sochichiu noncommutative tachyonic solitons interaction gauge field 08 * ( 2000 ) 026 /hep-th/0007217sochichiu notes dynamics noncommutative solitons m(atrix theory yang - mills model http://www.arxiv.org/abs/hep-th/0104076 sochichiu vacua iib jhep 05 ( 2000 ) 026/abs/hep-th/0004062 t. banks fischler shenker l. susskind m theory matrix model conjecture . d55 * ( 1997 ) 51125128.arxiv.org/abs/hep-th/9610043 baseian matinian savvidy nonlinear plane waves massless yang - mills theory russian zh 29 * ( 1979 ) 641 matinian savvidy arutunian stochasticity yang - mills classical mechanics elimination higgs mechanism efi-512 - 55 - 81-yerevan medvedev dynamic stochasticity quantization theor . 60 * ( 1985 ) 782797 arefeva koshelev b. medvedev chaos - order transition matrix theory mod lett a13 * ( 1998 ) 2481 .arxiv.org/abs/hep-th/9804021. arefeva. medvedev. rytchkov v. volovich chaos in m(atrix theory http://www.arxiv/abs/hep-th/9710032 | 0.496295 | 3,079 | 1,635 | , Saving $0.1 in GPT-4. |
67,683 | we reformulate the statement of the feit - thompson theorem in terms of diagrams in the category of finite groups , namely iterations of the quillen lifting property with respect to particular morphisms . | we reformulate several notions in finite group theory in terms of diagrams in the category of finite groups , namely being solvable , nilpotent , @xmath0-group and prime - to-@xmath0 group , abelian , perfect , subnormal subgroup , injective and surjective homomorphism ( see fig . 1 and fig . 2 ) .
these notions are enough to formulate the feit - thompson theorem . these properties are obtained by iterating the same diagram chasing `` trick '' in the category of groups , often starting from a single morphism - counterexample ; the `` trick '' is known as the quillen lifting property and was introduced by quillen [ qui ] to axiomatise algebraic topology in terms of categories .
our motivation was to formulate part of finite group theory in a form amenable to automated theorem proving while remaining human readable ; [ g1 ] tried to do the same thing for the basics of general topology . in more detail
, we reformulate these notions using the following four operations @xmath1 , @xmath2 , @xmath3 , @xmath4 on the classes ( properties ) @xmath5 of morphisms in a category . _ left , resp .
right , quillen negation _
@xmath1 , resp .
@xmath2 , of a property @xmath5 is the class of all morphisms @xmath6 , resp . @xmath7 , such that @xmath8 , resp .
@xmath9 , for all @xmath10 ; see fig . 1 for the definition of the quillen lifting property @xmath11 : @xmath12
@xmath13 the classes @xmath1 , @xmath2 are subcategories which contain no morphisms from @xmath5 but isomorphisms ; hence the terminology .
classes @xmath14 , @xmath4 denote the subclass consisting of morphisms from / to the terminal object @xmath15 , i.e. the trivial group .
classes @xmath16 , and @xmath17 can described as classes of objects admitting no non - trivial map from / to the objects corresponding to @xmath14 .
we also say that a morphism @xmath18 is _ left , resp .
right , quillen unlike @xmath5 _ or _
_ iff @xmath19 , resp .
@xmath20 , and that @xmath18 is _ left , resp .
right , exemplified by @xmath5 _
iff @xmath21 , resp .
@xmath22 . in this notation
our main observations are that a finite group @xmath23 is soluble iff @xmath24 lies in the class @xmath25 or , equivalently , right exemplified by taking the commutator subgroup , i.e. lies in @xmath26\lra s\,:\ , s\text { an arbitrary group}\,\}^{\rtt lr}$ ] .
a finite group @xmath23 is nilpotent
iff the diagonal map @xmath27 , @xmath28 is exemplified by morphisms from the trivial group , i.e. iff the diagonal map lies in @xmath29 where @xmath30 denotes the class of morphisms from the trivial group @xmath15 . a finite group @xmath23 is perfect iff @xmath24 is exemplified by @xmath31 or , equivalently , left unlike taking the commutator subgroup , i.e. @xmath24 lies in @xmath32 or , equivalently , in the class @xmath26\lra s\,:\ , s\text { an arbitrary group}\,\}^{\rtt l}$ ] .
a group @xmath23 is a prime - to-@xmath0-group , resp . a @xmath0-group , iff @xmath24 lies in the class @xmath33 , resp .
@xmath34 ; in words , it is right @xmath35-unlike , resp .
right unlike a @xmath35-unlike morphism .
classes @xmath36 and @xmath37 are the classes of surjective and injective morphisms , i.e. a map is surjective , resp .
injective , iff it is @xmath38-unlike , resp .
little attempt has been made to go beyond these examples .
hence open questions remain : are there other interesting examples of lifting properties in the category of ( finite ) groups ? can a complete group - theoretic argument be reformulated in terms of diagram chasing , say the classification of ca - groups or @xmath40-groups , or elementary properties of subgroup series ; can category theory notation be used to make expositions easier to read ? can these reformulations be used in automatic theorem proving ?
is there a decidable fragment of ( finite ) group theory based on the quillen lifting property and , more generally , diagram chasing , cf .
[ gls ] ? examples of lifting properties outside of group theory may be found in a short note [ dmg ] in _ the de morgan gazette _ and [ g1 ] .
this approach was motivated in part by hopes to express some statements of finite group theory in terms of diagrams in the category of finite groups , and then use automated diagram chasing to construct short formal proofs . particularly important for our motivation
was the fact that in the few examples of lifting properties we did find , it was the lifting property with respect to a simple counterexample in some intuitive sense , and that the lifting properties are closely related to the usual definitions . # 1#2#3#4#5#6 @xmath41^{i } \ar@{->}[d]_f & x \ar[d]^g \\ b \ar[r]_-{j } \ar@{-->}[ur]^{{\tilde j } } & y } $ ] @xmath42 \ar@{->}[d]_{(p ) } ` & x \ar[d]^{\therefore ( q ) } \\ b \ar[r ] \ar@{-->}[ur ] & y } $ ] @xmath43 \ar@{->}[d]_{\therefore(p ) } ` & x \ar[d]^ { ( q ) } \\ b \ar[r ] \ar@{-->}[ur ] & y } $ ]
figure 2 lists the diagrams representing the properties of groups .
below we describe the same diagrams , and make a couple of remarks about the diagram chasing and the notion of an inner automorphism and sylow theory .
there is no non - trivial homomorphism from a group @xmath44 to @xmath23 , write @xmath45 , iff @xmath46 a group @xmath47 is _ abelian _ iff @xmath48 where @xmath49 is the abelianisation morphism sending the free group into the abelian free group on two generators ; a group @xmath23 is _ perfect _
, @xmath50 $ ] , iff @xmath51 for any abelian group @xmath47 , i.e. @xmath52 equivalently , for an arbitrary homomorphism @xmath53 , @xmath54 yet another reformulation is that , for each group @xmath55 , @xmath56\lra s .\ ] ] in the category of finite or algebraic groups , a group @xmath57 is _ soluble _ iff @xmath58 for each perfect group @xmath23 , i.e. @xmath59 alternatively , a group @xmath57 is _ soluble _ iff for every homomorphism @xmath18 it holds @xmath60\lra g \text { for each group } g\ \implies\ f \,\rtt 0\lra\ , h .\ ] ] a prime number @xmath0 does not divide the number elements of a finite group @xmath23 iff @xmath23 has no element of order @xmath0 , i.e. no element @xmath61 such that @xmath62 yet @xmath63 , equivalently @xmath64 , i.e. @xmath65 a finite group @xmath23 is a @xmath0-group , i.e. the number of its elements is a power of a prime number @xmath0 , iff in the category of finite groups @xmath66 a group @xmath57 is the normal closure of the image of @xmath67 , i.e. no proper normal subgroup of @xmath57 contains the image of @xmath67 , iff for an arbitrary group @xmath23 @xmath68 a group @xmath69 is a subnormal subgroup of a finite group @xmath23 iff @xmath70 i.e. @xmath71 right - lifts wrt any map @xmath72 such that @xmath57 is the normal closure of the image of @xmath67 ; the lifting property implies that @xmath71 is injective . recall that @xmath69 is a subnormal subgroup of a finite group @xmath23 iff there is a finite series of subgroups @xmath73 such that @xmath74 is normal in @xmath75 , @xmath76 .
this is probably the only claim which requires a proof .
first notice that if @xmath69 is normal in @xmath23 then the lifting property holds . given a square corresponding to @xmath77 ,
the preimage of @xmath69 in @xmath57 is a normal subgroup of @xmath57 containing the image of @xmath67 , hence the preimage of @xmath69 contains @xmath57 and the lifting property holds .
the lifting property is closed under composition , hence it holds for subnormal subgroups as well .
now assume @xmath69 is not subnormal in @xmath23 .
as @xmath23 is finite , there is a minimal subnormal subgroup @xmath78 of @xmath23 . by construction
no proper normal subgroup of @xmath79 contains @xmath69 but the lifting property @xmath80 fails .
finally , a finite group @xmath23 is nilpotent iff the diagonal group @xmath23 is subnormal in @xmath81 [ nilp ] , i.e. iff the diagonal map @xmath82 , @xmath28 right - lifts wrt any @xmath72 such that @xmath57 is the normal closure of the image of @xmath67 , @xmath83 sylow theorem implies in a finite group @xmath23 , each @xmath0-subgroup is contained in a maximal one @xmath84 , @xmath85 is prime to @xmath0 , and the maximal @xmath0-subgroups are conjugated by an inner automorphism .
it is not clear how to express this in a satisfactory manner in terms of category theory ( diagram chasing ) .
perhaps something along the following lines : ( in the category of finite groups ) each arrow @xmath24 decomposes as @xmath86 uniquely up to conjugation . here
label @xmath87 may mean something like @xmath88 is a power of @xmath0 , and label @xmath89 may mean something like @xmath90 is prime to @xmath0 .
we remark that the notion of an inner automorphism can be reformulated in a diagram chasing manner .
an inner automorphism @xmath91 of a group @xmath23 extends to an automorphism @xmath92 of a group @xmath57 for any embedding @xmath93 .
[ inn , sch ] show this is a characterisation : an automorphism @xmath94 is inner iff it extends to an automorphism of @xmath57 for any embedding @xmath95 .
see [ inn ] and references therein for several more similar reformulations .
the feit - thompson theorem can be expressed as a combination of lifting properties : the theorem says says that each ( finite ) group of odd order is soluble , i.e. for each perfect finite group @xmath23 and each finite group @xmath57 , @xmath96 note that all these examples but the last one have a flavour of negation a notion being defined by the lifting property with respect to the simplest counterexample .
1#2#3#4#5#6#7 @xmath97@xmath98 @xmath99@xmath100@xmath101 @xmath102 @xmath103 + @xmath104 @xmath105@xmath106 @xmath107 @xmath108
it seems embarrassing to thank anyone for ideas so trivial , and we do that in the form of historical remarks .... this work is a continuation of [ dmg ] ; early history is given there .
examples here were motivated by a discussion with s.kryzhevich .
i thank paul schupp for pointing out the characterisation of inner automorphisms of [ sch ] .
i thank m.bays , k.pimenov , v.sosnilo and s.synchuk for proofreading , and several students for encouraging and helpful discussions . | reformulate notions finite group theory diagrams solvable nilpotent @xmath0-group prime - to-@xmath0 group abelian perfect subnormal subgroup injective surjective homomorphism see fig . 1 2 notions formulate feit - thompson theorem properties obtained iterating same diagram chasing trick '' starting from single morphism - counterexample quillen lifting property introduced by quillen algebraic topology motivation formulate finite group theory amenable automated theorem proving human readable tried basics general topology reformulate notions using four operations @xmath1 @xmath2 @xmath4 on classes @xmath5 of morphisms category quillen negation @xmath1 property @xmath5 class of all morphisms @xmath6 @xmath7 @xmath10 see fig . 1 definition quillen lifting property @xmath11 @xmath12 @xmath13 classes @xmath1 @xmath2 subcategories no morphisms from @xmath5 but isomorphisms classes @xmath14 @xmath4 denote subclass morphisms from to terminal object @xmath15 trivial groupclasses @xmath16 @xmath17 classes objects admitting no non - trivial map from objects @xmath14 morphism @xmath18 is left resp right unlike @xmath5 or @xmath19 @xmath20 @xmath18 left right exemplified by @xmath5 @xmath21 @xmath22 finite group @xmath23 soluble @xmath24 in class @xmath25 or right exemplified commutator subgroup @xmath26 finite group @xmath23 nilpotent diagonal map @xmath27 @xmath28 exemplified by morphisms from trivial group diagonal map in @xmath29 @xmath30 denotes class of morphisms from trivial group @xmath15 finite group @xmath23 perfect @xmath24 exemplified by @xmath31 or left unlike commutator subgroup @xmath24 in @xmath32 or class @xmath26 group @xmath23 is prime - to-@xmath0-group @xmath0-group @xmath24 in class @xmath33 resp @xmath34 right @xmath35-unlikeunlike @xmath35-unlike morphism classes @xmath36 @xmath37 surjective injective morphisms map is surjective injective @xmath38-unlike little attempt beyond examples open questions other examples of lifting properties finite ) groups ? group - theoretic argument reformulated diagram chasing classification ca - groups @xmath40-groups elementary properties subgroup series ; category theory notation expositions ? reformulations automatic theorem proving ? decidable fragment of finite ) group theory based on quillen lifting property diagram chasing gls ] ? examples lifting properties outside group theory in short note [ dmg ] de morgan gazette _ [ g1 ] approach motivated by express statements finite group theory diagrams groups use automated diagram chasing construct short formal proofs important few examples of lifting properties lifting simple counterexample closely related to usual definitions1#2#3#4#5#6 @xmath41 \ar@{->}[d]_f x \ar[d]^g\ar[r]_-{j \ar@{-->}[ur]^{{\tilde j y @xmath42 \ar@{->}[d]_{(p ) x \ar[d]^{\therefore q )\ar[r ] \ar@{-->}[ur ] y @xmath43 \ar@{->}[d]_{\therefore(p ) x \ar[d]^ {\ar[r\ar@{-->}[ur ] y figure 2 diagrams properties groups diagrams diagram chasing inner automorphism sylow theory no non - trivial homomorphism from @xmath44 to @xmath23 @xmath47 abelian _ @xmath48 @xmath49 abelianisation morphism free group abelian group @xmath23 perfect _ @xmath50 @xmath51 abelian group @xmath47 arbitrary homomorphism @xmath53 , @xmath54 group @xmath55 @xmath56 group @xmath57 _ soluble _ @xmath58 each perfect group @xmath23@xmath59 group @xmath57 soluble homomorphism @xmath18 holds @xmath60 each group prime number @xmath0 divide elements finite group @xmath23 no element order element @xmath61 @xmath62 @xmath63 @xmath64 @xmath65 finite group @xmath23 @xmath0-group number elements power prime number @xmath0 finite groups @xmath66 group @xmath57 normal closure image @xmath67 no normal subgroup contains image @xmath67 arbitrary group @xmath23 @xmath68 group @xmath69 subnormal subgroup finite group @xmath23 @xmath70 @xmath71 right - lifts map @xmath72 @xmath57 normal closure image @xmath67 lifting property implies @xmath71 injective @xmath69 subnormal subgroup finite group @xmath23 finite series subgroups @xmath73 @xmath74 normal @xmath75 @xmath76 claim requires proof @xmath69 normal @xmath23 lifting property holdssquare @xmath77 preimage @xmath69 in @xmath57 normal subgroup @xmath67 contains @xmath57 lifting property holds lifting property closed under composition holds for subnormal subgroups @xmath69 not subnormal in @xmath23 @xmath23 finite minimal subnormal subgroup @xmath78 of @xmath23 no normal subgroup @xmath79 contains @xmath69 property @xmath80 fails finite group @xmath23 nilpotent diagonal group @xmath23 subnormal in @xmath81 diagonal map @xmath82 @xmath28 right - lifts @xmath72 @xmath57 normal closure image @xmath67 @xmath83 sylow theorem finite group @xmath23 @xmath0-subgroup maximal @xmath84 @xmath85 prime to @xmath0 maximal @xmath0-subgroups conjugated by inner automorphism express category theory arrow @xmath24 decomposes as @xmath86 conjugation label @xmath87 @xmath88 power of @xmath0 @xmath89 @xmath90 prime to @xmath0 notion inner automorphism reformulated diagram chasinginner automorphism @xmath91 @xmath23 extends @xmath92 @xmath57 embedding @xmath93 characterisation automorphism @xmath94 inner extends @xmath57 embedding @xmath95 see inn ] similar reformulations feit - thompson theorem combination lifting properties each finite group odd order soluble perfect finite group @xmath23 @xmath57 @xmath96 examples last one flavour negation defined lifting property simplest counterexample 1#2#3#4#5#6#7 @xmath97@xmath98 @xmath99@xmath100@xmath101 @xmath102 @xmath103 @xmath104 @xmath105@xmath106 @xmath107 @xmath108 embarrassing to thank ideas trivial historical remarks work continuation of [ dmg ] early history examples motivated discussion with s.kryzhevich thank paul schupp characterisation inner automorphisms sch thank m.bays k.pimenov v.sosnilo s.synchuk proofreading students discussions | 0.429937 | 2,786 | 1,541 | , Saving $0.1 in GPT-4. |
67,688 | we calculate flavour dependent lepton asymmetries within the @xmath0 inspired supersymmetric standard model ( @xmath1 ) .
our analysis reveals that the substantial lepton cp asymmetries can be induced even if @xmath2 . | in models with heavy right handed neutrinos lepton asymmetry can be dynamically generated via the out of equilibrium decay of the lightest right handed neutrino @xmath3 .
this asymmetry subsequently gets converted into the baryon asymmetry due to sphaleron interactions . in the standard model and its minimal supersymmetric extension the appropriate amount of baryon asymmetry
can be induced only if the mass of the lightest right handed neutrino @xmath4 . in this case reheat temperature
has to be relatively high , i.e. @xmath5 , that results in an overproduction of gravitinos . to avoid gravitino problem
@xmath6 should be low enough @xmath7 . here
we study the generation of lepton cp asymmetries within the the exceptional supersymmetric standard model ( e@xmath8ssm ) @xcite .
the e@xmath8ssm is based on the @xmath9 gauge group which is a subgroup of @xmath0 . by definition @xmath10 where @xmath11 , @xmath12 . to ensure anomaly cancellation the particle content of the e@xmath8ssm
is extended to include three complete * @xmath13 * representations of @xmath0 @xcite .
the * @xmath14 * multiplets contain sm family of quarks and leptons , right handed neutrino @xmath15 , sm singlet field @xmath16 which carry non
zero @xmath17 charge , a pair of @xmath18doublets @xmath19 and @xmath20 which have the quantum numbers of higgs doublets and a pair of colour triplets of exotic quarks @xmath21 and @xmath22 which can be either diquarks ( model i ) or leptoquarks ( model ii ) . @xmath19 and @xmath20 form either higgs or inert higgs multiplets .
in addition to the complete @xmath14 multiplets the low energy particle spectrum of the e@xmath8ssm is supplemented by @xmath18 doublet @xmath23 and anti - doublet @xmath24 from extra @xmath25 and @xmath26 to preserve gauge coupling unification @xcite . in the e@xmath8ssm the right - handed neutrinos
do not participate in the gauge interactions and therefore are expected to gain masses at some intermediate scale , while the remaining matter survives down to the electroweak scale .
the part of the e@xmath8ssm superpotential describing the interactions of @xmath27 with other fields can be written as @xmath28 where @xmath29 while @xmath30 . in the model i @xmath31 whereas in the model ii
all terms in eq .
( [ 1 ] ) can be present . +
the process of the lepton asymmetry generation is controlled by the flavour cp ( decay ) asymmetries .
the cp asymmetries associated with the decays of @xmath3 and @xmath32 into leptons and sleptons are given by @xmath33 where @xmath34 and @xmath35 correspond to either lepton or slepton field while @xmath36 and @xmath37 are partial decay widths of @xmath38 and @xmath39 .
at the tree level cp asymmetries vanish .
the non zero contributions to the cp asymmetries arise from the interference between the tree
level amplitudes of the lightest right handed neutrino decays and one loop corrections to them .
supersymmetry ensures that @xmath40 the tree level and one
loop diagrams that contribute to the @xmath41 are shown in fig . 1 . when @xmath42 we get ( see @xcite ) @xmath43 \left(\frac{m_1}{m_j}\right)\,,\ ] ] where @xmath44 if @xmath45 while @xmath46 .
+ + the cp asymmetries associated with the decays of @xmath3 and @xmath32 into exotic quarks ( squarks ) can be defined similarly to the lepton ones . the interference of the tree
level decay amplitude with the one loop corrections ( see fig .
2 ) yields @xmath47\left(\frac{m_1}{m_j}\right)\,,\ ] ] where @xmath48 .
to simplify our analysis we impose @xmath49 symmetry under which all superfields except @xmath50 , @xmath51 and @xmath52 are odd . the @xmath49 symmetry allows to avoid unacceptably large non - diagonal flavour transitions and reduces the part of the e@xmath8ssm superpotential that describes the interactions of @xmath27 with other superfields @xmath53 we also assume that the right handed neutrino mass scale is relatively low ( @xmath54 ) , so that @xmath55 are negligibly small ( @xmath56 ) , and @xmath57 . in this approximation @xmath58 is much larger than other cp asymmetries . in the considered case
the maximal absolute value of @xmath58 is given by @xmath59 is a coupling of the second lightest right handed neutrino to @xmath60 and @xmath23 . from eq .
( [ 7 ] ) one can see that in the considered case the substantial lepton decay asymmetry @xmath61 ( @xmath62 ) can be induced even for @xmath2 if @xmath63 vary from @xmath64 to @xmath65 . at low energies
the asymmetry generated in the @xmath23 sector gets converted into the ordinary lepton asymmetries via the decays of @xmath23 .
this suggests that in the e@xmath8ssm successful thermal leptogenesis can be achieved without encountering the gravitino problem
. 9 s.f .
king , s. moretti , r. nevzorov , `` theory and phenomenology of an exceptional supersymmetric standard model '' , phys .
d * 73 * ( 2006 ) 035009 ; s.f .
king , s. moretti , r. nevzorov , `` exceptional supersymmetric standard model '' , phys .
b * 634 * ( 2006 ) 278 ; s.f .
king , s. moretti , r. nevzorov , `` gauge coupling unification in the exceptional supersymmetric standard model , '' phys .
b * 650 * ( 2007 ) 57 . | models heavy right handed neutrinos lepton asymmetry generated via decay lightest neutrino @xmath3 asymmetry converted into baryon asymmetry sphaleron interactions standard model supersymmetric extension baryon asymmetry induced mass lightest neutrino @xmath4 reheat temperature high @xmath5 overproduction gravitinos avoid problem @xmath6 low @xmath7 study generation lepton cp asymmetries supersymmetric standard model e@xmath8ssm based on @xmath9 gauge group subgroup @xmath0 cancellation particle content extended three @xmath13 representations of @xmath0 @xcite @xmath14 multiplets contain sm family quarks leptons right handed neutrino @xmath15 sm singlet field @xmath16 zero @xmath17 charge @xmath18doublets @xmath19 @xmath20 higgs doublets colour triplets of exotic quarks @xmath21 @xmath22 diquarks leptoquarks @xmath19 @xmath20 form higgs or inert higgs multiplets@xmath14 multiplets low energy particle spectrum e@xmath8ssm supplemented by @xmath18 anti - doublet @xmath24 @xmath25 gauge coupling unification e@xmath8ssm right - handed neutrinos participate interactions gain masses intermediate scale remaining matter survives electroweak scale@xmath8ssm superpotential interactions @xmath27 fields written @xmath28 @xmath29 @xmath30 model @xmath31 all terms present lepton asymmetry generation controlled by flavour cp decay asymmetries asymmetries decays @xmath3 @xmath32 into leptons sleptons @xmath33 @xmath34 @xmath35 correspond lepton slepton field partial decay widths @xmath38 @xmath39 tree level cp asymmetries vanish non zero contributions interference between tree level amplitudes right handed neutrino decays one loop corrections supersymmetry ensures @xmath40 tree level one loop diagrams @xmath41 shown fig . 1 @xmath42 @xmath43 @xmath44 @xmath45 @xmath46cp asymmetries decays @xmath3 @xmath32 into exotic quarks defined to lepton ones interference tree level decay with one loop corrections yields @xmath47\left(\right @xmath48 @xmath49 symmetry superfields except odd symmetry large non - diagonal flavour transitions reduces e@xmath8ssm superpotential interactions @xmath27 with superfields @xmath53 right handed neutrino mass scale low @xmath54 @xmath55 small @xmath58 larger than cp asymmetries maximal absolute value of @xmath58 by @xmath59 coupling of second lightest right handed neutrino to @xmath60 @xmath23 substantial lepton decay asymmetry @xmath61 ( @xmath62 ) induced for @xmath2 if @xmath63 vary from @xmath64 to @xmath65 at low energies asymmetry in @xmath23 sector converted into ordinary lepton asymmetries via decays @xmath23 e@xmath8ssm successful thermal leptogenesis without gravitino problem king morettinevzorov theory exceptional supersymmetric standard model phys . d * 73 2006 ) 035009 king moretti nevzorov supersymmetric standard model phys . b 634 2006 ) 278 king gauge coupling unification standard model . b * 650 2007 ) 57 . | 0.258852 | 1,402 | 737 | , Saving $0.0 in GPT-4. |
67,689 | analysis of transit times in exoplanetary systems accurately provides an instantaneous orbital period , @xmath0 , of their member planets .
a long - term monitoring of those transiting planetary systems puts limits on the variability of @xmath0 , which are translated into the constraints on the time variation of the gravitational constant @xmath1 .
we apply this analysis to @xmath2 transiting systems observed by the _ kepler
_ spacecraft , and find that @xmath3 for 20092013 , or @xmath4 if @xmath5 is constant . while the derived limit is weaker than those from other analyses , it is complementary to them and can be improved by analyzing numerous transiting systems that are continuously monitored . | are the fundamental _ constants _ in nature really _ constant _ over the cosmological time scale ?
this question has a long history in physics and cosmology , and has been discussed intensively in different contexts .
one of the most famous examples includes the large number hypothesis by @xcite , who raises a possibility of the gravitational constant @xmath1 being proportional to @xmath6 .
recently , @xcite reported a curious oscillatory trend in the values of @xmath1 measured over the last three decades , @xmath7 $ ] .
more intriguingly , they claimed that the period and phase of the modulation are in agreement with the variation of the length of the day of the earth @xcite .
the proposed modulation , however , is unlikely to reflect the real time variation of @xmath1 @xcite .
indeed , the amplitude of @xmath8 is shown to be inconsistent with the dynamics of the solar system @xcite .
in addition , the subsequent studies have shown that the trend reported by @xcite is more likely to be an artifact .
nevertheless , it is important and interesting to discuss and compare with other independent constraints on @xmath9 on such short time scales since most of the previous literature focused on @xmath9 averaged over the cosmological time scale . for that purpose
, we consider the orbital periods of transiting exoplanetary systems in the present paper . so far , more than 4000 candidate systems have been reported by the _ kepler _ mission @xcite , which have been monitored over several years , and approximately 1000 of them have been confirmed to host planets .
the orbital period @xmath10 of such planets can be accurately determined by their central transit times .
furthermore , a systematic search for any time variability of the period has been performed primarily in order to probe the gravitational interaction between multiple planets , which is referred to as transit timing variation ( ttv , * ? ? ?
* ; * ? ? ?
the ttv analysis is conventionally used to determine the mass of planets without radial velocity follow - up observations and/or to infer the presence of undetected perturbers .
instead , we attempt to put a constraint on the time variation of @xmath1 from the same analysis , but focusing on those systems that exhibit no clear ttv signature . in particular , _ hot jupiters _ , planets orbiting around host stars within a week or so , are particularly suited for constraining the variation of @xmath1 on time scales of months to years .
as will be shown below , our sample yields the constraint @xmath4 _ if @xmath5 is constant _ , which is weaker by six orders of magnitude than those based on the pulsar timing ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) and the lunar ranging ( e.g. , * ? ? ?
we would like to emphasize , however , that the conventional assumption of the constant @xmath11 is not general , but has been introduced just for simplicity .
if @xmath1 would vary periodically as was claimed by @xcite , planetary systems with different orbital periods would be ideal to search for the possible resonant effect close to the variation period of days to years , and thus the resulting constraint on the amplitude of @xmath11 could be stronger depending on the expected period of the oscillation
. therefore it would be interesting to know the current limit from the existing systems at this point in any case .
the precise data of our solar system also put more stringent constraints on the variation of @xmath1 , but they are mostly sensitive to the variability on time scales of years , and the perturbative effect of eight planets and other bodies need to be carefully separated , as has been performed by @xcite for instance . on the contrary , the result from different transiting systems can be simply added for tighter constraints because the effect of @xmath9 should change the period of any system in a coherent fashion . for these reasons , ttvs of transiting planetary systems offer a complementary and straightforward method to explore the variation of @xmath1 on shorter time scales .
in the two - body problem , motion of a planet around its host star is exactly periodic , and so are its transits .
the transit times of a planet in a multiple planetary system , however , sometimes deviate from the exact periodicity due to the gravitational perturbation from other planets in the system .
this phenomenon , known as the `` transit timing variation '' ( ttv , * ? ? ?
* ; * ? ? ?
* ) , has been successfully modeled in tens of multi - transiting planetary systems discovered by _
kepler _ to determine the mass ratios of their member planets ( e.g. , * ? ? ?
* ; * ? ? ?
* ; * ? ? ?
* ) as well as to infer the existence of non - transiting planets ( e.g. , * ? ? ?
* ; * ? ? ?
figure [ fig : ttv ] displays an example of the ttv signals for the kepler-9 system , which is the first multiple planetary system detected through the transit method .
it hosts two transiting planets , kepler-9b ( @xmath12 ) and kepler-9c ( @xmath13 ) with orbital periods of @xmath14 and @xmath15 , respectively @xcite . in this system ,
the orbital periods of the two planets are not exactly constant because of the strong mutual gravitational interaction between the two in addition to the dominant gravity due to the host star ( 1.07@xmath16 , * ? ? ?
this results in systematic deviations of the central transit times with respect to the mean period as exhibited in filled circles .
such ttv signals have often been used to constrain the system parameters , in particular to estimate the mass of planets without radial velocity measurement . in the following ,
however , we use the _ absence _ of the ttv signal to put an upper limit on the period variation @xmath17 , which translates into @xmath9 through kepler s third law : @xmath18 where @xmath19 is the semi - major axis and @xmath20 is the total mass of the two - body system . although equation ( [ eq : kepler-3rd ] ) is exactly correct only when @xmath1 is constant , we assume that it still holds when @xmath1 varies adiabatically , as we consider here
. then it leads to @xmath21 note that the variation of @xmath1 , in principle , simultaneously induces non - vanishing @xmath22 and @xmath23 .
thus it may be more useful to rewrite equation ( [ eq : kepler-3rd ] ) in terms of the specific angular momentum @xmath24 and the eccentricity @xmath25 : @xmath26 then we have @xmath27 equation ( [ eq : dotp-2 ] ) simply reduces to @xmath28 if the orbit is circular and both the specific angular momentum and mass are conserved under the variation of @xmath1 ( e.g. , * ? ? ? * ) .
while the assumption of @xmath29 is perfectly justified for non - relativistic stars and planets considered here , it is not the case for compact objects including neutron stars . among the confirmed transiting planets observed with _ kepler _ , we select kepler-1b ( or tres-2 ) with @xmath30 and @xmath31 period , and kepler-2b ( or hat - p-7b ) with @xmath32 and @xmath33 period .
the two planets have the highest transit signal - to - noise ratio , while exhibiting no identifiable feature of ttvs . in figure
[ fig : individual ] , we show the fractional variations in the orbital periods of these planets against the observed date . for each planet
, we follow the procedure of @xcite to determine the central times of individual transits @xmath34 , where @xmath35 stands for the number of transits counted from a fixed epoch . here
we use only the data sampled at a short cadence ( one minute ) , which yield the transit times with higher precision than the long cadence data sampled at the @xmath36-minute interval .
we compute the orbital period @xmath37 between @xmath34 and @xmath38 for each @xmath35 and plot @xmath39 against @xmath40 , where @xmath41 is an average of all @xmath37 .
the corresponding root - mean - square @xmath42 are @xmath43 and @xmath44 for kepler-1b and kepler-2b , respectively .
these values translate into @xmath45 and @xmath46 over four years , which are smaller than the amplitude of the proposed variation of @xmath1 @xcite by an order of magnitude .
the constraints from the above two systems almost reach the limit of the _ kepler _ photometry for the existing systems .
however , the further improvement can be achieved by combining many systems in a statistical fashion .
we select 10 confirmed transiting planetary systems with the highest transit signal - to - noise ratios : kepler-1 , kepler-2 , kepler-13 , kepler-12 , kepler-6 , kepler-7 , kepler-423 , kepler-17 , kepler-5 , and kepler-3 .
none of these systems exhibit any clear ttvs .
although the transit times in some of these systems ( e.g. , kepler-3 or alias hat - p-11 ) are affected by the strong star - spot activities that deform the transit signals , we do not exclude them because our purpose here is simply to illustrate the advantage of combining the constraints from many independent systems .
we apply the same analysis as in the previous subsection to the above 10 systems .
we plot the resulting @xmath42 as the gray dots in figure [ fig : statistical ] , and their averages in @xmath2-day bins as the black circles with error bars . now
the standard deviation of the binned @xmath42 is @xmath47 , which is a few times smaller than the constraints from the individual systems .
note that the choice of the smoothing bin size is completely arbitrary at this point .
the binning would smooth out a possible variation of @xmath1 less than the bin size and the resulting @xmath42 would depend on the bin size . here
we choose the 10-day bin just for definiteness , and could use a different value if a specific model of the variation of @xmath1 is given .
the periodogram of the binned @xmath42 in figure [ fig : periodogram ] delivers a rough idea of the expected constraints for the different choice of the smoothing bin size .
all of the peaks in the periodogram have the amplitudes less than the significance level of @xmath48 ( horizontal dashed line ) and are consistent with the gaussian noise , i.e. , the data exhibit no significant modulation over the period range down to @xmath49 .
a larger bin size therefore should put a somewhat stronger constraint on @xmath42 , and hence on @xmath50 .
the proper interpretation of the constraint , however , depends on the model of @xmath51 .
we have shown that the methodology of using exoplanets as a precision clock provides reaonably interesting limits on the time variation of @xmath1 .
our current analysis finds @xmath52 for 20092013 , which corresponds to @xmath53 if @xmath5 is constant .
in contrast , constraints from lunar laser ranging experiments and the binary pulsar system psr b1913 + 16 correspond to @xmath54 @xcite and @xmath55 @xcite , respectively . although our current constraint is significantly less stringent than the binary pulsar timing result that is based on the similar principle , it is complementary in many aspects .
first , we can safely neglect @xmath23 for exoplanetary systems unlike binary pulsar systems where the self - gravitational energy significantly contributes to the total mass .
second , we may constrain the @xmath22-term independently of @xmath56 by combining the precise measurement of the radial velocity ( proportional to @xmath57 ) with the photometric transit timing data .
third , the timing analysis of the secondary eclipse ( opposite phase of the planetary transit in front of the host star , i.e. , the occultation of the planet by the stellar disk ) of transiting planets may also constrain the @xmath22-term through the variation of the expected arrival time difference @xmath58 from the photometric data alone .
fourth , the growing number of continuously monitored transiting planets promises to statistically improve the constraint beyond the value that we discussed in this paper .
incidentally , if @xmath1 varies periodically with an oscillation period of days to years , the dynamics of the planetary system with the similar orbital period would be perturbed in a resonant fashion .
while we have not studied such a possible resonant effect in this paper , constraints from systems with different periods are important once the simple assumption of constant @xmath11 is abandoned .
finally , we would like to emphasize that the current data for numerous exoplanetary systems are already precise enough to put complementary and meaningful constraints on the variation of the fundamental constants , as has been feasible only through the solar system data and/or a few binary pulsars .
thus future data of the exoplanetary systems should definitely improve the situation and might even bring an unexpected surprise .
we thank teruaki suyama for calling our attention to @xcite .
is supported by jsps research fellowships for young scientists ( no .
26 - 7182 ) , and also by the leading graduate course for frontiers of mathematical sciences and physics .
y.s . acknowledges the support from the grant - in aid for scientific research by jsps ( no . 24340035 ) .
mean motion resonance ( kepler-9b with @xmath59 , top ; kepler-9c with @xmath60 , bottom ) transit the same host star .
the residuals of the transit times from the linear fit are plotted against the _ kepler _ observational time spanning about four years .
the error of each transit time is smaller than the point size .
[ fig : ttv],scaledwidth=50.0% ] | fundamental constants in nature constant over cosmological time scale ? question long history in physics cosmology discussed in different contexts famous large number hypothesis by @xcite gravitational constant @xmath1 proportional to @xmath6 @xcite reported oscillatory trend in values @xmath1 over last three decades claimed period phase modulation with length of day earth proposed modulation unlikely to reflect real time variation of @xmath1 amplitude of @xmath8 inconsistent with dynamics solar system studies trend @xcite likely artifact important to discuss compare other independent constraints on @xmath9 on short time scales previous literature focused on @xmath9 over time scale consider orbital periods of transiting exoplanetary systems in present paper than 4000 candidate systems reported by kepler mission @xcite monitored over years approximately 1000 confirmed to host planets orbital period @xmath10 of planets determined by central transit timessystematic search for time variability performed gravitational interaction between planets referred as transit timing variation ( ttv ttv analysis used to determine mass of planets without radial velocity infer presence of undetected perturbers attempt to constraint on time variation of @xmath1 focusing on systems no clear ttv signature hot jupiters _ planets orbiting around host stars within week suited for constraining variation @xmath1 on scales months to years sample yields constraint @xmath4 _ if @xmath5 is constant weaker by six orders of magnitude than based on pulsar timing and lunar ranging conventional assumption of constant @xmath11 not general introduced for simplicity if @xmath1 vary periodically planetary systems with different orbital periods ideal to search for resonant effect close to variation period of days to years constraint on amplitude of @xmath11 could be stronger depending on expected period of oscillation interesting to know current limit from existing systemsprecise data solar system constraints on variation @xmath1 sensitive to variability on time scales years perturbative effect of eight planets other bodies separated by @xcite result from different transiting systems can added for tighter constraints effect of @xmath9 period system ttvs of transiting planetary systems offer method to explore variation @xmath1 on shorter time scales in two - body problem motion planet around host star periodic transits transit times planet in multiple planetary system sometimes deviate from periodicity due to gravitational perturbation from other planets phenomenon `` transit timing variation '' modeled in multi - transiting planetary systems discovered by kepler to determine mass ratios of planets infer existence of non - transiting planets figure displays ttv signals for kepler-9 system first multiple planetary system detected through transit method hosts two transiting planets kepler-9b ( @xmath12 and kepler-9c ( @xmath13 ) with orbital periods of @xmath14 and @xmath15 @xcitesystem orbital periods two planets not constant strong mutual gravitational interaction dominant gravity due host star 1.07@xmath16 results in deviations central transit times mean period circles ttv signals used constrain system parameters estimate mass planets without radial velocity measurement of ttv signal upper limit on period variation @xmath17 translates into @xmath9 through kepler s third law @xmath18 @xmath19 semi - major axis @xmath20 total mass two - body system equation kepler-3rd correct when @xmath1 constant holds when @xmath1 varies adiabatically leads to @xmath21 variation of @xmath1 induces non - vanishing @xmath22 and @xmath23 rewrite equation kepler-3rd specific angular momentum @xmath24 eccentricity @xmath25 : @xmath26 @xmath27 equation reduces to @xmath28 if orbit circular specific angular momentum mass conserved under variation of @xmath1 assumption of @xmath29 justified for non - relativistic stars planets not for compact objects including neutron starsconfirmed transiting planets kepler select kepler-1b @xmath30 @xmath31 period kepler-2b @xmath32 @xmath33 two planets highest transit signal - to - noise ratio no identifiable feature ttvs figure fractional variations orbital periods against observed date follow procedure @xcite determine central times transits @xmath34 @xmath35 number transits fixed epoch use data short cadence one minute transit times higher precision long cadence-minute interval compute orbital period @xmath37 between @xmath34 @xmath38 each @xmath35 plot @xmath39 against @xmath40 @xmath41 average of @xmath37 root - mean - @xmath42 are @xmath43 @xmath44 for kepler-1b kepler-2b values translate into @xmath45 @xmath46 over four years smaller than amplitude proposed variation @xmath1 @xcite order of magnitude constraints from systems reach limit kepler _ photometry improvement combining many systems select 10 transiting planetary systems with highest transit signal - to - noise ratios kepler-1 kepler-2 kepler-3systems exhibit clear ttvs transit times some kepler-3 or alias hat - p-11 affected by star - spot activities transit signals exclude them illustrate advantage of combining constraints from systems apply same analysis to above 10 systems plot resulting @xmath42 as gray dots in figure averages in @xmath2-day bins as black circles with error bars standard deviation of binned @xmath42 is @xmath47 few times smaller than constraints smoothing bin size arbitrary variation of @xmath1 less @xmath42 on bin size choose 10-day bin for definiteness use different value if specific model variation @xmath1 periodogram of binned @xmath42 in figure delivers idea expected constraints for smoothing bin size peaks have amplitudes less than significance level of @xmath48 consistent with gaussian noise data exhibit no significant modulation over down to @xmath49 larger bin size should stronger constraint on @xmath42 @xmath50 interpretation of constraint depends on model of @xmath51 using exoplanets as precision clock provides limits on time variation of @xmath1analysis finds @xmath52 for 20092013 corresponds to @xmath53 if @xmath5 constant constraints from lunar laser experiments binary pulsar system psr b1913 + 16 correspond to @xmath54 and @xmath55 @xcite current constraint less stringent than binary pulsar timing result complementary in can neglect @xmath23 for exoplanetary systems unlike binary pulsar systems self - gravitational energy contributes to mass constrain @xmath22-term independently of @xmath56 by combining radial velocity proportional to @xmath57 with photometric transit timing data timing analysis of secondary eclipse of transiting planets may constrain @xmath22-term through variation expected arrival time difference @xmath58 from photometric data growing number of monitored transiting planets improve constraint if @xmath1 varies periodically to dynamics of planetary system similar perturbed constraints from systems with different periods important once assumption of constant @xmath11 abandoned current data for exoplanetary systems precise enough to put complementary constraints on variation of fundamental constants feasible only through solar system data binary pulsarsfuture data exoplanetary systems improve situation bring unexpected surprise thank teruaki suyama @xcite supported jsps research fellowships scientists 26 - 7182 graduate course mathematical sciences physics acknowledges support grant aid scientific research jsps 24340035 ) motion resonance kepler-9b @xmath59 kepler-9c @xmath60 transit same host star residuals transit times plotted against kepler observational time four years error transit smaller than point size fig : ttv=50.0% | 0.499673 | 3,029 | 1,555 | , Saving $0.1 in GPT-4. |
67,693 | using recent d data on the dijet mass spectrum , we present a limit on flavor - universal colorons . at 95% cl
we find @xmath0 gev .
we discuss the implications of this limit for models of quark compositeness , non - standard gluon interactions , and dynamical electroweak symmetry breaking .
in addition , we place a lower bound @xmath1 tev on the scale of color - octet axial - vector contact interactions among quarks which could arise in models of quark compositeness .
( 0,0)(0,0 ) ( 295,280)buhep-98 - 24 ( 295,265)hep - ph/9809472 | the flavor - universal coloron model @xcite was originally proposed to explain the apparent excess of high-@xmath2 jets in the inclusive jet spectrum measured at the fermilab tevatron by the cdf collaboration @xcite .
this model is a flavor - universal variant of the coloron model of hill and parke @xcite which involves a minimal extension of the standard description of the strong interactions , including the addition of one gauge interaction and a scalar multiplet , but no new fermions .
the flavor - universal coloron model of the strong interactions can be grafted onto the standard one - higgs - doublet model of electroweak physics , yielding a simple , complete , and renormalizable theory .
alternatively , it can provide the basis for dynamical generation of electroweak symmetry breaking and the generation of the top quark s mass in models @xcite akin to top - color - assisted technicolor @xcite .
previous work on the phenomenology of the colorons has considered effects on the @xmath3 parameter @xcite , the inclusive jet spectrum @xcite , the dijet spectrum and angular distributions @xcite , and b - tagged dijets @xcite .
the most recent of these analyses @xcite put a limit @xmath4 gev on the the coefficient of the four - fermion contact interaction to which heavy coloron exchange would give rise .
this letter explores the effects colorons would have on the dijet mass spectrum measured at the tevatron by the d collaboration @xcite and establishes a still stronger limit on contact interactions arising from colorons : @xmath0 gev at 95% cl . in section 2 ,
we briefly review the model .
section 3 explains how our limit was derived .
section 4 discusses our conclusions and the implications for models of quark compositeness , non - standard gluon interactions , and dynamical electroweak symmetry breaking .
we also present a separate limit on the scale of color - octet axial - vector contact interactions among quarks .
in the flavor - universal coloron model @xcite , the strong gauge group is extended to @xmath5 .
the gauge couplings are , respectively , @xmath6 and @xmath7 with @xmath8 .
each quark transforms as a ( 1,3 ) under this extended strong gauge group .
the model also includes a scalar boson @xmath9 transforming as a @xmath10 under the two @xmath11 groups . for a range of couplings in the scalar potential
, @xmath9 develops a vacuum expectation value @xmath12 which breaks the two strong groups to their diagonal subgroup @xcite .
we identify this unbroken subgroup with qcd . when the extended color symmetry breaks , the original gauge bosons mix to form an octet of massless gluons and an octet of massive colorons .
the gluons interact with quarks through a conventional qcd coupling with strength @xmath13 .
the colorons @xmath14 interact with quarks through a new qcd - like coupling @xmath15 where @xmath16 is the color current @xmath17 and @xmath18 .
the mass of the colorons may be written @xmath19 in terms of the parameters of the model .
below the scale @xmath20 , coloron - exchange may be approximated by the effective four - fermion interaction @xmath21 this can be re - written in the form commonly used in studies of quark compositeness @xcite @xmath22 where the scale @xmath23 is defined by @xmath24 .
contact interactions of this kind tend to increase quark - quark scattering at high invariant mass above the standard qcd prediction .
the d collaboration recently @xcite measured the inclusive dijet mass spectrum at @xmath25= 1.8 tev for dijet masses above 200 gev and jet pseudorapidity @xmath26 . the collaboration also measured the ratio of spectra at @xmath27 and @xmath28 and used this to place a limit on certain models of quark compositeness .
we use this same ratio of spectra to place limits on the effects of flavor - universal colorons .
we calculated the leading - order ( lo ) dijet spectrum @xmath29 { d\sigma\over d\hat{t}}(ab \to cd ) \nonumber\end{aligned}\ ] ] in terms of the jets pseudorapidities ( @xmath30 , @xmath31 ; @xmath32 ) , combined invariant mass ( @xmath33 ) and momentum fractions ( @xmath34 ) ; the parton distribution functions ( @xmath35 ) ; and the two - body parton scattering cross section @xmath36 the leading qcd contributions to @xmath37 may be found in @xcite and the contributions from heavy coloron exchange are given . ] in @xcite .
note that we included only production of gluons and light quarks in our calculations since produced top quarks would not contribute appreciably to the d dijet sample jet . ]
taking a series of different values for the coloron interaction strength @xmath38 , we then determined the ratio @xmath39 at values of @xmath33 corresponding to the weighted center of each mass bin for 800 gev @xmath40 1400 gev ; this yielded the same ratio as keeping @xmath33 fixed at the weighted center value of 873.2 gev . ]
measured by d in ref.@xcite .
we evaluated the fractional difference between the dijet spectra for pure qcd and for the various values of @xmath38 . to simulate a next - to - leading order ( nlo ) prediction of the effects of colorons
, we multiplied a nlo qcd prediction obtained using the jetrad program @xcite by the lo fractional differences .
the results are illustrated in figure 1 .
we extract a limit on the coloron interaction strength from the ddata @xcite using bayesian techniques with a gaussian likelihood function @xmath41^t\ , s^{-1}\ , [ d - f(x ) ] \right)\ ] ] where @xmath42 is the vector of data points for the different mass bins , @xmath43 is the vector of theory points for the different masses at different values of @xmath44 , and @xmath45 is the covariance matrix . motivated by the form of @xmath46 ( eqn .
( [ lamta ] ) ) , the prior probability is assumed to be flat when @xmath47 .
since the ratio of spectra at nlo is sensitive to the choice of @xmath48 and parton distribution function , each possible choice is treated as a different theory . the 95% confidence limit ( cl ) on @xmath49
is calculated by requiring that @xmath50 the limit in @xmath51 is then transformed back into a limit on @xmath49 and thence into a limit on @xmath38 .
the most conservative lower bound we obtain at the 95% cl is @xmath0 gev for the cteq3 m pdf and @xmath52 ( where is the maximum jet @xmath2 in the event ) as illustrated in figure 2 .
this limit is incompatible with the suggestion of a coloron interaction strength of order 700 gev @xcite in earlier measurements of the high @xmath2 jet inclusive cross - section @xcite .
our limit places a new exclusion bound in the @xmath53 vs. @xmath54 parameter space of the flavor - universal coloron model . as shown in figure 3 , this improves on the recent d limit based on the dijet angular distribution @xcite .
note that the region at @xmath55 where our limit appears to provide a direct lower bound on @xmath53 has already been excluded by cdf s search for new particles decaying to dijets @xcite .
this is fortunate , because our limit actually becomes less reliable here : the condition @xmath56 , under which ( [ conta ] ) is a reasonable approximation to coloron exchange , would no longer hold for the highest - energy data point .
an updated search for new resonances decaying to dijets would be a useful complement to the bounds we report here . in the context of the dynamical electroweak symmetry breaking model of @xcite in which flavor - universal colorons help
produce the mass of the top quark , the value of @xmath54 is approximately 4 . in other words ,
the interaction strength is near its upper limit for the higgs phase of the model @xcite .
our bound implies that @xmath57 tev in such models , placing them at the upper right of the allowed region in figure 3 .
our findings also set a limit on a broader array of new strong interaction physics .
writing eqn.([conta ] ) in the more conventional form for compositeness studies ( [ lamta ] ) shows that our limit is equivalent to a lower bound @xmath58 tev on the scale of new color - octet vectorial current - current interactions .
such interactions could arise from quark compositeness or from non - standard gluon interactions ( e.g. gluon compositeness ) @xcite .
finally , we have used the same methods to set a limit on a color - octet axial - vector current - current interaction among quarks .
this has the form @xmath59 where @xmath60 .
the contributions of this contact interaction to parton scattering are given .
] in @xcite .
the most conservative bound we find is @xmath61 tev at 95@xmath62 cl . note that we can not self - consistently interpret this as providing a lower bound on the mass of an axigluon @xcite whose exchange underlies the contact interaction ( [ axta ] ) .
the relation @xmath63 ( since @xmath64 for axigluons ) shows that the mass of the supposed axigluon would not satisfy the condition @xmath65 . instead
, our limit should be interpreted as bounding the scale of contact interactions arising in models of quark compositeness .
we thank the d collaboration for making the measurements that made this work possible .
we also thank r.s .
chivukula for comments on the manuscript .
e.h.s . acknowledges the support of the nsf faculty early career development ( career ) program and the doe outstanding junior investigator program .
thanks the d collaboration for their support and contributions to this work and , w.t .
giele , e.w.n .
glover , and d.a .
kosower for help with jetrad .
_ this work was supported in part by the national science foundation under grant phy-95 - 1249 and by the department of energy under grant de - fg02 - 91er40676 with boston university and grant de - fg02 - 91er40684 at northwestern university .
_ j. pati and a. salam , phys .
lett . * 58b * ( 1975 ) 333 ; j. preskill , nucl . phys . *
b177 * ( 1981 ) 21 ; l. hall and a. nelson , phys .
153b * ( 1985 ) 430 ; p.h .
frampton and s.l .
glashow , phys .
b190 * ( 1987 ) 157 and phys . rev .
* 58 * ( 1987 ) 2168 .
`` search for new particles decaying to dijets , @xmath68 and @xmath69 at cdf '' the cdf collaboration ( r.m .
harris for the collaboration ) , fermilab - conf-95/152-e , june 1995 . published in proceedings of the 10th topical workshop on proton - antiproton collider physics , fermilab ,
may 9 - 13 , 1995 .
hep - ex/9506008 . | flavor - universal coloron model proposed excess high-@xmath2 jets in jet spectrum fermilab tevatron collaboration model variant of model hill minimal extension description strong interactions addition one gauge interaction scalar multiplet no new fermions grafted standard one - higgs - doublet model electroweak physics simple renormalizable theory basis for dynamical electroweak symmetry breaking top quark s mass - color - assisted technicolor previous work colorons considered effects on @xmath3 parameter inclusive jet spectrum dijet spectrum angular distributions b - tagged dijets recent @xmath4 gev on coefficient four - fermion contact interaction heavy coloron exchange letter explores effects colorons on dijet mass spectrum tevatron establishes stronger limit on contact interactions colorons @xmath0 gev at 95% cl section 2 model section 3 explains limit section 4 discusses conclusions implications for quark compositeness non - standard gluon interactions dynamical electroweak symmetry breaking separate limit on color - octet axial - vector contact interactions among quarksflavor - universal coloron model strong gauge group extended to @xmath5 gauge couplings @xmath6 @xmath7 @xmath8 quark transforms ( 1,3 ) under gauge group scalar boson @xmath9 transforming @xmath10 under @xmath11 groups @xmath9 develops vacuum expectation value @xmath12 breaks strong groups diagonal subgroup @xcite unbroken subgroup with qcd color symmetry breaks original gauge bosons mix massless gluons massive colorons gluons interact quarks conventional qcd coupling @xmath13 colorons @xmath14 interact quarks new qcd - coupling @xmath15 @xmath16 color current @xmath17 mass colorons written @xmath19 parameters model below @xmath20 coloron - exchange approximated by four - fermion interaction @xmath21 - written quark compositeness defined @xmath24 interactions increase quark - scattering high invariant mass collaboration measured dijet mass spectrum at @xmath25= 1.8 tev above 200 gev jet pseudorapidity @xmath26 measured ratio spectra @xmath27 @xmath28 quark compositeness effects flavor - universal coloronscalculated leading - order dijet spectrum @xmath29 d\sigma cd jets pseudorapidities combined invariant mass momentum fractions parton distribution functions two - body parton scattering cross section @xmath36 leading qcd contributions to @xmath37 in @xcite contributions heavy coloron exchange included production gluons light quarks top quarks contribute dijet sample jet values coloron interaction strength @xmath38 determined ratio @xmath39 values @xmath33 weighted center each mass bin for 800 gev @xmath40 1400 gev yielded same ratio @xmath33 fixed at weighted center value 873.2 gev measured by d@xcite evaluated fractional difference between dijet spectra for pure qcd values @xmath38 - leading order prediction effects colorons multiplied nlo qcd prediction jetrad program by lo fractional differences results illustrated in figure 1extract limit coloron interaction strength from ddata @xcite using bayesian techniques gaussian likelihood function @xmath41^t s d - f(x @xmath42 vector data points mass bins theory points masses values @xmath44 @xmath45 covariance matrix motivated form @xmath46 prior probability flat @xmath47 ratio spectra nlo sensitive choice @xmath48 parton distribution function each choice different theory 95% confidence limit @xmath49 calculated @xmath50 limit @xmath51 transformed into limit @xmath49 @xmath38 conservative lower bound 95% cl @xmath0 gev cteq3 m pdf @xmath52 maximum jet @xmath2 event figure 2 limit incompatible with coloron interaction strength 700 gev @xcite measurements high @xmath2 jet cross - section limit new exclusion bound in @xmath53 vs. @xmath54 parameter space flavor - universal coloron model improves recent d limit dijet angular distribution @xcite region at @xmath55 limit lower bound @xmath53 excluded by cdf s search new particles decaying dijetslimit less reliable condition @xmath56 reasonable approximation to coloron exchange hold for highest - energy data point updated search for new resonances decaying to dijets complement to bounds dynamical electroweak symmetry model @xcite flavor - universal colorons produce mass top quark value of @xmath54 approximately 4 interaction strength near upper limit for higgs phase implies @xmath57 tev upper right of allowed region in figure 3 findings set limit on new strong interaction physics eqn.([conta ] ) limit equivalent to lower bound @xmath58 tev new color - octet vectorial current - current interactions interactions from quark compositeness or non - standard gluon interactions used same methods limit on color - octet axial - vector current - current interaction among quarks form @xmath59 @xmath60 contributions contact interaction to parton scattering conservative bound is @xmath61 tev at 95@xmath62 cl not interpret lower bound on mass of axigluon @xcite exchange underlies contact interaction relation @xmath63 shows mass axigluon satisfy condition @xmath65 limit bounding scale contact interactions models quark compositeness thank d collaboration measurements r.s chivukula comments manuscript support nsf faculty career program doe junior investigator program thanks d collaboration w.t. giele glover d kosower jetrad work supported national science foundation grant phy-95 - 1249 department of energy grant - fg02 - 91er40676 boston university 91er40684 northwestern university j. pati a. salam phys . 58b 1975 ) 333 preskill phys b177 1981 ) 21 l. hall a. nelson phys 153b 1985 ) 430 frampton s glashow phys b190 1987 ) 157 58 1987 ) 2168 search new particles decaying dijets @xmath68 @xmath69 cdf collaboration harris fermilab - conf-95/152-e june 1995 proceedings 10th workshop proton - antiproton collider physics fermilab may 9 - 13 1995 hep - ex/9506008 | 0.358138 | 2,562 | 1,323 | , Saving $0.1 in GPT-4. |
67,697 | path - integral - monte - carlo simulation has been used to calculate the properties of a two - dimensional ( 2d ) interacting bose system .
the bosons interact with hard - core potentials and are confined to a harmonic trap .
results for the density profiles , the condensate fraction , and the superfluid density are presented . by comparing with the ideal gas
we easily observe the effects of finite size and the depletion of the condensate because of interactions .
the system is known to have no phase transition to a bose - einstein condensation in 2d , but the finite system shows that a significant fraction of the particles are in the lowest state at low temperatures .
pacs numbers:03.75.fi,02.70.lq,05.30.jp,05.70.fh,32.80.pj,67.40.db | recent experiments on alkali atoms cooled by laser methods and evaporation in a magnetic trap have allowed the observation of bose - einstein condensation ( bec ) in a harmonic potential formed by an external magnetic field . by varying the trapping field so that it is very narrow in one dimension , it is possible to separate the single - particle states in the oscillator potential into well - defined bands . by occupying only states in the lowest band
one has an effective two - dimensional system .
in contrast to such a quasi-2d system , the system considered here is genuinely two dimensional . in this paper
we will investigate the behaviour of harmonic bose systems in 2d using the very powerful finite - temperature path - integral monte carlo ( pimc ) simulation technique@xcite .
the pimc technique is in principle capable of describing systems of arbitrary interaction strength and density and allows one to study the static properties of the condensed gases .
the only fundamental uncertainty arises from the choice of the interaction potential . here a hard - core potential appropriate for the s - wave scattering length of @xmath0rb was chosen . in this case
the hard - core parameter is @xmath1 in the dimensionless units of ref .
the analogous three dimensional case has been studied previously@xcite .
the hamiltonian for the system under consideration is @xmath2 the density matrix for @xmath3 bose particles at inverse temperature @xmath4 can be written as a convolution with @xmath5 intermediate density matrices or `` time slices '' at inverse temperature @xmath6 .
then the probability density for finding a many - particle configuration @xmath7 is @xmath8 where @xmath9 denotes a vector with permuted particle labels . the metropolis algorithm is used to sample from this distribution . with larger m or correspondingly larger temperatures for each time slice , the intermediate density matrices approach the classical limit which leads to the _ primitive approximation_. in order to make the computation for many particles feasible , @xmath5 can be reduced by several orders of magnitude by calculating the density matrix @xmath10 for the interaction involving just two particles and approximating each time slice by@xcite @xmath11 as long as only two particles interact , this is equivalent to using the primitive approximation . to calculate @xmath10 , we note that in the harmonic potential the hamiltonian for two particles decouples into a center of mass and a relative motion term with the latter given by @xmath12 where @xmath13 is the reduced mass . with the trotter breakup , the quotient in ( [ eq : rho - prod ] ) , which plays the role of a correction term , can be transformed to @xmath14 with @xmath15 and the relative coordinates @xmath16 and @xmath17 .
@xmath18 is the density matrix for the hamiltonian @xmath19 ( eq .
( [ eq : rel - hamilt ] ) ) describing the interaction between two interacting particles without confining potential and @xmath20 is the density matrix for a single particle of reduced mass @xmath21 in the harmonic potential@xcite . calculating @xmath18 for a hard - core potential using an eigenfunction expansion yields @xmath22 with the radial wave functions @xmath23 where @xmath24 and @xmath25 are the bessel and neumann functions and @xmath26 is the angle between @xmath27 and @xmath28 .
since the numerical evaluation of this function is quite time consuming , values distributed over a mesh with parameters @xmath29 and @xmath26 are computed for each value of @xmath30 .
a simple linear interpolation method proved to be sufficient to reach the required degree of accuracy for evaluation in the simulation .
the full correction factor ( eq . ( [ eq : corr ] ) ) with dependence on the confining potential can then be calculated very efficiently during the monte - carlo simulation . to further speed up the calculation ,
a boxing algorithm@xcite is used that divides the space of the system into boxes with a size of at least the `` healing length '' of the pair interaction .
when we compute the effects of the interaction of a given particle , it is necessary to consider only a small number of interactions with particles in boxes neighbouring that of the particle in question .
furthermore , attempting only permutations with particles from the same box , increases the acceptance probability for these moves and results in a more effective sampling of permutation space.@xcite the value of @xmath30 has to be chosen carefully for each particle density to accommodate both the trotter breakup and the approximation of pair interactions ( eq .
( [ eq : rho - prod ] ) ) . for the density used ,
tests with values for @xmath30 ranging over orders of magnitude have been performed , showing that a further decrease below @xmath31 yield the same results within statistical errors .
in 2d , with interactions , there is no phase transition to a bose condensed state@xcite even in a trap , nevertheless there should be a macroscopic fraction of particles in the ground state at finite temperature when the particle number is finite .
we tentatively assume that , in 2d , the condensation fraction can be determined , as in 3d@xcite , by observing a macroscopic number of particles with long exchange cycles in the lowest state of their subsystem .
for each temperature a characteristic length of permutation cycles @xmath32 can be chosen so that the density profiles for particles on cycles longer than @xmath32 are essentially the same . particles on shorter cycles will have a broader density profile .
particles on cycles longer than @xmath32 are identified with the condensate ( fig . 1 ) .
what is plotted is the average distribution in one coordinate after integrating over the other coordinate .
we have fit the lowest temperature curve in fig . 1 with the infinite - n solution of the zero - temperature gp equation@xcite .
the interaction coefficient has been determined by the fit since , in 2d , there is no straightforward connection between a hard - core interaction and a contact pseudopotential .
we observe no obvious bimodality , due to separate distributions of condensed and non - condensed particles in the overall density in fig . 1 ;
if a kink did exist in the density it would likely be washed out by the integration over one coordinate . in two dimensions interactions
seem to lead to a comparatively stronger depletion of the condensate ( fig .
2 ) than in three dimensions ( cf . , ref . 3 ) .
this is expected , since there is no condensate at finite temperature in the thermodynamic limit for the interacting system and therefore the critical temperature has to approach zero ( for large numbers of particles ) when interactions are turned on.@xcite
recently one of the authors @xcite showed that the hartree - fock - bogoliubov equations indicated that there is a phase transition in 2d , although it can not be to the bec state .
perhaps this transition is of the kosterlitz - thouless ( kt ) type @xcite and involves superfluidity .
the superfluid fraction can be related to the mean square surface area enclosed by feynman paths .
sindzingre _ et .
_ have shown that@xcite @xmath33 with @xmath34 the area swept out by the paths and @xmath35 the classical moment of inertia .
the average is taken over configurations in the simulation .
the resulting superfluid densities are very small ( fig . 2 )
. a smaller @xmath36 than in the translationally invariant system is expected because the formation of the superfluid begins in the middle of the potential well where the contribution to the moment of inertia is small .
paths with different orientations will contribute with different signs to the area making cancellation possible .
the simulation results for the 2d system without confining potential@xcite are in agreement with the kt theory for the superfluid transition .
if this description is also appropriate with confining potential , the vortex picture of the kt - transition suggests an additional mechanism leading to a decrease of @xmath36@xcite : superfluidity is destroyed by dissipation through vortices which are not paired . at low temperatures vortices form pairs which unbind at higher temperatures . in the non - uniform system
the two vortices forming a pair will in general experience a slightly different potential leading to imperfect pairing and thereby a lowering of the superfluid fraction .
= 5 in fig . 1 .
density profiles for 1000 particles at temperatures @xmath37 for @xmath38 .
the condensate parts are displayed on the left and the profiles for all particles on the right .
all profiles are normalised to unity .
temperatures are given with reference to the critical temperature @xmath39 of the ideal system in the thermodynamic limit .
the smooth dotted curve through the 0.21 data is the infinite - n solution of the gp equation with interaction strength determined by the fit .
= 4.75 in fig . 2 . left figure : condensate fraction for the interacting system with n=1000 and @xmath38 and comparison with theoretical prediction for the ideal gas in the thermodynamic limit ( tdl ) and with finite size corrections for 1000 particles .
right figure : dependence of the superfluid fraction on temperature for 1000 particles .
we thank werner krauth for providing his pimc program for the three dimensional case@xcite . | experiments on alkali atoms cooled laser in magnetic trap bose - einstein condensation in harmonic potential external magnetic field varying trapping field single - particle states oscillator potential into bands occupying states lowest band two - dimensional system system two dimensional behaviour harmonic bose systems in 2d using finite - temperature path - integral carlo simulation technique arbitrary interaction strength density static properties condensed gases uncertainty from choice interaction potential hard - core potential for s - wave scattering length @xmath0rb chosen hard - core parameter is @xmath1 dimensionless units analogous three dimensional case studied previously hamiltonian for system is @xmath2 density matrix for @xmath3 bose particles at inverse temperature @xmath4 convolution with @xmath5 intermediate density matrices slices at inverse temperature @xmath6 probability density for many - particle configuration @xmath7 is @xmath8 @xmath9 vector with permuted particle labels metropolis algorithm distribution larger temperatures time slice intermediate density matrices approach classical limit to primitive approximationcomputation particles @xmath5 reduced calculating density matrix @xmath10 for interaction two particles approximating by @xmath11 primitive approximation @xmath10 harmonic potential hamiltonian for two particles decouples into center mass relative motion term @xmath12 @xmath13 reduced mass trotter breakup quotient rho - prod correction term transformed to @xmath14 with @xmath15 relative coordinates @xmath16 @xmath17 @xmath18 density matrix for hamiltonian @xmath19 - hamilt interaction two particles without confining potential @xmath20 density matrix for single particle reduced mass @xmath21 harmonic potential@xcite calculating @xmath18 hard - core potential eigenfunction expansion yields @xmath22 with radial wave functions @xmath23 @xmath24 @xmath25 bessel and neumann functions @xmath26 angle between @xmath27 @xmath28 numerical evaluation time consuming values distributed over mesh parameters @xmath29 @xmath26 computed for each value @xmath30 simple linear interpolation method accuracy simulation full correction factor corr confining potential calculated during monte - carlo simulationspeed calculation boxing algorithm used divides space system into boxes size healing length of pair interaction effects interaction particle consider small number interactions with particles in boxes neighbouring attempting permutations with particles from same box increases acceptance probability effective sampling permutation space value of @xmath30 chosen carefully for each particle density accommodate trotter breakup approximation pair interactions tests with values for @xmath30 over orders of magnitude performed decrease below @xmath31 same results statistical errors 2d no phase transition to condensed state macroscopic fraction of particles in ground state at finite temperature when particle number finite condensation fraction determined by observing macroscopic number of particles with long exchange cycles in lowest state subsystem for each temperature characteristic length of permutation cycles @xmath32 chosen density profiles for particles on cycles longer than same shorter cycles broader density profile particles cycles longer than identified with condensate ( fig . 1 ) plotted average distribution in one coordinate after integrating other lowest temperature curve in fig .infinite - n solution zero - temperature gp equation interaction coefficient determined by fit no connection between hard - core interaction contact pseudopotential no bimodality separate distributions condensed non - condensed particles density fig . 1 kink likely washed out by integration over one coordinate two dimensions interactions lead stronger depletion condensate 2 than three dimensions expected no condensate at finite temperature thermodynamic limit system critical temperature zero large numbers when interactions turned on hartree - fock - bogoliubov equations phase transition in 2d bec state transition kosterlitz - thouless ( type involves superfluidity superfluid fraction related to mean square surface area enclosed by feynman paths sindzingre shown @xmath33 @xmath34 area swept out by paths classical moment of inertia average taken over configurations simulation superfluid densities small ( fig . 2 ) smaller @xmath36 invariant system expected formation superfluid begins middle potential well contribution moment of inertia small paths different orientations contribute different signs area cancellation possiblesimulation results 2d system without confining potential with kt theory superfluid transition confining potential vortex picture kt - transition suggests mechanism decrease superfluidity destroyed by dissipation through vortices not paired low temperatures vortices form pairs unbind higher non - uniform system two vortices pair different potential imperfect pairing lowering superfluid fraction fig 1 density profiles 1000 particles at temperatures condensate parts left profiles all particles right profiles normalised to unity temperatures critical temperature @xmath39 ideal system smooth dotted curve 0.21 data infinite - n solution gp equation interaction strength determined fig 2 condensate fraction interacting system with n=1000 @xmath38 comparison with theoretical prediction ideal gas limit finite size corrections 1000 particles dependence superfluid fraction on temperature 1000 particles werner krauth pimc program three dimensional case@xcite | 0.330063 | 2,005 | 992 | , Saving $0.1 in GPT-4. |
67,706 | the cluster synchronization ( cs ) is a very important characteristic for the higher harmonic coupling kuramoto system .
a novel transformation is provided , and it gives cs by the periodic properties of the density function .
the periodic properties of the density function also make the cluster sections boundaries barrier - like , which helps to explain the sensitiveness of cs on the initial conditions of the oscillators .
detailed numerical studies confirm the theoretical predictions from this new view of the symmetry transformation .
the work is very beneficial to the further study on cs in various systems . | the symmetries play important role in varies branches of theoretical physics , both in classical and modern areas .
for example , in the classical mechanics , the kepler problem is easily solved if one unified the conservation of energy , angular momentum which are the results of the symmetry of the kepler problem . even without solving the problem ,
the conservation of the angular momentum will tell one many information , like the motion of the planet in the sun system being of plane , etc . in the paper
, we will exploit the symmetry method to study the generalized kuramoto system and give the answer to the question of cluster synchrony state without solving the problem directly .
the kuramoto model ( km ) captures the main property of the collective synchronization with the first harmonic coupling as @xmath0 and revealed the second continuous transition at the critical coupling strength @xmath1 .
km is applied in many physical , biological and social systems , including electrochemical oscillators , josephson junction arrays , cardiac pacemaker cells , circadian rhythms in mammals , network structure and neural network@xcite-@xcite .
km have been generalized in many aspects@xcite-@xcite .
one of them is the introduction of the globally higher harmonic coupling @xmath2 , where many new and interesting phenomena appear , like the cluster synchronization ( cs ) , and switching of the oscillators between different clusters with the external force , etc.@xcite-@xcite .
higher harmonic coupling ( hhc ) is dominating in @xmath3-josephson junction @xcite , in the electrochemical oscillators in higher voltage@xcite , in neuronal networks with learning and network adaption@xcite-@xcite .
cs is the most outstanding feature of this higher harmonic coupling kuramoto model(hhc - km ) . here
we will investigate hhc - km from the point of symmetry , and provide a group transformation , and give cs a thoroughly novel interpretation , and answer the question on the same threshold for cs in different parameters @xmath4 .
the generalized kuramoto model with the higher harmonic coupling is @xmath5 in the case of small strength @xmath6 , the term @xmath7 dominates the change of the phase @xmath8 and the whole phase system is in the incoherent state . whenever @xmath9 exceeds @xmath1 , the second terms in eq.([kuramoto mth ] )
predominate and cs emerges@xcite-@xcite .
it has been also known that cs is sensitive to the initial conditions of the oscillators in ref.@xcite .
we study cs from completely new view .
we will try to find the relation between the generalized and standard kuramoto models , and penetrate the phenomena of cs to study their essence . in the standard kuramoto model
@xmath10 the coupling strength @xmath11 is assumed . by introduction of the transformation @xmath12 together with @xmath13 chang into @xmath14 , eq.([kuramoto mth ] ) takes the form @xmath15 which is the same as that of the standard kuramoto model case in the fast study model . ] .
the transformation ( [ trans ] ) is crucial to obtain the information on eq.([kuramoto mth ] ) and give the explanation to cs . for eq.([kuramoto 1-nth ] ) , the density function @xmath16 in the large @xmath17 limit satisfies the continuous equations @xmath18=0,\label{f for trans}\\ & & r=\int \int f(\phi,\omega , t)e^{i\phi}d\phi d\omega.\label{f for order}\end{aligned}\ ] ] generally , the dynamical information for cs is obtained through solve eqs.([f for trans])-([f for order ] ) .
nevertheless , the transformation ( [ trans ] ) make it possible to alternatively investigate cs without assorting to the direct solutions to eqs.([f for trans])-([f for order ] ) .
see details in the following .
suppose initially uniform distribution in @xmath19 for the phases @xmath20 , the transformation indicates the corresponding initial phases uniform distribution is in @xmath21 for the phases @xmath22 . because eq.([f for trans ] ) is periodic in @xmath3 , with the initial periodic condition in @xmath3 ,
the solution @xmath16 is also periodic in @xmath3 .
so one has @xmath23 which results in the following outstanding properties for the corresponding density function @xmath24 @xmath25 hence the cluster phenomenons appear , and the phases @xmath20 cluster into @xmath4 sections . from eq.([symmetry in theta ] ) , it is easy to see that the order parameter is zero no matter the phases @xmath20 are in cs state or not , that is , @xmath26 so , the order parameter no longer works in the generalized one , as is shown in fig.[fig1 ] for the cases of @xmath27 and is substituted by the generalized order parameter @xmath28 defined as @xmath29 eq.([symmetry in theta ] ) guarantees the generalized order parameters @xmath28 being the same for all parameters @xmath4 , which also could be obtained from the fact of the same density function @xmath16 for different @xmath4 in calculation @xmath30 by eq.([f for order ] ) .
we also numerically calculate @xmath28 against @xmath9 forwardly for different @xmath31 with the same initial random distributions in @xmath19 and the numerical results confirm the conclusion .
see the second panel in fig.[fig1 ] for detail . [
cols="^,^ " , ]
cs has been investigated by the method of self - consistent approach in refs.@xcite , @xcite-@xcite .
neural network actually studied the combination of the first and second harmonic couplings in the generalized kuramoto model@xcite-@xcite , which is also treated in ref.@xcite , @xcite . in the @xmath17
identical oscillatorscase , the symmetry viewpoint is applied and cs of the two groups of @xmath4 and @xmath32 oscillators is connected with their symmetry groups of the dynamics @xmath33 @xcite , @xcite .
the symmetry group @xmath34 is only suited for the identical oscillators in the kuramoto model .
however , it is still very difficult to obtain clear analytical results by the self - consistent approach and detailed understanding of cs @xcite . in the nice work @xcite ,
cs have been investigated by the self - consistent approach .
the density function for the second harmonic coupling case is decomposed into the symmetric and asymmetric parts in ref.@xcite , and the ott - antonsen ( oa ) mechanism is utilized to analyze the symmetric case . however , the asymmetric one is not accessible to the analytical study , and numerical methods are needed to the full solution of the density function @xcite . for higher harmonic coupling
than the second , the density function @xmath35 is decomposed in to @xmath4 parts as @xmath36 and @xmath37 .
oa mechanism could be utilized for @xmath38 and the critical strength @xmath39 is obtained for the the lorentz s distribution of the natural frequency @xmath40 .
however , it is not easy to obtain other @xmath41 and numerical methods are used for @xmath24@xcite . however , our study is completely different from that in ref.@xcite .
we mainly rely on the transformation ( [ trans ] ) and the periodic properties of the density function to study the most typical phenomena cs in the generalized kuramoto model . by the transformation ( [ trans ] )
, it is possible to relate cs with the periodic properties of the density function @xmath42 or @xmath35 . to hold the periodic properties for @xmath35
, the initial distribution of the oscillators in terms of @xmath43 must range randomly in @xmath19 .
because of the periodic properties of @xmath42 or @xmath35 and the relation between @xmath44 and @xmath3 , the @xmath4 cluster synchrony states appear corresponding to @xmath44 , which are in one of the sections @xmath45 for @xmath46 . corresponding the cluster sections ,
there naturally exist boundaries for them , which function as the potential barriers to forbid the synchrony oscillators to pass through .
the existence of the barrier - like boundary can also explain the sensitiveness of cs to the initial conditions .
the initial distribution of the phases @xmath44 in @xmath47 with @xmath48 will break the periodic condition for @xmath42 or @xmath35 and the violation will result in the sensitiveness of cs to the initial distribution ( 0,a ) .
the explanation to cs in the letter is novel and simple , and has both the profound mathematical insight and clear physical understanding .
our detailed numerical studies confirm the symmetric analysis .
the work was partly supported by the national natural science of china ( no .
10875018 ) and the major state basic research development program of china ( 973 program : no.2010cb923202 ) .
99 kuramoto y. , chemical oscillations , waves , and turbulence ( springer , berlin,1984 ) .
acebron j. a. , bonilla l. l. , perez vicente c. j. , ritort f. , and spigler r. , rev .
77 2005 137 .
strogatz s. h. , physica d 143 2000 1 .
kuramoto y. and nishikawa i. , j. stat .
49 1987 569 .
gupta s. , campa a. , and ruffo s. , j. stat .
mech . : theory exp .
r08001 2014 1 .
tanaka h. , lichtenberg a. j. and oishi s. , phys .
78 1997 ( 2104 .
acebron j. a. and spigler r. , phys
81 19982229 .
acebron j. a. , bonilla l. l. and spigler r. , phys .
e 62 2000 3437 .
wang h. and li x. , phys . rev.e .
83 2011 066214 .
zhang x. , hu x. , kurths j. , and zonghua liu .
, phys . rev.e .
88 2013 010802(r ) .
pinto rafael s. and saa a. , phys.rev .
e 91 2015022818 .
zou y. , pereira t. , small m. , liu z. , and kurths j. , phys .
112 2014 114102 .
gomez - gardenes j. , gomez s. , arenas a. , and moreno y. , phys .
106 2011 128701 .
abrams d. m. andstrogatz s. h. , phys . rev .
lett.,93 2004 174102 .
fu c. , deng z. , huang l. , and wang x. , phys .
e 87 2013 032909 .
wu y. , xiao j. , g. hu and m. zhan , europhys .
97 2012 40005 .
zhu y. , zheng z. and yang j. , phys .
reve.89 2014 022914 .
ju p. , dai q. , cheng h. , and yang j. , phys.rev.e 90 2014 019903 .
skardal p. s. , ott e. and restrepo j. g. , phys .
e 84 2011 036208 .
komarov m. and pikovsky a. , phys .
prl 111 2013 204101 .
goldobin e. , koelle d. , kleiner r. and r. g. mints .
107 2011 227001 .
goldobin e. , kleiner r. , koelle d. and r.g . mints .
111 2013 057004 .
kiss i. z. , zhai y. and hudson j. l. , phys .
94 2005 248301 .
kiss i. z. , zhai y. and hudson j. l.,prog .
161 2006 99 .
seliger p. , young s. c. and tsimring l. s. , phys .
e 65 2002 041906 .
hansel d. , mato g. and meunier c. , europhys .
23 1993 367 .
hansel d. , mato g. and meunier c. , phys .
e 48 1993 3470 .
ashwin p. and borresen j. , phys .
e 70 2004026203 .
banaji m. , phys .
e 71 2005 016212 .
niyogi r. k. and english l. q. , phys .
e 80 2009 066213 .
okuda k. , physica d 63 1993 424 .
golomb d. , hansel d. , shraiman b. and sompolinskyh .
rev . a 45 1992 3516 .
komarov m. and pikovsky a. , arxiv:1404.7292 2014 .
ott e. and antonsen t. m. , chaos 18 2008 037113 .
rosenblum m. and pikovsky a. , phys .
98 2007 064101 .
pikovsky a. and rosenblum m. , physica d : nonlinear phenomena 238 2009 27 .
baibolatov y. , rosenblum m. , zhanabaev z. z. , kyzgarina m. and pikovsky a. , phys . rev .
e 80 2009 046211 .
g. tian , s. hu , s. zhong , arxiv:1505.03660 , ( 2015 ) | symmetries in theoretical physics classical modern in classical mechanics kepler problem solved if conservation of energy angular momentum results of symmetry conservation angular momentum information like motion planet in sun system paper symmetry method to study generalized kuramoto system answer cluster synchrony state without solving problem kuramoto model ( km captures collective synchronization with first harmonic coupling @xmath0 second continuous transition at critical coupling strength @xmath1 km applied in physical biological social systems electrochemical oscillators josephson junction arrays cardiac pacemaker cells circadian rhythms network structure neural network km generalized in aspects introduction of globally higher harmonic coupling @xmath2 new phenomena like cluster synchronization switching oscillators between clusters with external force higher harmonic coupling in @xmath3-josephson junction electrochemical oscillators neuronal networks with learning adaption cs outstanding feature of higher harmonic coupling kuramoto model(hhc - km )investigate hhc - km symmetry provide group transformation give cs novel interpretation answer question same threshold cs different parameters @xmath4 generalized kuramoto model higher harmonic coupling @xmath5 small strength @xmath6 term @xmath7 dominates change phase @xmath8 phase system incoherent @xmath9 exceeds @xmath1 second terms eq.[kuramoto mth predominate cs emerges@xcite cs sensitive initial conditions oscillators.@xcite study cs new view find relation generalized standard kuramoto models penetrate phenomena cs study essence standard kuramoto model @xmath10 coupling strength @xmath11 assumed transformation @xmath12 @xmath13 @xmath14 eq.([kuramoto mth takes form @xmath15 same standard kuramoto model fast study model transformation ( trans crucial obtain information eq.([kuramoto mth explanation cs density function @xmath16 large @xmath17 limit satisfies continuous equations @xmath18=0 for trans=\phi dynamical information cs obtained through solve eqs.([f for trans)-([f for ordertransformation ( trans ) investigate cs without assorting direct solutions eqs.([f for trans)-([f order ] ) uniform distribution in @xmath19 phases @xmath20 transformation indicates @xmath21 phases @xmath22 eq.([f for trans ] ) periodic in @xmath3 solution @xmath16 periodic @xmath3 @xmath23 results outstanding properties for density function @xmath24 @xmath25 cluster phenomenons appear phases @xmath20 cluster into @xmath4 sections order parameter zero no phases @xmath20 in cs state order parameter no longer works generalized one substituted by order parameter @xmath28 @xmath29 guarantees generalized order parameters @xmath28 same for all parameters @xmath4 same density function @xmath16 for different @xmath4 calculation @xmath30 by eq.([f for order ] ) numerically calculate @xmath28 against @xmath9 for different @xmath31 same initial random distributions in @xmath19 numerical results confirm conclusion second panel fig.[fig1 detail cs investigated self - consistent approach in refs.@xcite-neural network studied first second harmonic couplings in kuramoto model@xcite treated in ref.@xcite @xmath17 identical oscillatorscase symmetry viewpoint applied cs @xmath4 @xmath32 oscillators connected with symmetry groups @xmath33 @xcite symmetry group @xmath34 suited for identical oscillators difficult to obtain clear analytical results by self - consistent approach understanding cs @xcite cs investigated by self - consistent approach density function second harmonic coupling decomposed into symmetric asymmetric parts ott - antonsen ) mechanism analyze symmetric case asymmetric not accessible numerical methods needed solution density function higher harmonic coupling density function @xmath35 decomposed @xmath4 parts oa mechanism for @xmath38 critical strength @xmath39 for lorentz s distribution natural frequency @xmath40 not easy to obtain other @xmath41 numerical methods for @xmath24@xcite our study different.@xcite rely on transformation ( trans ] ) periodic properties density function study typical phenomena cs relate cs with periodic properties density function @xmath42 or @xmath35periodic properties @xmath35 initial distribution oscillators @xmath43 range randomly in @xmath19 periodic properties @xmath42 @xmath35 relation @xmath44 @xmath3 @xmath4 cluster synchrony states @xmath44 sections @xmath45 @xmath46 boundaries potential barriers synchrony oscillators barrier boundary sensitiveness cs to initial conditions initial distribution phases @xmath44 @xmath47 @xmath48 periodic condition @xmath42 or @xmath35 violation sensitiveness cs initial distribution explanation cs novel simple profound mathematical insight physical understanding numerical studies confirm analysis work supported by national natural science china 10875018 major state basic research development program china kuramoto y. chemical oscillations waves turbulence springer berlin,1984 acebron j bonilla l perez vicente ritort spigler r. rev . 77 2005 strogatz s. physica d 143 2000 kuramoto y. nishikawa i.. stat . 49 1987 569 gupta s campa ruffo s. stat . mech theory r08001 2014 tanaka h. lichtenberg a oishi s. phys .1997 2104 acebron j spigler 19982229 bonilla spigler 2000 3437 wang h li x 83 2011 066214 zhang hu x kurths zonghua liu 88 2013 010802 pinto saa a 91 2015022818 zou y pereira small m liu kurths j 112 2014 114102 gomez gardenes j gomez s arenas moreno y 106 2011 128701 abrams d. s phys 2004 174102 fu c deng z huang wang x 87 2013 032909 wu y xiao hu zhan 97 2012 40005 zhu y zheng z yang j 2014 022914 ju p. dai q cheng h yang j 90 2014 019903 skardal p. ott e restrepo j phys 84 2011 036208 komarov m pikovsky a. 111 2013 204101 goldobin e koelle d. kleiner r r g mints 107 2011 227001goldobin kleiner koelle 2013 057004 kiss zhai hudson j phys 94 2005 248301 kiss zhai hudson 161 2006 seliger p young s tsimring l 65 2002 041906 hansel d mato g meunier c europhys 23 1993 367 hansel meunier 1993 3470 ashwin p borresen j. 2004026203 banaji m 2005 016212 niyogi r english l. 80 2009 066213 okuda k 63 1993 424 golomb d hansel d shraiman b sompolinskyh 1992 3516 komarov m pikovsky a. arxiv:1404.7292 ott e antonsen t 18 2008 037113 rosenblum m pikovsky a phys 98 2007 064101 physica nonlinear phenomena 238 2009 baibolatov y rosenblum zhanabaev z kyzgarina m pikovsky a phys 80 2009 046211 tian hu zhong arxiv:1505.03660 | 0.44551 | 3,101 | 1,572 | , Saving $0.1 in GPT-4. |
67,727 | the existence of a new ultraviolet scale @xmath0 for effective theories with gravity and @xmath1 gauge fields has recently been conjectured as a possible criterion for distinguishing parts of the swampland from the string landscape . here
we discuss a possible phenomenological signature of this scale , for electromagnetic fields , in astrophysical observations .
ba - th/06 - 555 + cern - ph - th/2006 - 031 + hep - th/0611227 _ dipartimento di fisica , universit di bari , + via g. amendola 173 , 70126 bari , italy _ , _ istituto nazionale di fisica nucleare , sezione di bari , bari , italy _ and _ cern , theory unit , physics department , + ch-1211 geneva 23 , switzerland _
+ + _ ( gravity research foundation , wellesley hills , ma , 02481 - 0004 , usa ) + and awarded with honorable mention " _ to appear in * int . j. mod .
phys .
d * ( december 2006 , special issue ) cern - ph - th/2006 - 031 + february 2006 0.2 cm the presence of a new ultraviolet scale @xmath2 for effective theories with gravity and @xmath1 gauge fields , with coupling @xmath3 , has recently been conjectured in the first hep - th paper of the year @xcite .
theoretical evidence for this conjecture has been presented , mainly based on the properties of known ( heterotic ) string theory backgrounds , and it has been argued that this conjecture could be related to a new criterion for distinguishing parts of the swampland " @xcite from the string landscape " . according to @xcite , the conjectured scale @xmath2 is hidden , and completely unexpected , in the context of the effective field theory ; here we will suggest that , rather surprisingly , there could be a hint at such a scale in present astrophysical observations . we are referring , in particular , to the electro - gravity " scale @xmath4 .
the physical relevance of such a scale might emerge from observed regularities concerning the mass , the angular momentum , and the magnetic moment of various classes of astrophysical structures .
such regularities were noticed long ago but , up to now , have not been unambiguously explained in the context of standard ( effective ) models of gravity and electromagnetic interactions .
here we discuss the possibility that they be related to the general weakness of gravity with respect to gauge forces , according to the conjecture presented in @xcite .
we start by recalling an empirical relation which seems to connect the mass @xmath5 and the angular momentum @xmath6 of many classes of astronomical objects , ranging from planets to superclusters .
the relation was originally proposed in the form @xcite @xmath7 , and later extended to a generic slope @xmath8 as follows @xcite : j= p m^ , [ 1 ] where @xmath9 is some universal " phenomenological parameter to be determined by observations .
detailed compilations of data have shown the presence of a considerable scattering of points in the @xmath10 plane within the various astrophysical classes , but the overall fit of the plots of @xmath11 versus @xmath12 , for all classes of objects , has confirmed that eq .
( [ 1 ] ) may be representative of the general trend of the data . in that case , the slope @xmath8 has been determined to be @xmath13 according to a first compilation @xcite , and @xmath14 in a subsequent , larger compilation of data running over about @xmath15 powers of ten in the mass @xcite ( see fig . 1 , left panel ) .
both results are compatible with a power @xmath16 .
the experimental value of the ( dimensional ) coefficient @xmath9 is strongly correlated to the slope @xmath8 , and is thus affected by a large uncertainty .
if we fix @xmath16 , however , the corresponding mean value of @xmath9 is very similar in the two cases , as one obtains ( in cgs units ) @xmath17 according to @xcite , and @xmath18 from the data of @xcite .
it must be remarked , at this point , that the case with slope @xmath16 is qualitatively different from the generic case @xmath19 , because for @xmath16 the dimensions of @xmath9 are such that we can express @xmath9 in terms of only two ( dimensional ) fundamental parameters , the newton constant @xmath20 and the ligth velocity @xmath21 ( indeed , @xmath22=[g_n c^{-1}]$ ] ) . if , instead , @xmath19 , then the definition of @xmath9 requires , besides @xmath20 and @xmath21 , another parameter with mass dimensions , since @xmath22=[g_nm^{2-\da } c^{-1}]$ ] , and the possibility of a universal parametrization is lost .
we are thus lead to the conjecture that the small deviations of @xmath8 from @xmath23 may represent systematic " perturbations ( due to local - physics effects ) of some underlying , scale - free " fundamental relation , that can be empirically parametrized by setting @xmath16 and p^-1= g_n^-1c , = 2 , [ 2 ] where @xmath24 is a dimensionless number . with
this assumption one obtains from the data a mean value of @xmath9 ranging from @xmath25 , according to @xmath26 , to @xmath27 , from the results of @xcite .
a second empirical relation concerns the approximated proportionality between the angular momentum and the magnetic moment @xmath28 of astronomical bodies . by setting j= q m^[4 ]
one then finds that for @xmath29 the numerical value of the ( dimensional ) parameter @xmath30 is of the same order of magnitude for many stars and planets .
the recurrence of this direct proportionality coefficient ( sometimes referred as the magnetic bode s law " @xcite ) was pointed out long ago by blackett @xcite for the earth , the sun and the star @xmath31 virginis , and was later extended to other planets , white dwarfs , and pulsars in @xcite .
after blackett , new and direct tests of this relation have been proposed also for macroscopic spinning bodies in the laboratory @xcite , and the required sensitivity for these experiments seems now to be within the reach of modern technology .
it is unclear , at present , whether such an empirical correlation may retain universal validity when extrapolated to include other ( and larger ) classes of astronomical objects ( the analyses are also complicated , in general , by the possible variability in time of the magnetic fields , by the presence of toroidal and polar magnetic components , ) .
a recent data analysis @xcite , spanning a range of about @xmath32 orders of magnitude in both @xmath6 and @xmath28 , has indeed confirmed the trend suggested in @xcite for the planetary sector , and extended it to a large compilation of stellar objects , in the sense that the plots of @xmath33 versus @xmath11 , for the centers of gravity " of the data of the various sub - samples , have been fitted by an overall slope @xmath34 , which is marginally compatible with @xmath29 ( see fig .
[ f1 ] , right panel ) .
however , it has been shown that there are significant deviations from this mean result inside the sub - samples of data corresponding to different astrophysical classes ( deviations from a linear @xmath35 correlation are most evident for hot stars , isolated white dwarfs , and isolated pulsars @xcite ) .
and @xmath36 spanned within the various sub - sambles ( adapted from ref .
@xcite).__,title="fig:",width=245 ] and @xmath36 spanned within the various sub - sambles ( adapted from ref .
@xcite).__,title="fig:",width=298 ] the statistical significance of the proposed correlation fitting the complete set of data , averaged over all sub - samples , is likely to remain controversial for a long time , as also noted in @xcite . leaving this problem to a forthcoming , more detailed analysis @xcite ,
in this paper we will take seriously the results of @xcite , and we shall concentrate the following discussion on the case @xmath29 which is ( at least marginally ) compatible with the data , and which is dimensionally preferred , as before , in the sense of allowing a scale - free parametrization of @xmath30 in terms of @xmath21 and @xmath20 only .
when @xmath37 the definition of @xmath30 requires indeed an extra length parameter , as @xmath38= [ c^{3 - 2\ep } l^{2 - 2\ep } g_n^{-1+\ep/2}]$ ] ; with @xmath29 we can instead adopt a universal parametization , and set q^-2= g_n c^-2 , = 1 , [ 5 ] where @xmath39 is a dimensionless coefficient . using the data reported in @xcite , and the mean value of @xmath30 associated to the slope @xmath29 , one then finds @xmath40 .
the result is very close to the value of @xmath39 one would obtain from the old analyses of @xcite , mainly restricted to the planetary sector .
now we come to the new and central point of this paper , namely to the observation that the experimental values of the two dimensionless coefficients @xmath24 and @xmath39 turn out to be not only very close to each other , within observational uncertainties , but also very close to the numerical value of the fine - structure constant , @xmath41 . barring an accidental numerical coincidence ,
we are thus led to our central assumption , namely to the speculation that the two coefficients @xmath24 and @xmath39 actually represent _ the same _ physical parameter ( modulo numerical factors of order @xmath42 ) , and that this parameter actually corresponds to the fine - structure constant : _
em = e^2/c 1/137 7.3 10 ^ -3 .
[ 7 ] this simple assumption , suggested by the astrophysical data , leads immediately to the disclosure of a direct connection between the new cut - off scales conjectured in @xcite and the two empirical laws ( [ 1 ] ) , ( [ 4 ] ) . according to eqs .
( [ 2 ] ) , ( [ 7 ] ) , in fact , the parameter @xmath9 determines an energy scale @xmath43 given ( in natural units @xmath44 ) by p^-1/2 ~= = e , [ 8 ] which is exactly the new uv scale suggested in @xcite . according to eqs .
( [ 5 ] ) , ( [ 7 ] ) , the parameter @xmath30 determines instead the dual " magnetic scale q ~= = e= g_mag , [ 9 ] also suggested in @xcite .
it seems remarkable that these two related scales emerge _ separately _ and quite _ independently _ from two different ( and apparently unrelated ) empirical laws .
one might see in this result a signal supporting the conjectures of @xcite .
it should be pointed out , however , that the numerical value of @xmath45 is not very different from a realistic asymptotic value of the square of the gut gauge - coupling parameter , @xmath46 .
it is a curious coincidence that the replacement of @xmath47 with @xmath48 in eq .
( [ 8 ] ) would lead to extract from the data another energy scale , @xmath49 , which also may have the meaning of uv cut - off scale ( numerically similar but conceptually very different from @xmath50 ) , as recently stressed on the grounds of the principle of asymptotic saturation of all couplings , induced by loop corrections @xcite .
the additional relation ( [ 9 ] ) , and the presence of macroscopic magnetic fields in the relevant macroscopic observations , however , would seem to point at the electro - gravity scales ( [ 8 ] ) , ( [ 9 ] ) as a possibly more appropriate interpretation of the data .
it should be mentioned that the near coincidence of @xmath24 and @xmath51 was previously noticed in @xcite , without any reference , however , to the empirical law concerning magnetic moments .
also , a theoretical interpretation of the parameter @xmath30 , as a possible imprint of early geometric spin - torsion interactions at the epoch of structure formation , was attempted in @xcite , providing the result @xmath52 , in agreement with eq .
( [ 9 ] ) ( leaving unexplained , however , the mass - angular momentum relation ) . the interpretation of @xmath43 , @xmath53 as effective cut - off scales , following the conjectures of @xcite , seems instead to connect two different empirical relations , and to predict a limiting scale for the angular momentum and the magnetic moment of macroscopic bodies of mass @xmath5 , fixing a lower bound on the _ maximal _ allowed values of @xmath6 and @xmath28 : j ( me ) ^2 , m [ 10 ] ( in natural units , from the combination of eqs .
( [ 1 ] ) , ( [ 8 ] ) and eqs . (
[ 4 ] ) , ( [ 9 ] ) , respectively ) .
as remarked in @xcite for other restrictions implied by the uv scale @xmath43 , this _ does not _
mean that it is impossible to have states with @xmath6 and @xmath28 smaller than the bound ( [ 10 ] ) , but it means that there must exist states satisfying the above inequalities .
the lesson we could learn , eventually , from the data is that the astrophysical structures , in the absence of perturbations , try to accumulate along a preferred " configuration corresponding to the state with the highest values of @xmath6 and @xmath28 compatible with the saturation of the above bound , required for a universe safely placed inside the string landscape @xcite . | it is a pleasure to thank gabriele veneziano for clarifying discussions , and dharamvir ahulawalia - khalilova for pointing out some interesting papers to my attention . | pleasure thank gabriele veneziano clarifying dharamvir ahulawalia khalilova interesting papers | 0.066411 | 39 | 23 | , Saving $0.0 in GPT-4. |
67,731 | the calculation of the cosmological relic density of the dark matter candidate within supersymmetric models is an interesting possibility to obtain additional constraints on the supersymmetric parameter space with respect to collider , electroweak precision , and low - energy data . considering future cosmological precision measurements
, radiative corrections can play an important role in the analysis .
we present full qcd and susy - qcd corrections to neutralino pair annihilation into quark - antiquark pairs and analyze their impact on the neutralino annihilation cross section and the relic density . by comparing to the relic density obtained from a pure leading order calculation ,
we show that the corrections strongly influence the extraction of susy mass parameters from cosmological data . | beginning with the rotational curves of galaxies , the astronomical and astrophysical evidence gathered in recent years supports the hypothesis of dark matter in our universe . moreover
, the models of structure formation favor the so - called cold dark matter which consists of weakly interacting particles with non - relativistic velocities .
based mainly on the cosmic microwave background measurements , one obtains the following limits on the relic density of cold dark matter cross - section @xmath0 where the @xmath1 interval was obtained from the three - year data of the wilkinson microwave anisotropy probe ( wmap ) , combined with the results of sloan digital sky survey ( sdss ) , supernovae legacy survey ( snls ) experiment and from baryonic acoustic oscillations data @xcite .
one of the main motivations for physics beyond the standard model ( sm ) is the inability of sm to provide candidates for cold dark matter . in the minimal supersymmetric standard model ( mssm ) with r - parity conservation , the lightest supersymmetric particle ( lsp ) is always stable .
if it is also a colour singlet , electrically neutral and massive , it is a good candidate for cold dark matter . in many scenarios , in particular those where supersymmetry breaking is gravity mediated ,
the lightest neutralino is the lsp and a suitable dark matter candidate .
we can then calculate its number density @xmath2 by solving the boltzmann equation @xmath3 where the first term on the right - hand side corresponds to a dilution due to the expansion of the universe , and the second one to a decrease due to annihilations and co - annihilations . here
, @xmath4 denotes the hubble expansion parameter , @xmath5 the density of the relic particle in thermal equilibrium and @xmath6 is the relative velocity of the annihilating pair .
the boltzmann equation can be integrated numerically and the energy density of the dark matter particle is approximately inversely proportional to the thermally averaged annihilation cross - section @xmath7 , @xmath8 the latter includes all annihilation and co - annihilation processes of the dark matter particle into standard model particles .
these by now standard calculations are incorporated in program packages such as darksusy @xcite and micromegas @xcite which allow different scenarios and different models to be studied . using the experimental limits ( [ cwmap ] ) , we can put constraints on the mssm parameter space .
if one is confined to the mssm with gravity mediated supersymmetry breaking ( msugra ) , the constraints allow for a rather narrow allowed band in the space of the high - scale parameters @xmath9 .
moreover , the constraints are expected to become even more strict in the future when the planck mission will provide additional data . to match and fully exploit the current and expected experimental precision , one has to improve the theoretical prediction of the relic density by including the radiative corrections .
we calculated the qcd and susy - qcd corrections to neutralino annihilation into quarks . in the work presented here ,
we investigate the effects of radiative corrections on the constraints of the mssm parameter space .
neutralino annihilation into fermions is one of the most important annihilation channels .
it is dominant for light neutralinos where all other final states are not kinematically allowed .
the leading order cross - section was discussed in much detail in @xcite .
the cross - section in the non - relativistic limit is conventionally written as ( see @xcite ) @xmath10 where @xmath6 is the relative velocity of the neutralino pair . in the case of annihilation into fermions ,
the coefficient @xmath11 is proportional to the mass of the fermion @xcite . as a consequence ,
the dominant final states are the third generation fermions , i.e. in our case the top and bottom quarks .
note that the contribution from the coefficient @xmath12 is suppressed with respect to the one coming from @xmath11 .
the annihilation into fermions has to be enhanced by a resonance in order to yield a small enough relic density satisfying the experimental constraints ( [ cwmap ] ) .
it proceeds mainly through a higgs boson resonance ( in particular the cp - odd higgs resonance ) .
this implies that yukawa coupling of fermions to the higgs boson becomes an essential parameter in this calculation along with @xmath13 .
the bottom yukawa coupling receives large susy corrections for large @xmath13 @xcite , which is another reason to incorporate radiative corrections in the predictions .
the approximation used in @xcite is already included in the micromegas package alongside with the qcd corrections for higgs decays into fermions ( see @xcite ) .
.the minimal supergravity ( msugra ) scenarios that lead to a neutralino relic density agreeing with the current limits of eq .
( [ cwmap ] ) , satisfy the current susy - particle mass limits , and present important contributions of neutralino annihilation into bottom or top quark - antiquark pairs .
we indicate the msugra parameters at the high scale , the relic density obtained with micromegas , and the contributions of the different quark final states to the annihilation cross - section @xmath7 . [ cols="^,^,^,^,^,^,^,^,^",options="header " , ] ( 450,290)(0,0 ) ( 5,150 ) ( 155,152 ) ( 5,0 ) ( 155,2 ) ( 310,2 )
we calculate the complete qcd and susy - qcd corrections not only to the higgs boson exchange but also to the @xmath14-channel z - boson exchange and the @xmath15-channel and @xmath16-channel squark exchange .
we also include parts of the correction for the higgs exchange which are not treated in @xcite .
we find that the additional correction when including the full one - loop radiative result are 10 - 15% relative to the corrections already included in micromegas .
we present numerical results for four typical msugra parameter points shown in tab .
[ tpoints ] .
the high - scale parameters were evolved down to the electroweak scale using spheno @xcite and the neutralino relic density has been calculated with micromegas , where we have included our calculation of the annihilation cross section as discussed above . for each point , the relic density fulfills the limits ( [ cwmap ] ) and additionally , the branching ratio of the decay @xmath17 lies within the experimental limits @xmath18 obtained from combined measurements from babar , belle , and cleo @xcite .
this limit leads to constraints on parameters connected to sfermions and gauginos .
the selected points with a dominant bottom quark final state are either near the bulk region ( point 1 ) with low fermion mass parameter @xmath19 for low @xmath13 ( the scalar mass is rather large in order to avoid the constraint by @xmath20 ) or in the higgs - funnel region for large @xmath13 where our point 2 is positioned exactly on the cp - odd higgs resonance .
both points 3 and 4 for which the top quark final state is important , lie in the focus point region where the scalar mass parameter @xmath21 is large . in figs .
[ fig1 ] and [ fig2 ] we show the numerical results for the relevant annihilation cross section and the resulting neutralino relic density for the points presented in tab . [ tpoints ] .
we show the leading order result ( green dash - dotted ) , the approximation already implemented in micromegas ( red dashed ) , and the result including our full one - loop susy - qcd corrections ( blue solid ) . in the graphs showing the cross section ( left row )
we also indicate the boltzmann distribution function involved in the calculation of the thermal average ( couloured area ) . as already stated above , the one - loop contributions that are not implemented in micromegas lead to an increase of about 10 15 % in the annihilation cross section both for bottom and top quark final states . in consequence ,
the neutralino relic density decreases by about the same amount for the different points , as can be seen in the center and right panels showing the neutralino relic density as a function of the high - scale mass parameters @xmath19 and @xmath21 , respectively . for fixed values of @xmath13 , @xmath22 and @xmath23 ,
the favoured band in the @xmath21-@xmath19 plane is shifted to lower gaugino masses and higher scalar masses ( focus point region ) .
9 j. hamann , s. hannestad , m. s. sloth and y. y. y. wong , phys . rev .
d * 75 * ( 2007 ) 023522 [ arxiv : astro - ph/0611582 ] .
p. gondolo , j. edsjo , p. ullio , l. bergstrom , m. schelke and e. a. baltz , jcap * 0407 * ( 2004 ) 008 [ arxiv : astro - ph/0406204 ] , g. belanger , f. boudjema , a. pukhov and a. semenov , arxiv:0803.2360 [ hep - ph ] , g. belanger , f. boudjema , a. pukhov and a. semenov , comput .
commun .
* 176 * ( 2007 ) 367 [ arxiv : hep - ph/0607059 ] , g. belanger , f. boudjema , a. pukhov and a. semenov , comput .
commun .
* 174 * ( 2006 ) 577 [ arxiv : hep - ph/0405253 ] , g. belanger , f. boudjema , a. pukhov and a. semenov , comput .
. commun .
* 149 * ( 2002 ) 103 [ arxiv : hep - ph/0112278 ] . m. drees and m. m. nojiri , phys . rev .
d * 47 * ( 1993 ) 376 [ arxiv : hep - ph/9207234 ] .
g. jungman , m. kamionkowski and k. griest , phys .
* 267 * ( 1996 ) 195 [ arxiv : hep - ph/9506380 ] . m. s. carena , d. garcia , u. nierste and c. e. m. wagner , nucl .
b * 577 * ( 2000 ) 88 [ arxiv : hep - ph/9912516 ] .
m. drees and k. i. hikasa , phys .
b * 240 * ( 1990 ) 455 [ erratum - ibid .
b * 262 * ( 1991 ) 497 ] .
w. porod , comput .
commun .
* 153 * ( 2003 ) 275 [ arxiv : hep - ph/0301101 ] .
e. barberio _ et al .
_ [ heavy flavor averaging group ( hfag ) ] , arxiv : hep - ex/0603003 . | rotational curves of galaxies astronomical astrophysical evidence supports hypothesis dark matter universe models structure formation favor cold dark matter weakly interacting particles non - relativistic velocities cosmic microwave measurements limits on relic density of cold dark matter cross - section @xmath1 interval from three - year data wilkinson microwave anisotropy probe results sloan digital sky survey supernovae legacy survey baryonic acoustic oscillations data physics beyond standard model inability provide candidates for cold dark matter minimal supersymmetric standard model lightest supersymmetric particle lsp stable colour singlet electrically neutral massive good candidate for cold dark matter supersymmetry breaking gravity mediated lightest neutralino lsp suitable dark matter candidate calculate number density @xmath2 boltzmann equation @xmath3 first term dilution due expansion second decrease due to annihilations co - annihilations @xmath4 hubble expansion parameter @xmath5 density of relic particle in thermal equilibrium @xmath6 relative velocity of annihilating pairboltzmann equation integrated numerically energy density dark matter particle inversely proportional to thermally averaged annihilation cross - section @xmath7 includes annihilation co - annihilation processes into standard model particles standard calculations in darksusy @xcite micromegas @xcite allow different scenarios models experimental limits constraints on mssm parameter space confined to gravity mediated supersymmetry breaking constraints allow narrow allowed band high - scale parameters @xmath9 constraints expected strict future planck mission additional data improve theoretical prediction relic density including radiative corrections calculated qcd susy - qcd corrections to neutralino annihilation into quarks effects radiative corrections on constraints mssm parameter space neutralino annihilation into fermions important dominant for light neutralinos other final states not allowed leading order cross - section discussed in @xcite cross - section non - relativistic limit written as @xmath10 is relative velocity of neutralino pair annihilation into fermions coefficient @xmath11 proportional to mass fermion dominant final states are third generation fermionstop bottom quarks contribution from @xmath12 suppressed @xmath11 annihilation into fermions enhanced by resonance small relic density experimental constraints through higgs boson resonance cp - odd higgs resonance yukawa coupling of fermions to higgs boson essential @xmath13 bottom yukawa coupling large susy corrections for @xmath13 @xcite incorporate radiative corrections predictions approximation @xcite included in micromegas package qcd corrections for higgs decays into fermions minimal supergravity msugra scenarios neutralino relic density current limits satisfy susy - particle mass limits contributions of neutralino annihilation into bottom top quark - antiquark pairs msugra parameters high scale relic density with micromegas contributions quark final states to annihilation cross - section @xmath7 450,290)(0,0 5,150 155,152 5,2 310,2 calculate complete qcd susy - qcd corrections to higgs boson exchange @xmath14-channel z - boson exchange @xmath15-channel @xmath16-channel squark exchangeinclude correction higgs exchange not treated in @xcite additional correction full one - loop radiative result 10 - 15% relative corrections micromegas present numerical results for four msugra parameter points in tab . high - scale parameters evolved to electroweak scale using spheno @xcite neutralino relic density calculated with micromegas included calculation annihilation cross section each point relic density fulfills limits branching ratio decay @xmath17 within experimental limits @xmath18 measurements babar @xcite limit leads to constraints on parameters sfermions gauginos selected points dominant bottom quark final state near bulk region low fermion mass parameter @xmath19 or higgs - funnel region for large @xmath13 point 2 on cp - odd higgs resonance points 3 and 4 in focus point region scalar mass parameter @xmath21 large figs fig1 fig2 show numerical results annihilation cross section neutralino relic density for points tab . leading order result approximation implemented in micromegas result including full one - loop susy - qcd correctionsgraphs showing cross section left indicate boltzmann distribution function thermal average one - loop contributions not implemented in micromegas lead increase 10 15 % in annihilation cross section bottom top quark final states neutralino relic density decreases same different points center right panels function high - scale mass parameters @xmath19 @xmath21 fixed values of @xmath13 favoured band in @xmath21-@xmath19 plane shifted to lower gaugino masses higher scalar masses focus point j. hamann s. hannestad sloth wong phys . rev . d * 75 * ( 2007 ) 023522 arxiv astro - ph/0611582 ] p. gondolo j. edsjo ullio bergstrom schelke baltz 0407 ( 2004 ) 008 astro - ph/0406204 g. belanger f. boudjema a. pukhov. semenov arxiv:0803.2360 hep belanger semenov commun 176 ( 2007 ) 367 hep - ph/0607059 belanger pukhov semenov 174 ( 2006 ) 577 arxiv hep - ph/0405253 ] belangersemenov comput commun 149 2002 ) 103 arxiv hep - ph/0112278 m drees nojiri phys . 47 1993 ) 376 hep - ph/9207234 ] jungman kamionkowski griest. 267 1996 ) 195 hep - ph/9506380 ] carena garcia nierste wagner nucl 2000 ) 88 hep - ph/9912516 ] drees hikasa. 240 1990 ) 455 erratum ibid 262 1991 ) 497 porod comput commun 153 2003 ) 275 arxiv hep - ph/0301101 ] barberio heavy flavor averaging group hfag ) hep - ex/0603003 | 0.357108 | 2,488 | 1,256 | , Saving $0.1 in GPT-4. |
67,732 | by comparing lattice data for the two heavy - light meson system ( @xmath0 ) with a standard many - body approach employing only interquark potentials , it is shown that the use of _ unmodified two - quark _ potentials leads to a gross overestimate of the binding energy . | in this conference the emphasis is on the study of _ single _ hadrons , with virtually nothing being said about systems involving more than three quarks .
this is unfortunate , since it avoids much of particle physics and all of nuclear physics .
it is , therefore , desirable to attempt to bridge this gap . since the ability to perform accurate lattice simulations for systems containing many quarks is severely limited with four quarks being the current maximum
the challenge is to see how ( if ) standard many - body techniques can be developed for the new situation of interacting quarks and gluons . here
we immediately meet a cultural difference . for particle physicists ,
the use of simply two - quark potentials in multiquark problems is not a discussion point , since they know ( believe ) that it is not possible . in spite of this , there are many - body physicists , who still believe ( hope ) that this is not so . over the last few years attempts have been made to clarify this situation by comparing the _ exact _ energies of four - quark systems as calculated on a lattice with standard many - body models using only two - quark potentials [ 1 ] .
the four - quark system was studied for two reasons : + i ) accurate lattice simulations are still possible .
+ ii ) it is the first step in the description of interacting hadrons , since the system can be partitioned into two colour singlets .
the outcome from those pilot comparisons was that
`` beyond all reasonable doubt '' the resulting binding energies are grossly overestimated by the models [ figure [ fig:1 ] ] and that a four - quark form factor is necessary .
however , this conclusion was based on lattice data for four static quarks in quenched su(2 ) . in the present work @xcite most of these approximations
have been removed by using su(3 ) with two heavy - light mesons i.e. @xmath0 compared with the earlier @xmath1 .
furthermore , the gauge field is now treated in the unquenched approximation .
but this latter improvement only has a small effect .
the corresponding many - body model now needs some additional assumptions before a comparison with the lattice data is possible .
perhaps the most serious is the use of a non - relativistic kinetic energy , even though the light quark mass is approximately that of the strange quark . however , in the binding energy there is a considerable cancellation between the four- and two - quark kinetic energies , which then removes much of the kinetic energy effect from the binding energy .
also the model is developed only for the spin - independent contribution to the binding energy .
this requires an averaging of the lattice data . both of these assumptions
are not thought to qualitatively effect the outcome .
the main defect of the calculation is that , at present , comparison can only be made for small distances between the two static quarks
the region where lattice data could suffer from a lack of rotational invariance .
however , in spite of these shortcomings , the result is that the use of simply two - quark potentials again overestimates the lattice binding by upto a factor of three a result that is not expected to be qualitatively changed by other models without multi - quark interactions [ figure [ fig:2 ] solid line ] .
inclusion of a four - quark interaction term can then remove the discrepancy [ figure [ fig:2 ] dashed line ] .
the reason why the `` data - model '' comparison is good for small @xmath2 in figure [ fig:1 ] and not in figure [ fig:2 ] is simply because the @xmath0 energy even for small @xmath3 is dominated by potential energy contributions , where the @xmath4 and @xmath5 separations are @xmath6 . as seen in figure [ fig:1 ] , at such separations the model already becomes too attractive in the absence of the form factor @xmath7 .
in refs . @xcite a model was developed for understanding the lattice energies of four static quarks + @xmath8 in terms of two - quark potentials .
this model , in its simpliest form , was constructed in terms of the two basis states that can be made by partitioning the four quarks into two color singlets - namely - @xmath9[q(2)\bar{q}(4 ) ] \nonumber \\
b&=&[q(1)\bar{q}(4)][q(2)\bar{q}(3 ) ] , \end{aligned}\ ] ] where @xmath10 $ ] denotes a color singlet .
these two states are not orthogonal and have a normalisation matrix of the form . @xmath11 in the extreme weak coupling limit the parameter @xmath12 and in the strong coupling limit @xmath13 .
however , for intermediate situations it is parametrised as @xmath14,\ ] ] where @xmath15 is the string energy , @xmath16 is an area defined by the positions of the quarks and @xmath17 is a free parameter .
a single value for @xmath17 was then capable of giving a reasonable understanding for 100 pieces of data the ground and first excited states of configurations from six different four - quark geometries calculated on a @xmath18 lattice . in this model
the interaction between the quarks is expressed as a potential matrix of the form @xmath19 where @xmath20 has the form expected in the weak coupling limit with the one - gluon - exchange - potential @xmath21 away from the weak coupling limit , @xmath7 is no longer unity and in addition @xmath22 is taken to be the full two quark potential .
the energy of the four static quarks a function of the @xmath23 is then given by diagonalising latexmath:[\[\label{vn } easy to generalise to @xmath0 by integrating over the positions of the two light quarks using a suitable variational wavefunction @xcite .
this results in the need to diagonalise @xmath25 is the kinetic energy and @xmath3 is the distance between the two static quarks . in order to carry out the radial integrals in @xmath26 and @xmath27 , the function @xmath7 in eq .
[ fexp ] was now reduced to the more symmetric form @xmath28.\ ] ] again @xmath17 is a free parameter , which should be adjusted to fit the four - quark lattice energies .
however , now only the lattice data for @xmath29 is sufficiently accurate to extract a value of @xmath17 see figure [ fig:2 ] .
the present model is now being developed to include : + 1 ) a semi - relativistic approximation , in which for the two light quarks the kinetic energy operator in momentum space is simply replaced by @xmath30 .
+ 2 ) a spin dependence in the potential between the two light quarks . this model ,
although very simple , contains the same basic assumptions made in the more elaborate many - body models that incorporate kinetic energy e.g. the resonating group method @xcite .
it is , therefore , reasonable that this simplified model can to some extent check the validity of its more elaborate counterparts .
the comparison between lattice data and a model for the @xmath0 system supports ( confirms ) the earlier result with four static quarks in quenched su(2 ) that four - quark energies can not be described simply in terms of two - quark potentials and that attempts to do so could lead to a large overestimate of the binding energy .
one way of overcoming this problem is to include into the model a four - quark form factor as seen in figure 2 .
9 a.m. green , c. michael and j. paton , nucl.phys . * a554 * , 701 ( 1993 ) . c. michael and p. pennanen , phys.rev .
d * 60 * , 54012 ( 1999 ) , hep - lat/9901007 .
a.m. green , j. koponen and p. pennanen , hep - ph/9902249 .
a.m. green and p. pennanen , phys .
c * 57 * , 3384 ( 1998 ) , hep - lat/9804003 .
ukqcd collaboration : c. michael and p. pennanen , work in progress .
m. oka and yazaki , `` quarks and nuclei '' , int .
review of nuclear physics - vol . 1 ( 1984 )
p.489 , eds . t.t.s.kuo and e.osnes ( world scientific publishing co. singapore ) | conference emphasis on study single _ hadrons nothing said about systems more than three quarks unfortunate avoids particle physics nuclear physics desirable to bridge gap accurate lattice simulations for systems many quarks limited four quarks current maximum challenge standard many - body techniques for new situation interacting quarks and gluons cultural difference for particle physicists use of two - quark potentials in multiquark problems not discussion point not possible many - body physicists still believe not attempts to clarify by comparing exact energies of four - quark systems with standard many - body models using two - quark potentials four - quark system studied for two reasons accurate lattice simulations possible first step in description of interacting hadrons system partitioned into two colour singlets outcome pilot comparisons binding energies overestimated by models four - quark form factor necessary conclusion based on lattice data for four static quarks in quenched su(2 ) in present work approximations removed by using su(3 ) with two heavy - light mesons @xmath0 gauge field now treated in unquenched approximation improvement small effectmany - body model needs additional assumptions before comparison with lattice data serious is use of non - relativistic kinetic energy light quark mass approximately strange quark in binding energy considerable cancellation between four- and two - quark kinetic energies removes kinetic energy effect from binding energy model developed only for spin - independent contribution to binding energy requires averaging of lattice data assumptions not qualitatively effect outcome defect calculation comparison only for small distances between two static quarks use of two - quark potentials overestimates lattice binding by upto factor of three not by without multi - quark interactions inclusion of four - quark interaction term can discrepancy data - model comparison good for small @xmath2 not because @xmath0 energy for dominated by potential energy contributions @xmath4 and @xmath5 separations are @xmath6 at model becomes too attractive in absence of form factor @xmath7 model developed for understanding lattice energies of four static quarks + @xmath8 in two - quark potentialsmodel constructed two basis states partitioning four quarks into two color singlets @xmath9[q(2)\bar{q}(4 ) ][q(1)\bar{q}(4)][q(2)\bar{q}(3 ) ]{aligned @xmath10 color singlet two states not orthogonal normalisation matrix form @xmath11 weak coupling limit @xmath12 strong coupling limit @xmath13 intermediate situations parametrised as @xmath14 @xmath15 string energy @xmath16 area defined by positions quarks @xmath17 free parameter single value for @xmath17 understanding for 100 pieces data ground states configurations from six four - quark geometries @xmath18 lattice interaction between quarks expressed potential matrix of form @xmath19 @xmath20 form weak coupling limit - exchange potential @xmath21 @xmath7 no unity @xmath22 full two quark potential energy of four static quarks function @xmath23 diagonalising latexmath @xmath0 integrating over positions two light quarks variational wavefunction diagonalise @xmath25 kinetic energy @xmath3 distance between two static quarkscarry out radial integrals in @xmath26 @xmath27 function @xmath7 reduced to symmetric form @xmath28. @xmath17 free parameter adjusted to fit four - quark lattice energies lattice data for @xmath29 accurate extract value @xmath17 figure model developed to include semi - relativistic approximation two kinetic energy operator replaced by @xmath30 spin dependence in potential between light quarks model simple contains assumptions elaborate many - body models resonating group method simplified model can check validity of elaborate counterparts comparison between lattice data model for @xmath0 system supports confirms result with four static quarks four - quark energies not described in two - quark potentials to overestimate of binding energy include model four - quark form factor in figure 2 a.m. green c. michael j. paton nucl.phys . a554 701 ( 1993 ) p. pennanen phys.rev . 60 54012 ( 1999 ) hep - lat/9901007. green j. koponen p. pennanen hep - ph/990224957 3384 1998 hep lat/9804003 collaboration c michael p pennanen work progress oka yazaki quarks nuclei review nuclear vol 1 1984 p.489 eds t.t.s.kuo e.osnes world scientific publishing | 0.316051 | 1,865 | 930 | , Saving $0.1 in GPT-4. |
67,735 | we suggest a general rule for the shift quantum numbers @xmath0 of the relative ground states of antiferromagnetic heisenberg spin rings .
this rule generalizes well - known results of marshall , peierls , lieb , schultz , and mattis for even rings .
our rule is confirmed by numerical investigations and rigorous proofs for special cases , including systems with a haldane gap for @xmath1 .
implications for the total spin quantum number @xmath2 of relative ground states are discussed as well as generalizations to the @xmath3 model . | rigorous results on spin systems like the marshall - peierls sign rule @xcite and the famous theorems of lieb , schultz , and mattis @xcite have sharpened our understanding of magnetic phenomena .
they as well serve as theoretical input for quantum computing with spin systems.@xcite exact diagonalization methods allow to evaluate eigenvalues and eigenvectors of @xmath4 for small spin rings of various numbers @xmath5 of spin sites and spin quantum numbers @xmath6 where the interaction is given by antiferromagnetic nearest neighbor exchange @xcite .
one quantity of interest is the shift quantum number @xmath7 associated with the cyclic shift symmetry of the rings . using the sign rule of marshall and peierls @xcite or equivalently the theorems of lieb , schultz , and mattis
@xcite one can explain the shift quantum numbers for the relative ground states in subspaces @xmath8 of total magnetic quantum number @xmath9 for rings with even @xmath5 .
in the case of @xmath10 one knows the shift quantum numbers of the total ground states for all @xmath5 via the bethe ansatz.@xcite the sign rule of marshall and peierls as well as the theorems of lieb , schultz , and mattis only apply to bipartite rings , i. e. rings with even @xmath5 .
nevertheless , even for frustrated rings with odd @xmath5 astonishing regularities are numerically verified .
this creates the need for a deeper insight or at best an analytic proof for the simple k - rule [ c-1 ] which comprises all these results . unifying the picture for even and odd @xmath5
, we find for the ground state without exception:@xcite 1 .
the ground state belongs to the subspace @xmath11 with the smallest possible total spin quantum number @xmath2 ; this is either @xmath12 for @xmath13 integer , then the total magnetic quantum number @xmath9 is also zero , or @xmath14 for @xmath13 half integer , then @xmath15 .
2 . if @xmath13 is integer , then the ground state is non - degenerate .
3 . if @xmath13 is half integer , then the ground state is fourfold degenerate .
4 . if @xmath6 is integer or @xmath13 even , then the shift quantum number is @xmath16 .
if @xmath6 is half integer and @xmath13 odd , then the shift quantum number turns out to be @xmath17 .
if @xmath13 is half integer , then @xmath18 and @xmath19 is found .
@xmath20 symbolizes the greatest integer less than or equal to @xmath21 .
.numerically verified shift quantum numbers for selected @xmath5 and @xmath6 in subspaces @xmath22 . instead of @xmath9 the quantity @xmath23 is used .
the shift quantum number for the magnon vacuum @xmath24 is always @xmath16 .
the shift quantum numbers are invariant under @xmath25 and hence only displayed for @xmath26 .
extraordinary shift quantum numbers given in bold do not comply with eq .
[ e-1 - 1 ] .
[ cols="^,^ , < , < , < , < , < , < , < , < , < , < , < " , ] the question ( [ e-4 - 12 ] ) whether @xmath27 holds in @xmath28 with minimal @xmath29 was investigated numerically .
for some of the investigated rings the respective energies are given in table [ t-2 ] .
figure [ f-1 ] shows the ratio @xmath30 for rings with @xmath31 and various @xmath5 . this ratio is smaller than one for @xmath32 for all investigated @xmath5 . only for
@xmath33 the ratio reaches values above one .
nevertheless , as discussed in the previous section , in the cases of integer @xmath6 this ratio must approach zero like @xmath34 since @xmath35 approaches the haldane gap . but also in the cases of half integer spin one
is led to anticipate that the curves rising with @xmath5 for small @xmath5 will bend down later and approach zero for large @xmath5 .
dmrg calculations could help to clarify this question .
it is a legitimate question whether the k - rule holds for heisenberg spin rings only or whether it is valid for a broader class of spin hamiltonians . in order to clarify this question we investigate the following xxz - hamiltonian @xmath36 for various values of @xmath37 .
the case @xmath38 corresponds to the original heisenberg hamiltonian ( [ e-2 - 5 ] ) , @xmath39 results in the antiferromagnetic ising model , @xmath40 in the ferromagnetic ising model , and @xmath41 describes the xy - model .
we have numerically investigated the cases of @xmath42 for @xmath43 and @xmath44 . for @xmath45
no violation of the k - rule was found , whereas the k - rule is violated for @xmath46 . in the limiting case of the ising model the k - rule [ c-1 ]
is in general violated . any product state @xmath47 will be an eigenstate of the ising hamiltonian and the shifted states @xmath48 belong to the same eigenvalue @xmath49 .
the set of the corresponding shift quantum numbers then depends on the degree of symmetry of @xmath47 : let @xmath50 denote the smallest positive integer such that @xmath51 .
clearly , @xmath50 divides @xmath5 .
then the corresponding shift quantum numbers will be of the form @xmath52 .
in most cases , @xmath53 and hence all possible shift quantum numbers will occur , which violates [ c-1 ] . on the other hand
consider the total ground state @xmath54 of an even @xmath10 antiferromagnetic ising spin ring .
here we have @xmath55 and only the shift quantum numbers @xmath56 occur , also contrary to [ e-1 - 1 ] .
figure [ f-2 ] summarizes our findings as a graphics .
it is not clear at which @xmath37 exactly the k - rule breaks down , this quantum phase transition might very much depend on @xmath5 and @xmath6 .
it is then an open question whether another k - rule takes over .
finally we would like to mention that the exactly solvable @xmath10
@xmath57-model @xcite satisfies the k - rule ( [ e-1 - 1 ] ) .
this model is essentially equivalent to a system of @xmath58 non - interacting fermions .
more precisely , for odd @xmath58 its energy eigenvalues are of the form @xmath59 with corresponding shift quantum numbers @xmath60 relative ground state configurations @xmath61 for @xmath62 and odd @xmath5 are , for example , @xmath63 this leads to the shift quantum numbers @xmath64 in accordance with ( [ e-1 - 1 ] ) . similarly , the values @xmath65 are realized . in the case of
we have @xmath66 with corresponding shift quantum numbers @xmath67 and the k - rule [ c-1 ] follows analogously .
we thank shannon star for motivating discussions and ian affleck for pointing out some useful literature . | results on spin systems marshall - peierls sign rule theorems of lieb schultz mattis sharpened understanding magnetic phenomena theoretical input for quantum computing diagonalization methods eigenvalues eigenvectors @xmath4 for small spin rings interaction by antiferromagnetic nearest neighbor exchange interest shift quantum number @xmath7 with cyclic shift symmetry rings sign rule of marshall peierls theorems of lieb schultz mattis explain shift quantum numbers relative ground states in subspaces total magnetic quantum number @xmath9 for rings with even @xmath5 shift quantum numbers total ground states for @xmath5 via bethe ansatz sign rule of marshall peierls theorems of lieb schultz mattis apply to bipartite rings even frustrated rings with odd regularities numerically verified need for deeper insight proof for k - rule [ c-1 ] unifying picture for even odd @xmath5 ground state belongs to subspace @xmath11 with smallest total spin quantum number @xmath2 integer total magnetic quantum number zero half integer@xmath13 integer ground state non - degenerate half integer ground state fourfold degenerate @xmath6 integer @xmath13 even shift quantum number @xmath16 @xmath6 half integer @xmath13 odd shift quantum number @xmath17 @xmath13 half integer @xmath18 @xmath19 found @xmath20 symbolizes greatest integer less than or equal @xmath21 verified shift quantum numbers @xmath5 @xmath6 subspaces @xmath22 @xmath9 @xmath23 used shift quantum number magnon vacuum @xmath24 @xmath16 shift quantum numbers invariant under @xmath25 displayed for @xmath26 extraordinary shift quantum numbers bold comply with eq @xmath27 holds @xmath28 with minimal @xmath29 investigated energies given in table [ t-2 ] figure f-1 ratio @xmath30 for rings @xmath31 various @xmath5 ratio smaller than one for @xmath32 all @xmath5 @xmath33 ratio above one integer @xmath6 ratio approach zero like @xmath34 @xmath35 approaches haldane gaphalf integer spin curves rising with @xmath5 small bend approach zero for large dmrg calculations clarify question k - rule heisenberg spin rings or broader spin hamiltonians investigate xxz - hamiltonian @xmath36 values @xmath37 @xmath38 original heisenberg hamiltonian e-2 - 5 @xmath39 antiferromagnetic ising model ferromagnetic ising model xy - model investigated @xmath42 @xmath43 @xmath44 @xmath45 no violation k - rule rule violated for @xmath46 limiting case ising model k - rule [ c-1 ] violated product state @xmath47 eigenstate ising hamiltonian shifted states @xmath48 same eigenvalue @xmath49 shift quantum numbers on symmetry @xmath47 @xmath50 smallest positive integer divides @xmath5 shift quantum numbers @xmath52 @xmath53 all possible shift quantum numbers occur violates [ c-1 ] consider total ground state @xmath54 of @xmath10 antiferromagnetic ising spin ring @xmath55 shift quantum numbers @xmath56 occur contrary to [ e-1 - 1 ] figure f-2 ] summarizes findings @xmath37 k - rule breaks down quantum phase transition depend on @xmath5 @xmath6 open question another k - rule over solvable @xmath10 @xmath57-model satisfies k - rule e-1 - 1 equivalent to @xmath58 non - interacting fermions @xmath58 energy eigenvalues @xmath59 shift quantum numbers @xmath60 ground state configurations @xmath61 @xmath62 @xmath5 @xmath63 leads to shift quantum numbers @xmath64 e-1 - 1 values @xmath65 realized @xmath66 shift quantum numbers @xmath67 k - rule c-1 ] follows thank shannon star ian affleck literature | 0.266214 | 1,648 | 858 | , Saving $0.0 in GPT-4. |
67,738 | the paper revisits the investment simulation based on strategies exhibited by generalized @xmath0-zipf law to present an interesting characterization of the wildness in financial time series .
the investigations of dominant strategies on each specific time series shows that longer words dominant in larger time scale exhibit shorter dominant ones in smaller time scale and vice versa .
moreover , denoting the term wildness based on persistence over short term trend and memory represented by particular length of words , we can see how wild historical fluctuations over time series data coped with the zipf strategies .
@xmath1 + .25 in | in order to choose financial products for investment ( e.g. : stock indexes , future exchange indexes , foreign exchange indexes ) one often aware of how fast a particular financial values change over time by referring to its historical time series .
there are commonsensical thoughts that some markets are very `` wild '' for price varies so fast and the other is not of price moves very slowly . from the bottom - up perspective , it is apparent that the price movement will eventually remain the market depth liquidity , i.e. : order flow necessary to move price by a given amount over time . in return
, other thing that also contributed to this issue is the market climate surrounding investors or traders triggering the herding behavior over the market @xcite .
ausloos and bronlet @xcite previously have described the way of texifying the time series data and perform statistical aspects revealing the zipf law . in advance
, they also propose an interesting investment strategy based on their findings on power - laws inspired by the so - called dfa method .
the zipf law of the financial time series is presented by transforming the ups and downs of financial index fluctuations into `` words '' of _ m_-length and _ k_-types of `` letters '' @xcite .
obviously , there are a lot of possibilities on representing the financial indexes into texts , but the rest of the paper will use the letter _ `` u '' _ for higher value over time respect to the price in the opening and closing session of the market and _ `` d '' _ for downward price movement - whether they can be weekly , daily , sessional ( morning and afternoon market sessions ) , and even smaller time intervals , e.g. : hour , minutes , seconds @xcite .
however , concerning the investment simulation , we will use the smallest interval of hourly @xcite .
the paper wants to open a new possible analytical door on financial market by bridging the investment - based paradigm and the adherence of the generalized zipf law .
we see a possibility on a new and important way to approach , characterize and extract information within the time series data by analysis on the texified fluctuations .
the next section of the paper elaborates the zipf law in financial data in the way we have the simulated investment results by using the generalized zipf law .
this is followed by discussions on `` wildness '' of the time series data concerning the appropriate value of _ k _ used in the simulation .
by representing the price fluctuations into , say two alphabets , we have possible words in the time series data equal to @xmath2 possible words . they are _
`` @xmath3 '' _ , _
`` @xmath4 '' _ , _
`` @xmath5 '' _ , and _ `` @xmath6''_. this is applied up to several numbers of possible words .
thus , we sort frequency of occurrence of the words along our data to have the illustration of the zipf plot - rank the sequence started from @xmath7 for the most frequent words .
according to zipf law , we will have relation of : @xmath8 in accordance with r / s analysis , bronlet , et .
@xcite showed the conjecture relation : @xmath9 it is important to note that the probability of ups _ ( u ) _ and downs _
( d ) _ of price in most of the case is not the same for every time series data . here ,
we found bias : @xmath10 where _
p(u ) _ and _ p(d ) _ each is the probability of ups and downs in time series data . in this case , we need to put into account the frequency of expectation _ f _ , which represent random situation ( un - correlated ) . in this case , in our calculation we use _
f _ as the substitution from the original form of zipf ( _ f _ ) method with coefficient of exponent _
a_. this mathematically can clearly prevent the presence of exponent @xmath11 which is possible to occur since those biases . the value of _
f _ can be calculated as @xmath12 by now , we can start investment simulation for some financial data from various markets .
this is aimed for us to have basic of prediction based on scaling character from each financial data in case of investment .
technically , the question is : what would likely to happen if we have words sequence with alphabet sequence length of @xmath13 ? if we have words sequence of @xmath14 , how big the probability of _
c(t ) _ to be `` up '' or `` down '' , with _ c(t )
_ is showing the characters of time _
t_. from this point , we calculate the prediction based on the words we have ranked in zipf plot to find the level of probability of up , @xmath15 , that is the occurrence of sequence @xmath16 , and the probability of down , @xmath17 , or the occurrence of the sequence @xmath18 .
it is suggested @xcite to consider variable of strength parameter that is how far we can trust a calculation result , which is through relative probability represented by : @xmath19 this _
d(t ) _ parameter shows how big the probability of the prediction result to have value within interval _ @xmath20$]_. in the simulation , we use parameter _
d(t ) _ as a form of fraction of how many we _ `` buy '' _ or _
`` sell '' _ the index value . as we know
, we will pose to _ `` buy '' _ when @xmath21 and vice versa . by the simulation
, simply we can write the outcome of investment after n time as @xmath22 while in each time - steps we have @xmath23d(t)\ ] ] an example of our simulation result is depicted in figure 2 preceding the zipf law in jakarta composite index ( ihsg ) , djia , and nasdaq on each daily data in figure 1 .
obviously , the figure shows that the dominant investment strategy in ihsg is _ zipf ( 6,2 ) _ while in djia and nasdaq are _ zipf ( 4,2 ) _ and _ zipf ( 5,2 ) _ respectively .
after transforming a series of financial data into text and having simulation result based on zipf law in hand , plenty of analytical thoughts may appear .
the notion of relative wildness of financial data can be treated as the way we cope with the possibilities of ups and downs of the existing fluctuations . obviously , when a time series data can be read well by the generalized zipf law , we have a good view to cope with the wildness of the time series .
furthermore , the length of the words suitable to gain better along the certain time window of investment enriches our notion of this relative wildness. here we can see how persistent a word along certain time . for example , if a particular time series data get better gain while we use the rule of investment of _ zipf ( 7,2 ) _ relative to the one of _ zipf ( 2,2 ) _ , then we can say that the wildness of the one with _
zipf ( 7,2 ) _ is wilder than the other one . in other words
, we can say that concerning the short term trend persistence , time series represented by @xmath24 words in investment is less wild relative to the one with @xmath25 words . in advance
, we can also assume that the persistence over particular words as a memory effect of time series data .
in this perspective , apparently a less wild time series has longer memory since the persistent trend is longer .
this is very interesting when we try to see the effect in several specific financial data , e.g. : financial indexes and foreign exchange rates . in figure @xmath26 and @xmath27
, we show the result of our investigation in daily yen / usd and euro / usd data showing good gain in _ zipf ( 3,2 ) _ and _ zipf ( 2,2 ) _ respectively . comparing to the previous one in figure ( 2 )
, we can say relatively that both foreign exchange rates exhibit wilder fluctuations regarding persistence over trends and memory represented by the specific words .
certainly , we can understand this since foreign exchange involves more traders with higher sensitivity to fundamental issue relative to the market indexes .
the later question is whether or not the better strategies in a certain financial products are persistent over different time scales .
we do the similar analysis to hangseng index , nikkei225 index , and gpb / usd rate over different time orders .
figure ( 9 ) and ( 12 ) shows our results with hangseng and nikkei225 indexes on respective time series data depicted in figure ( 7 ) and ( 10 ) . interestingly , a wilder time series with shorter dominant words in small time scale have longer words of dominant strategies in the larger time scale and vice versa .
this is showed by hangseng index whose dominant strategy of _ zipf ( 7,2 ) _ in daily data shrinks upto _
zipf ( 4,2 ) _ in sessional data and _ zipf ( 3,2 ) _ in hourly in our investment simulation . in return , the dominant _ zipf ( 3,2 ) _ in daily nikkei225 investment simulation exhibits longer _ zipf ( 6,2 ) _ in sessional and hourly trading .
this is very interesting while gbp / usd rate shows similar situations of _ zipf ( 7,2 ) _ in daily data and _ zipf ( 8,2 ) _ in hourly simulations . intuitively , by referring to the wildness of a time series , we could say that there could be a pattern presented here .
the time series shows less wildness in a certain time scale could have wilder in bigger or narrower time scale .
this opens further investigations in spite of the fact that the generalized zipf analysis could yield a good prediction of time series fluctuations .
we show the simulation result of investment by using the strategies laid upon our understanding of zipf law in texified financial fluctuation . we evaluate the different strategies regarding
the length of words used in each simulation and discover interesting properties over different time series data and different time scale . in our simulations
we can see that longer words dominant in larger time scale exhibit shorter dominant ones in smaller time scale and vice versa . in the other hand
, we present that some financial indexes are wilder than other respect to the persistence of short term trend ( represented by the words or sequence of letters ) and memory over fluctuation patterns .
we expect to have more properties on doing several further investigations by involving more letter ( @xmath28 ) and by trying to relate the findings with terminologies often used to analyze the memory , distributions , and correlations of the time series . here ,
notwithstanding , we have seen some possible and interesting findings of the textual analysis upon texified time series data more than the advantage of the analysis as a good and promising prediction method .
authors thank m. ausloos for some important literatures , yohanis and jackson silaban for financial data , tiktik dewi sartika for some typeset corrections , and bfi colleagues for discussions .
noone but authors are responsible for possible errors and omissions .
bronlet , ph . , and ausloos , m. ( 2004 ) .
_ generalized @xmath29-zipf law for fractional brownian motion - like time series with or without effect of an additional linear trend_. pre - print : arxiv : cond - mat/0209306 vanderwalle , n. , brisbois , f. , and lefebvre , p.h .
`` managing both sign and size of fluctuations within the n - zipf framework '' . _ international journal of theoretical and applied finance _ 3(3):409 - 414 . | choose financial products for investment ( stock indexes future exchange indexes foreign exchange indexes ) aware fast financial values change over time historical time series commonsensical thoughts some markets wild '' for price varies fast other not price slowly bottom - up perspective price movement remain market depth liquidity order flow necessary to move price given amount over time contributed issue market climate surrounding investors triggering herding behavior market ausloos and bronlet described texifying time series data statistical aspects revealing zipf law propose investment strategy based on findings on power - laws dfa method zipf law financial time series transforming ups downs financial index fluctuations into words letters '' possibilities representing financial indexes into texts paper use letter _ `` u '' _ for higher value over time respect price opening closing session market _ `` d '' _ for downward price movement weekly daily sessional smaller time intervals investment simulation use smallest interval of hourly paper new analytical door on financial market bridging investment - based paradigm generalized zipf law possibility new way to approach characterize extract information time series data by analysis on texified fluctuationsnext section paper elaborates zipf law in financial data simulated investment results generalized zipf law followed by discussions on wildness time series data appropriate value of _ k _ simulation representing price fluctuations into two alphabets possible words in time series data equal to @xmath2 @xmath3 @xmath4 @xmath5 @xmath6'' applied to several numbers possible words sort frequency occurrence words zipf plot - rank sequence from @xmath7 for most frequent words according to zipf law relation of : @xmath8 analysis relation : @xmath9 probability of ups downs of price not same for every time series data found bias : @xmath10 where p(u and p(d ) _ probability of ups and downs series data frequency of expectation _ f _ random situation un - correlated calculation use _ f _ as substitution from original form zipf method with coefficient of exponent _ a_ prevent presence of exponent @xmath11 value of _ f _ calculated as @xmath12 start investment simulation for financial data from various markets aimed for basic prediction based on scaling character from each financial data investmentquestion is what words sequence length @xmath13 ? sequence @xmath14 probability _ c(t ) _ up or down showing characters time _ t_ calculate prediction words ranked zipf plot probability up @xmath15 @xmath16 down @xmath17 @xmath18 suggested consider variable strength parameter trust calculation result relative probability represented by @xmath19 _ d(t ) _ parameter shows probability prediction result value within interval @xmath20$]_ simulation use parameter _ d(t ) _ as of buy or sell index value buy '' _ when @xmath21 vice versa simulation write outcome investment after n time as @xmath22 each time - steps @xmath23d(t) example simulation result figure 2 zipf law in jakarta composite index ( ihsg ) djia nasdaq daily data figure 1 dominant investment strategy in ihsg is _ zipf ( 6,2 ) _ djia nasdaq _ zipf ( 4,2 ) _ and zipf ( 5,2 ) _ after transforming financial data into text simulation result based on zipf law analytical thoughts may appearrelative wildness of financial data with fluctuations time series data read by zipf law good view cope with wildness length of words suitable to gain along time window enriches relative wildness persistent word time if time series data get better gain rule investment of _ zipf ( 7,2 ) _ relative to _ zipf ( 2,2 ) _ wildness of with _ zipf ( 7,2 ) _ wilder short term trend persistence time series represented by @xmath24 words in investment less wild to with @xmath25 words persistence over words memory effect of time series data less wild time series has longer memory persistent trend longer interesting effect in specific financial data financial indexes foreign exchange rates figure @xmath26 and @xmath27 daily yen / usd euro / usd data good gain in _ zipf ( 3,2 ) _ and _ zipf ( 2,2 ) _ both foreign exchange rates exhibit wilder fluctuations regarding persistence over trends memory by specific words foreign exchange involves more traders with higher sensitivity to fundamental issue market question better strategies in financial products persistent over different time scalessimilar analysis to hangseng index nikkei225 index gpb / usd rate over time orders figure ( 9 ) 12 ) results with hangseng nikkei225 indexes on time series data figure 7 ) 10 ) wilder time series shorter dominant words small longer words strategies larger showed by hangseng index dominant strategy ( 7,2 ) _ daily shrinks ( 4,2 ) sessional 3,2 ) _ hourly dominant _ zipf ( 3,2 ) _ daily nikkei225 simulation longer _ zipf ( 6,2 ) _ in sessional hourly gbp / usd rate shows similar situations _ zipf ( 7,2 ) _ daily zipf ( 8,2 ) _ hourly simulations wildness time series pattern less wildness time scale wilder bigger time scale opens investigations generalized zipf analysis prediction time series fluctuations simulation result investment strategies zipf law financial fluctuation evaluate strategies length of words simulation discover properties over different time series time scale longer words dominant in larger time scale exhibit shorter smaller time scale some financial indexes wilder persistence of short term trend memory over fluctuation patternsexpect more properties investigations involving letter @xmath28 relate findings terminologies memory distributions correlations time series seen findings textual analysis time series data advantage prediction method authors thank ausloos literatures yohanis jackson silaban financial data tiktik dewi sartika typeset corrections bfi colleagues discussions authors responsible errors omissions bronlet ph ausloos m. 2004 ) generalized @xmath29-zipf law brownian motion time series additional linear trend pre - print arxiv cond - mat/0209306 vanderwalle brisbois f lefebvre p managing sign size fluctuations n - zipf framework international journal theoretical applied finance 3(3):409 - 414 | 0.362298 | 2,498 | 1,257 | , Saving $0.1 in GPT-4. |
67,759 | in 2011 , bailes reported on the discovery of a detached companion in a 131 minute orbit around psr j17191438 , a 173 hz millisecond pulsar .
the combination of the very low mass function and such a short orbital period is unique .
the discoverers suggested that the progenitor system could be an ultracompact x - ray binary ( ucxb ) , which is a binary with a sub - hour orbital period in which a ( semi-)degenerate donor fills its roche lobe and transfers mass to a neutron star .
the standard gravitational - wave driven ucxb scenario , however , can not produce a system like psr j17191438 as it would take longer than the age of the universe to reach an orbital period of 131 min .
we investigate two modifications to the standard ucxb evolution that may resolve this discrepancy .
the first involves significant heating and bloating of the donor by pulsar irradiation , and in the second modification the system loses orbital angular momentum via a fast stellar wind from the irradiated donor , additional to the losses via the usual gravitational wave radiation . in particular
a donor wind is effective in accelerating orbital expansion , and even a mild wind could produce the 131 minute period within the age of the universe .
we note that ucxbs could be an important class of progenitors of solitary millisecond radio pulsars . | psr j17191438 , with a 5.8 ms spin period , was found to be orbited every 131 min by a companion ( * ? ? ? * ( bailes 2011 ) ) .
the mass function of @xmath0 implies that in the ( a priori ) very likely case that we do not observe the psr j17191438 system close to face - on , the companion must have a mass of the order of @xmath1 .
this system is very interesting because almost all millisecond radio pulsars either have a stellar - mass companion or no companion at all .
there are black widow x - ray binaries with similar orbital periods but those are semi - detached and have much more massive companions .
the system is detached , and from its orbital period a minimum average density of the companion of @xmath2 can be derived .
neutron star ) and open circles ( @xmath3 neutron star ) indicate the 1 , 0.1 and 0.01 a priori probabilities that the companion mass is higher than indicated .
figure from @xcite.,width=364 ] in fig .
[ fig : wdrad ] we show the constraints on the companion set by its minimum density ( i.e. , its roche - lobe radius at a given mass , which is the maximum radius ) , and the radii of zero - temperature white dwarfs of different compositions , which give the minimum radius .
we see that a carbon - oxygen composition is the most likely , followed by helium , but that requires an inclination with an a priori probability near @xmath4 , or less in case of a finite - temperature companion .
hydrogen is possible if we see the systems nearly face - on , with an a priori probability near @xmath5
@xcite suggested that this system could have evolved from an ultracompact x - ray binary ( ucxb ) , which consists of a ( semi-)degenerate donor transferring mass to a neutron star , driven by angular momentum loss via gravitational wave radiation ( * ? ? ? * ( e.g. savonije 1986 ) ) .
the mean donor density is one - to - one related to the orbital period , as it always the case for a roche - lobe filling star with a more massive companion .
therefore , as the donor loses mass and becomes less dense , the orbit expands , and the mass transfer rate decreases .
however , ucxbs are expected to reach orbital periods of at most 90 min within the age of the universe ( * ? ? ?
* ( deloye & bildsten 2003 ) ) .
we discuss two modifications to this standard ucxb scenario that could explain how they could reach an orbital period of 131 min .
the first is a relatively large donor size , the second is a stellar wind from the donor .
both are caused by heating and irradiation of the donor by the accretion disk , accretor , and magnetosphere . .
the dotted parts of the curves take longer than the age of the universe to reach .
the vertical line is located at the orbital period of psr j17191438 , and the triangles mark the a priori probabilities that the donor mass is higher than indicated , based on the mass function .
figure adapted from @xcite.,width=364 ]
the size of the donor in an ucxb influences the mass transfer rate .
a larger ( ` bloated ' ) donor with respect to the zero - temperature radius implies a higher mass is required to have a given average density , which corresponds to an orbital period , as mentioned before .
a higher donor mass in turn implies a higher mass transfer rate and faster evolution . in fig .
[ fig : mp ] is shown that ucxbs with relatively large donors can reach a period of 131 within the age of the universe . however , the donor mass would be about ten times larger than suggested by the mass function .
this is unlikely to be the main explanation of the 131 min period , but it could contribute .
in a binary , the specific angular momentum is highest in the lowest - mass component . at the extreme mass ratio found in ucxbs
, a large amount of angular momentum can be lost by systems if the donor loses matter in a fast , isotropic wind .
such a wind can be caused by irradiation of the donor .
[ fig : time ] shows that a relatively low wind mass loss rate during the entire evolution is already enough to speed the orbital expansion to a period of 131 min within the age of the universe .
long - term observations by the _ rossi xte _ all - sky monitor show that most of the ucxbs with orbital periods longer than 40 min are one to two orders of magnitude brighter than expected when their mass transfer was driven only by angular momentum loss via gravitational wave radiation ( * ? ? ?
* ( van haaften 2012c ) ) . also , @xcite found that cataclysmic variables below the period gap evolve faster than expected from gravitational wave emission only .
this suggests that they indeed evolve faster , which supports the hypothesis that psr j17191438 is an ucxb descendant .
strong evidence comes from the psr j13113430 system ( * ? ? ?
* ( pletsch 2012 ) ) , an ucxb with an unusually long orbital period of 93.8 min ( * ? ? ? * ; * ? ? ? * ( romani 2012 , kataoka 2012 ) ) and an evaporating helium donor ( * ? ? ? * ( romani 2012 ) ) .
@xcite explained the formation of psr j17191438 via an originally hydrogen - rich donor star .
the present detached state of the system may be caused by the gradually decreasing irradiation , which could result in the donor shrinking inside its roche lobe .
then the accretion rate would decrease , further decreasing the level of donor irradiation .
psr j17191438 very likely has a low - mass companion , composed mainly of carbon / oxygen or possibly helium .
its orbital period of 131 min could have been reached by an ultracompact x - ray binary whose late - time evolution was driven by angular momentum loss via a stellar wind from the donor , rather than the emission of gravitational waves , which subsequently became detached .
a wind of @xmath6 would be sufficient .
the composition of the companion is most likely carbon - oxygen , though helium can not be ruled out .
old ucxbs are expected to have very low donor masses ( below @xmath7 ) .
if psr j17191438 has indeed formed out of an ucxb , then we would expect to see more binary millisecond radio pulsars with very low mass companions .
such systems are very rare though , whereas solitary millisecond radio pulsars are relatively common .
this suggests the possibility that low - mass companions of ucxbs are completely evaporated ( * ? ? ?
* ( ruderman 1989 ) ) , leaving a population of solitary millisecond radio pulsars . | psr j17191438 5.8 ms spin period orbited every 131 min by companion bailes 2011 mass function of @xmath0 implies companion must mass @xmath1 system interesting all millisecond radio pulsars have stellar - mass companion or no companion black widow x - ray binaries similar orbital periods semi - detached more massive companions system detached from orbital period minimum average density of companion of @xmath2 derived neutron star open circles ( @xmath3 neutron star indicate 1 0.1 0.01 probabilities companion mass higher figure from @xcite fig constraints on companion by minimum density roche - lobe radius radii of zero - temperature white dwarfs minimum radius carbon - oxygen composition most likely followed by helium requires inclination probability near @xmath4 or less finite - temperature companion hydrogen possible if systems nearly face - on probability near @xmath5 system evolved from ultracompact x - ray binary ucxb semi-)degenerate donor transferring mass to neutron star driven by angular momentum loss via gravitational wave radiation savonije 1986mean donor density related to orbital period roche - lobe filling star massive companion donor loses mass less dense orbit expands mass transfer rate decreases ucxbs expected to reach orbital periods most 90 min within age universe deloye & bildsten 2003 two modifications to ucxb scenario could explain orbital period 131 min first large donor size second stellar wind from donor both caused by heating irradiation donor by accretion disk magnetosphere dotted parts of curves take longer than age universe to reach vertical line at orbital period of psr j17191438 triangles mark probabilities donor mass higher mass function size donor influences mass transfer rate larger donor implies higher mass average density corresponds to orbital period higher donor mass implies higher mass transfer rate faster evolution fig ucxbs with large donors can reach period of 131 within age universe donor mass ten times larger than suggested by mass function unlikely to main explanation of 131 min period could contribute in binary specific angular momentum highest in lowest - mass component extreme mass ratio in ucxbs large angular momentum lost if donor loses matter in fast isotropic wind wind caused by irradiation donor low wind mass loss rate to orbital expansion to 131 min long - term observations by _ rossi xte _ all - sky monitor show ucxbs with orbital periods longer than 40 min are one to two orders of brighter when by angular momentum loss via gravitational wave radiation @xcite found cataclysmic variables below period gap evolve faster from gravitational wave emission supports hypothesis psr j17191438 is ucxb descendant evidence from psr j13113430 system ucxb with long orbital period of 93.8 min evaporating helium donor @xcite explained formation of psr j17191438 via hydrogen - rich donor star present detached state system may caused by decreasing irradiation in donor shrinking inside roche lobe accretion rate decreasing donor irradiation psr j17191438 likely has low - mass companion of carbon / oxygen or possibly heliumorbital period 131 min reached by ultracompact x - ray binary evolution driven by angular momentum loss via stellar wind gravitational waves wind of @xmath6 sufficient composition companion likely carbon - oxygen helium old ucxbs low donor masses below @xmath7 if psr j17191438 formed ucxb expect more binary millisecond radio pulsars with low mass companions systems rare solitary millisecond radio pulsars common suggests possibility low - mass companions ucxbs evaporated leaving population solitary millisecond radio pulsars | 0.265193 | 1,488 | 719 | , Saving $0.0 in GPT-4. |
67,769 | we compare several methods for the efficient generation of correlated random sequences ( colored noise ) by filtering white noise to achieve a desired correlation spectrum .
we argue that a class of iir filter - design techniques developed in the 1970s , which obtain the global chebyshev - optimum minimum - phase filter with a desired magnitude and arbitrary phase , are uniquely suited for this problem but have seldom been used .
the short filters that result from such techniques are crucial for applications of colored noise in physical simulations involving random processes , for which many long random sequences must be generated and computational time and memory are at a premium .
the generation of correlated random sequences , or `` colored noise '' , is important for many physical simulations involving random processes @xcite , and often the required computational time and memory is a critical concern .
although many filter - based techniques have been applied to this problem @xcite , in this brief manuscript we point out that these methods are suboptimal : a global chebyshev - optimum stable iir filter for this problem may be designed based on techniques developed in the 1970s @xcite .
colored - noise generation is required for many types of numerical simulations , such as thermodynamics @xcite , laser noise and first - passage time problems @xcite , and chaotic dynamics @xcite . in general , any numerical model involving stochastic differential equations
in which there is some background distribution , nonlinearity , or external noise associated with the quantities driving the fluctuations will require the use of colored noise @xcite .
such colored noise is most commonly described by a _ linear process _ : the output of a time - invariant linear filter applied to white noise @xcite .
the key implementation question , as we discuss below , is to determine how to apply this filter process . with rare exceptions @xcite , the underlying random process is almost always gaussian in nature , if it arises by a physical process governed by the central - limit theorem @xcite .
it is this nearly ubiquitous case that we focus on here , in which only the correlation function of zero - mean colored noise is of concern , and not any higher - order statistics . from a computational standpoint ,
the central problem is that many of these applications require many random sequences to be generated in parallel , and/or very long sequences , imposing performance and memory constraints on the noise generation .
this occurs , for example , in the generation of power - law ( @xmath0 ) `` pink noise '' for applications ranging from hydrology to music , where very long sequences ( whose length may not be known _ a priori _ ) are often required @xcite .
multiple long random sequences also occur in the generation of rayleigh random variates for simulation of nonisotropic scattering , macrocellular systems , and physical models of mobile multipath radio channels @xcite .
perhaps the most stringent requirements for colored - noise generation occur for spatio - temporal noise : where a different random sequence may need to be generated simultaneously for every point in space .
this occurs , for example , in modeling of brownian motion @xcite , or for simulation of thermal radiation @xcite .
vs. @xmath1 ( red ) .
the periodogram ( spectrum ) of a finite - length correlated random sequence generated by the methods in this paper is also shown ( black ) , computed by bartlett s method @xcite .
, scaledwidth=50.0% ] the fundamental technique for generating correlated random sequences is to start with white noise ( uncorrelated random numbers ) and to apply a filter whose power spectrum matches the desired correlation spectrum @xcite .
that is , suppose that we want colored - noise @xmath2 with some desired correlation function @xmath3 , the ( discrete - time ) fourier transform of which is the correlation spectrum @xmath4 .
we then start with white noise @xmath5 , usually uncorrelated gaussian random numbers , with zero mean and @xmath6 , and apply a filter @xmath7 : in frequency domain , @xmath8 .
the desired correlation spectrum is achieved if @xmath9 .
although this filtering operation can be performed entirely in the frequency domain via a fast fourier transform ( fft ) of the data sequences @xcite , such an approach is often too inefficient .
filtering entirely in the frequency domain requires the entire data sequence @xmath2 to be computed and stored in advance , and if many long sequences are required the storage becomes prohibitive .
the alternative is to perform the filtering in the time domain , using either finite - impulse response ( fir ) filters or infinite - impulse response ( iir ) filters ( also called recursive or arma filters ) . since the generation of colored noise is of interest to many researchers without a background in signal processing ,
let us briefly review the basics of fir and iir filtering , which are covered in more detail by numerous textbooks , e.g. ref . .
fir filters consist of a finite - length sequence @xmath10 that is convolved with the @xmath5 .
the fourier transform of the @xmath10 is the filter @xmath7 , but in general a finite - length sequence can only approximate an arbitrary desired spectrum @xmath4 .
in particular , an fir filter yields a spectrum @xmath7 which is a polynomial in @xmath11 with coefficients @xmath10 . a better approximation may be obtained by generalizing to a ratio of two polynomials ( i.e. , rational functions ) , which leads to iir filters .
an iir filter is determined by two finite - length sequences , @xmath12 ( @xmath13 , @xmath14 ) and @xmath10 ( @xmath15 ) , that determine @xmath2 via the recurrence ( which can be written in several equivalent forms ) : @xmath16 that is , @xmath2 is a convolution of @xmath10 with @xmath5 and of @xmath12 with the previous values of @xmath2 .
the filter design problem is then , given filter orders @xmath17 and @xmath18 , to find the @xmath12 and @xmath10 that best approximate the desired spectrum .
therefore , we must choose what type of filter to apply ( iir or fir ) and a filter - design method .
several choices have been previously proposed in the context of colored - noise generation .
the simplest method , as we mentioned above , is to just perform a fast fourier transform ( fft ) of the entire desired spectrum multiplied by random phases @xcite .
this is equivalent to an fir filter of the same length as the data , designed by the `` window '' method @xcite .
one can also employ fir filters of shorter lengths , designed by a variety of standard methods such as parks - mcclellan @xcite . for certain noise problems
, an fir filter may also be designed by analytical methods @xcite . in order to shorten the length of the required filter , and thus the memory and time requirements , iir
recursive filters have been proposed @xcite . in general , the design of iir filters is a difficult problem @xcite , and past approaches to colored noise generation by iir filtering have used local - optimization @xcite or yule - walker methods @xcite that are not guaranteed to yield the global - optimum filter coefficients .
another difficulty is that not all iir - filter design techniques are guaranteed to yield a stable filter ( one which does not lead to a diverging process ) .
one important exception is exponentially correlated noise , for which an exact first - order iir filter is known analytically and is commonly used ( although typically derived from a stochastic differential equation @xmath19 and not recognized as an iir filter _ per se _ ) @xcite . however , there is a key property of the iir filter design problem for colored - noise generation that makes optimal filter design practical : the phase of the filter @xmath7 is irrelevant , since it is multiplied in any case by white noise @xmath20 with a random phase .
in particular , we can exploit results by dudgeon @xcite and rabiner @xcite , who demonstrated that the _ global _ chebyshev - optimum magnitude - only minimum - phase stable iir filter - design problem can be efficiently solved by a sequence of linear - programming problems @xcite .
alternative methods with similar properties have also been proposed @xcite , and the dudgeon and rabiner technique was recently generalized to multidimensional iir filters @xcite . however
, its applicability to the problem of colored - noise generation does not seem to have been appreciated , and in this manuscript we demonstrate that it can yield dramatically more efficient filters than previous approaches . to demonstrate the efficacy of the various filter - design approaches for colored - noise generation , we consider an example drawn from thermodynamic simulations of gray - body thermal radiation @xcite . in this problem , thermal effects are modeled as random fluctuating current sources everywhere in space , with a correlation spectrum @xmath21 , for a constant @xmath22 determined by the temperature , based on the planck distribution .
this distribution is shown in fig . 1 , along with the periodogram of a finite - length correlated random sequence generated by the methods in this paper .
an iir filter is defined by a rational polynomial function in @xmath11 : @xmath23 where @xmath17 and @xmath18 define the filter order and @xmath24 for convenience .
a stable minimum - phase iir filter has all of its poles and zeros within the unit circle in @xmath25 ( i.e. , for @xmath26 ) @xcite .
given a sequence of uncorrelated random numbers @xmath5 , the output colored - noise sequence @xmath2 is then given by the recurrence ( eq .
1 ) above .
the required memory , along with the computation time per output , is therefore @xmath27 .
an fir filter is the special case @xmath28 .
a sequence @xmath2 of length @xmath29 from an fir filter can be generated in @xmath30 time instead of @xmath31 by use of fast fourier transforms ( via overlap - add or overlap - save techniques @xcite ) , but the memory requirements are not improved . ,
vs. @xmath32 for the planck distribution @xmath4 given in fig . 1 .
the chebychev error is plotted for six different filter methods : iir yule - walker ( red squares ) , fir parks - mcclellan ( black crosses ) , fir least - squares ( blue dots ) , iir global - optimum filter ( gray circles ) , and two nonlinear conjugate - gradient methods optimizing two different @xmath33 norms : ( a ) @xmath34 ( green diamonds ) and ( b ) @xmath35 ( magenta triangles ) . , scaledwidth=50.0% ] [ fig : abserr ] in fig . 2 , we plot the @xmath36 ( chebyshev ) error , @xmath37 , as a function of the total filter order @xmath32 , for filters @xmath7 designed by several techniques .
( here , we employ the chebyshev error over the entire frequency bandwidth ; for other applications , only a subset of the bandwidth may be of interest . ) the best method , i.e. smallest _ chebychev error _ for any given - order , is the optimal magnitude - only iir filter design ( implemented using the differential - correction algorithm @xcite ) . the other design methods plotted consist of two fir and two iir filter techniques .
the two linear - phase fir filter techniques are ( from the matlab signal - processing toolbox ) : first , the parks - mcclellan algorithm , which finds the global chebyshev optimum @xcite ( black ) ; second , a least - squares fir optimization ( blue ) , as described in ref . .
the two iir filter designs plotted are : first , a nonlinear conjugate - gradient minimization of a least - squares norm proposed by ref . , and suggested by ref . for use in generation of rayleigh - correlated noise ( green diamonds ) ; second , another ( non - global ) optimization technique based on the modified yule - walker algorithm in the matlab signal - processing toolbox @xcite .
the conjugate - gradient method only finds a local optimum , and is highly sensitive to the starting point of the optimization for this problem @xcite . here
, we use the yule - walker iir filter as the starting point , and conjugate gradient is able to improve upon it significantly ( in fact , it happens to nearly match the parks - mcclellan fir performance ) .
we examined the conjugate - gradient minimization of two different @xmath33 norms [ @xmath38 of the error : @xmath39 @xcite and @xmath35 , although it turns out to make little difference in the result .
error , @xmath34 , vs. @xmath32 for the planck distribution @xmath4 given in fig . 1 .
the @xmath34 error is plotted for six different filter methods : iir yule - walker ( red squares ) , fir parks - mcclellan ( black crosses ) , fir least - squares ( blue dots ) , iir global - optimum filter ( gray circles ) , and two nonlinear conjugate - gradient methods optimizing two different @xmath33 norms : ( a ) @xmath34 ( green diamonds ) and ( b ) @xmath35 ( magenta triangles).,scaledwidth=50.0% ] [ fig : l2a ] error , @xmath40 , vs. @xmath32 for the planck distribution @xmath4 given in fig . 1 . the @xmath35 error is plotted for six different filter methods : iir yule - walker ( red squares ) , fir parks - mcclellan ( black crosses ) , fir least - square ( blue dots ) , iir global - optimum filter ( gray circles ) , and two nonlinear conjugate - gradient methods optimizing two different @xmath33 norms : ( a ) @xmath34 ( green diamonds ) and ( b ) @xmath35 ( magenta triangles).,scaledwidth=50.0% ] [ fig : l2b ] we suspect that the chebyshev norm is typically the most appropriate one for physical simulations involving random processes .
the reason is that a large error in a narrow bandwidth , which might be allowed by a least - square ( @xmath33 ) norm , could result in a spectral feature that might be mistaken for a spurious physical phenomenon ; a large `` spike '' in error may also interact adversely with nonlinearities in the physics . nevertheless , in fig . 3 and fig . 4 ,
we show two different @xmath33 errors , @xmath34 and @xmath35 , for the same filter designs as in fig . 2 , and
the results demonstrate that the chebychev - optimal iir filter is at least comparable , and often superior to , the other methods , even in norms that it does not strictly optimize
. in particular , the @xmath35 norm of the chebyshev - optimum iir filter does better than all other local optimization techniques for most @xmath32 .
further gains could potentially be made , if this is the desired error norm , by starting with the chebyshev filter and then performing a local optimization by some other method . | we are grateful to alan v. oppenheim , stephen boyd , and almir mutapcic for helpful discussions .
this work was supported in part by the materials research science and engineering center program of the national science foundation under award dmr 02 - 13282 , by a department of energy ( doe ) computational science fellowship under grant de - fg02 - 97er25308 , and also by the paul e. gray undergraduate research opportunities program fund at mit .
y. k. chan and r. w. edsinger , `` a correlated random numbers generator and its use to estimate false alarm rates of airplane sensor failure detection algorithms , '' _ ieee trans . automatic control _
26 , pp . 676680 , 1981 .
l. r. rabiner , n. y. graham , and h. d. helms , `` linear programming design of iir digital filters with arbitrary magnitude function , '' _ ieee trans .
acoust , speech and sig .
_ , vol . 22 , no . 2 , pp .
117124 , 1974 . | grateful to alan oppenheim stephen boyd almir mutapcic discussions work supported materials research science engineering center program national science foundation award 02 - 13282 department energy computational science fellowship grant - 97er25308 paul. gray research fund. chan w edsinger correlated random numbers generator estimate false alarm rates airplane sensor failure detection algorithms automatic control 26 pp 1981. rabiner n y graham h d. helms linear programming design digital filters arbitrary magnitude function acoust speech vol 22 no . 2 pp 117124 1974 | 0.072792 | 238 | 118 | , Saving $0.0 in GPT-4. |
67,770 | the antarctic impulsive transient antenna ( anita ) completed its second long duration balloon flight in january 2009 , with 31 days aloft ( 28.5 live days ) over antarctica .
anita searches for impulsive coherent radio cherenkov emission from 200 to 1200 mhz , arising from the askaryan charge excess in ultra - high energy neutrino - induced cascades within antarctic ice .
this flight included significant improvements over the first flight in payload sensitivity , efficiency , and flight trajectory .
analysis of in - flight calibration pulses from surface and sub - surface locations verifies the expected sensitivity . in a blind analysis
, we find 2 surviving events on a background , mostly anthropogenic , of @xmath0 events .
we set the strongest limit to date for @xmath1 ev cosmic neutrinos , excluding several current cosmogenic neutrino models . | the existence of cosmic - ray particles of energies above @xmath2 ev , first established in the early 1960 s , has become a problem of the first rank in particle astrophysics .
models for their production must generate particle energies many orders of magnitude higher than achievable on earth , and these models in turn require extreme source physics that has not yet been formulated in a self - consistent manner . even more problematic , the propagation of such particles is limited by strong energy loss due to the greisen - zatsepin - kuzmin ( gzk ) process @xcite .
hadrons are produced via the delta - photoproduction resonance by interactions with cosmic microwave background ( cmb ) photons .
this gzk cutoff in energy limits the propagation distance of the ultra - high energy cosmic rays ( uhecrs ) to within 100 - 200 mpc in the current epoch , and severely distorts the observed energy spectrum .
astronomy using charged - particle uhecrs is thus limited to the local universe , suffering from both the loss of source spectral information , and the difficulty in back - tracing uhecrs through intergalactic magnetic fields . at distances of several gpc
, corresponding to the star - formation maximum at redshift @xmath3 , the higher energy and density of the cmb photons leads to even greater restrictions on cosmic - ray propagation , and a more rapid energy loss to delta photoproduction .
however , information about the source particles does survive in the form of secondary neutrinos in the decay chain , known as the ultra - high energy ( uhe ) cosmogenic neutrinos , first described by berezinsky & zatsepin ( bz ) @xcite .
their momenta are unaffected by magnetic fields , and they propagate without energy loss directly to earth , retaining information about the cosmic distribution of uhecrs and their sources .
the anita long duration balloon experiment was designed to search for cosmogenic neutrinos via electromagnetic cascades initiated by the neutrinos in antarctic ice , the most massive body of accessible , solid , radio - transparent dielectric material on earth .
we previously placed limits on the uhe cosmic neutrino flux from the first flight of anita @xcite , and provided a separate detailed description @xcite of the anita instrument , flight system , data acquisition , and analysis methods . in this article
we detail upgrades and augmentations beyond the instrument and methodology previously reported .
the second flight of anita ( anita - ii ) launched from williams field , antarctica on december 21 , 2008 and landed near siple dome after 31 days aloft , resulting in 28.5 live days .
[ payload08 ] shows an image of the payload on ascent after it had deployed to its full flight configuration , and an inset of the balloon and payload at float altitude . the mean ice depth in the field - of - view was 1.4 km , approximately one attenuation length at sub - ghz radio frequencies @xcite .
anita - ii flew at an altitude of 35 - 37 km above sea level ( 33 - 35 km above the ice surface ) , and was thus able to synoptically view a volume of @xmath4 m @xmath5 of ice .
anita - ii s sensitivity to cosmogenic neutrinos was improved substantially compared to anita - i : the front - end system noise temperature was reduced by 40 k , a 20% improvement in temperature @xcite ; 8 additional quad - ridged horn antennas were added to the previous total of 32 ; and the efficiency of the hardware trigger was optimized for impulsive signals . also , the instrument was made much more robust to the effects of bursts of anthropogenic radio - frequency ( rf ) interference with the ability to mask channels from the trigger in the azimuthal sectors of the payload pointing at the noise source .
masking occurred on time scales of a minute .
this upgrade significantly improved the livetime when in view of strong sources such as mcmurdo and amundsen - scott stations , and trigger thresholds remained at thermal - noise levels throughout the flight .
the combined effect of all of these modifications led to an increase of about a factor of four improvement in the expected signal from typical cosmogenic neutrino models as compared to anita - i .
8 meters , and each antenna face is 0.95 m across .
the inset shows the balloon and payload viewed telescopically at float altitude of 35 km.,width=273 ] the anita - ii hardware trigger selects impulsive radio signals with broadband frequency content and temporal coherence over nanosecond timescales in the vertical polarization .
the broadband nature of triggered signals was achieved by requiring power in multiple frequency bands , and temporal coherence was ensured by requiring impulsive power in a full - band channel .
all simulations and laboratory measurements of askaryan signals @xcite from ice sheet neutrino interactions show that the rf signal at the payload is predominantly vertically polarized , due to the surface fresnel coefficient and cherenkov geometry @xcite .
no strict pulse shape requirements are enforced in the trigger to allow for variations in the shape of any individual neutrino - induced cascade , and the trigger is thus very inclusive .
the trigger threshold rides at the ambient thermal noise level to maintain an approximately constant trigger rate of @xmath610 hz , which is dominated ( 98.5% ) by incoherent thermal noise fluctuations .
these thermal - noise triggers have well - modeled statistical probabilities for producing random impulses , and are highly suppressed in analysis with requirements of spatial and temporal coherence .
most other triggers are from anthropogenic sources .
such signals can mimic neutrino - like impulses , and may arise from high - voltage discharges in electrical equipment or from metallic structures charged by blowing snow or related effects . to remove such backgrounds , we identify active and prior human activity in antarctica and optimize pointing resolution to reliably associate anthropogenic signals with known sources .
we know of no expected particle - physics backgrounds .
extensive calibration and validation of the system response and trigger efficiency for a range of impulse signal - to - noise ratios ( snrs ) was done both prior to flight and with ground - to - payload impulse generating antennas during flight .
these in - flight pulser systems were located at the launch site and at a remote field station at taylor dome on the edge of the antarctic plateau .
in both cases , impulse generators were fed through an antenna immersed in the ice as deep as 90 m @xcite . in -
flight measurements of the impulses from taylor dome provided validation that refraction effects on signal propagation through the ice surface do not significantly affect the coherence of the received signal , to distances of 400 km .
measurements of trigger efficiency in - flight were consistent with expectations from the ground calibrations , considering the narrower frequency content of impulses from taylor dome .
ground - to - payload signals provided an equally critical function in detector alignment and in determining the precision of directional reconstruction .
anita s antenna signals are combined via pulse - phase interferometric methods @xcite , resulting in a radio map ( `` interferometric image '' ) for each polarization of the intensity as a function of payload elevation and azimuth .
the largest peak in either map determines the direction of the signal source .
the payload coordinate frame is tied to the geodetic frame via onboard gps and sun - sensor measurements to a precision of about 0.1@xmath7 .
the pointing resolution in our analysis is @xmath8 in elevation and @xmath9 in azimuth , depending on the snr of the event .
our signal region comprises events that do not come from known sources of human activity ( `` camps '' ) and have no geospatial partner event within distance and angular separation criteria .
we blinded the signal region in two ways simultaneously .
first , we inserted an undisclosed number ( 12 ) of neutrino - like calibration events at random , concealed times throughout the flight .
second , we blinded ourselves to events in the signal region _
i.e. _ events that have no spatial partner . the analysis efficiency and background
are estimated before opening this hidden signal box .
prior to unblinding , the complete set of in - flight calibration pulser data is used to optimize the pointing reconstruction of the analysis . in this section ,
we describe the sequence of analysis cuts that we applied to the sample until the final step , opening the blind signal box .
table [ event - table ] shows the the total event sample , the sequential number of surviving events as each analysis cut is applied , and the efficiency of each cut .
the number of * hardware - triggered events * is the total number of events recorded .
* quality events * exclude on- and off - payload calibration pulser events , unbiased auto - trigger events , and a small fraction ( @xmath61% ) of events with data corruption .
99 k. greisen , phys .
16 , 748 ( 1966 ) ; g. t. zatsepin , v. a. kuzmin , jetp lett . 4 , 78 ( 1966 ) .
v. s. beresinsky , g. t. zatsepin , phys .
b 28 , 423 ( 1969 ) .
anita collab . : p. w. gorham _
et al . _ ,
103 , 051103 ( 2009 ) .
anita collab .
: p. w. gorham _
et al . _ ,
32 , 10 ( 2009 )
. s. barwick _
et al . _ , j. glaciol .
51 , 231 ( 2005 ) .
g. a. askaryan , jetp 14 , 441 ( 1962 ) ; jetp 21 , 658 ( 1965 ) .
j. alvarez - muniz , e. zas , phys .
b 411 , 218 ( 1997 ) .
d. saltzberg _ et al .
lett . , 86 , 2802 ( 2001 ) .
anita collab . : p. w. gorham _
et al . _ ,
99 , 171101 ( 2007 ) .
r. engel , d. seckel , t. stanev , phys . rev .
d 64 , 093010 ( 2001 ) .
a. vieregg , ucla ph.d .
dissertation ( 2010 ) .
bedmap consortium : m. b. lythe , d. g. vaughan _
et al . _ , j. geophys .
106 , b6 , 11335 ( 2001 ) .
anita collab : s. hoover _
_ , to be submitted to phys . rev .
, arxiv:1005.0035 .
g. feldman , r. cousins , phys .
d 57 , 3873 ( 1998 ) ; j. conrad , _ et al .
_ , phys . rev .
d 67 , 012002 ( 2003 ) .
_ , astropart .
, 5 , 81 ( 1996 ) . l. a. anchordoqui _ et al . _ ,
d 66 , 103002 ( 2002 ) .
n. lehtinen _
_ , phys . rev .
d 69 , 013008 ( 2004 ) .
s. hoover , ucla ph.d .
dissertation ( 2010 ) .
icecube collab . :
m. ackermann _
j. 675 , 1014 ( 2008 ) .
i. kravchenko , _ et al .
_ , phys . rev .
d 73 , 082002 ( 2006 ) .
pierre auger collab . : j. abraham _
et al . _ ,
100 , 211101 ( 2008 ) .
hires collab . :
r. u. abbassi _
et al . _ ,
j. 684 , 790 ( 2008 ) .
r. j. protheroe , p. a. johnson , astropart .
phys . 4 , 253 ( 1996 ) .
o. e. kalashev _ et al .
d 66 , 063004 ( 2002 ) .
o. e. kalashev _
et al . _ ,
d 65 , 103003 ( 2002 ) .
c. aramo , _ et al .
_ , astropart .
23 , 65 ( 2005 ) .
_ , astropart .
23 , 19 ( 2005 ) . v. barger , p. huber , d. marfatia , phys .
b 642 , 333 ( 2006 ) .
t. stanev _
_ , phys . rev .
d 73 , 043003 ( 2006 ) .
v. berezinsky , arxiv : astro - ph/0509675 ( 2005 ) .
h. yuksel , m. d. kistler , phys . rev .
d 75 , 083004 ( 2007 ) . s. yoshida _
et al . _ ,
j. 479 , 547 ( 1997 ) .
e. waxman , j. bahcall , phys .
d 59 , 023002 ( 1999 ) . | cosmic - ray particles above @xmath2 ev established 1960 s problem in particle astrophysics models production generate energies higher earth require extreme source physics propagation limited by energy loss due to greisen - zatsepin - kuzmin gzk process @xcite hadrons produced delta - with cosmic microwave photons gzk cutoff limits propagation ultra - high energy cosmic rays to 100 - 200 mpc distorts energy spectrum astronomy using charged - particle uhecrs limited to local universe loss of source spectral information difficulty back - tracing uhecrs through intergalactic magnetic fields at distances several gpc star - formation maximum at redshift @xmath3 higher energy density cmb photons restrictions propagation rapid energy loss to delta photoproduction information source particles secondary neutrinos ultra - high energy cosmogenic neutrinos described by berezinsky & zatsepin @xcite unaffected by magnetic fields propagate without energy loss to earth retaining information about cosmic distribution sourcesanita balloon experiment for cosmogenic neutrinos via cascades antarctic ice massive - transparent dielectric material limits on cosmic neutrino flux from first flight description of anita instrument flight system data acquisition analysis methods article upgrades augmentations beyond methodology second flight anita ii launched from williams field antarctica december 21 2008 landed near siple dome after 31 days 28.5 live days payload08 shows image payload on ascent balloon payload at float altitude mean ice depth 1.4 km one attenuation length at sub - ghz radio frequencies anita - ii flew at 35 - 37 km above sea level 33 - 35 km above ice surface synoptically view volume m of ice - ii s sensitivity to cosmogenic neutrinos improved front - end system noise temperature reduced by 40 k 20% improvement 8 quad - ridged horn antennas added 32 efficiency hardware trigger optimized for impulsive signals instrument more robust to anthropogenic radio - frequency interference mask channels from trigger in payload noise source masking on time scales minuteupgrade improved livetime strong sources mcmurdo amundsen - scott stations trigger thresholds at thermal - noise levels flight factor four improvement expected signal from cosmogenic neutrino models anita - i 8 meters antenna face 0.95 m across shows balloon payload viewed telescopically at float altitude 35 km anita - ii trigger selects radio signals broadband frequency temporal coherence nanosecond vertical polarization broadband power multiple frequency bands temporal coherence power full - band channel simulations measurements neutrino show rf signal at payload vertically polarized surface fresnel coefficient cherenkov geometry no strict pulse shape requirements trigger inclusive trigger threshold rides at ambient thermal noise level constant trigger rate @xmath610 hz dominated 98.5% by incoherent thermal noise fluctuations thermal - noise triggers random impulses suppressed in analysis spatial temporal coherence other triggers from anthropogenic sources signals mimic neutrino - like impulses arise from high - voltage discharges electrical equipment metallic structures charged by blowing snowidentify active human activity in antarctica optimize pointing resolution associate anthropogenic signals with known sources no expected particle - physics backgrounds calibration validation system response trigger efficiency for impulse signal - to - noise ratios done prior flight ground - - payload impulse antennas during flight pulser systems located at launch site field station taylor dome antarctic plateau impulse generators fed through antenna immersed in ice 90 m measurements refraction effects affect coherence signal distances 400 km trigger efficiency consistent with ground calibrations narrower frequency content impulses ground - to - payload signals detector alignment precision directional reconstruction antenna signals combined via pulse - phase interferometric methods radio map interferometric for each polarization intensity payload elevation azimuth largest peak determines direction signal source payload coordinate frame tied to geodetic frame via onboard gps sun - sensor measurements precision 0.1@xmath7 pointing resolution @xmath8 elevation @xmath9 azimuth depending snr event signal region comprises events from known sources human activity no geospatial partner event within distance angular separation criteriablinded signal region inserted undisclosed 12 neutrino - like calibration events random flight blinded ourselves to events no spatial partner analysis efficiency background estimated before opening hidden signal box complete in - flight calibration pulser data used pointing reconstruction sequence analysis cuts until opening blind signal box shows total event sample surviving events efficiency hardware - triggered events total recorded quality events exclude on- off - payload calibration pulser events unbiased auto - trigger events small data corruption k. greisen , phys . 16 , 748 ( 1966 ) g t. zatsepin v. kuzmin jetp 4 , 78 ( 1966 ). beresinsky zatsepin phys 28 , 423 ( 1969 ) anita collab p. w. gorham 103 , 051103 ( 2009 ) 32 , 10 ( 2009 ) barwick glaciol . 51 , 231 ( 2005 ) askaryan jetp 14 , 441 ( 1962 ) jetp 21 , 658 ( 1965 ) j. alvarez - muniz e. zas , phys .411 , 218 1997 ) saltzberg 86 , 2802 2001 ) anita gorham 99 , 171101 ( 2007 ) engel seckel stanev . rev 64 , 093010 ( 2001 ) vieregg ucla ph.d dissertation 2010 ) bedmap consortium b lythe vaughan geophys. 106 , b6 , 11335 ( 2001 ) anita s. hoover phys. rev arxiv:1005.0035 cousins phys . 57 , 3873 ( 1998 ) j conrad phys rev 67 , 012002 ( 2003 astropart 5 , 81 ( 1996 ) anchordoqui 66 , 103002 ( 2002 ) lehtinen phys. 69, 013008 ( 2004 ) s. hoover ucla ph.d dissertation 2010 icecube m. 675 , 1014 ( 2008 ) kravchenko phys . rev. 73 , 082002 ( 2006 pierre auger j. abraham 100 , 211101 ( 2008 ) abbassi 684 , 790 ( 2008 ) protheroejohnson astropart phys . 4 , 253 1996 ) 66 , 063004 ( 2002 ) 65 , 103003 2002 ) aramo astropart 23 , 65 ( 2005 ), 19 2005 barger huber marfatia phys 642 , 333 ( 2006 ) stanev . 73, 043003 ( 2006 ) berezinsky astro - ph/0509675 ( 2005 ) yuksel kistler phys . 75 , 083004 ( 2007 ) yoshida 479 , 547 ( 1997 ) waxman bahcall . 59 , 023002 ( 1999 ) | 0.437508 | 2,905 | 1,454 | , Saving $0.1 in GPT-4. |
67,773 | both the choice made by the observer and consciousness are discussed in terms of cyclical time .
that is , while the process of classical choice evolves forward in time , that is , from @xmath0 to @xmath1 , the quantum reference frame evolves backward in time , that is , from @xmath1 to @xmath0 , to equate itself with the classical choice made by the observer , such that at the end , this corresponds to the case of self - observation in consciousness in linear time .
this indicates that discrete and finite information is accompanied by a continuous or infinite `` semantic '' quantum part .
in particular , the continuous semantic aspect is considered to be related to universal grammar , a suggested innate structure in languages .
this paper also argues that the cyclical time model can be considered to have both small and large cycles and will also argue that at the most basic level , consciousness is strongly connected to time .
this means that another aspect is added , that is , a more detailed description of the ongoing proposal of the subjective model , in which the classical is just as fundamental as the quantum part . | one of the biggest mysteries in physics has been the apparent discrepancy between classical and quantum theories . this is because quantum theory involves many elements not seen in classical physics , such as wave / particle duality , superposition , entanglement , nonlocality etc .
@xcite . there is also a historical reason for this confusion , because quantum theory was developed in the early twentieth century , and its descriptions of certain behaviors were presumed to replace many previously considered classical behaviors .
this explains why one of the largest efforts in the theoretical physics research community is now being directed toward trying to come up with a quantum version of classical theory , namely , gravitation @xcite . however , contradictory to the common perception , the classical aspect is also an integral part of quantum theory .
that is , measurement in quantum theory is provided in classical space - time through the eigenvalue outcome .
this view is also consistent with the current difficulty of coming up with a quantum theory of gravity regardless of decades of effort . in @xcite ,
there has been a suggestion that reality ought to be subjective .
this suggestion has been drawn from the argument of consciousness , that is , that only in the case of consciousness are the object and the observer the same .
this phenomenon is unique , considering the observation or measurement results from the relative difference between the object being observed and the observer .
therefore , it has been suggested that because of consciousness , one should stop considering the observer and the object as separate entities .
in @xcite , more detailed argument was provided as to what it means that the observer and the object are inseparable .
it was argued that given equally accessible microstates , the observer s freely chosen will corresponds to the reality : @xmath2 note that there have been a number of discussions suggesting the relation between entropy and gravity @xcite .
therefore , if we assume the entropy in terms of the observer s choice , as in @xcite , this implies the relation of gravity or einstein s equation to ( [ reality ] ) .
recently , @xcite has introduced the concept of cyclical time with consciousness . in the cyclical model ( see @xcite for a review ) , time proceeds from @xmath0 to @xmath1 , from @xmath1 to @xmath3 , from @xmath3 to @xmath4 etc . , and at a certain @xmath5 , @xmath6 proceeds back to @xmath0 ( figure [ cycle ] ) . in particular , cyclical time
was discussed in terms of self - reference in logic and self - observation in consciousness .
it was argued that a paradoxical or noncomputable element may be removed in the cyclical model . in the following
, self - observation will be discussed in terms of cyclical time , particularly using both classical and quantum elements . to @xmath1 , from @xmath1 to @xmath3 , etc .
but at some @xmath7 , @xmath6 proceeds back to @xmath0 .
the figure shows the simplest case , where @xmath8 .
, title="fig : " ]
the cyclical time model may seem strange and counterintuitive , since we are used to time moving forward only ; thus , it seems strange that time circulates .
nevertheless , it will be illustrated that this cyclical model is in fact consistent not only in explaining the experience of time but also in terms of consistency between classical and quantum theories .
the argument in @xcite was that classical probability theory may be considered equivalent to quantum theory , except for the continuity part .
indeed , when the axioms of classical theory are compared with the axioms of quantum theory , the difference between the two is the extra continuity axiom in quantum theory .
meanwhile , @xcite argued that one can exploit this idea such that classical probability theory can be considered to involve a classical choice made by the observer , while continuity is associated with the choice of quantum reference frame , an observable .
below , it will be discussed more explicitly what is meant by a classical choice made by the observer accompanied by continuity , or a dirac - type negative sea in the cyclical model .
-direction , and the classical reference frame @xmath9 is also positioned similarly .
so ( i ) the observer chooses @xmath10 with the classical vector @xmath9 , where time proceeds from @xmath0 to @xmath1 .
and ( ii ) in the cyclical time model , time @xmath1 then proceeds back to @xmath0 , while the quantum reference frame @xmath11 evolves to equate itself with @xmath9 .
, title="fig : " ] on the other hand , @xcite showed that the cyclical model can be adopted in association with consciousness . before moving on to discuss the cyclical model ,
let us first review self - observation in linear time .
in general , quantum measurement involves a state vector and a reference frame . in the schrdinger picture
, it would be the state vector that is rotated , and in the case of the heisenberg picture , the reference frame is rotated counterclockwise . in both cases , the observation would be the same . in the case of self - observation ,
the state vector , the object , is the same as the reference frame , or the observable . in that case , in the schrdinger picture , the reference frame
would be rotated , whereas in the heisenberg picture , the reference frame would be rotated counterclockwise by the same amount .
the two are not generally equivalent @xcite .
therefore , self - observation is not computable in a linear time model .
a cyclical model may be initially considered with a reference frame , a classical one .
that is , the vector is defined in classical space - time rather than in complex hilbert space , as in quantum theory .
one could imagine this classical vector as follows : when one performs a measurement of a quantum state , the state vector and the observable , or the reference frame , are defined in complex vector space .
however , in order to carry out the actual measurement , the objects and thereference frame are needed not only in quantum space but also in classical space as well ( figure [ sea ] ) .
in particular , in order to discuss self - observation in the cyclical model , the objects in quantum and classical spaces are not needed , but only the reference frames , that is to say , the classical reference frame @xmath9 and the quantum reference frame @xmath11 . from @xmath0 to @xmath1 , the classical reference frame @xmath9 points at @xmath10 ( figure [ time ] ( i ) ) .
a cyclical model is now considered where @xmath1 proceeds back to @xmath0 . in this case
, the quantum reference frame @xmath11 evolves from 0 to @xmath10 , such that it aligns itself with the classical reference frame @xmath9 ( figure [ time ] ( ii ) ) .
the above argument may be understood as follows .
when the observer makes a choice of @xmath10 , it is done in classical space . however , understanding or being aware of the choice occurs in the quantum domain . in other words , from @xmath0 to @xmath1 , the observer observes something to be at @xmath10 , that is , the observer chooses @xmath10 . from @xmath1 to @xmath0 , that something was the observer s own reference frame .
in this section , we wish to discuss how small time cycles relate to the large cycle in the cyclical model ( figure [ time1 ] ) .
let us begin with a quantum reference frame @xmath11 and the classical reference frame @xmath9 , which is represented in classical space equivalent to the @xmath12-direction of the bloch sphere of a qubit .
as before , time moves forward , and the classical choice made is @xmath13 ; the quantum reference frame also evolves to equate itself with @xmath13 but with time moving backwards .
that would be the first case .
the second small cycle may be considered with the classical vector @xmath9 pointing at @xmath14 with time moving forward .
the reference frame @xmath11 evolves backward in time so that it equates itself with the classical reference frame at @xmath14 ( figure [ cyclic2 ] ( i ) ) .
is positioned at @xmath13 with time moving forward , and then the quantum reference frame equates itself with the classical vector at @xmath13 with time circulating back , which completes the first small cycle
. then the vector @xmath9 chooses @xmath14 with time moving forward , and eventually , the quantum reference frame equates itself with @xmath9 with time circulating back , which completes the second small cycle .
( ii ) a large time cycle is achieved with the evolution of another quantum reference frame , @xmath15 , to equate itself with @xmath9 .
, title="fig : " ] let us now consider a larger cycle with the introduction of another quantum reference frame , @xmath15 , initially pointing in the @xmath12-direction .
the larger cycle is accomplished with the evolution of @xmath15 to equate itself with @xmath9 at @xmath16 ( figure [ cyclic2 ] ( ii ) ) . in order to further elaborate on the relation between small and large cycles ,
let us consider a special case where @xmath17 that is , the observer is not choosing anything but is only experiencing time evolution in a cyclical manner .
in such a case , the observer is conscious of time in small cycles as well as in a larger time cycle .
therefore , the experience of time evolution , such as seconds ( small cycle ) and minutes ( large cycle ) occurs as the observer perceives these cycles . at the fundamental level
, this subjective nature of time may be stated as follows : @xmath18 the experience of time is perceived by a dirac - type negative sea or consciousness .
in this paper , we have discussed the relation between classical choice and the continuous or infinite semantics of the choice in terms of quantum theory . in particular , it is noted that the cyclical model is consistent with our previous discussion @xcite on subjective reality and universal grammar , which argues that all languages have a universal structure that is not learned but is innate in human nature @xcite .
that is , based on the argument presented in this paper , the innateness or universal grammar may be related to this continuous or infinite part associated with formal language , represented through the finite classical choice made by the observer .
we have also argued that the association of infinity or continuity with classical choice provides a more specific description of ongoing efforts in the subjective model , particularly regarding the subjective nature of time .
we have shown the connection of consciousness with time in a way such that the most fundamental experience of consciousness is in fact associated with time .
it is interesting to note the relation between small and large cycles of time using the proposed subjective model .
chomsky n. aspects of the theory of syntax , mit press , 1965 . chomsky n. rules and representations , new york : columbia university press , 1980 .
cook vj and newson m. chomsky s universal grammar : an introduction , wiley - blackwell ; 3rd edition , 2007 .
green mb , schwarz jh , witten e. superstring theory .
cambridge university press , cambridge , 1987 .
green mb , schwarz jh , witten e. superstring theory : volume 2 , loop amplitudes , anomalies and phenomenology .
cambridge university press , 2012 .
hardy l. quantum theory from five reasonable axioms .
arxiv : quant - ph/0101012 .
jacobson t. thermodynamics of space - time : the einstein equation of state .
phys rev lett 1995 ; 75 : 1260 - 1263 .
padmanabhan t. thermodynamical aspects of gravity : new insights .
rep prog phys 2010 ; 73 : 046901 .
peres a. quantum theory : concepts and methods , kluwer , dordrecht , 1997 .
reynolds m. axiomatisation and decidability of f and p in cyclical time .
j phil logic 1994 ; 23 : 197 - 224 .
song d. non - computability of consciousness .
neuroquant 2007 ; 5 : 382 - 391 .
arxiv:0705.1617 [ quant - ph ] .
song d. einstein s moon .
phys usp 2012 ; 55 : 942 - 943
. arxiv:1008.2892 [ physics.gen-ph ] .
song d. physical version of singularity resolution in the observable universe .
arxiv:1602.01450 [ physics.gen-ph ] .
song d. aspects of reference and information . submitted to neuroquant 2016 .
verlinde ep . on the origin of gravity and the laws of newton .
j high energy phys 2011 ; 1104 : 029 . | biggest mysteries in physics apparent discrepancy between classical and quantum theories quantum theory involves elements not seen in classical physics wave / particle duality superposition entanglement nonlocality historical reason for confusion quantum theory developed in early twentieth century descriptions presumed to replace classical behaviors explains largest efforts theoretical physics now directed toward quantum version of classical theory gravitation contradictory to classical aspect integral part of quantum theory measurement in quantum theory provided in classical space - time through eigenvalue outcome view consistent with current difficulty of quantum theory of gravity regardless decades effort suggestion that reality subjective suggestion drawn from argument of consciousness object and observer same unique observation measurement results from relative difference between object and observer suggested should stop considering observer and object as separate entities more detailed argument provided observer and object inseparable argued given equally accessible microstates observer s freely chosen will corresponds to reality discussions suggesting relation between entropy and gravityassume entropy observer s choice in @xcite implies relation of gravity einstein s equation to reality @xcite introduced concept cyclical time with consciousness cyclical model @xcite time proceeds from @xmath0 to @xmath1 to @xmath3 to @xmath4 @xmath5 @xmath6 proceeds back to @xmath0 cyclical time discussed self - reference in logic self - observation in consciousness paradoxical noncomputable element removed in cyclical model self - observation discussed cyclical time using classical quantum elements to @xmath1 @xmath1 to @xmath3 @xmath7 @xmath6 proceeds back to @xmath0 figure shows simplest case @xmath8 cyclical time model seem strange counterintuitive time moving forward time circulates cyclical model consistent explaining experience of time consistency between classical and quantum theories argument in @xcite classical probability theory equivalent to quantum theory except continuity part axioms classical theory compared quantum theory difference is extra continuity axiom in quantum theory@xcite argued exploit idea classical probability theory involve classical choice observer continuity associated with choice quantum reference frame observable discussed classical choice observer accompanied by continuity or dirac - type negative sea in cyclical model -direction classical reference frame @xmath9 positioned similarly observer chooses @xmath10 with classical vector @xmath9 time proceeds from @xmath0 to @xmath1 cyclical time model time @xmath1 proceeds back to @xmath0 quantum reference frame @xmath11 evolves equate with @xmath9 @xcite showed cyclical model adopted association with consciousness before model review self - observation in linear time quantum measurement involves state vector reference frame schrdinger picture state vector rotated heisenberg picture reference frame rotated counterclockwise both observation same self - observation state vector object same as reference frame observable schrdinger picture reference frame rotated heisenberg picture rotated counterclockwise same not generally equivalent self - observation not computable in linear time model cyclical model may initially considered with reference frame classical onevector defined in classical space - time complex hilbert space quantum theory classical vector measurement quantum state state vector observable reference frame defined in complex vector space measurement objects thereference frame needed in quantum classical space ( figure [ self - observation in cyclical model objects in quantum classical spaces not needed only reference frames classical reference frame @xmath9 and quantum reference frame @xmath11 from @xmath0 to @xmath1 classical reference frame @xmath9 points at @xmath10 ( figure cyclical model considered @xmath1 proceeds back to @xmath0 quantum reference frame @xmath11 evolves from 0 to @xmath10 aligns with classical reference frame @xmath9 observer choice of @xmath10 in classical space understanding choice occurs in quantum domain from @xmath0 to @xmath1 observer observes something at @xmath10 chooses @xmath10 from @xmath1 to @xmath0 something observer s own reference frame discuss small time cycles relate to large cycle in cyclical model figurebegin quantum reference frame @xmath11 classical reference frame @xmath9 classical space equivalent @xmath12-direction bloch sphere qubit time moves forward classical choice @xmath13 quantum reference frame evolves equate @xmath13 time moving backwards first case second small cycle classical vector @xmath9 @xmath14 time moving forward reference frame @xmath11 evolves backward equates with classical @xmath14 figure positioned @xmath13 time moving forward quantum reference frame equates classical vector @xmath13 time circulating back completes first small cycle vector @xmath9 chooses @xmath14 time moving forward quantum reference frame equates with @xmath9 time circulating back completes second small cycle large time cycle achieved evolution quantum reference frame @xmath15 equate @xmath9 consider larger cycle another quantum reference frame @xmath15 @xmath12-direction larger cycle evolution @xmath15 equate with @xmath9 @xmath16 figure cyclic2 ( ii elaborate relation small large cycles consider special case @xmath17 observer not choosing experiencing time evolution cyclical observer conscious of time small cycles larger time cycleexperience of time evolution seconds minutes occurs observer perceives cycles subjective nature of time stated experience time perceived by dirac - type negative consciousness discussed relation between classical choice continuous infinite semantics quantum theory cyclical model consistent with previous discussion on subjective reality universal grammar languages have universal structure not learned innate in human nature innateness universal grammar related to continuous infinite part with formal language represented through finite classical choice association of infinity continuity with classical choice provides specific description of ongoing efforts subjective model nature time shown connection of consciousness with time fundamental experience associated with time relation between small large cycles of time using proposed subjective model chomsky n. theory of syntax mit press 1965 rules representations new columbia university press 1980 cook vj newson m chomsky s universal grammar introduction wiley - blackwell 3rd 2007 green mb schwarz jh witten e superstring theory cambridge university press 1987 superstring theory volume 2 loop amplitudes anomalies phenomenology cambridge university press 2012 hardy l. quantum theory from five reasonable axiomsquant - ph/0101012 jacobson thermodynamics space time einstein equation phys 1995 75 1260 - 1263 padmanabhan thermodynamical gravity phys 2010 73 046901 peres quantum theory reynolds axiomatisation f p cyclical time phil logic 1994 23 197 - 224 song non - computability consciousness neuroquant 2007 5 382 - 391 arxiv:0705.1617 quant - ph song einstein s moon phys 2012 55 : 942 - 943 physics.gen-ph singularity resolution universe arxiv:1602.01450.gen-ph reference information neuroquant 2016 verlinde origin gravity laws newton high energy phys 2011 1104 : 029 | 0.437839 | 2,768 | 1,391 | , Saving $0.1 in GPT-4. |
67,774 | we review the observational evidence for geodetic precession in psr b1913 + 16 and present the latest observations and results from modelling the system geometry and beam .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in | after the discovery of psr b1913 + 16 by hulse & taylor in 1974 ( hulse & taylor 1975 ) , it was immediately realized that this system represents a highly stable and accurate clock orbiting in the gravitational field of a compact star .
psr b1913 + 16 has indeed fulfilled all promises by finally allowing tests of theories of gravity in the strong field limit which can not be realized in the solar system ( see weisberg , these proceedings ) .
indeed , no participant at this conference needs to be reminded about the role which this pulsar has played in the confirmation of the existence of gravitational waves . while these tests are based on timing observations , damour & ruffini ( 1974 ) pointed out very soon after the discovery of this pulsar , that by studying also its emission properties one can test another prediction made by general relativity , ie .
that of geodetic precession . in general relativity ,
the proper reference frame of a freely falling object suffers a precession with respect to a distant observer , called geodetic precession . in a binary pulsar system this geodetic precession leads to a relativistic spin - orbit coupling , analogous of spin - orbit coupling in atomic physics . as a consequence , the pulsar spin precesses about the total angular momentum , changing the relative orientation of the pulsar towards earth .
due to such a change in geometry , we should also expect a change in the radio emission received from the pulsar .
the precession rate ( e.g. boerner et al . 1975 ) depends on the period and the eccentricity of the orbit as well as the pulsar and companion mass . as joel weisberg demonstrates in these proceedings ,
all these values can be obtained accurately from timing observations . with these
we obtain a precession rate of @xmath0 deg yr@xmath1 .
since the orbital angular momentum is much larger than the pulsar spin , the orbital spin practically represents a fixed direction in space , defined by the orbital plane of the binary system .
given the calculated precession rate , it takes 297.5 years for the pulsar spin vector to precess around it . as a result of the precession the angle between the pulsar spin axis and
our line - of - sight should change with time , so that different portions of the emission beam are observed .
consequently , one expects changes in the measured pulse shape , in particular in the profile width , as a function of time . in the extreme case
, the precession may move the beam out of our line - of - sight and the pulsar may disappear from the sky until it becomes visible again .
the pulse profiles were naturally studied closely in order to detect possible changes .
finally , weisberg et al .
( 1989 , hereafter wrt89 ) , discovered a change in the relative amplitude of the two prominent profile components . while these changes can indeed be considered as the first signs of the effects of geodetic precession , a change in the component separation or profile width as expected from a cone - like pulsar beam was not detected .
due to precession , the distance of our line - of - sight to the magnetic axis should also change with time , so that a change in the position angle ( pa ) swing of the linearly polarised emission component is expected .
cordes , wasserman & blaskiewicz ( 1990 , hereafter cwb90 ) studied polarisation data to compare profiles and pa swings obtained from 1985 to 1988 .
cwb90 did neither detect very clear changes in the pulse shape , nor could they find any significant change in the pa swing .
cwb90 pointed out , however , that the existence of a core component , which is very prominent at lower frequencies , complicates the interpretation of the polarisation data .
they noted similar to wrt89 that the core may also be responsible for the change in relative component amplitude with time .
psr b1913 + 16 is also monitored with the 76-m lovell telescope and with the 100-m effelsberg telescope .
the analysis of effelsberg profiles measured between 1994 and 1998 by kramer ( 1998 ) revealed that the profile components were still changing their relative amplitude , consistent with the rate first determined by wrt89 .
even more interesting , however , was the first detection of changes in the separation of the components . in order to model this long - expected decreasing width of the profile ,
two simple assumptions were made , i.e. those of a circular hollow cone - like beam and the precession rate as predicted by general relativity .
both assumptions lead to a model which has only four free parameters : the misalignment angle @xmath2 between the pulsar spin and the orbital angular momentum , the inclination angle between the pulsar spin axis and its magnetic axis , @xmath3 , the radius of the emission beam , @xmath4 , and the precession phase given by the reference epoch @xmath5 . with the post - keplerian parameters measured by pulsar timing
, general relativity allows one to compute the value of @xmath6 , i.e. the sine of the orbital inclination angle . for psr b1913 + 16 ,
we compute a value of @xmath7 , or equivalently @xmath8 whereas the ambiguity can not be resolved from timing alone .
the best fit of this model therefore allows four equivalent solutions .
one pair of solutions corresponds to @xmath7 , the other pair to @xmath9 , respectively .
the remaining choice is given by the unknown relative orientation of the pulsar spin and the orbital angular momentum , i.e. as to whether the pulsar rotation is pro - grade or retro - grade .
it can be argued that a retro - grade case is less likely ( kramer 1998 ) , so that the polarisation information can be used to separate the remaining two solutions , as only one gives the correct observe sense of pa swing .
the finally obtained misalignment angle of @xmath10 obtained by kramer ( 1998 ) is in excellent agreement with earlier simulations by bailes ( 1988 ) who studied the effects of asymmetric supernova explosions and predicted @xmath11 as a typical value for psr b1913 + 16-like systems .
the obtained best fit also lead to the prediction that the pulsar will disappear from the sky around the year 2025 ! moreover , it also implies that the component separation remains almost unchanged for about 60 yr .
it is now easy to understand why wrt89 were not lucky to detect changes in the component separation .
similarly , computing the change in pa swing which had to be measured by cwb90 for a positive detection of a geometry change , produces a value which is only slightly larger than their estimated detection limit .
it should also be noted that based on an emission model and the relative change in component ratio alone , istomin ( 1991 ) also suggested a disappearance of the pulsar around 2020 .
the full model as presented by kramer ( 1998 ) also predicts a reappearance around the year 2220 .
psr b1913 + 16 will , in total , only be observable for about a third of the precession period .
while this seems to affect possible detection rates of double neutron star systems and hence the detection rate of gravitational wave detectors like ligo or geo600 , averaged over time existing numbers do not change , as discussed in more detail by kramer ( 1998 , 2002 ) .
in figure 1a we show the latest measurements for the component separation as a function of time , demonstrating that the profile continues to narrow as predicted by the model .
weisberg & taylor ( 2000 , 2002 , also these proceedings ) also obtained new measurements after the arecibo upgrade , confirming the results reported above .
thanks to the superior sensitivity of the telescope , they did not only measure the same decrease in component separation , but could also measure a general decrease in profile width at several intensity levels . using these data
they derive a geometry which is in agreement with that of kramer ( 1998 ) , and they also obtain a map of a pulsar emission beam for the first time . since our line - of - sight moves through the emission beam , each profile represents a slightly different cut through the beam structure . during their data analysis , weisberg &
taylor separate the measured profiles into odd and even parts and use the width information for all intensity levels of the even profiles , combined with a mapping function , to derive a model for geometry and beam shape .
the results of this mapping process are surprising as the beam seems not only to be elongated but even hour - glass shaped .
we use our data for an complementary , alternative approach .
we propose to use the original profiles in order to maintain information about the features causing the profile asymmetry , namely the off - set core .
we then compute profiles observable at different epochs taking full spherical geometry into account , and compare these model profiles to the observed data ( see figure 1b ) . while this work is still in progress
, initial results suggest that the beam may indeed be slightly elongated although an hour - glass beam shape may not be necessary to explain the data .
further modelling will be necessary but it may provide us with a beam map which can then be used as input for tests of the precession rate .
the observations of other precessing pulsars like psr b1828@xmath1211 ( stairs et al .
2000 ) or perhaps psr b1931 + 24 ( kramer et al . in prep . ) may help in this process to understand the pulsar beam pattern . at the moment ,
the detection of effects of geodetic precession in psr b1913 + 16 is a successful qualitative test of general relativity , but it may become possible to perform also a quantitative test by measuring the precession rate using our beam models .
bailes , m. , 1998 , a & a , 202 , 109 boerner , g. , _ et al _ , 1975 , a & a , 44 , 417 cordes , j.m . , _ et al .
_ , 1990 , apj , 349 , 546 , ( cwb90 ) damour , t. , ruffini , r. , 1974 , academie des sciences paris comptes rendus ser.scie.math . , 279 esposito , l.w .
, harrison , e.r . , 1975 , apjl , 196 , l1 hulse , r.a . , taylor , j.h . , 1975 ,
apj , 195 , l51 istomin , y.n . , 1991 , soviet astronomy letters , 17 , 301 kramer , m. , 1998 , apj , 509 , 856 kramer , m. , 2002 , proc .
of mg9 stairs , i.h . , _
_ , 2000 , nature , 406 , 484 weisberg , j.m . ,
_ et al _ , 1989 , apj , 347 , 1030 ( wrt89 ) weisberg , j.m . ,
taylor , j.h .
, aspconf.ser . , 202 , 127 weisberg , j.m . , taylor , j.h . , 2002 , 576 , 942 | discovery of psr b1913 + 16 by hulse & taylor in 1974 ( 1975 ) realized system represents stable accurate clock orbiting in gravitational field of compact star psr b1913 + 16 tests of theories gravity in strong field limit not in solar system weisberg ) no role pulsar in confirmation of gravitational waves tests based on timing observations damour & ruffini ( 1974 ) pointed out studying emission properties test prediction general relativity geodetic precession reference frame of freely falling object suffers precession distant observer geodetic precession in binary pulsar system geodetic precession leads to relativistic spin - orbit coupling analogous pulsar spin precesses about total angular momentum changing orientation towards earth change expect change in radio emission pulsar precession rate ) depends on period eccentricity of orbit pulsar companion mass values obtained from timing observations obtain precession rate of @xmath0 deg yr@xmath1 orbital angular momentum larger than pulsar spin orbital spin represents fixed direction in space defined by orbital plane of binary systemcalculated precession rate takes 297.5 years for pulsar spin vector to precess around precession angle between pulsar spin axis line - of - sight change with time different portions emission beam observed expects changes in pulse shape profile width function of time extreme precession may move beam out pulsar disappear until visible pulse profiles studied detect changes weisberg et al . 1989 discovered change in relative amplitude of two prominent profile components geodetic precession change in component separation or profile width not detected to precession distance of line - of - sight to magnetic axis change with time change in position angle pa ) swing of linearly polarised emission component expected cordes wasserman blaskiewicz 1990 studied polarisation data profiles pa swings 1985 to 1988 clear changes in pulse shape significant change in pa swing core component prominent at lower frequencies complicates interpretation polarisation data core may be responsible for change in relative component amplitude with time psr b1913 + 16 monitored with 76-m lovell telescope 100-m effelsberg telescopeanalysis of effelsberg profiles 1994 1998 by kramer revealed profile components changing amplitude consistent with rate determined by wrt89 detection of changes in separation model decreasing width profile two assumptions circular hollow cone - beam precession rate predicted by general relativity assumptions lead to model four free parameters misalignment angle between pulsar spin orbital angular momentum inclination angle magnetic axis radius of emission beam precession phase reference epoch post - keplerian parameters measured by pulsar timing general relativity allows compute value @xmath6 sine of orbital inclination angle for psr b1913 + 16 value of @xmath7 @xmath8 ambiguity not resolved from timing model allows four equivalent solutions one corresponds to @xmath7 other @xmath9 remaining choice by unknown orientation of pulsar spin orbital angular momentum pulsar rotation pro - grade or retro - grade retro - grade case less likely kramer 1998 polarisation information separate two solutions one gives correct sense swingobtained misalignment angle of @xmath10 by kramer ( 1998 ) with simulations by bailes ( 1988 ) predicted @xmath11 typical for psr b1913 + 16-like systems best fit to pulsar around 2025 implies component separation almost unchanged for 60 yr . why wrt89 not detect changes separation computing change in pa swing by cwb90 for produces value slightly larger than estimated detection limit istomin ( 1991 ) suggested disappearance of pulsar around 2020 full model by kramer ( 1998 ) predicts reappearance around 2220 psr b1913 + 16 observable for third of precession period detection rates of double neutron star systems wave detectors like ligo geo600 averaged over time existing numbers do not change discussed by kramer ( 1998 2002 ) figure 1a latest measurements for component separation as function of time profile to as predicted by model weisberg & taylor ( 2000, 2002 obtained new measurements after arecibo upgrade confirming results superior sensitivity telescope decrease in component separation general decrease in profile width at intensity levelsdata derive geometry with kramer ( 1998 ) obtain map of pulsar emission beam first line - of - sight moves through emission beam each profile represents different beam structure weisberg & taylor separate profiles into odd even use width information intensity levels mapping function derive model for geometry beam shape results surprising beam elongated hour - glass shaped use data for alternative approach propose use original profiles maintain information profile asymmetry off - set core compute profiles epochs compare model profiles to observed data figure 1b ) initial results suggest beam may be slightly elongated hour - glass beam shape not necessary further modelling necessary may provide beam map input for tests precession rate observations of precessing pulsars like psr b1828@xmath1211 or psr b1931 + 24 may help understand pulsar beam pattern detection of geodetic precession in psr b1913 + 16 successful qualitative test of general relativity may possible quantitative test measuring precession rate using beam models bailes , m. , 1998 , , 202 , 109 boerner , g.et al 1975 44 , 417 cordes j.m 1990 apj 349 , 546 cwb90 damour ruffini 1974 academie paris comptes rendus 279 esposito harrison 1975 196 , l1 hulse taylor j.h 1975 apj 195 , l51 istomin 1991 soviet astronomy letters 17 , 301 kramer 1998 apj 509 , 856 kramer 2002 mg9 stairs 2000 nature 406 , 484 weisberg j.m et al 1989 apj 347 , 1030 ( wrt89 ) weisberg j.m taylor j aspconf 202 , 127 weisberg taylor 2002 576 , 942 | 0.358798 | 2,433 | 1,251 | , Saving $0.1 in GPT-4. |
67,783 | we propose to interpret the stellar imf as a property of the turbulence in the star forming gas .
gravitationally unstable density enhancements in the turbulent flow collapse and form stars .
their mass distribution can be derived analytically from the power spectrum of the turbulent flow and the isothermal shock jump conditions in the magnetized gas . for a power spectrum index @xmath0 ,
consistent with larson s velocity dispersion
size relation as well as with new numerical and analytic results on supersonic turbulence , we obtain a power law mass distribution of dense cores with a slope equal to @xmath1 , consistent with the slope of salpeter s stellar imf .
below one solar mass , the mass distribution flattens and turns around at a fraction of a solar mass , as observed for the stellar imf in a number of stellar clusters , because only the densest cores are gravitationally unstable . the mass distribution at low masses is determined by the log normal distribution of the gas density .
the intermittent nature of this distribution is responsible for the generation of a significant number of collapsing cores of brown dwarf mass . | the origin of the stellar initial mass function ( imf ) is a fundamental problem in astrophysics because the stellar imf determines photometric properties of galaxies and the dynamical and chemical evolutions of their interstellar medium . in this contribution
we address the relation between statistical properties of turbulence and the origin of the stellar imf as discussed in padoan & nordlund ( 2002 , 2004 ) .
the main result of these works is that the power law slope , @xmath2 , of the stellar imf measured by @xcite is the consequence of the turbulent nature of the star forming gas and is directly related to the turbulent power spectrum slope , @xmath3 , @xmath4 . from this point of view
it is not surprising that the origin of salpeter s result has remained mysterious for half a century , as our understanding of turbulence has not improved much since the seminal work by @xcite .
the situation has changed during the last decade , because ever increasing computer resources have recently allowed significant progress in both fields of turbulence and star formation . since salpeter s work , the stellar imf has been measured successfully in many systems .
we now know that the imf in young clusters ( e.g. chabrier 2003 ) reaches a maximum at a fraction of a solar mass , and then turns around with a relative abundance of brown dwarfs ( bds ) that may vary from cluster to cluster ( e.g. luhman et al .
the work we present here explains also this feature of the imf as a consequence of the supersonic turbulence in the star
forming gas . using the properties of supersonic turbulence we have derived from recent numerical simulations , we predict the stellar imf essentially without free parameters .
this predicted imf is shown to depend on the rms mach number , mean density and temperature of the turbulent flow .
it is also shown to agree well with salpeter s imf for large stellar masses and with the low mass imf derived for young stellar clusters .
our view of the imf as a natural property of supersonic turbulence in the magnetized and isothermal star
forming gas provides an explanation for the origin of bds as well .
according to this picture , bds may be formed in the same way as hydrogen burning stars .
the velocity power spectrum in the inertial range of turbulence ( between the scales of energy injection and viscous dissipation ) is a power law , @xmath5 , where @xmath6 is the wave number .
the spectral index is @xmath7 for incompressible turbulence ( kolmogorov 1941 ) , and @xmath8 for pressureless turbulence ( burgers 1974 ; gotoh & kraichnan1993 ) . in recent numerical simulations of isothermal , super alfvnic and highly supersonic magneto
hydrodynamic ( mhd ) turbulence we have obtained a power spectrum intermediate between the burgers and the kolmogorov power spectra ( boldyrev , nordlund & padoan 2002 ) and consistent with the prediction by @xcite , @xmath9 . the probability density function ( pdf ) of gas density in isothermal turbulent flows is well approximated by a log normal distribution with moments depending on the rms mach number of the flow ( nordlund & padoan 1999 ; ostriker , gammie & stone 1999 ) .
the density structure is characterized by a complex system of interacting shocks resulting in a fractal network of dense cores , filaments , sheets and low density voids
, with a large density contrast .
most of the mass concentrates in a small volume fraction ( according to the log normal pdf ) , a manifestation of the intermittent nature of the turbulence .
an example of a projected turbulent density field is shown in figure [ fig0 ] .
a simple model of the expected mass distribution of dense cores generated by supersonic turbulence has been proposed in @xcite .
the model is based on two assumptions : i ) the power spectrum of the turbulence is a power law ; ii ) the typical size of a dense core scales as the thickness of the postshock gas .
the first assumption is a basic result for turbulent flows and holds also in the supersonic regime ( boldyrev et al .
the second assumption is suggested by the fact that postshock condensations are assembled by the turbulent flow in a dynamical time .
condensations of virtually any size can therefore be formed , independent of their jeans mass . with these assumptions , together with the jump conditions for mhd shocks
, the mass distribution of dense cores can be related to the power spectrum of turbulent velocity , @xmath5 : @xmath10 if the turbulence spectral index @xmath3 is taken from the analytical prediction by @xcite , which is consistent with the observed velocity dispersion size larson relation ( larson 1979 , 1981 ) and with our numerical results ( boldyrev et al .
2002 ) , then @xmath11 and the mass distribution is @xmath12 , almost identical to salpeter s stellar imf ( salpeter 1955 ) .
the exponent of the mass distribution is rather well constrained , because the value of @xmath3 for supersonic turbulence can not be smaller than the incompressible value , @xmath13 ( slightly larger with intermittency corrections ) , and the burgers case , @xmath14 . as a result
, the exponent of the mass distribution is predicted to be within the range of values of 1.3 and 1.5 . while massive cores are usually larger than their critical bonnor
ebert mass , @xmath15 , the probability that small cores are dense enough to collapse is determined by the statistical distribution of core density .
in order to compute this collapse probability for small cores , we assume i ) the distribution of core density can be approximated by the log normal pdf of gas density and ii ) the core density and mass are statistically independent . because of the intermittent nature of the log - normal pdf , even very small ( sub stellar ) cores have a finite chance to be dense enough to collapse . based on the first assumption
, we can compute the distribution of the critical mass , @xmath16 , from the log normal pdf of gas density assuming constant temperature ( padoan , nordlund & jones 1997 ) .
the fraction of cores of mass @xmath17 larger than their critical mass is given by the integral of @xmath18 from 0 to @xmath17 .
using the second assumption of statistical independence of core density and mass , the mass distribution of collapsing cores is @xmath19\,{\rm d}\ln m ~. \label{imfpdf}\ ] ] this mass distribution is a function of the rms mach number , mean density and temperature of the turbulent flow , as these parameters enter the pdf of gas density and thus @xmath18 .
figure [ fig1 ] ( left panel ) shows five mass distributions computed from equation ( [ imfpdf ] ) with three different values of the sonic rms mach number and two different values of density .
@xcite have suggested that bds may originate from the process of turbulent fragmentation like hydrogen burning stars .
this is illustrated by the analytical imf in the left panel of figure [ fig1 ] , which shows a relatively large abundance of bds is predicted for sufficiently large values of the mean density or rms mach number .
the imf of the cluster ic 348 in perseus , obtained by @xcite , is plotted in the right panel of figure [ fig1 ] ( solid line histogram ) . the imf of this cluster has been chosen for the comparison with the theoretical model because it is probably the most reliable observational imf including both bds and hydrogen burning stars . in figure [ fig1 ]
we have also plotted the theoretical mass distribution computed for @xmath20 @xmath21 , @xmath22 k and @xmath23 .
these parameters are appropriate for the central @xmath24 arcmin of the cluster ( @xmath25 pc ) , where the stellar density corresponds to approximately @xmath26 @xmath21 .
the figure shows that the theoretical distribution of collapsing cores , computed with parameters inferred from the observational data , is roughly consistent with the observed stellar imf in the cluster ic 348 .
similar imfs have been obtained for several other young clusters ( chabrier 2003 ) .
we have proposed to explain the stellar imf as a property of supersonic turbulence .
this scenario is very different from previous theories of star formation ( see shu , adams & lizano 1987 ) , where it is assumed that stars of small and intermediate mass are formed from sub
critical cores evolving quasi
statically , on the time
scale of ambipolar drift .
these theories do not account for the ubiquity of turbulence in star forming clouds and therefore ignore the effect of turbulence in the fragmentation process .
a naive interpretation of our results may lead to the conclusion that the stellar imf at large masses should be a universal power law , with a slope very close to salpeter s value .
the statistics of turbulence controlling the origin of the stellar imf are indeed universal ( they depend on _ flow _ properties such as the rms mach number , not _ fluid _ properties , and are insensitive to initial conditions that are soon `` forgotten '' due to the chaotic nature of the turbulence ) . however , such statistics are derived from ensemble averages . according to our derivation ,
massive stars originate from shocks on relatively large scales . in any given system ( for example a molecular cloud )
the number of large scale compressions ( thus the number of massive stars ) is relatively small . because of the small size of the statistical sample and because of the intermittent nature of the turbulence , large deviations from salpeter s imf are possible in individual systems .
salpeter s imf is therefore a universal property of supersonic turbulence , but it is not necessarily reproduced precisely in every star forming region . | origin of stellar initial mass function imf fundamental problem in astrophysics determines photometric properties galaxies dynamical chemical evolutions interstellar medium relation between turbulence origin stellar imf discussed in padoan & nordlund ( 2002 2004 power law slope @xmath2 of stellar imf measured by @xcite consequence turbulent nature star forming gas related to turbulent power spectrum slope @xmath3 @xmath4 origin of salpeter s result remained mysterious for half a century understanding of turbulence improved since situation changed decade computer resources allowed progress in turbulence star formation since salpeter s work stellar imf measured successfully in many systems imf in young clusters reaches maximum at fraction of solar mass turns around with abundance brown dwarfs work explains imf as consequence supersonic turbulence in star forming gas using properties supersonic turbulence predict stellar imf without free parameters predicted imf on rms mach number mean density temperature of turbulent flow with salpeter s imf for large stellar masses low mass imf for young stellar clustersimf natural property supersonic turbulence in magnetized isothermal star forming gas explanation origin bds bds formed as hydrogen burning stars velocity power spectrum in inertial range turbulence energy injection dissipation is power law @xmath5 @xmath6 wave number spectral index @xmath7 for incompressible turbulence @xmath8 pressureless turbulence burgers 1974 simulations isothermal super supersonic magneto hydrodynamic turbulence obtained power spectrum intermediate between burgers kolmogorov power spectra consistent with prediction @xcite , @xmath9 probability density function of gas density in isothermal turbulent flows approximated by log normal distribution moments depending on rms mach number flow density structure complex system interacting shocks fractal network of dense cores filaments sheets low density voids large density contrast mass concentrates in small volume fraction intermittent nature turbulence example projected turbulent density field shown in figure fig0 model expected mass distribution of dense cores supersonic turbulence proposed in @xcite .model based on two assumptions power spectrum turbulence is power law typical size of dense core scales as thickness postshock gas first assumption basic result for turbulent flows supersonic regime second assumption postshock condensations assembled by turbulent flow in dynamical time condensations any size can be formed independent of mass jump conditions for shocks mass distribution of dense cores related to power spectrum turbulent velocity if turbulence spectral index @xmath3 taken from prediction @xcite consistent with observed velocity dispersion size larson relation numerical results mass distribution is @xmath12 identical to salpeter s stellar imf exponent mass distribution constrained value of @xmath3 for supersonic turbulence not smaller than incompressible value @xmath13 and @xmath14 exponent mass distribution predicted within range values 1.3 and 1.5 massive cores usually larger than critical mass probability small cores dense to collapse determined by statistical distribution core density collapse probability for distribution core density approximated by normal gas density core density and mass statistically independentintermittent log - normal pdf small sub stellar cores finite chance collapse first assumption compute distribution critical mass @xmath16 from log normal pdf gas density constant temperature padoan nordlund jones 1997 fraction of cores mass @xmath17 larger than critical mass by integral of @xmath18 from 0 to @xmath17 second assumption statistical independence core density mass mass distribution of collapsing cores is @xmath19\ mass distribution function of rms mach number mean density temperature of turbulent flow gas density @xmath18 figure fig1 shows five mass distributions from equation three values sonic rms mach number two values density bds may originate from turbulent fragmentation like hydrogen burning stars illustrated analytical imf large abundance of bds for large values mean density rms mach number imf of cluster ic 348 in perseus obtained by @xcite plotted right panel cluster chosen most reliable observational imf including bds hydrogen burning stars plotted theoretical mass distribution for @xmath20 @xmath21 @xmath22 @xmath23 parameters appropriate for central @xmath24 arcmin cluster ( @xmath25 pc ) stellar density corresponds to @xmath26 @xmath21 figure shows theoretical distribution of collapsing cores observational data consistent with observed stellar imf in cluster ic 348 similar imfs obtained for other young clusters ( chabrier 2003 ) proposed to explain stellar imf as property supersonic turbulence different from previous theories star formation stars small mass formed from sub critical cores evolving statically ambipolar theories account for ubiquity turbulence in star forming clouds ignore effect in fragmentation process naive interpretation stellar imf at large masses universal power law slope close to salpeter s value statistics of turbulence controlling origin stellar imf universal depend on flow properties not fluid properties insensitive to initial conditions statistics derived from ensemble averages massive stars originate from shocks on large scales in system large scale compressions massive stars small small size statistical sample intermittent nature turbulence large deviations from salpeter s imf possible in individual systemssalpeter s imf universal property supersonic turbulence not reproduced star region | 0.35042 | 2,111 | 1,014 | , Saving $0.1 in GPT-4. |
67,787 | crank - nicolson ( c - n ) method combined with @xmath0-spline basis set can be used in both time - dependent and time - independent calculations of the photoionization cross sections of atoms and molecules . for time - independent systems
, it is found that c - n method in imaginary - time propagation ( itp ) can converge directly to not only the ground state , but also excited and continuum states by controlling the time step size .
the c - n method can also be directly applied in time - dependent calculations . both time - dependent and time - independent calculations
agree very well with previous results .
this method can also be extended to two - electron systems directly . | photoionization of atoms and molecules is a fundamental physical process , which has been studied for many years .
accurate solution of continuum states is an essential task in the various quantum mechanical problems , such as the calculation of multiphoton ionization , above - threshold ionization ( ati ) , and electron scattering .
however , the wavefunctions of continuum states oscillate to infinity . as a result
, it is not as easy as the calculations of bound states .
there are a number of methods to calculate the resonance states , such as complex scaling method , r - matrix method .
however , they are not used to calculate continuum state with any positive energy .
although variational method @xcite and least - squares schemes @xcite have been introduced in the calculation of continuum wavefunctions , efficient and accurate solutions are still necessary . in general
, a variational approach can be used to calculate all the eigenstates of any quantum system using some appropriate basis set . after direct diagonalization of a hamiltonian matrix for appropriate potentials , all the eigenvalues and wavefunctions are obtained .
negative eigenvalues correspond to bound states , while positive eigenvalues correspond to the continuum states . however , this method is not efficient for large scale matrix problems involving large basis sets for convergence and high accuracy . in any standard matrix diagonalization program ,
the demand on computer memory grows as @xmath1 and the cpu time as @xmath2 ( @xmath3 is the size of the hamiltonian matrix)@xcite .
when @xmath4 , the computation becomes very slow , even impractical in some cases .
in addition , the obtained discrete continuum states may not be what we want because we do nt know their boundary conditions . for certain chemical and physical problems , not all the eigenstates of a system
are required .
for example , to numerically solve time - dependent schrdinger equations ( tdse ) , one only needs the initial wavefunction , and then propagates it under a hamiltonian including time - dependent interaction terms by using general split - operator methods ( som)@xcite , crank - nicolson methods ( c - n method)@xcite , or efficient large vector matrix operations@xcite .
if the initial state in a simulation is the ground state , one does not need to diagonalize a basis set dependent hamiltonian , but one propagates a tdse for an arbitrary initial vector by an imaginary - time - propagation method ( itp method)@xcite .
complex time steps also allow to obtain higher order accuracy in som@xcite . in this paper , we show that itp method based on c - n scheme can be generalized to converge to bound exited and unbound continuum states by changing time step . to our knowledge , this has not been reported . the itp calculations and time - dependent calculations based on c - n method
agree very well .
the paper is organized as follows : the principle of itp method for excited and continuum states is presented in sec .
the application of the itp method in photoionization of one - electron diatomic molecular ion h@xmath5 and heh@xmath6 is shown in sec .
[ iii ] . in sec .
[ iv ] , we illustrate the application of the proposed itp method in two - electron systems .
the paper concludes in sec .
in this section , we briefly introduce next the principle of usual itp method for ground state and our proposed itp method based on c - n scheme for all the bound and continuum states .
the principle of the itp method can be found as follows . for a general tdse ( atomic units , a.u .
, @xmath7 are used unless otherwise specified ) , @xmath8 introducing the imaginary time @xmath9 in an itp method , and if the hamiltonian @xmath10 is time - independent , then the new time - dependent wavefunction after a time step @xmath11 can be written as : @xmath12 expanding an initial arbitrary wavefunction @xmath13 as a linear combination of the time - independent eigenstates @xmath14 of @xmath10 , @xmath15 where @xmath16 , then eq . ( [ e2 ] )
can be written as : @xmath17 it is obvious that higher energy states will exponentially decay as a function of @xmath18 ( see fig .
[ fig1 ] ( a ) ) , at a rate which is faster than for the lower energy states .
after sufficiently long time evolution , any arbitrary initial wavefunction will thus converge to the ground state .
if we are interested in an excited state @xmath19 , we have to at first obtain the converged ground state @xmath20 and other lower excited states @xmath21 , with @xmath22 , then project out all the lower energy states by @xmath23 from the initial state before evolution@xcite . for higher excited states ,
this method becomes slowly convergent and difficult to apply .
it is even not practical for continuum states due to the large degeneracy of those states . ) and eq .
( [ e6 ] ) as a function of @xmath24 .
, title="fig:",width=264,height=188 ] ) and eq .
( [ e6 ] ) as a function of @xmath24 . ,
title="fig:",width=264,height=188 ] [ cols="^,^,^,^",options="header " , ]
in summary , based on the crank - nicolson numerical method , we present in this work an extension of the itp method @xmath9 for directly converging an arbitrary initial vector to desired bound excited states by controlling the time step size .
it is different from the usual itp methods which can only converge directly to the ground state . to obtain converged excited state , the usual itp methods require filtering out all lower bound states from the initial state . for continuum states with positive energies
, we generalize the imaginary propagation method by @xmath25 , and show convergence of any arbitrary state to desired continuum states .
this method is proven to be simple and accurate for small number of iterations @xmath26 or total time @xmath27 .
the application of this itp method in the photoionization of diatomic molecules is illustrated .
this method can also be used in two - electron systems .
we emphasize that this c - n method is a systematic method to study ultrafast dynamics . after obtaining the initial bound state by c - n itp ,
then the c - n method can be used to solve tdse in real time space .
after the interactions between intense ultrashort laser pulses and atoms and molecules , the energy and momentum information of the photoelectrons can be extracted from the time - dependent wave functions@xcite by projecting on the continuum states obtained by itp .
this new method can be directly extended to complex system@xcite , confined systems@xcite , dirac equations@xcite , multiparticle states@xcite and also in tddft methods@xcite for complex molecular systems .
the author thanks prof .
andr d. bandrauk very much for helpful discussions .
m. brosolo , and p. decleva , chem . phys . * 159 * , 185 ( 1992 ) .
m. brosolo , p. decleva , and a. lisini , j. phys .
b * 25 * , 3345 ( 1992 ) .
m. brosolo , p. decleva , and a. lisini , comput .
. comm . * 71 * , 207 ( 1992 ) .
x.b . bian , h.p .
liu , and t.y .
shi , chin .
. lett . * 25 * , 2008 ( 2008 ) .
hermann , and j.a .
fleck , jr .
a * 38 * , 6000 ( 1988 ) .
bandrauk , and h. shen , chem .
. lett . * 176 * , 428 ( 1991 ) .
j. crank , and p. nicolson , proc .
. soc . * 43 * , 50 ( 1947 ) .
z.g . sun , and w.t .
yang , j. chem . phys . *
134 * , 041101 ( 2011 ) .
g. avila , and t. carrington jr .
, in _ quantum dynamic imaging _ , edited by a.d .
bandrauk , and m. ivanov , ( springer , new york , 2011 ) , p.1 .
l.y . peng , and a.f .
starace , j. chem . phys .
* 125 * , 154311 ( 2006 ) .
bandrauk , e. dehghanian , and h.z .
lu , chem .
. lett . * 419 * , 346 ( 2006 ) .
r. kosloff and h. tal - ezer , chem .
. lett . * 127 * , 223 ( 1986 ) .
h. bachau , e. cormier , p. decleva , j.e .
hansen , and f. martn , rep .
. phys . * 64 * , 1815 ( 2001 ) .
bian , and a.d .
bandrauk , phys .
a * 83 * , 023414 ( 2011 ) ; ibid . *
83 * , 041403 ( 2011 ) .
bian , h.x .
qiao , and t.y .
shi , chin .
* 23 * , 2403 ( 2006 ) . c. froese fischer and m. idress , computer phys . * 3 * , 53 ( 1989 ) .
madsen , l.a.a .
nikolopoulos , t.k .
kjeldsen , and j. fernndez , phys .
a * 76 * , 063407 ( 2007 ) .
mercouris , y. komninos , s. dionissopoulou and c. a. nicolaides , j. phys .
b * 30 * , 2133 ( 1997 ) ; j. phys .
b * 29 * , l13 ( 1996 ) . th .
mercouris , y. komninos , and c. a. nicolaides , advances in quantum chemistry * 60 * , 333 ( 2010 ) .
bates , u. pik , and g. poots , proc .
a * 66 * , 1113 ( 1953 ) .
madsen , and j.m .
peek , at .
data * 2 * , 171 ( 1971 ) .
davydkin , and l.p .
rapoport , j. phys .
b * 7 * , 1101 ( 1974 ) .
s. kawai , and a.d .
bandrauk , phys .
a * 75 * , 063402 ( 2007 ) .
ponomarev , and l.n .
somov , j. comput . phys . * 20 * , 183 ( 1976 ) .
h. bachau , j. phys .
b * 35 * , 509 ( 2002 ) .
guan , e.b .
secor , k. bartschat , and b.i .
schneider , phys .
a * 84 * , 033420 ( 2011 ) .
guan , k. bartschat , and b.i .
schneider , phys .
a * 83 * , 043403 ( 2011 ) .
h. miyagi , t. morishita , and s. watanabe , phys .
a * 85 * , 022708 ( 2012 ) .
richards , and f.p .
larkins , j. phys .
b * 19 * , 1945 ( 1986 ) .
l. tao , c. w. mccurdy , and t. n. rescigno , phys .
a * 79 * , 012719 ( 2009 ) .
j. colgan , m. s. pindzola , and f. robicheaux , phys .
a * 68 * , 063413 ( 2003 ) .
bian , and a.d .
bandrauk , phys .
* 105 * , 093903 ( 2010 ) .
kamta , and a.d .
bandrauk , phys .
a * 75 * , 041401 ( 2007 ) ; ibid . *
76 * , 053409 ( 2007 ) .
kamta , and a.d .
bandrauk , phys .
lett . * 94 * , 203003 ( 2005 ) .
campos , d.l .
nascimento , d.t .
cavalcante , a.l.a .
fonseca , and a.o.c .
nunes , int .
j. quantum chem . * 106 * , 2587 ( 2006 ) .
l.j . zhang , j.m .
feagin , v. engel , and a. nakano , phys .
a * 49 * , 3457 ( 1994 ) .
pindzola , _ et al .
_ , j. phys .
b * 40 * r39 ( 2007 ) .
mccurdy , and f. martn , j. phys .
b * 37 * 917 ( 2004 ) .
drake , and z.c .
yan , chem .
* 229 * 486 ( 1994 ) .
pekeris , phys . rev . * 126 * 1470 ( 1962 ) . s. a. ndengu , o. motapon , r. l. melingui melono and a. j. etindele , j. phys .
b * 47 * 015002 ( 2013 ) .
e. lorin , a. bandrauk , nonlinear anal .- real world appl . * 12 * , 190 ( 2011 ) .
k. lopata , and n. govind , j. chem .
theory comput .
* 7 * , 1344 ( 2011 ) .
fowe , and a.d .
bandrauk , phys .
a * 81 * , 023411 ( 2010 ) . | photoionization of atoms molecules fundamental physical process studied years accurate solution of continuum states essential in quantum mechanical problems multiphoton ionization above - threshold ionization electron scattering wavefunctions of continuum states oscillate to infinity not as easy as bound states methods calculate resonance states complex scaling r - matrix method not used to continuum state with positive energy variational method @xcite least - squares schemes introduced in continuum wavefunctions efficient accurate solutions necessary variational approach to calculate all eigenstates quantum system after direct diagonalization hamiltonian matrix eigenvalues wavefunctions obtained negative eigenvalues correspond to bound states positive to continuum states method not efficient for large scale matrix problems large basis sets high accuracy matrix diagonalization demand on computer memory grows cpu time as @xmath2 @xmath4 computation becomes slow obtained discrete continuum states may not be want know boundary conditions for certain chemical physical problems not all eigenstates requiredsolve time - dependent schrdinger equations needs initial wavefunction propagates under hamiltonian time - dependent interaction terms split - operator methods crank - nicolson methods large vector matrix operations if initial state is ground state diagonalize basis dependent hamiltonian propagates tdse for arbitrary initial vector by imaginary - time - propagation method complex time steps allow higher order accuracy in som itp method generalized to converge to bound exited unbound continuum states by changing time step not reported itp calculations time - dependent calculations c - n method agree paper organized principle of itp method for excited continuum states presented application in photoionization of one - electron diatomic molecular ion shown application proposed itp method in two - electron systems paper concludes introduce principle of itp method for ground state proposed itp method based on c - n scheme for bound continuum states principle of itp method general tdse ( atomic units@xmath7 used unless specified @xmath8 imaginary time @xmath9 itp method hamiltonian @xmath10 time - independent new time - dependent wavefunction after time step @xmath11 written @xmath12 expanding initial arbitrary wavefunction @xmath13 linear combination time - independent eigenstates @xmath14 @xmath10 @xmath15 @xmath16 eq e2 written @xmath17 higher energy states decay function @xmath18 fig faster lower energy states after long evolution arbitrary initial wavefunction to ground state interested excited state @xmath19 obtain converged ground state @xmath20 lower excited states @xmath21 @xmath22 project lower energy states by @xmath23 from initial state before evolution higher excited states method becomes convergent difficult apply not practical for continuum states large degeneracy eq . e6 function of @xmath24 eq . e6 function @xmath24 based crank - nicolson numerical method present extension itp method @xmath9 for converging arbitrary initial vector to bound excited states controlling time step sizedifferent from usual itp methods converge to ground state converged state require lower bound states from initial for continuum states positive energies generalize imaginary propagation method by @xmath25 show convergence of arbitrary state to desired continuum states method simple accurate for small iterations total time @xmath27 application in photoionization of diatomic molecules illustrated used in two - electron systems c - n method systematic ultrafast dynamics after initial bound state by c - n itp solve tdse in real time interactions between ultrashort laser pulses atoms molecules energy momentum information photoelectrons extracted from time - dependent wave functions by projecting on continuum states obtained by itp method extended to complex confined systems dirac equations multiparticle states tddft methods for complex molecular systems author thanks prof . andr d. bandrauk for discussions m. brosolo p. decleva chem . . 159, 185 ( 1992 ) j. phys . 25 3345 ( 1992 ) comput. 71 207 ( 1992 )25 2008 hermann fleck jr 38 6000 ( 1988 ) bandrauk h. shen chem lett 176 428 ( 1991 ) j. crank p. nicolson soc 43 50 ( 1947 ) z sun w yang chem 134 041101 ( 2011 ) g. avila t. carrington quantum dynamic imaging bandrauk m. ivanov springer new 2011 ) l.y peng a.f starace chem phys 125 154311 ( 2006 ) bandrauk dehghanian h lu chem 419 346 ( 2006 ) r. kosloff h. tal - ezer chem lett 127 223 ( 1986 ) h. bachau cormier p. decleva hansen f. martn rep 64 1815 ( 2001 ) bian bandrauk phys 83 023414 ( 2011 ) 83 041403 ( 2011 ) bian qiao shi chin 23 2403 ( 2006 ). froese fischer m. idress computer phys . 3 53 ( 1989 )madsen nikolopoulos kjeldsen fernndez phys 76 * 063407 ( 2007 ) mercouris komninos dionissopoulou nicolaides b 30 * 2133 ( 1997 ) 29 l13 ( 1996 ) mercouris komninos nicolaides advances quantum chemistry 60 * 333 ( 2010 ) bates pik poots a 66 * 1113 ( 1953 ) madsen peek data 2 * 171 ( 1971 ) davydkin rapoport phys . b 7 * 1101 ( 1974 ) s kawai bandrauk phys 75 * 063402 ( 2007 ) ponomarev somov phys . 20 183 ( 1976 ) bachau phys b 35 * 509 ( 2002 ) guan secor bartschat schneider phys 84 033420 ( 2011 ) phys 83 * 043403 ( 2011 ) miyagi morishita watanabe phys a 85 * 022708 ( 2012 ) richards flarkins phys . 19 1945 ( 1986 ) tao mccurdy rescigno phys 79 012719 ( 2009 ) colgan pindzola robicheaux . 68 063413 ( 2003 ) bian bandrauk . 105 093903 ( 2010 ) kamta bandrauk 75 041401 ( 2007 ) ibid 76 * 053409 ( 2007 ) kamta bandrauk phys 94 * 203003 ( 2005 ) campos nascimento cavalcante fonseca nunes quantum chem. 106 * 2587 ( 2006 ) zhang feagin engel nakano phys . 49 3457 ( 1994 ) pindzola . 40 r39 ( 2007 ) mccurdy f. martn phys . 37 * 917 ( 2004 ) drake yan chem . 229 * 486 ( 1994 ) pekeris phys . 126 * 1470 ( 1962 ) ndengu motapon melingui melono etindele phys . 47 * 015002 ( 2013 )lorin bandrauk nonlinear anal real world appl 12 190 ( 2011 ) lopata govind chem theory comput 7 1344 ( 2011 ) fowe bandrauk. 81 023411 ( 2010 ) | 0.502156 | 3,131 | 1,570 | , Saving $0.1 in GPT-4. |
67,794 | we perform perturbative qcd factorization of infrared radiations associated with an energetic @xmath0 quark from a polarized top quark decay , taking the semi - leptonic channel as an example .
the resultant formula is expressed as a convolution of an infrared - finite heavy - quark kernel with a @xmath0-quark jet function .
evaluating the heavy - quark kernel up to leading order in the coupling constant and adopting the jet function from qcd resummation , we predict the dependence of the spin analyzing power for a polarized top quark on the invariant mass of the @xmath0-quark jet . it is observed that the spin analyzing power could be enhanced by a factor 2 compared to the inclusive case with the jet mass being integrated over .
it is worthwhile to test experimentally the enhancement of the spin analyzing power due to the inclusion of jet dynamics . | a study of the top quark physics contributes to the understanding of the origin of the electro - weak symmetry breaking in the standard model and its extensions .
progress in this field from the large hadron collider ( lhc ) , a top quark factory , has been reviewed in @xcite .
higher - order calculations of top production and decay processes were summarized in @xcite ( see also references therein ) .
especially , observables related to the spin of a top quark may reveal various new physics effects , which can be explored at lhc ( for recent references , refer to @xcite ) .
since a top quark decays almost @xmath1 into a @xmath0 quark and a @xmath2 boson before hadronization , one can extract the information on the spin of a `` bare top quark '' from the angular distribution of its decay products .
for this purpose , the spin analyzing power @xmath3 for a final state @xmath4 in a polarized top quark decay has been defined via @xmath5 where @xmath6 is the partial decay width , @xmath7 is the polarization vector of a polarized ensemble of top quarks at rest with @xmath8 , and @xmath9 is the angle of the particle momentum measured from the top quark spin .
that is , @xmath3 relates the top quark spin to the angular distribution of the decay product @xmath4 .
qcd corrections to the polarized @xmath10 decay were investigated in @xcite , and those to the spin analyzing power were done for semi - leptonic decays in @xcite , and for hadronic decays in @xcite .
the results have been summarized in table 3 of ref .
@xcite : charged leptons and down - type quarks exhibit the largest power ( @xmath11 ) , a bottom quark has a second - largest power ( @xmath12 ) , neutrinos and up - type quarks have a value of @xmath13 , and the less energetic non-@xmath0 jet shows @xmath14 @xcite . it was observed that next - to - leading - order ( nlo ) corrections to the spin analyzing power are not important , and the @xmath0 quark mass could be safely ignored in the above analysis of the polarized top quark decay @xcite .
although the light - particle jets from the @xmath15 and @xmath16 quarks have been considered , the spin analyzing power associated with the @xmath0-quark jet is not known yet . in particular , a jet mass can be measured , so it is interesting to examine how the spin analyzing power for a polarized top quark depends on the @xmath0-quark jet mass .
this is the motivation of the present work .
a fundamental framework for studying high - energy processes is the perturbative qcd ( pqcd ) factorization theorem @xcite , in which cross sections and decay widths are factorized into convolutions of several subprocesses .
take top quark decays as an example .
the subprocesses include jet functions which contain infrared radiations associated with energetic hadronic final states , soft functions which collect infrared gluons exchanged among the top quark and hadronic final states , and heavy - quark kernels from the difference between the qcd diagrams for the top quark decays and the effective diagrams for the jet and soft functions . in this paper
we shall demonstrate the factorization of the @xmath0-quark jet from the semi - leptonic top quark decay , that requires the eikonal approximation for particle propagators in infrared regions and on the ward identity for organizing all diagrams with attachments of infrared gluons @xcite .
the soft functions will be neglected here due to the cancellation between virtual and real corrections , and to the fact that they have minor effects on jet mass distributions . the pqcd factorization formula for the semi - leptonic decay of a polarized top quark is then expressed as a convolution of the infrared - finite heavy - quark kernel with the @xmath0-quark jet function .
it has been known that the overlap of collinear and soft dynamics produces large double logarithms in a jet function , which should be summed up to all orders .
references on the resummation of various double logarithms can be found in @xcite .
recently , the qcd resummation technique for light - particle jets have been developed @xcite , by means of which jet substructures , such as jet mass distributions and energy profiles , can be calculated . substituting the quark jet function derived in @xcite into the pqcd factorization formula for the polarized top quark decay ,
we show that the spin analyzing power increases with the jet mass quickly , and is possibly enhanced by a factor 2 in a wide range of the jet mass .
this result is attributed to the observation that the shape of the jet function puts more weight on the contribution from the kinematic region with higher @xmath0-quark jet momentum , where the spin analyzing power is larger .
it is worthwhile to test experimentally the enhancement of the spin analyzing power due to the inclusion of jet dynamics .
the pqcd factorization of the @xmath0-quark jet function from the semi - leptonic decay of a polarized top quark is performed up to nlo by employing the eikonal approximation in sec .
the procedures that rely on the application of the ward identity for the all - order factorization are outlined .
the doubly differential width for the angular distribution of the @xmath0-quark jet with a specified invariant mass is presented in sec .
we propose a parametrization for the quark jet function derived in @xcite in order to simplify numerical analysis .
the choices of physical parameters and our predictions are summarized .
section [ conclusion ] is devoted to a conclusion .
we consider the semi - leptonic top quark decay @xmath17 where @xmath18 , @xmath19 , @xmath20 , and @xmath21 are the momenta of the top quark , the @xmath0 quark , the charged lepton , and the neutrino , respectively , as indicated in fig . [ fig1 ] for the leading - order ( lo ) amplitude . in this section
we shall identify the leading infrared contributions , and perform their factorization into the @xmath0-quark jet function .
the averaged decay amplitude squared @xmath22 at lo in both the electro - weak coupling and the strong coupling , depicted in fig .
[ fig2 ] , is written as @xmath23 with the leptonic and hadronic tensors @xmath24,{\nonumber}\\ t^{\mu\rho } & = & \mbox{tr}\left[\gamma^{\mu}p_{l}\hat{w}_{t}(k{\hspace{-1.5mm}\!/}_{t}+m_{t } ) \gamma^{\rho}p_{l}(k{\hspace{-1.5mm}\!/}_{b}+m_{b } ) \right ] .
\label{eq.m0sq}\end{aligned}\ ] ] in the above expression @xmath25 is the gauge coupling of the weak interaction , @xmath26 is the cabibbo - kobayashi - maskawa matrix element , @xmath27 , @xmath28 and @xmath29 are the momentum , the mass , and the decay width of the @xmath2 boson , respectively , @xmath30 arises from the summation over the polarizations of the @xmath2 boson , @xmath31 is the projection matrix , @xmath32 is the spin projector for the top quark , @xmath33 and @xmath34 are the masses of the top quark and the @xmath0 quark , respectively , and the lepton masses have been dropped .
the three - body phase - space integral for the top quark decay is given by @xmath35 with @xmath36 being the energy of the particle @xmath4 , @xmath37 .
the angle @xmath38 is the polar angle of the @xmath0 quark with respect to the top quark spin as indicated in fig .
[ fig4 ] , the azimuthal angle @xmath39 has been integrated out , and @xmath40 denotes the angle between the @xmath41-@xmath42 plane and the @xmath42-@xmath43 plane . [ cols="^ " , ]
in this paper we have performed pqcd factorization of the @xmath0-quark jet function from the semi - leptonic decay of a polarized top quark by means of the eikonal approximation and the ward identity .
the resultant formula is expressed as the convolution of the infrared - finite heavy - quark kernel with the jet function , which is dominated by infrared radiations . adopting the lo heavy - quark kernel and parameterizing the jet function from qcd resummation ,
we have predicted the dependence of the spin analyzing power associated with the @xmath0-quark jet on its invariant mass .
it has been verified that our formalism reproduces the fixed - order results by integrating over the jet mass .
our investigation indicated that the spin analyzing power could be possibly enhanced by a factor 2 as measuring the angular distribution of the @xmath0-quark jet with a specified invariant mass .
the mechanism responsible for the enhancement has been elaborated , which is attributed to the broadening of a jet in the mass distribution with the jet energy .
this broadening puts more weight on the contribution from the region with higher jet energy , that intends to give a larger spin analyzing power .
the above observation is rather insensitive to the choice of the jet radius , and worth of experimental confrontation in view of deeper understanding of the effect from including jet dynamics into the polarized top quark decay . as future works , nlo corrections to the heavy - quark kernel
need to be taken into account in order to improve the precision of our predictions . at this level ,
the dependence of the top quark mass definition on renormalization schemes ( for instance , the @xmath44 mass @xcite or the pole mass ) , becomes an essential issue .
it is possible to extend the current formalism to hadronic decays of a polarized top quark , whose analysis is more complicated .
another important direction is to develop the formalism for a boosted top quark at lhc @xcite , that mainly decays into a single jet .
substructures for a polarized top quark jet , which improve the extraction of the information on the top quark polarization , can be calculated .
we thank z.g .
si for useful discussions . this work was supported in part by the national science council of r.o.c . under grant
nsc-101 - 2112-m-001 - 006-my3 , and by the national center for theoretical sciences of r.o.c .. | study top quark physics contributes to understanding origin electro - weak symmetry breaking standard model extensions progress from large hadron collider lhc top quark factory reviewed in @xcite calculations of top production decay processes summarized in @xcite observables related to spin top quark may reveal new physics effects explored at lhc @xcite top quark decays into @xmath0 quark @xmath2 boson before hadronization extract information on spin from angular distribution decay products spin analyzing power @xmath3 for final state polarized top quark decay defined via @xmath5 partial decay width polarization vector angle particle momentum from top quark spin @xmath3 relates top quark spin to angular distribution decay product qcd corrections to polarized @xmath10 decay investigated in @xcite spin analyzing power for semi - leptonic decays hadronic decays results summarized in table 3charged leptons down - type quarks exhibit largest power @xmath11 bottom quark second - largest power @xmath12 ) neutrinos up - type quarks value @xmath13 less energetic non-@xmath0 jet shows @xmath14 next - - order corrections spin analyzing power not important @xmath0 quark mass ignored polarized top quark decay light - particle jets from @xmath15 @xmath16 quarks considered spin analyzing power @xmath0-quark jet not known jet mass measured examine spin analyzing power polarized top quark depends on @xmath0-quark jet mass motivation present work framework high - energy processes perturbative qcd ) factorization theorem cross sections decay widths factorized into convolutions subprocesses top quark decays subprocesses include jet functions infrared radiations soft functions infrared gluons heavy - quark kernels factorization @xmath0-quark jet from semi - leptonic top quark decay requires eikonal approximation for particle propagators infrared regions ward identity organizing diagrams attachments infrared gluonssoft functions neglected due cancellation corrections minor effects on jet mass distributions pqcd factorization formula for semi - leptonic decay polarized top quark convolution infrared - finite heavy - quark kernel with @xmath0-quark jet function overlap collinear soft dynamics produces large double logarithms in jet function references resummation double logarithms in @xcite qcd resummation technique for light - particle jets developed @xcite jet substructures mass distributions energy profiles calculated substituting quark jet function into pqcd factorization formula decay spin analyzing power increases with jet mass enhanced by factor 2 jet mass shape jet function weight contribution kinematic region higher-quark jet momentum spin analyzing power larger test enhancement spin analyzing power inclusion jet dynamics pqcd factorization of @xmath0-quark jet function decay performed up to nlo eikonal approximation in sec procedures ward identity all - order factorization outlined doubly differential width for angular distribution @xmath0-quark jet with invariant mass presented in sec .propose parametrization quark jet function in @xcite simplify numerical analysis choices physical parameters predictions summarized section [ conclusion ] devoted conclusion consider semi - leptonic top quark decay @xmath17 @xmath18 are momenta of top quark @xmath0 quark charged lepton neutrino indicated in fig . fig1 leading - order lo ) amplitude identify leading infrared contributions perform factorization into @xmath0-quark jet function averaged decay amplitude squared @xmath22 at lo electro - weak coupling strong coupling depicted in fig . fig2 written as @xmath23 with leptonic hadronic tensors @xmath24\mu\rho{tr\left\gamma^{}p_}{eq{aligned @xmath25 gauge coupling weak interaction @xmath26 cabibbo - kobayashi - maskawa matrix element @xmath27 momentum mass decay width @xmath2 boson @xmath30 summation polarizations @xmath2 boson @xmath31 projection matrix spin projector top quark @xmath33 @xmath34 masses top quark @xmath0 quark lepton masses dropped three - body phase - space integral top quark decay @xmath35 @xmath36 energy particle @xmath4 angle @xmath38 polar angle @xmath0 quark top quark spin fig azimuthal angle @xmath39 integrated @xmath40 angle @xmath41-@xmath42-@xmath43 plane performed pqcd factorization @xmath0-quark jet function from semi - leptonic decay polarized top quark eikonal approximation ward identity resultant formula convolution infrared - finite heavy - quark kernel jet function dominated infrared radiations heavy - quark kernel parameterizing jet function predicted dependence spin analyzing power @xmath0-quark jet on invariant massverified formalism reproduces fixed - order results integrating over jet mass investigation indicated spin analyzing power enhanced by factor 2 angular distribution @xmath0-quark jet with invariant mass mechanism enhancement elaborated attributed to broadening jet mass distribution with energy weight on contribution region higher jet energy larger spin analyzing power observation insensitive to jet radius worth experimental confrontation jet dynamics into polarized top quark decay future works nlo corrections to heavy - quark kernel need improve precision predictions dependence top quark mass definition on renormalization schemes essential possible extend formalism to hadronic decays polarized top quark direction develop formalism for boosted top quark at lhc @xcite decays into single jet substructures for polarized top quark jet improve extraction information top quark polarization can be calculated thank z.g . si for discussions work supported by national science council of r.o grant-101 - 2112-001 - national center for theoretical sciences | 0.354735 | 2,325 | 1,156 | , Saving $0.1 in GPT-4. |
67,796 | classical novae are potential gamma - ray emitters , because of the disintegration of some radioactive nuclei synthesized during the explosion . some short - lived isotopes ( such as @xmath0n and @xmath1f ) , as well as the medium - lived @xmath2na , decay emitting positrons , which annihilate with electrons and thus are responsible for the prompt emission of gamma - rays from novae .
this emission consists of a 511 kev line plus a continuum between 20 and 511 kev , and is released before the maximum in visual luminosity , i.e. , before the discovery of the nova .
the main characteristics of this prompt emission , together with the related uncertainties ( both of nuclear and hydrodynamical origin , with a particular emphasis on the influence of the envelope properties ) and prospects for detectability are analyzed in this paper .
gamma - ray astronomy , gamma - ray lines , nucleosynthesis , novae , cataclysmic variables | the explosive phenomenon of classical novae occurs on the surface of accreting white dwarfs , in close binary systems of the cataclysmic variable type , whenever certain initial conditions of white dwarf mass and luminosity and mass - accretion rate are met .
the hydrogen - rich accreted matter , mixed with matter from the underlying white dwarf core , burns in degenerate conditions , leading to a thermonuclear runaway and the ensuing explosion .
an important increase in visual luminosity and the ejection of a fraction of the accreted envelope are some of the consequences of the explosion .
the outburst is accompanied by the release of gamma - rays , with different time scales .
the prompt @xmath3-ray emission from classical novae has its origin in e@xmath4-e@xmath5 annihilation , with the positrons coming mainly from @xmath0n and @xmath1f decays ( see the pioneering work from @xcite ) .
the positrons emitted by @xmath2na decay also contribute to the prompt emission , but with a much lower flux than those from @xmath0n and @xmath1f , because of its much longer decay time .
in general , the released positrons annihilate and produce a line at 511 kev and a continuum below it .
the continuum is produced both by the positronium emission ( when it is formed in triplet state ) and by the comptonization of the photons emitted in the line ; it has a cut - off at around 20 - 30 kev , because of photoelectric absorption ( this mechanism has larger cross - sections than compton scattering at low energies ) .
the isotopes @xmath0n and @xmath1f have relatively short lifetimes , which makes the corresponding emission of short duration and tightly related to the conditions in the expanding envelope ( opacity to gamma - rays ) .
the prompt gamma - rays from novae are emitted very early in the explosive phase , i.e. , before the maximum in visual luminosity and , therefore , before optical discovery for the majority of novae .
there is another type of gamma - ray emission , related to line emission from the decay of medium - lived radioactive nuclei ( such as 478 kev from @xmath6be and 1275 kev from @xmath2na ) , lasting for months and years after the explosion , which has not been detected in any nova up to now @xcite , but its study is out of the scope of this paper ( see for instance the pioneering papers from @xcite , @xcite , and the recent ones from @xcite , @xcite and references therein ) .
complete evolution of various nova models ( both of the co and one type , depending on the chemical composition of the underlying white dwarf , which , in turn , depends on its core mass ) from the accretion phase up to the ejection one , has been computed , with a hydrodynamical code ( see @xcite , for details about the code ) .
it is important to stress that some nuclear reaction rates play a very important role on the synthesis of the most relevant radioactivities in nova explosions . in particular ,
@xmath1f synthesis is affected mainly by @xmath1f+p reactions ( @xmath1f(p,@xmath3 ) and @xmath1f(p,@xmath7 ) , which are still quite uncertain ( see @xcite and @xcite ) .
therefore , the final amount of @xmath1f , which almost directly translates into the flux of positron annihilation gamma rays emitted promptly , will be as close to reality as the measured nuclear cross sections are .
it is also important to treat accurately the initial composition of the underlying white dwarf core , specially for the case of the one novae , as well as the amount of mixing between accreted and white dwarf matter ( see discussion in @xcite ) . in table
[ radioac ] we show the radioactivities , relevant for the positron annihilation gamma rays , for a handful of computed nova models .
one important result is that similar amounts of @xmath1f are produced in both nova types , contrary to what happens with other longer lived radioctivities ( i.e. , @xmath6be , mainly produced in co novae , and @xmath2na , mainly produced in one novae ) .
the main properties of the emission of positron annihilation gamma rays from novae are displayed in figure [ lcannihil ] , where the time axis origin is at peak temperature ( which occurs before visual luminosity maximum ) . first we show ( figure [ lcannihil ] left ) the light curve of the 511 kev line , for the four models from table [ radioac ] .
although the amount of ejected @xmath1f is similar in all the models , the less massive co nova ( 0.8 m@xmath8 ) emits a smaller flux than the other ones ; that s because its outer expanding shells move at smaller velocities and , therefore , are more opaque to gamma - rays .
it is important to notice that there is also an initial peak in the light curves , related to @xmath0n decay , of even shorter duration than that produced by @xmath1f decay ( because of the shorter lifetime of @xmath0n as compared with @xmath1f ) .
another interesting aspect is that in one novae the 511 kev line has a longer duration , because of the contribution of the positrons from @xmath2na - decay : a low - level `` plateau '' becomes visible at the tail of the light curve ( see figure [ lcannihil ] ) .
however , when the envelopes finally become transparent ( one or two weeks after t@xmath9 ) e@xmath4 escape without annihilating .
a similar behavior is displayed by the continuum light curves ( see figure [ lcannihil ] right ) ; here one can see that there is much more flux in the continuum than in the 511 kev line ( which has widths fwhm between 3 and 8 kev ) .
other important factors , beyond the critical nuclear reaction rates mentioned above , concern aspects of the envelope . as an illustrative example
, the low mass co nova ( 0.8 m@xmath8 ) emits a flux in the 511 kev line , which is smaller than that emitted by other more massive novae by a factor larger than their ratio of @xmath1f yields ( see figure [ lcannihil ] left and table [ radioac ] ) . in order to study these factors ,
we varied ejected masses and velocity profiles in some of our models , leaving the abundances of radioactive material unchanged .
these models are interesting for illustrative purposes , although they are not self consistent .
an additional interest comes from the fact that all current theoretical nova models still fail to reproduce some observed ejected masses ( observed larger than theoretical ) .
the effect of ejected mass is shown in figure [ paramm ] for a co ( left ) and an one ( right ) nova . in both cases ,
the effect of ejected mass at early epochs is opposite to that at later times . at early times , the larger the ejected mass the larger the opacity of the envelope to gamma - rays ; therefore , novae with larger ejected masses emit smaller fluxes . on the contrary , at later times ( after @xmath101 day , see figure [ paramm ] ) , the envelope is more transparent and the fluxes almost directly reflect the amount of @xmath1f ( i.e. , the larger the ejected mass the larger the flux ) . in the one novae ,
the fluxes emitted in the `` plateau '' phase directly reflect the amount of @xmath2na ( thus the ratio of fluxes equals the ratio of ejected masses ) .
the influence of the velocity of the ejecta is displayed in figure [ paramv ] . at early epochs ,
the gamma - ray flux increases with envelope velocities because of increased transparency .
later on ( around 1 day after peak temperature ) , one novae differ from co novae , because of their much larger @xmath2na content : the decline of the gamma - ray flux in one novae is delayed by @xmath2na positrons annihilating in the envelope , until the moment when the envelope becomes transparent to positrons and also these annihilations fade away .
this happens some weeks later , the exact figure depending on the transparency of the envelope ; therefore , the smaller the velocity , the smaller the transparency at a given time and the longer the duration of this phase ( see figure [ paramv ] right ) .
we have shown that the continuum ( 20 - 511 kev ) and 511 kev line light curves provide a direct insight into the dynamics of the expanding envelope , as well as information about the nova type and its @xmath1f ( co and one novae ) and @xmath2na ( one novae ) content .
the fluxes in various continuum bands and in the 511 kev line are more intense than those in the 478 kev line and 1275 kev lines , but their short duration and early appearence makes them impossible to detect with standard pointed observations .
only all - sky monitors , able to do surveys in the range from 20 to 600 kev , are suited for such type of detection .
attempts have been made with the tgrs instrument onboard the wind satellite @xcite , the batse onboard the compton gamma - ray observatory @xcite .
the negative results obtained up to now are consitent with our predicted fluxes .
therefore , we have to wait for more sensitive instruments ( exist , advanced compton telescope ) , or hopefully the shield of the future integral spectrometer spi @xcite , to confirm the theoretical predictions and to extract all the crucial information about the nova phenomenon that only the gamma - rays can provide .
clayton d.d . , 1981 ,
apj , 244 , l97 clayton d.d . , hoyle f. , 1974 , apj , 187 , l101 coc a. , hernanz m. , jos j. , thibaud j.p .
, 2000 , a&a , 357 , 561 fishman g.j .
et al . , 1991 , in durouchoux p. , prantzos n. , eds .
, gamma - ray line astrophysics .
aip , new york , p. 190 gmez - gomar j. , hernanz m. , jos j. , isern j. , 1998 , mnras , 296 , 913 harris m.j . , leising m.d .
, share g.h .
, 1991 , apj , 375 , 216 harris m.j . ,
1996 , a&as , 120 , 343 harris m.j . ,
et al . 1999 ,
apj , 522 , 424 hernanz m. et al . , 1999 , in `` astronomy with radioactivities '' , eds .
r. diehl & d. hartmann , mpe report 274 , p. 111
hernanz m. , jos j. , coc a. , gmez
gomar j. , isern j. , 1999 , apj .
, 526 , l97 hernanz m. et al . , 2000 , in 5@xmath11 compton symposium .
aip , new york , 82 iyudin a.f .
, et al . 1995 , a&a , 300 , 422 jean p. et al . , 1999 ,
. lett . & comm . , 38 , 421 jos j. , hernanz m. , 1998 , apj , 494 , 680 leising m.d . , clayton d. , 1987 , apj , 323 , 159 leising m.d . , share g.h . ,
chupp e.l .
, kanbach g. , 1988 , apj , 328 , 755 | explosive phenomenon classical novae occurs on surface accreting white dwarfs close binary systems cataclysmic variable type initial conditions white dwarf mass luminosity mass - accretion rate met hydrogen - rich accreted matter mixed with white dwarf core burns leading to thermonuclear runaway explosion increase in visual luminosity ejection of fraction accreted envelope consequences outburst by release gamma - rays different time scales prompt @xmath3-ray emission origin in e@xmath4-e@xmath5 annihilation positrons from @xmath0n @xmath1f decays positrons by @xmath2na decay contribute to emission lower flux than longer decay time released positrons annihilate produce line at 511 kev continuum below continuum produced by positronium emission comptonization of photons cut - off at around 20 - 30 kev photoelectric absorption isotopes @xmath0n @xmath1f short lifetimes emission short duration related to conditions expanding envelope opacity to gamma - rays prompt gamma - rays from novae emitted early in explosive phasebefore maximum visual luminosity optical discovery for majority novae another type gamma - ray emission related to line emission from decay of medium - lived radioactive nuclei 478 kev from @xmath6be 1275 kev from @xmath2na lasting months years after explosion not detected in nova study out of this paper see pioneering papers @xcite recent ones references complete evolution of nova models co one type depending on chemical composition underlying white dwarf depends core mass from accretion phase to ejection computed with hydrodynamical code see @xcite details nuclear reaction rates synthesis relevant radioactivities in nova explosions @xmath1f synthesis affected by @xmath1f+p reactions ((p,@xmath3 )(p,@xmath7 ) uncertain @xcite final amount of @xmath1f translates into flux positron annihilation gamma rays close to reality measured nuclear cross sections important to treat accurately initial composition underlying white dwarf core novae amount mixing between accreted white dwarf matter see discussion @xcitetable [ radioac ] radioactivities for positron annihilation gamma rays computed nova models similar amounts @xmath1f produced in both nova types contrary to other longer lived radioctivities @xmath6be co @xmath2na one novae main properties emission positron annihilation gamma rays from novae displayed in figure [ lcannihil ] time axis origin at peak temperature before visual luminosity maximum light curve of 511 kev line for four models ejected @xmath1f similar less massive co nova ( 0.8 m@xmath8 ) emits smaller flux outer expanding shells move smaller velocities more opaque to gamma - rays initial peak in light curves related to @xmath0n decay shorter duration than @xmath1f decay shorter lifetime of @xmath0n in one novae 511 kev line longer duration positrons from @xmath2na - decay low - level plateau '' visible at tail of light curve envelopes transparent after t@xmath9 ) e@xmath4 without annihilatingsimilar behavior by continuum light curves figure lcannihil right more flux in continuum than 511 kev line widths between 3 and 8 kev important factors beyond nuclear reaction rates concern envelope low mass co nova ( 0.8 m@xmath8 ) emits flux in 511 kev line smaller than other massive novae larger than ratio of @xmath1f yields figure table varied ejected masses velocity profiles in models leaving radioactive material unchanged models interesting for illustrative not self consistent current theoretical nova models fail to reproduce some observed ejected masses larger than effect of ejected mass shown in figure [ paramm for co and one ) nova effect at early epochs opposite to later times early larger ejected mass larger opacity of envelope to gamma - rays novae with larger masses emit smaller fluxes at later times after @xmath101 day envelope more transparent fluxes reflect @xmath1f larger ejected mass larger flux in one novae fluxes in plateau phase reflect @xmath2na ratio of fluxes equals ratio of ejected massesinfluence velocity ejecta displayed in figure paramv early epochs gamma - ray flux increases with envelope velocities increased transparency later 1 day after peak temperature one novae differ from co novae larger @xmath2na content decline gamma - ray flux one novae delayed by @xmath2na positrons envelope until envelope transparent annihilations fade happens weeks later transparency envelope smaller velocity smaller transparency longer duration phase continuum ( 20 - 511 kev 511 kev line light curves provide insight dynamics expanding envelope information nova type @xmath1f @xmath2na content fluxes in continuum bands 511 kev line more intense than 478 kev 1275 kev lines short duration early appearence impossible to detect with standard observations all - sky monitors 20 to 600 kev suited for detection attempts with tgrs instrument wind satellite @xcite batse compton gamma - ray observatory @xcite negative results consitent with predicted fluxeswait sensitive instruments advanced compton telescope future integral spectrometer confirm predictions extract crucial information phenomenon gamma - rays clayton d.d 1981 apj 244 , l97 hoyle f 1974 apj 187 , l101 coc hernanz jos j thibaud j.p 2000 a 357 , 561 fishman g.j 1991 durouchoux p. prantzos n. gamma - ray line astrophysics new york p. 190 gmez - gomar j hernanz m jos isern j. 1998 296 , 913 harris m leising m share g.h 1991 apj 375 , 216 1996 120 , 343 1999 apj 522 , 424 1999 astronomy with radioactivities diehl hartmann report 274 p. 111 hernanz jos j coc gomar isern j. 1999 apj 526 , l97 hernanz 2000 5@xmath11 compton symposium new york 82 iyudin a.f .1995 a&a 300 422 jean p. 1999 lett comm 38 421 jos j. hernanz m 1998 apj 494 680 leising m.d clayton 1987 323 , 159 leising share chupp kanbach 1988 328 755 | 0.431367 | 2,554 | 1,323 | , Saving $0.1 in GPT-4. |
67,802 | we describe a data format currently in use amongst european institutions for exchanging and archiving pulsar data .
the format is designed to be as flexible as possible with regard to present and future compatibility with different operating systems .
one application of the common format is simultaneous multi - frequency observations of single pulses .
a data archive containing over 2500 pulse profiles stored in this format is now available via the _ internet _ , together with a small suite of computer programs that can read , write and display the data . | the * e*uropean * p*ulsar * n*etwork ( `` * epn * '' ) is an association of european astrophysical research institutes that co operate in the subject of pulsar research .
all institutes have up until now developed their own individual hardware and software facilities tailored to their own requirements and will , of course , continue to do so in future .
contact and co operation has always existed between the scientists of the member institutes and outside , but the lack of a common standard format for pulsar data has hampered previous collaborative research efforts . in this paper , we describe a flexible format that we have developed for exchanging data between epn pulsar groups .
the format has some generic similarities to the widely used fits format ( wells et al .
1981 ) but has been designed to meet the specific needs of the epn .
the format has proved so successful that we now advocate its use as a useful world wide utility for pulsar data exchange . to aid implementation of the format , we have written a suite of freely available _ fortran77 _ sub
routines which can be easily incorporated within existing software to read and write data in this format .
astrophysical applications of such a format currently being pursued by epn groups include the establishment of a data bank of pulse profiles as well as simultaneous observations of pulsars by several european observatories .
the underlying principles of the format result from a number of requirements .
this was essentially a balance between the need for efficient data storage and providing sufficient information about the data for potential users .
specifically , the following requirements had to be met : * * operating system independence : * to make the data format as portable as possible between present and future operating systems , we have opted to use only ascii data throughout .
we have arrange these data so that words are aligned over 80-byte boundaries , this simplifies inspection and printing of the files . * * completeness : * the data should contain all information for the identification of the source and the observing circumstances useful for further analyses of the data by others . * * compactness : * descriptive information should not dominate the format .
the measured values that form the bulk of a block of data are given as scaled four - character hexadecimal numbers , giving a dynamic range of up to 65536:1 .
* * versatility : * the format should be suitable for continuously sampled multi - channel filterbank search data , synchronous integrated and single
pulse data as well as processed data . in addition , we have designed the format , so that it can be used for observations of pulsars outside the radio regime _ i.e. _ variable units for the observing frequency and bandwidth , as well as topocentric telescope coordinates which are time variable for satellite observatories .
space is left for more descriptors , future adaptations and expansions . * * simplicity and ease of access : * we describe a data format consisting of a standardised fixed - length header with a variable length data structure attached to it .
the header fully describes the structure of the data , which is not changed within one file but can vary between files . in this way , it is possible to calculate the length of a data block within each file after reading its header .
the file can then be opened for random access with fixed block length , faster than a sequential read . many of the above mentioned requirements were already met by a format in use at jodrell bank to which we made suitable modifications and extensions to make it more flexible .
each epn file consists of one or more epn blocks .
the basic structure of an epn block is shown in fig .
1 . each file has a common fixed length _ header _ followed by a number of individual _ data streams _ of equal length .
the header describes the data , containing information on the pulsar itself , the observing system used to make the observation as well as some free - form information about the processing history of the data .
the onus is on the site
specific conversion process to ensure correct conversion to the standardised entries and reference to common catalogues ( e.g. the taylor et al . 1993 catalogue of pulsar parameters ) .
the full list of header variables is given in tables 1 and 2 .
the data streams themselves may be outputs of different polarisation channels , or individual channels ( bands ) of a filterbank or a combination thereof . in total , there may be @xmath0 data streams of i.e. different frequencies for each polarisation .
each data stream starts with a small , fixed length sub - header in front of the actual data values .
the number of data streams and their length may vary between different epn files , but is constant within each file .
a character field and an ordinal number is provided for each stream for its identification .
[ cols=">,^,^,^,<",options="header " , ]
pulsars are , in general , very weak sources , typically requiring the addition of several thousand individual pulses with a large radio telescope equipped with sensitive receivers in order to attain a sufficiently large signal - to - noise ratio .
the brightest pulsars are , however , strong enough so that individual pulses can be observed .
these pulses are known to exhibit great variety in morphology and polarimetric properties from one pulse to the next ( see for example lyne & smith 1990 ) .
it is presently unclear whether the same features in the individual pulses are present at different observing frequencies .
one of the current research topics being carried out by epn is a multi - frequency study of single pulses .
the project requires the pulses observed at different telescopes to be time - aligned and thus the format described above has an ideal application in this project .
after conversion of the data into this format , the time alignment of the pulses and subsequent statistical analyses is a relatively straight - forward procedure . as an example , a set of pulses from psr b0329 + 54 observed simultaneously at bologna ( 410 mhz ) , jodrell bank ( 1.404 ghz ) and effelsberg ( 4.850 ghz ) are shown in fig .
2 . the pulses show a remarkable similarity at these three frequencies , although counter examples are also observed .
full results of this study will be published shortly .
( 0,2.7 ) ( -.2,6.5 ) ( -.2,5.0 )
as well as being used for data interchange between epn members , the common format forms the basis of a pulse profile archive presently being maintained at the max - planck - institut fr radioastronomie in bonn .
the idea of the archive is to build up a useful collection of pulse profiles which anybody with access to the _ internet _ can use .
presently , around 2500 pulse profiles are stored in this format .
the profiles themselves have usually already been , or about to be , published so that full credit for any subsequent use via the database can go to the contributing authors .
the archive has the following url address : _ http://www.mpifr - bonn.mpg.de / pulsar / data/_. authors are encouraged to make their data available to this archive and should contact duncan lorimer ( email : _
dunc@mpifr-bonn.mpg.de_ ) if they wish to do this .
we thank the referee , dick manchester , for useful comments on an earlier version of the manuscript .
the epn was funded under brussels human capital and mobility grant number chrx - ct94 - 0622 .
lyne a. g. , smith f. g. , 1990 , pulsar astronomy .
cambridge university press taylor j. h. , manchester r. n. , lyne a. g. , 1993 , apjs , 88 , 529 wells d. c. , greisen e. w. , harten r. h. , 1981 , a&a supp .
ser , 44 , 363
to incorporate the capability to read and write data in this format within existing analysis software , a simple routine exists which can read and write data in this format . in addition , we have written some sample programs which can plot the data and display the header parameters .
the software are written in _
fortran77 _ and have been packaged into a single unix tar file which is freely available via the _ internet_. to down - load the package , log into the anonymous ftp area : * ftp.mpifr-bonn.mpg.de * , with the username * anonymous * using your complete e
mail address as the password .
once logged in , issue the following commands : the present package contains some sample data and two example programs `` plotepn '' and `` viewepn '' which plot and view epn files respectively .
the ascii file * 00readme * in this packages gives further details of the software and how to use it . | e*uropean * p*ulsar * n*etwork ( epn ) is association of european astrophysical research institutes co operate in pulsar research all institutes developed own hardware software facilities will continue future contact co operation existed between member institutes lack of common standard format for pulsar data hampered collaborative research efforts this paper flexible format developed for exchanging data between epn pulsar groups format similarities to fits format ( wells et 1981 ) designed to meet specific needs epn format successful advocate use as world wide utility for pulsar data exchange written suite of freely available _ fortran77 _ sub routines incorporated within software to read write data format astrophysical applications of format include data bank of pulse profiles simultaneous observations of pulsars by several european observatories principles of format result from balance between efficient data storage sufficient information for potential users requirements operating system independence format portable between systems only ascii data data words aligned over 80-byte boundaries simplifies inspection printing completeness data should contain information for identification of source observing circumstances for further analyses compactness descriptive information should not dominate formatmeasured values bulk block data as scaled four - character hexadecimal numbers dynamic range up to 65536:1 versatility format for continuously sampled multi - channel filterbank search data synchronous integrated single pulse data processed data designed format for observations of pulsars outside radio regime variable units for observing frequency bandwidth topocentric telescope coordinates time variable for satellite observatories space left for more descriptors future adaptations expansions simplicity ease of access data format standardised fixed - length header with variable length data structure header describes structure data vary between files possible to calculate length data block each after header file opened for random access with fixed block length faster than sequential read requirements met by format at jodrell bank modifications extensions flexible each epn file of one or more epn blocks structure shown in fig . 1 each file common fixed length header followed by individual data streams of equal length header describes data information on pulsar observing system processing history onus on site specific conversion process correct conversion to standardised entries reference to common catalogues full list of header variables in tables 1 and 2data streams be outputs of different polarisation channels or filterbank or total @xmath0 data streams of different frequencies for each polarisation each stream starts with small fixed length sub - header front of data values number of data streams length vary between epn files but constant within each file character field ordinal number provided for each stream for identification cols=" pulsars weak sources several thousand individual pulses with large radio telescope large signal - to - noise ratio brightest pulsars strong enough individual pulses be observed pulses exhibit variety in morphology polarimetric properties unclear whether same features in pulses at different observing frequencies current research multi - frequency study of single pulses project requires pulses at time - aligned format described ideal application after conversion data into format time alignment pulses statistical analyses straight - forward set of pulses from psr b0329 + 54 observed at bologna jodrell bank effelsberg shown in fig . 2 pulses show similarity at these three frequencies counter examples also observed full results will be published shortly0,2.7 ) -.2,6.5 ) -.2,5.0 ) used for data interchange between epn members common format basis pulse profile archive at max - planck - institut fr radioastronomie bonn idea collection pulse profiles internet around 2500 pulse profiles stored in format profiles published credit for use to contributing authors archive http://www.mpifr - bonn.mpg.de / pulsar / data/ authors data available contact duncan lorimer dunc@mpifr-bonn.mpg.de thank referee dick manchester for comments epn funded under brussels human capital mobility grant chrx - ct94 - 0622 lyne a. g. smith f. g. 1990 pulsar astronomy cambridge university press taylor j.. manchester r. lyne a.. 1993 apjs , 88 , 529 wells d greisen harten r.. 1981 a&a supp . ser , 44 , 363 capability read write data format analysis software routine write written sample programs plot data display header parameters software written in fortran77 packaged into single unix tar file freely available via internetdown - load package log anonymous ftp area ftp.mpifr-bonn.mpg.de username anonymous complete e mail address password logged in issue commands present package contains sample data two example programs plotepn viewepn '' plot view epn files ascii file 00readme gives details software use | 0.317611 | 1,868 | 904 | , Saving $0.1 in GPT-4. |
67,812 | despite a wealth of experimental data for high-@xmath0 processes in heavy - ion collisions , discriminating between different models of hard parton - medium interactions has been difficult .
a key reason is that the pqcd parton spectrum at rhic is falling so steeply that distinguishing even a moderate shift in parton energy from complete parton absorption is essentially impossible .
in essence , energy loss models are effectively only probed in the vicinity of zero energy loss and , as a result , at rhic energies only the pathlength dependence of energy loss offers some discriminating power . at lhc
however , this is no longer the case : due to the much flatter shape of the parton @xmath1 spectra originating from 2.76 @xmath2gev collisions , the available data probe much deeper into the model dynamics .
a simultaneous fit of the nuclear suppression at both rhic and lhc energies thus has great potential for discriminating between various models that yield equally good descriptions of rhic data alone . | high-@xmath0 processes taking place in the background of the medium produced in ultrarelativistic heavy - ion ( a - a ) collisions are a cornerstone of the experimental a - a program at the lhc .
the aim is to use these processes to do `` jet tomography '' , i.e. to study both the short - distance physics of the bulk medium ( i.e. its relevant degrees of freedom ) and the geometry of its expansion .
however , so far attempts to extract even solid qualitative statements about the nature of parton - medium interaction have not been successful .
two main reasons can be identified : 1 ) for steeply falling primary parton spectra , medium - induced shifts in parton energy can not be distinguished from parton absorption , hence observables lose sensitivity to model details ; 2 ) computed observable quantities depend both on assumptions made about the bulk medium evolution and the parton - medium interaction . in this work
, we propose to resolve the second ambiguity by a systematic investigation of multiple observables , while we demonstrate that the first is significantly lessened for the harder lhc parton kinematics .
we test different combinations of medium evolution and parton - medium interaction models against a large body of high @xmath0 observables .
in particular , for the medium evolution we use a 3 + 1d ideal @xcite , a 2 + 1d ideal @xcite and a 2 + 1d viscous hydro code @xcite with both cgc and glauber initial conditions . on the parton - medium interaction side ,
we test a radiative energy loss model @xcite , a parametrized @xcite and a monte - carlo ( mc ) model @xcite for incoherent energy loss , a strong - coupling phenomenological model based on ads / cft ideas @xcite and the mc in - medium shower code yajem @xcite with its variant yajem - d @xcite which introduces an explicit pathlength / energy dependence into the minimum virtuality scale down to which the shower is evolved . in a first run ,
we pose the question to what degree the underlying medium model is able to influence high @xmath0 observables . in fig .
[ f-1 ] we present an example of results from a systematic investigation of both the influence of smooth hydrodynamical models @xcite and event - by - event hydrodynamics @xcite with initial state density fluctuations @xcite . summarizing the result , we find that the medium model has a considerable ( factor @xmath3 ) influence on observables such as @xmath4 at high @xmath0 or extracted parameters such as the transport coefficient @xmath5 .
chiefly responsible is the location of the freeze - out hypersurface the agreement with data in general improves if the freeze - out hypersurface is large , but noticeable effects are also caused by the initialization time or the presence / absence of viscosity .
fluctuations in the hydrodynamical initial state play a minor role for extracted parameters ( @xmath6 20% ) , due to a cancellation of competing effects ( see @xcite ) , but for non - central collisions the cancellation is incomplete , leading to a decrease in suppression for small fluctuation size scale . if hard probes are used to constrain the fluctuation size , a scale of @xmath7 fm is preferred .
in a second run , we test the pathlength dependence of the available parton - medium interaction models against the data @xcite . in a static medium with constant density , we expect incoherent processes to scale with pathlength @xmath8 ( elastic ) , radiative energy loss ( asw ) with @xmath9 due to coherence effects and the strong coupling model ( ads ) with @xmath10 due to the drag - force like interaction of the virtual gluon cloud with the medium .
the shower code yajem is known to have in principle an @xmath9 dependence due to coherence , but which effectively reduces to @xmath8 by finite - energy corrections , whereas yajem - d has a complicated non - linear pathlength dependence . in an evolving hydrodynamical medium ,
the pathlength dependence is effectively much more complicated due to effects like spatial inhomogeneities in the medium , longitudinal and transverse flow and viscous entropy production .
we find that the data allow to unambiguously rule out linear pathlength dependence , leading to the conclusions that a large component ( @xmath11% ) of elastic energy loss is not favoured by the data and that finite - energy corrections to coherence arguments need to be taken seriously .
the other models we tested ( asw , ads and yajem - d ) remain viable with the data , although in each case only in combination with a particular hydrodynamical evolution model .
using the ekrt saturation model , we can extrapolate one of our default hydro runs to lhc energies @xcite and thus significantly reduce the uncertainty in the hydrodynamical modelling , as the well - defined extrapolation procedure allows to compare results for the same hydrodynamics at different @xmath12 . fig .
[ f - lhc ] shows results for @xmath13 in central and non - central collisions .
as expected , the sensitivity to model details in the @xmath0 dependence of the nuclear suppression factor is found to be much larger at lhc than at rhic . combined with pathlength - dependent observables such as @xmath14 or the dihadron suppression factor @xmath15
, precision lhc data thus have a high potential for distinguishing between different models .
we can quantify the quality of extrapolation in @xmath12 by the factor @xmath16 which quantifies the differences of parton - medium interaction parameters for a best fit to rhic and lhc data , using the same hydrodynamical models .
@xmath17 indicates an extrapolation without any tuning .
we find that the shower codes yajem - d ( @xmath18 ) and yajem ( @xmath19 ) extrapolate reasonably well , whereas the radiative energy loss scenario asw ( @xmath20 ) is not favoured by the data , and a strongly coupled scenario ads is strongly disfavoured with @xmath21 .
the latter effect can be readily understood by a dimensional analysis if the pathlength dependence is @xmath10 , then for dimensional reasons the model must probe the medium temperature as @xmath22 , i.e. the model responds much more strongly than all others to the higher initial medium density at the lhc , which leads to overquenching . given this finding , there is currently no reason to assume that the data would prefer a strongly coupled over a perturbative scenario of parton - medium interaction .
combining the constraints from pathlength dependence and @xmath12 extrapolation , assuming no systematic uncertainty on the published lhc data @xcite , out of the models tested here only yajem - d together with a hydrodynamics similar to the 3 + 1d code remains a viable description .
we take this as a strong indication that systematic studies along the lines discussed here are indeed a suitable tool to do jet tomography . *
acknowledgments : * this work was supported by the finnish academy ( projects 130472 and 133005 ) , the finnish graduate school of particle and nuclear physics , the jenny and antti wihuri foundation , the magnus ehrnrooth foundation and the u.s .
department of energy ( grants de - sc004286 and ( within the framework of the jet collaboration ) de - sc0004104 ) .
99 c. nonaka , s. a. bass , phys .
rev . * c75 * , 014902 ( 2007 ) . | high-@xmath0 processes heavy - ion collisions cornerstone experimental - program at lhc aim jet tomography study short - distance physics bulk medium geometry expansion attempts statements parton - medium interaction successful two reasons falling parton spectra medium - induced shifts parton energy distinguished from parton absorption observables lose sensitivity computed observable quantities depend on assumptions bulk medium evolution parton - medium interaction propose resolve second ambiguity by investigation multiple observables first lessened for lhc parton kinematics test medium evolution parton - medium interaction models against high @xmath0 observables medium evolution 3 + 1d ideal @xcite 2 + 1d ideal @xcite 2 + 1d viscous hydro code @xcite with cgc glauber initial conditions parton - medium interaction test radiative energy loss model @xcite parametrized @xcite monte - carlo mc model @xcite for energy loss strong - coupling phenomenological model ads / cft ideas mc in - medium shower code yajem @xcite variant - pathlength / energy dependence into minimum virtuality scalefirst run question medium model high @xmath0 observables fig . f-1 results influence smooth hydrodynamical models @xcite event - by - event hydrodynamics with initial state density fluctuations medium model observables @xmath4 high @xmath0 transport coefficient @xmath5 responsible location freeze - out hypersurface agreement data improves if large effects initialization time presence viscosity fluctuations hydrodynamical initial state minor role for extracted parameters @xmath6 20% ) cancellation competing effects non - central collisions cancellation incomplete decrease suppression for small fluctuation size scale constrain fluctuation size scale of @xmath7 fm preferred second run test pathlength dependence parton - medium interaction models against data @xcite static medium constant density incoherent processes scale with pathlength @xmath8 radiative energy loss with @xmath9 coherence strong coupling model with @xmath10 drag - force interaction virtual gluon cloud shower code yajem @xmath9 dependence due to coherence reduces to @xmath8 by finite - energy corrections yajem - d complicated non - linear pathlength dependenceevolving hydrodynamical medium pathlength dependence complicated due to spatial inhomogeneities longitudinal flow viscous entropy production data rule out linear pathlength dependence large component ( @xmath11% ) elastic energy loss not favoured finite - energy corrections coherence other models asw ads yajem - d viable with hydrodynamical evolution model ekrt saturation model extrapolate default hydro runs to lhc energies reduce uncertainty hydrodynamical modelling results @xmath12 fig . shows results @xmath13 in central non - central collisions sensitivity to model details @xmath0 dependence nuclear suppression factor larger at lhc than rhic with pathlength - dependent observables @xmath14 dihadron suppression factor @xmath15 precision lhc data high potential distinguishing models quantify quality extrapolation @xmath12 by factor @xmath16 differences parton - medium interaction parameters best fit rhic lhc data @xmath17 indicates extrapolation without tuningshower codes yajem - d @xmath18 yajem @xmath19 ) extrapolate well radiative energy loss scenario asw @xmath20 ) not favoured data strongly coupled scenario disfavoured with @xmath21 effect understood by dimensional analysis pathlength dependence @xmath10 model must probe medium temperature as @xmath22 responds to higher initial medium density leads to overquenching no reason to assume data strongly coupled over perturbative scenario parton - medium interaction constraints pathlength dependence @xmath12 extrapolation no systematic uncertainty lhc data yajem - d hydrodynamics similar to 3 + 1d code viable systematic studies suitable jet tomography work supported by finnish academy finnish graduate school of particle nuclear physics jenny antti wihuri foundation magnus ehrnrooth foundation u.s . department of energy grants de - sc004286 de - sc0004104 nonaka a bass phys . rev. c75 014902 ( 2007 | 0.267723 | 1,632 | 843 | , Saving $0.0 in GPT-4. |
67,817 | central collisions induced by high energy radioactive beams can be used as a novel means to obtain crucial information about the high density ( hd ) behaviour of nuclear symmetry energy .
this information is critical for understanding several key issues in astrophysics . within an isospin - dependent hadronic transport model using phenomenological equations of state ( eos ) for dense neutron - rich matter
, we investigate several experimental probes of the hd behavior of nuclear symmetry energy , such as , the @xmath0 to @xmath1 ratio , neutron - proton differential flow and its excitation function .
measurements of these observables will provide the first terrestrial data to constrain stringently the hd behaviour of nuclear symmetry energy and thus also the eos of dense neutron - rich matter . | nuclear reactions using rare isotopes has opened up several new frontiers in nuclear sciences@xcite . in particular ,
high energy heavy rare isotopes to be available at the future rare isotope accelerator ( ria ) and the new gsi accelerator facility provide a unique opportunity to explore novel properties of dense neutron - rich matter that was not in reach in terrestrial laboratories before .
this exploration will reveal crucial information about the @xmath2 of neutron - rich matter . to understand the latter and its astrophysical implications , such as , the origin of elements , structure of rare isotopes and properties of neutron stars are presently among the most important goals of nuclear sciences .
the @xmath2 of neutron - rich matter of isospin asymmetry @xmath3 can be written as @xmath4 within the parabolic approximation ( see e.g. , @xcite ) , where @xmath5 is the energy per nucleon in isospin symmetric nuclear matter and @xmath6 is the symmetry energy at density @xmath7 . to study the @xmath6 has been a longstanding goal of extensive research with various microscopic and/or phenomenological models over the last few decades .
the predicted results on the density dependence , especially at high densities , are extremely diverse and often contradictory .
theoretical results can be roughly classified into two groups , i.e. , a group where the @xmath6 rises monotonously and one in which it falls with the increasing density above about twice the normal nuclear matter density . above @xmath8 ,
even the sign of the symmetry energy is unclear .
the fundamental cause of the extremely uncertain hd behaviour of @xmath6 is the complete lack of terrestrial laboratory data to constrain directly the model predictions .
recently , several promising observables were proposed to extract the hd behavior of the symmetry energy using high energy radioactive beams@xcite .
in a neutron star , several of its main properties are affected significantly by the correlation between the density and the proton fraction @xmath9 at @xmath10 equilibrium .
the latter is determined by@xcite @xmath11 the equilibrium proton fraction is therefore entirely determined by the @xmath6 .
we use the following two representatives of the symmetry energy @xmath12 where @xmath13 .
the values of @xmath14 corresponding to these two forms of symmetry energy are shown in the insert of fig . 1 .
with the @xmath15 , the @xmath16 is @xmath17 for @xmath18 , indicating that the neutron star has become a pure neutron matter at these high densities . on the contrary , with the @xmath19 , the neutron star becomes more proton - rich as the density increases . to see the connection between the physics governing properties of both neutron stars and heavy - ion collisions , we show in fig .
1 the correlation between density @xmath7 and the isospin asymmetry @xmath20 ( effective isospin asymmetry taking into account different charge states of @xmath21 resonances ) over the entire reaction volume at the time of about the maximum compression in the central @xmath22 reactions with e / a=400 ( upper window ) and 2000 ( bottom window ) mev / nucleon , respectively .
these results were obtained using an isospin - dependent hadronic transport model@xcite .
the overall rise of @xmath23 at low densities is mainly due to the neutron skins of the colliding nuclei and the distillated neutrons .
effects due to the different symmetry energies are more clearly revealed especially at high densities .
an astonishing similarity is seen in the resultant @xmath24 correlations for the neutron star and the heavy - ion collision . in both cases ,
the symmetry energy @xmath25 makes the hd nuclear matter more neutron - rich than the @xmath26 and the effect grows with the increasing density .
of course , this is no surprise since the same underlying nuclear @xmath2 is at work in both cases .
the decreasing @xmath25 above @xmath27 makes it more energetically favorable to have the denser region more neutron - rich in both neutron stars and heavy - ion collisions .
it is well known that the @xmath28 ratio in heavy - ion collisions depends strongly on the isospin asymmetry of the reaction system .
it is also qualitatively easy to understand why this dependence can be used to extract crucial information about the eos of neutron - rich matter . on one hand , within the @xmath29 resonance model for pion production from first - chance independent nucleon - nucleon collisions@xcite ,
the primordial @xmath28 ratio is @xmath30 .
it is thus a direct measure of the isospin asymmetry @xmath31 of the dense matter in the participant region of heavy - ion collisions .
it was shown that the @xmath31 is uniquely determined by the high density behaviour of the nuclear symmetry energy@xcite .
therefore , the @xmath28 ratio can be used to probe sensitively the eos of neutron - rich matter . on the other hand , within the statistical model for pion production@xcite ,
the @xmath28 ratio is proportional to @xmath32 $ ] , where t is the temperature , @xmath33 and @xmath34 are the chemical potentials of neutrons and protons , respectively . at modestly high temperatures ( @xmath35 mev )
, the difference in the neutron and proton chemical potentials can be written as@xcite @xmath36,\ ] ] where @xmath37 is the coulomb potential for protons , @xmath38 is the thermal wavelength of a nucleon and @xmath39s are the inversion coefficients of the fermi distribution function .
the difference in neutron and proton symmetry potentials @xmath40 , where the function @xmath41 is completely determined by the density - dependence of the symmetry energy@xcite
. it is seen that the kinetic part of the difference @xmath42 relates directly to the isospin asymmetry @xmath43 or @xmath44 .
thus within the statistical model too , the @xmath28 ratio is sensitive to the @xmath31 .
moreover , the value of @xmath28 ratio is affected by the competition of the symmetry and coulomb potentials which all depend on the isospin asymmetry of the reaction system . at relatively low temperatures
there is a good chance to extract crucial information about the symmetry potential . shown in fig . 2 are the isospin asymmetry of the high density region @xmath45 ( left ) and the @xmath46 ratio ( right ) @xmath47 as a function of time for the central @xmath22 reaction at beam energies from 200 to 1000 mev / nucleon .
this ratio naturally becomes the final @xmath28 ratio at the freeze - out when the reaction time @xmath48 is much longer than the lifetime of the delta resonance @xmath49 .
the @xmath46 ratio is rather high in the early stage of the reaction because of the large numbers of neutron - neutron scatterings near the surfaces where the neutron skins of the colliding nuclei overlap .
it is seen that a variation of about 30% in the @xmath50 due to the different @xmath6 results in about 15% change in the final @xmath28 ratio .
it has thus an appreciable response factor of about 0.5 to the variation of hd
n / p ratio and is approximately beam energy independent .
therfore , the @xmath46 ratio is a sensitive probe of the hd behaviour of nuclear symmetry energy .
the neutron - proton differential collective flow is measured by@xcite @xmath51 where @xmath52 is the total number of free nucleons at the rapidity @xmath53 , @xmath54 is the transverse momentum of particle @xmath55 in the reaction plane , and @xmath56 is @xmath57 and @xmath58 for neutrons and protons , respectively . the free nucleons are identified as those having local nucleon densities less than @xmath59 .
the @xmath60 combines constructively the in - plane transverse momenta generated by the isovector potentials while reducing significantly influences of the isoscalar potentials of both neutrons and protons .
thus , it can reveal more directly the hd behaviour of @xmath6 in high energy heavy - ion collisions . a typical analysis for mid - central @xmath22 reactions at 400
mev / nucleon is made in the upper window of fig .
a clear signature of the hd behaviour of @xmath6 appears at both forward and backward rapidities .
the slope @xmath61 at mid - rapidity is different by about a factor of 2 in reactions at @xmath62 mev / nucleon as shown in the lower window of fig .
this large effect can be easily observed by using available detectors at several heavy - ion facilities in the world .
in summary , the hd behaviour of nuclear symmetry energy has been puzzling physicists for decades . in this work ,
central collisions induced by high energy radioactive beams are proposed as a novel means to solve this longstanding problem . within an isospin dependent hadronic transport model using two representative density dependent symmetry energy functions predicted by many body theories ,
several experimental probes for the hd symmetry energy are studied . among them , the @xmath28 ratio , neutron - proton differential collective flow and its excitation function are found most promising .
future comparisons between the experimental data and model calculations will constrain stringently the hd behaviour of nuclear symmetry energy . | nuclear reactions using rare isotopes opened new frontiers in nuclear sciences high energy heavy rare isotopes at future rare isotope accelerator new gsi accelerator facility provide opportunity explore properties dense neutron - rich matter exploration crucial information about @xmath2 of neutron - rich matter understand astrophysical implications origin of elements structure rare isotopes properties of neutron stars important goals nuclear sciences @xmath2 neutron matter isospin asymmetry written as @xmath4 parabolic approximation @xmath5 energy per nucleon in isospin symmetric nuclear matter @xmath6 symmetry energy at density study @xmath6 goal research predicted results on density dependence at high densities diverse often contradictory results classified into two groups @xmath6 rises monotonously falls with increasing density above normal density above @xmath8 symmetry energy unclear uncertain behaviour of @xmath6 lack of terrestrial laboratory data promising observables proposed to extract behavior symmetry energy using high energy radioactive beamsneutron star properties affected by correlation density proton fraction @xmath9 @xmath10 equilibrium latter determined by@xcite @xmath11 equilibrium proton fraction determined by @xmath6 two representatives symmetry energy @xmath12 @xmath13 values @xmath14 shown in fig . 1 @xmath15 @xmath16 @xmath17 for @xmath18 neutron star pure neutron matter at high densities @xmath19 star more proton - rich as density increases heavy - ion collisions fig . 1 correlation between density @xmath7 isospin asymmetry @xmath20 over reaction volume maximum compression in central @xmath22 reactions with e / a=400 2000 mev / nucleon results obtained using isospin - dependent hadronic transport model@xcite rise of @xmath23 at low densities due to neutron skins colliding nuclei distillated neutrons effects symmetry energies at high densities similarity in @xmath24 correlations for neutron star heavy - ion collision symmetry energy @xmath25 makes nuclear matter more neutron - rich grows with increasing densityno surprise nuclear @xmath2 both cases decreasing @xmath25 above @xmath27 favorable denser region neutron - rich in neutron stars heavy - ion collisions @xmath28 ratio in heavy - ion collisions depends on isospin asymmetry reaction system dependence information about eos neutron - rich matter @xmath29 resonance model for pion production collisions primordial @xmath28 ratio is @xmath30 direct measure of isospin asymmetry @xmath31 of dense matter region @xmath31 determined by high density behaviour nuclear symmetry energy@xcite @xmath28 ratio probe eos of neutron - rich matter statistical model for pion production@xcite @xmath28 ratio proportional to @xmath32 t temperature @xmath33 @xmath34 chemical potentials of neutrons protons at high temperatures difference in neutron proton chemical potentials written as@xcite @xmath36 @xmath37 coulomb potential for protons thermal wavelength nucleon inversion coefficients of fermi distribution function difference in neutron proton symmetry potentials determined by density - dependence symmetry energy@xcitekinetic difference @xmath42 relates to isospin asymmetry @xmath43 or @xmath44 @xmath28 ratio sensitive to @xmath31 value affected by symmetry coulomb potentials on isospin asymmetry reaction system low temperatures information symmetry potential . 2 isospin asymmetry high density region @xmath45 @xmath46 ratio function time central @xmath22 reaction at beam energies 200 to 1000 mev / nucleon ratio becomes final @xmath28 ratio at freeze - out reaction time @xmath48 longer than lifetime delta resonance @xmath49 @xmath46 ratio high early stage reaction large neutron scatterings variation 30% in @xmath50 @xmath6 results in 15% change in final @xmath28 ratio response factor 0.5 to variation hd n / p ratio beam energy independent @xmath46 ratio sensitive nuclear symmetry energy neutron - proton differential flow measured by @xmath51 @xmath52 free nucleons at rapidity @xmath53 @xmath54 transverse momentum of particle @xmath55 reaction is @xmath57 @xmath58 for neutrons protonsfree nucleons local nucleon densities less than @xmath59 @xmath60 combines - plane transverse momenta isovector potentials isoscalar potentials neutrons protons hd behaviour @xmath6 in high energy heavy - ion collisions analysis mid - central @xmath22 reactions at 400 mev / nucleon upper signature hd behaviour @xmath6 forward backward rapidities slope @xmath61 mid - rapidity different factor 2 in reactions @xmath62 mev / nucleon lower effect observed detectors at heavy - ion facilities hd behaviour nuclear symmetry energy physicists central collisions high energy radioactive beams isospin dependent hadronic transport model density dependent symmetry energy functions experimental probes hd symmetry energy studied @xmath28 ratio neutron - proton differential collective flow excitation function promising future comparisons data model calculations constrain hd behaviour nuclear symmetry energy | 0.315768 | 2,014 | 1,038 | , Saving $0.1 in GPT-4. |
67,818 | we study all possible observables in @xmath0 with new physics ( np ) , including new vector , scalar and tensor interactions , and investigate the prospects of extracting np wilson coefficients with optimal observables .
analysis of the full @xmath1 integrated branching fractions of @xmath0 show that the overall sensitivity of the observables of @xmath2 is more towards the scalar current , whereas the bin - by - bin analysis of @xmath1 distribution of the differential branching fraction points to regions of @xmath1 sensitive to tensor interactions .
interestingly , the observables in @xmath3 are more sensitive to tensor interactions , and bin - by - bin analysis of this mode shows the distinct regions of @xmath1 sensitive to vector or scalar interactions .
in addition to that , the @xmath4 polarisation asymmetry is found to be more sensitive to np compared to the other observables , in both decay modes . | measurements of branching fractions and other related observables in semileptonic decays of @xmath5 meson to @xmath4 can be interesting for an indirect probe of np .
earlier measurements on @xmath6 by belle @xcite and babar collaborations @xcite have shown some deviations from their standard model(sm ) predictions @xcite , indicating a possible signature of np in @xmath7 transitions .
several authors have tried to explain the observation in various np scenarios @xcite , as well as in a model - independent way @xcite . in order to distinguish between the possible signatures of np , the study of np in @xmath1 distribution of differential branching fractions in @xmath0 , various correlations among @xmath4 forward - backward asymmetry and
@xmath4 polarization asymmetry has been examined in the ref .
recently , a 2.1@xmath8 deviation in the measurement of @xmath9 has been reported by lhcb collaboration @xcite ; belle collaboration has also announced their most recent results on @xmath10 and measured values are consistent with the sm within error bars @xcite .
so far , the constraints on the new couplings are obtained assuming their presence one at a time @xcite .
if we consider all the interactions together , then it will be an impossible task to extract all the couplings from a single measurement .
however , if one can reduce the number of coupling parameters by imposing certain constraints on the full set of parameters , only then it is possible to obtain meaningful errors on the couplings , although the information lost due to various assumptions can not be retrieved .
therefore , it will be useful to have independent couplings , parametrised in such a way that the measured errors on different parameters are uncorrelated .
on the other hand , it is not necessary for a particular observable to have equal sensitivity to different types of np operators . therefore , it is useful to know how an observable can be optimised to guarantee the maximal sensitivity to a particular type of np interaction , which in turn will help us select observables suitable for the extraction of a particular type of coupling . hence , from a phenomenological point of view
, it is important to find out the significance of different types of np interaction , to an observable . to achieve this goal
, we use the optimal - observable analysis using the invariant mass squared @xmath1 , of the lepton - neutrino system .
we construct the optimal observable to identify the np structure that can be best extracted from a particular observable , with reasonable statistics .
it also provides a deeper understanding of the sensitivity that can be best obtained by any method , for a certain process .
this technique has been widely used in collider phenomenology @xcite . in this article , we analyse the @xmath1 distribution of the differential branching fractions , @xmath10 , the @xmath4 polarisation asymmetries , forward backward asymmetries of the decay @xmath0 , and @xmath11 polarisation asymmetry .
we include all possible non - standard four - fermi effective interactions of the lowest dimension , and estimate the expected statistical uncertainties in the extraction of various np wilson coefficients that can contribute to @xmath0 .
the most general effective hamiltonian describing the @xmath12 transitions with all possible four - fermi operators in the lowest dimension is given by @xcite @xmath13 , \label{eq1}\end{aligned}\ ] ] where the operator basis is defined as _
v_1 & = & ( |c_l ^b_l)(|_l _ _ l ) , o_v_2 = ( |c_r ^b_r)(|_l _ _ l ) , + o_s_1 & = & ( |c_l b_r)(|_r _ l ) , o_s_2 = ( |c_r b_l)(|_r _ l ) , + o_t & = & ( |c_r ^ b_l)(|_r _ _ l ) , [ eq2 ] and the corresponding wilson coefficients are given by @xmath14 . in this basis , neutrinos are assumed to be left handed .
our main focus is on the @xmath1 distribution of differential decay rate @xmath15 in @xmath0 .
the complete expressions are given in ref.@xcite .
as mentioned earlier , the optimal - observable analysis is a technique to systematically estimate the statistical uncertainties of the measurable parameters while extracting them from some observable .
elaborate discussions on this technique can be found in references @xcite . in order to apply this technique to @xmath0 ,
it is necessary to express the @xmath1 distribution of the differential decay rate as = _
i c_i f_i(q^2 ) , [ eq3 ] where @xmath16s are functions of @xmath17s .
the theoretical expressions for @xmath16s , along with the @xmath18s , can be extracted from a direct comparison between the similar terms on both sides of eq.([eq3 ] ) .
the coefficients @xmath16 , relevant for the branching fractions in @xmath0 , are given in table [ cird ] , and the corresponding @xmath18s are given in the appendix ( table [ fisrd ] ) .
i. adachi _ et al . _
[ belle collaboration ] , arxiv:0910.4301 [ hep - ex ] .
j. p. lees _ et al .
_ [ babar collaboration ] , phys .
d * 88 * , no . 7 , 072012 ( 2013 ) [ arxiv:1303.0571 [ hep - ex ] ] .
s. fajfer , j. f. kamenik and i. nisandzic , phys .
d * 85 * , 094025 ( 2012 ) [ arxiv:1203.2654 [ hep - ph ] ] .
m. tanaka , z. phys .
c * 67 * , 321 ( 1995 ) [ hep - ph/9411405 ] .
y. sakaki , m. tanaka , a. tayduganov and r. watanabe , phys .
d * 88 * , no .
9 , 094012 ( 2013 ) [ arxiv:1309.0301 [ hep - ph ] ] .
m. duraisamy , p. sharma and a. datta , phys .
d * 90 * , no .
7 , 074013 ( 2014 ) [ arxiv:1405.3719 [ hep - ph ] ] .
m. tanaka and r. watanabe , phys .
d * 87 * , no .
3 , 034028 ( 2013 ) [ arxiv:1212.1878 [ hep - ph ] ] .
y. sakaki , m. tanaka , a. tayduganov and r. watanabe , phys . rev .
d * 91 * , no .
11 , 114028 ( 2015 ) [ arxiv:1412.3761 [ hep - ph ] ] .
m. freytsis , z. ligeti and j. t. ruderman , phys .
d * 92 * , no . 5 , 054018 ( 2015 ) [ arxiv:1506.08896 [ hep - ph ] ] . s. fajfer , j. f. kamenik and i. nisandzic , phys .
d * 85 * , 094025 ( 2012 ) [ arxiv:1203.2654 [ hep - ph ] ] .
p. biancofiore , p. colangelo and f. de fazio , phys .
d * 87 * , no . 7 , 074010 ( 2013 ) [ arxiv:1302.1042 [ hep - ph ] ] .
r. aaij _ et al . _
[ lhcb collaboration ] , phys .
lett . * 115 * , no .
11 , 111803 ( 2015 ) [ arxiv:1506.08614 [ hep - ex ] ] . m. huschle _ et al . _ [ belle collaboration ] , arxiv:1507.03233 [ hep - ex ] . m. davier , l. duflot , f. le diberder and a. rouge , phys .
b * 306 * , 411 ( 1993 ) .
m. diehl and o. nachtmann , z. phys .
c * 62 * , 397 ( 1994 ) .
d. atwood and a. soni , phys .
d * 45 * , 2405 ( 1992 ) .
j. f. gunion , b. grzadkowski and x. g. he , phys .
lett . * 77 * , 5172 ( 1996 ) [ hep - ph/9605326 ] .
m. diehl and o. nachtmann , eur .
j. c * 1 * , 177 ( 1998 ) [ hep - ph/9702208 ] .
b. grzadkowski , z. hioki , k. ohkuma and j. wudka , phys .
b * 593 * , 189 ( 2004 ) [ hep - ph/0403174 ] .
s. davidson , d. c. bailey and b. a. campbell , z. phys .
c * 61 * , 613 ( 1994 ) [ hep - ph/9309310 ] .
e. ponton , arxiv:1207.3827 [ hep - ph ] . m. okamoto _ et al .
_ , nucl .
* 140 * , 461 ( 2005 ) [ hep - lat/0409116 ] ; j. a. bailey _
[ fermilab lattice and milc collaborations ] , pos lattice * 2010 * , 311 ( 2010 ) [ arxiv:1011.2166 [ hep - lat ] ] ; y. amhis _ et al .
_ [ heavy flavor averaging group collaboration ] , arxiv:1207.1158 [ hep - ex ] .
b. aubert _ et al . _ [ babar collaboration ] , phys .
lett . * 104 * , 011802 ( 2010 ) [ arxiv:0904.4063 [ hep - ex ] ] .
b. aubert _ et al . _
[ babar collaboration ] , phys .
* 100 * , 151802 ( 2008 ) [ arxiv:0712.3503 [ hep - ex ] ] . | measurements branching fractions observables in semileptonic decays @xmath5 to @xmath4 interesting for indirect probe np earlier measurements on @xmath6 by belle babar collaborations shown deviations from standard model predictions possible signature of np in @xmath7 transitions authors tried to explain observation in np scenarios model - independent distinguish possible signatures np study np in @xmath1 distribution differential branching fractions @xmath0 correlations among @xmath4 forward - backward asymmetry @xmath4 polarization asymmetry examined 2.1@xmath8 deviation in measurement @xmath9 reported by lhcb collaboration belle collaboration announced results on @xmath10 values consistent with sm error bars constraints on new couplings obtained assuming presence one at a time impossible to extract all couplings from single measurement reduce coupling parameters by imposing constraints on full parameters possible obtain meaningful errors on couplings information lost useful to independent couplings errors on parameters not necessary for observable equal sensitivity to np operatorsuseful to know observable maximal sensitivity to np interaction select observables for extraction coupling important to find significance of types np interaction to observable use optimal - observable analysis invariant mass squared @xmath1 lepton - neutrino system construct optimal observable identify np structure best extracted observable reasonable statistics provides deeper understanding of sensitivity best obtained method for process technique used in collider phenomenology analyse @xmath1 distribution of differential branching fractions @xmath10 @xmath4 polarisation asymmetries forward backward asymmetries of decay @xmath0 @xmath11 polarisation asymmetry include non - standard four - fermi interactions lowest dimension estimate expected statistical uncertainties in extraction np wilson coefficients @xmath0hamiltonian @xmath12 transitions four - fermi operators by @xcite @xmath13{eq1{aligned operator basis defined v_1 & |c_l ^b_l)( l o_v_2 =|c_r ^b _ l o_s_1 & |c_l b_r)( l o_s_2|c_r b_l)( l o_t &|c_r b_l)( _ l ) eq2 ] wilson coefficients by @xmath14 neutrinos left handed focus on @xmath1 distribution differential decay rate @xmath15 @xmath0 complete expressions in ref.@xcite optimal - observable analysis statistical uncertainties measurable parameters from observable discussions in references @xcite @xmath0 express @xmath1 distribution differential decay rate as = _ i c_i f_i(q^2 ) eq3 @xmath16s functions of @xmath17s theoretical expressions for @xmath16s @xmath18s extracted from comparison between similar terms both sides eq.([eq3coefficients @xmath16 branching fractions @xmath0 table @xmath18s appendix adachi belle collaboration arxiv:0910.4301 hep - ex lees babar collaboration phys . d * 88 * 7 072012 2013 ) arxiv:1303.0571 hep - ex fajfer kamenik nisandzic . d * 85 * 094025 2012 ) arxiv:1203.2654 hep - ph m tanaka. 67 * 321 1995 ) sakaki tanaka tayduganov watanabe phys d * 88 * 9 094012 2013 ) arxiv:1309.0301 hep - ph duraisamy sharma. datta phys . d * 90 * 7 074013 2014 arxiv:1405.3719 hep - ph tanaka watanabe phys d * 87 * 3 034028 ( 2013 arxiv:1212.1878 hep - ph sakaki tanaka tayduganov watanabe phys d * 91 * 11 114028 ( 2015 ) arxiv:1412.3761 hep - ph. freytsis ligeti ruderman phys . d * 92 *054018 2015 ) arxiv:1506.08896 hep - ph fajfer kamenik nisandzic phys d 85 094025 2012 ) arxiv:1203.2654 hep - ph p. biancofiore colangelo. de fazio . d 87 074010 2013 ) arxiv:1302.1042 hep - ph aaij collaboration . 115 11 111803 2015 ) arxiv:1506.08614 hep - ex huschle collaboration arxiv:1507.03233 hep - ex davier duflot. le diberder. rouge phys b 306 411 ( 1993 ) diehl. nachtmann c 62 397 ( 1994 ) atwood soni d 45 2405 ( 1992 ) gunion grzadkowski g he phys 77 5172 ( 1996 ) hep - ph/9605326. diehl nachtmann c 1 177 ( 1998 ) hep - ph/9702208 grzadkowski hioki ohkuma wudka phys . b 593 189 ( 2004 ) hep - ph/0403174. davidson. bailey campbell phys61 613 1994 ) hep - ph/9309310 ponton arxiv:1207.3827 hep - ph okamoto nucl 140 461 ( 2005 hep - lat/0409116 bailey fermilab lattice milc collaborations lattice 2010 311 ( 2010 arxiv:1011.2166 hep - lat amhis heavy flavor averaging group collaboration arxiv:1207.1158 hep - ex aubert babar collaboration 104 011802 2010 arxiv:0904.4063 hep - ex aubert babar collaboration phys 100 151802 ( 2008 ) arxiv:0712.3503 hep - ex | 0.351049 | 2,379 | 1,237 | , Saving $0.1 in GPT-4. |
67,824 | the aim of the na48 experiment at the cern sps is to measure direct cp violation in neutral kaon decays thus determining the parameter @xmath0 with an accuracy of @xmath1 .
the advantages of na48 with respect to previous experiments are high statistics and reduced systematic effects .
the principle of the experiment and the performance of the detector components are presented .
edinburgh - ppe/97 - 02 + 29th july 1997 for the na48 collaboration | in 1964 , cp violation was discovered in the decay of the long - lived kaon into two pions @xcite . since then
, considerable experimental and theoretical effort has been devoted to understanding its origin .
after thirty years , no evidence of cp - violation has been observed in particle physics outside the neutral kaon system .
the cause of violation of cp symmetry is still an open issue in physics .
the study of direct cp - violation has turned out to be difficult , both experimentally and theoretically , and the understanding of its magnitude and origin is still far from satisfactory .
the latest theoretical predictions for the value of @xmath2 are in the range @xmath3 @xcite .
the experimental measurement of @xmath0 should allow discrimination between different theoretical hypotheses .
e731 at fnal and na31 at cern have measured @xmath2 with a precision better than @xmath4 .
na31 has obtained a value of @xmath5 @xcite while e731 has found a value of @xmath6 @xcite .
the first one indicates evidence of direct cp violation and the second is compatible with a null result . to clarify
the situation requires new measurements , at cern , na48 and at fnal , ktev , aim for a more precise determination of @xmath7 with a accuracy on @xmath2 of @xmath8 .
experimentally the measurement of @xmath0 is obtained by evaluating the double ratio of the decay rates of @xmath9 and @xmath10 into two neutral and two charged pions : @xmath11 with simultaneous observation of all four decay modes in the same experimental setup . the basic na48 scheme @xcite features two nearly collinear beams of @xmath9 and @xmath10 , produced by protons hitting two different targets , and distinguished by tagging of the protons producing the @xmath10 component . a magnetic spectrometer with four drift chambers
measures the momentum of the charged particles , and a fast liquid krypton calorimeter measures the energy , the position and the time of a neutral decay .
provided that the position of the two beams coincide in the detector , the advantage of this method is that at any given energy and vertex position , the detection efficiencies cancel .
differences and variations in detection efficiencies for @xmath9 and @xmath10 decays become unimportant , as do rate dependent effects introduced by accidental activity in the detector elements . the statistical error on the double ratio r is dominated by the statistical precision of the rarest decay . in order to determine @xmath2 with a precision of @xmath12 ,
it is necessary to collect about @xmath13 @xmath14 decays , one order of magnitude more than the previous experiments , na31 and e731 .
it is mostly the systematic error which limits the precision attainable on the measurement of r. in fact , due to different beam divergences , energy spectra and decay vertex positions , the acceptances of @xmath9 and @xmath10 decays are not exactly the same . in order to have a negligible difference in momentum spectra ,
the production angles of the @xmath9 and @xmath10 beams have been carefully chosen ( 2.4 and 4.2 mrad respectively ) .
this way , the decay spectra are similar when the kaon momentum is between 70 and 170 gev . since the distribution of decay vertex are different ( one is flat ( @xmath9 ) and the other is exponential ( @xmath10 ) ) , a weight w , depending on kaon momentum p and on the longitudinal coordinate z of kaon decay vertex , will be applied to @xmath9 events . using this technique
, the spatial distribution of @xmath9 decay vertices reproduces that of the @xmath10 beam . another effect which can contribute to the systematic error
is due to the imperfect knowledge of the fiducial volume .
for the neutral decays , an accurate knowledge of the energy scale is required and this is achieved by adjusting the reconstructed longitudinal decay vertex position to a counter edge placed at the beginning of the @xmath10 beam .
background events , mainly three - body decays , are another important source of systematic errors . in case of neutral events ,
most of the decays @xmath15 are rejected by using anti - counter rings which veto photons outside the fiducial region , and an electromagnetic calorimeter with excellent resolution in energy and space .
the background of charged events is given by @xmath16 , @xmath17 and by @xmath18 .
the information of the muon veto counters , located at the rear end of the detector , is used at the trigger level to reject @xmath19 events .
the invariant mass is reconstructed at the trigger level as well , in order to discard @xmath18 and @xmath16 decays ( at @xmath20 level ) .
the na48 experiment compares the @xmath21 and @xmath22 decay rates from distinct @xmath9 and @xmath10 beams which enter a common decay region along a path entirely contained in vacuum .
the four decay modes are thus recorded at the same time and from the same fiducial length .
the two beams are nearly collinear , converging at an angle comparable with the beam divergence .
a 450 gev / c proton beam with a nominal flux of @xmath23 ppp strikes a beryllium target , producing the @xmath9 beam at an angle of 2.4 mrad . after a first collimator which limits acceptance , charged particles ( including the remaining primary protons )
are deviated from the @xmath9 line by a sweeping magnet .
a fraction of the primary protons ( @xmath24 ppp ) is channelled back towards the @xmath9 beam by a bent silicon crystal @xcite .
it then passes through a tagging counter , and is eventually deflected and focused on the @xmath9 line .
it then hits a second target , placed 72 mm above the first one and at a distance of 120 m from it , to produce a @xmath10 beam at an angle of 4.2 mrad .
this beam is defined by a collimator and converges with the @xmath9 one at an angle of 0.6 mrad .
the exit of the @xmath10 collimator coincides with the last of three collimators designed to define the @xmath9 beam , so that background from the collimators themselves can not reach the detector .
after the @xmath10 collimator , a detector , vetoes @xmath10 decays which occurred before the collimator exit and defines the fiducial region .
a schematic view of the the kaon beam lines and of the detector layout is shown in fig.[fig1 ] .
the @xmath9 and @xmath10 assignment for each decay is done by measuring the time difference between the passage of a proton in the tagging counter @xcite upstream of the @xmath10 target and the event time in the detector .
events with a time difference inside of a given interval @xmath25 will be called `` @xmath10 '' , any other events will be called `` @xmath9 '' .
it has been designed to cope with proton rates above 10mhz and to provide a detection efficiency close to 100% combined with a time resolution better than 500 ps . to ensure that the rate in each counter is less than 1mhz
the tagger is designed with two sets of staggered scintillation foils arranged alternately in the horizontal and vertical planes .
the depth of the foils in the beam direction is 4 mm and the width varies from 200 @xmath26 m on the beam - axis to 3000 @xmath26 m at the beam edges .
thus the entire beam profile is covered .
the time resolution is less than 200 ps , with a double pulse resolution below 4 ns .
inefficiency and accidental activity of the tagging counter result in a correction to r , but being decay mode independent they can not generate a fake non - zero value for @xmath2 , nevertheless the efficiency will be monitored at the @xmath12 level . a magnetic spectrometer @xcite is used to analyse the @xmath22 decay : two set of drift chambers on each side of a central dipole magnet provide a measurement of charged particle momenta .
each chamber has four views x , y , u and v to avoid ambiguities in the position measurement .
each view consist of two planes of 256 wires , at a distance of 1 cm from each other .
the spatial resolution of each view is @xmath27100@xmath26 m , which corresponds to @xmath28 = 0.6% accuracy on mean momentum measurement .
the field integral of the main field component ( @xmath29 ) is equivalent to a transverse momentum kick of 250 mev / c . during the 1995 and 1996 test runs ,
very good performance was obtained with the magnetic spectrometer for the @xmath30 decays .
the four drift chambers worked reliably and provided high detection efficiency and position resolution at rates close to 1mhz . the kaon invariant mass resolution obtained off - line from fully reconstructed @xmath22 events
is 3 mev/@xmath31 in both @xmath9 and @xmath10 beams ( fig .
[ fig2 ] ) .
the good resolution on the decay vertex position , in the transverse plane , provides a clear separation of the @xmath9 and @xmath10 sources , allowing a measurement of the tagging efficiency in the charged decay mode .
the timining information provided by this detector is used in conjunction with the tagger to identify the @xmath9 or @xmath10 origin of charged decays , and also in the trigger to select two - body decays .
the detector consist of two planes of scintillators counters , at a distance of 50 cm , located after the magnetic spectrometer along the beam line .
each plane consists of 64 scintillators , arranged vertically and horizontally .
the time resolution for a two pion event is better than 250 ps .
the characteristic of the decays to be detected , and the high rate of the experiment , impose same stringent requirements on the electromagnetic calorimeter .
in addition , a large sensitive area ( @xmath32 ) has to be covered .
good energy resolution is needed : @xmath33 at 10 gev with a constant term lower than 0.5% .
a spatial resolution @xmath34 is also needed .
the tranverse scale has to be accurate to 0.1/1000 mm to keep the systematic error induced by a possible difference in energy scale for charged and neutral decays at the desired level . for @xmath10 identification
a timing accurancy @xmath35 has to be achieved .
high single rate capability(@xmath36 ) is needed to collect the required statitiscs , high granularity and short sensitive time would reduce the effect of accidentals and background .
the na48 collaboration has built a fully sensitive liquid krypton ionization chamber with electrodes parallel to the shower direction and a tower structure segmentation into @xmath37 cells .
its transverse dimension is about 2.6 m and its thickness is 125 cm , corresponding to about 27 @xmath38 of liquid krypton .
the total krypton volume in the cryostat is about @xmath39 . to suppliment the time measurement of the detected photons ,
scintillating fibers have been inserted vertically in the liquid krypton at a depth of 9.5 @xmath38 .
the required performance has been achieved after several test beam runs with a prototype @xcite .
the energy resolution achieved with an electron test beam is @xmath40 during the 1996 test run period only the 8% of the lkr electronics was available ; the entire calorimeter was read - out by instrumenting `` supercells '' of 2x8 single cells . in this way
physics data could be taken with both beams .
the performance of the calorimeter was tested during the calibration period with an electron beam and the resolution obtained is in good agreement with prototype results .
the neutral decays in two @xmath41 with four clusters in the lkr calorimeter were identified and their rates were found to agree with the predicted ones .
the @xmath41 mass resolution is about 3.5 mev/@xmath31 ( fig.[fig2 ] ) due to the poor granularity of the supercells and their associated noise , but it is compatible with the proposal ( 1 mev/@xmath31 )
. the most important task of this subdetector is to provide an energy threshold signal which is part of trigger , and is used to reject background coming from the three body decays . the front and the back modules consist of alternate iron and scintillator planes oriented either horizontally or vertically .
the thickness corresponds to 7.2 interaction lengths .
the energy resolution of the hadron calorimeter is 65%/@xmath42 . to reject the background due to the @xmath17 it is used the information provided by the @xmath26-counters , located at the end of the beam line .
a total of 28 scintillator strips are arranged in three planes , preceded by an iron layer 80 cm thick .
the strips are alternatively vertical and horizontal , partially overlapping .
the efficiency of this system is better than 99% for muons above 5 gev .
the beam and detector for the na48 experiment are ready to take data for the study of direct cp violation in the @xmath43 system .
data taking will start in august .
the beams have achieved their design goals , including the novel use of a bent silicon crystal for deflecting protons at 450 gev / c and tagging them with time resolution of better than 500 ps at 10 mhz beam rates .
the detector is also complete , the required mass resolutions have been achieved in the magnetic spectrometer , and will be achieved in the liquid krypton calorimenter when the full electronics is commissioned in july 1997 .
christenson , j.w .
cronin , v.l .
fitch , r. turlay , _ phys .
lett . _ 13 , 214 ( 1964 )
. g. buchalla _
et all _ , _ nucl .
_ b337 , 313 ( 1990 ) ; a.j . buras _ et all _ , _ nucl . phys .
_ b408 , 209 ( 1993 ) ; m. ciuchini _
et all _ , _ phys .
_ b301 , 263 ( 1993 ) ; a.j .
et all _ , preprint hep - ph/9608365 , august 1996 . g.d .
et all _ , _ phys .
_ b317 , 233 ( 1993 ) .
et all _ , _ phys .
lett . _ 70 , 1203 ( 1993 ) .
et all _ , _ cern / spsc/90 - 22/p253_. n. doble _ et all _ , _ nucl . instr .
_ b119 , 181 ( 1996 ) .
p. grafstrom _
et all _ , _ nucl .
_ a344 , 487 ( 1994 ) .
d. bederede _
et all _ , _ nucl .
_ a367 , 88 ( 1985 ) .
et all _ , _ nucl .
_ a370 , 413 ( 1996 ) . | 1964 cp violation discovered in decay long - lived kaon into two pions @xcite experimental effort understanding origin after thirty years no evidence cp - violation observed particle physics outside neutral kaon system cause cp open issue study direct cp - violation difficult understanding magnitude origin satisfactory predictions value @xmath2 @xmath3 @xcite experimental measurement @xmath0 should discrimination hypotheses e731 fnal na31 cern measured @xmath2 better than @xmath4 na31 obtained value @xmath5 @xcite e731 @xmath6 @xcite first indicates direct cp violation second compatible with null result new measurements cern na48 fnal precise determination @xmath7 accuracy @xmath2 @xmath8 measurement @xmath0 double ratio decay rates @xmath9 @xmath10 into two neutral charged pions @xmath11 four decay modes basic na48 scheme @xcite features two collinear beams of @xmath9 @xmath10 produced protons hitting targets distinguished by tagging protons @xmath10magnetic spectrometer four chambers measures momentum charged particles fast liquid krypton calorimeter measures energy position time neutral decay two beams coincide detector energy vertex position detection efficiencies cancel differences efficiencies for @xmath9 @xmath10 decays unimportant rate dependent effects accidental activity detector statistical error on double ratio r dominated by precision rarest decay determine @xmath2 precision @xmath12 collect @xmath13 @xmath14 decays more than previous systematic error limits precision measurement r to different beam divergences energy spectra decay vertex positions acceptances of @xmath9 @xmath10 decays not same negligible difference in momentum spectra production angles of @xmath9 beams carefully chosen 2.4 4.2 mrad decay spectra similar when kaon momentum between 70 and 170 gev distribution of decay vertex different flat exponential weight w depending on kaon momentum p longitudinal coordinate z decay vertex applied to @xmath9 events spatial distribution of @xmath9 decay vertices reproduces @xmath10 beamsystematic error imperfect knowledge fiducial volume neutral decays accurate knowledge energy scale required achieved adjusting longitudinal decay vertex to counter edge beginning @xmath10 beam background events three - body decays source errors neutral events decays rejected anti - counter rings photons outside region electromagnetic calorimeter excellent resolution background charged events by @xmath16 @xmath18 muon veto counters rear end detector used trigger level reject @xmath19 events invariant mass reconstructed at trigger level discard @xmath18 @xmath16 decays na48 experiment compares @xmath21 @xmath22 decay rates from @xmath9 @xmath10 beams common decay region vacuum four decay modes recorded at same time same fiducial length two beams nearly collinear angle comparable divergence 450 gev / c proton beam flux @xmath23 ppp strikes beryllium target @xmath9 beam angle 2.4 mrad after first collimator charged particles deviated from @xmath9 line by sweeping magnet fraction primary protons channelled back towards @xmath9 beam by bent silicon crystal @xcitepasses through tagging counter on @xmath9 line hits second target 72 mm above first 120 m @xmath10 beam angle 4.2 mrad beam defined by collimator converges with @xmath9 angle 0.6 mrad exit @xmath10 collimator coincides with last three collimators beam background reach detector after collimator detector vetoes @xmath10 decays defines fiducial region schematic view beam lines detector layout shown in fig. ] @xmath9 and @xmath10 assignment for decay time difference between passage proton in tagging counter target event time in detector events with time difference interval @xmath25 called @xmath10 '' other @xmath9 '' designed to cope with proton rates above 10mhz detection efficiency close to 100% time resolution better than 500 ps rate counter less than 1mhz tagger designed with two sets staggered scintillation foils horizontal vertical planes depth foils beam direction 4 mm width varies from 200 m to 3000 m edges entire beam profile covered time resolution less than 200 ps double pulse resolution below 4 nsinefficiency accidental activity tagging counter correction to r decay mode independent generate fake non - zero value for @xmath2 efficiency monitored at @xmath12 level magnetic spectrometer @xcite @xmath22 decay two drift chambers central dipole magnet charged particle momenta each four views x y u v each two planes 256 wires distance 1 cm spatial resolution @xmath27100@xmath26 m @xmath28 = 0.6% accuracy mean momentum field integral @xmath29 ) equivalent to transverse momentum kick 250 mev / c 1995 1996 test runs good performance magnetic spectrometer for @xmath30 decays four drift chambers high detection efficiency position resolution close 1mhz kaon invariant mass resolution from @xmath22 events 3 mev/@xmath31 in @xmath9 @xmath10 beams good resolution on decay vertex position separation @xmath9 @xmath10 sources tagging efficiency charged decay mode timining information detector identify @xmath9 @xmath10 origin charged decays select two - body decays detector two planes scintillators counters distance 50 cm after magnetic spectrometerplane 64 scintillators vertically horizontally time resolution two pion event better than 250 ps decays high rate experiment impose stringent requirements electromagnetic calorimeter large sensitive area @xmath32 covered good energy resolution needed @xmath33 at 10 gev constant term lower than 0.5% spatial resolution @xmath34 needed tranverse scale accurate to 0.1/1000 mm timing accurancy @xmath35 high single rate capability@xmath36 needed high granularity short sensitive time reduce accidentals na48 collaboration built fully sensitive liquid krypton ionization chamber electrodes parallel shower direction tower structure into @xmath37 cells transverse dimension 2.6 m thickness 125 cm 27 @xmath38 liquid krypton total krypton volume @xmath39 scintillating fibers inserted vertically in liquid krypton depth 9.5 @xmath38 performance achieved after test beam runs prototype @xcite energy resolution electron test beam @xmath40 1996 test run 8% lkr electronics available entire calorimeter read - out by supercells of 2x8 single cells physics data taken with both beamsperformance calorimeter tested calibration electron beam resolution with prototype results neutral decays in two @xmath41 four clusters identified rates with predicted @xmath41 mass resolution 3.5 mev/@xmath31 due poor granularity supercells noise compatible with proposal 1 mev/@xmath31 task subdetector energy threshold signal reject background from three body decays front back modules iron scintillator planes oriented horizontally or vertically thickness to 7.2 interaction lengths energy resolution calorimeter 65%/@xmath42 reject background @xmath17 information @xmath26-counters end beam line 28 scintillator strips in three planes iron layer 80 cm thick strips vertical horizontal overlapping efficiency better than 99% for muons above 5 gev beam detector experiment ready data for direct cp violation in @xmath43 system data taking august beams achieved design goals bent silicon crystal for deflecting protons at 450 gev / c time resolution better than 500 ps at 10 mhz beam rates detector complete required mass resolutions achieved in magnetic spectrometer liquid krypton calorimenter full electronics commissioned july 1997 christenson , j.wcronin fitch turlay phys . 13 , 214 ( 1964 ) buchalla nucl b337 , 313 ( 1990 ) buras nucl phys b408 , 209 ( 1993 ) ciuchini phys b301 , 263 ( 1993 ) a.j preprint hep - ph/9608365 august 1996 g.d b317 , 233 ( 1993 ) phys lett 70 , 1203 ( 1993 ) cern spsc/90 - 22/p253 doble nucl instr b119 , 181 ( 1996 ) grafstrom nucl a344 , 487 ( 1994 ) bederede nucl a367 , 88 ( 1985 ) nucl a370 , 413 ( 1996 ) | 0.507033 | 3,351 | 1,733 | , Saving $0.1 in GPT-4. |
67,829 | spin - spin relaxation time ( @xmath0 ) and magnetic susceptibility ( @xmath1 ) of the second layer @xmath2he adsorbed on grafoil , exfoliated graphite , preplated with a monolayer @xmath3he are studied by pulsed - nmr in a density range of @xmath4 nm@xmath5 .
the temperature dependence of @xmath6 and @xmath7 show fermi fluid behaviour and no evidence of self - condensation are found even at the lowest density @xmath8 nm@xmath5 .
density dependence of @xmath0 at @xmath9 mhz shows a broad maximum of 5.7 ms around @xmath10 nm@xmath5 .
since the decrease of @xmath0 in dilute side can not be expected in the ideal 2d fluid , it can be understood as the relaxation caused by a small amount of solid @xmath2he at heterogeneity of the substrate .
we also measured the larmor frequency dependence of @xmath0 at @xmath11 nm@xmath5 .
@xmath12 has a @xmath13-linear dependence similarly to the earlier study on a first layer solid @xmath2he @xcite . from a comparison between our result and the earlier one ,
this linearity is almost independent of the particle motion .
now , it could be caused by a microscopic magnetic field inhomogeneity arisen from the mosaic angle spread and diamagnetism of the graphite substrate . | the second layer @xmath2he on graphite is an ideal example for studying two - dimensional ( 2d ) fermion system . by controlling the number of adsorption atoms
, we can widely change the areal density @xmath14 from nearly ideal fermi gas to highly compressed 2d solid @xcite . at the commensurate phase in the second layer ,
so called 4/7 phase " , a gap - less quantum spin liquid state is suggested .
the transition between a fermi fluid and the 4/7 phase is mott - hubbard type with divergence of the effective mass @xmath15 @xcite .
moreover , recent heat capacity study shows the existence of a low - density liquid state ( @xmath16 nm@xmath5 ) in the third layer @xcite .
but , this self - condensed state has not been confirmed in the second layer yet . to study the spin correlation in the quantum spin liquid , the self - condensed liquid , and the relaxation process in the liquid or
dilute fermi gas , pulsed - nmr is a useful probe . in this work , we report a preliminary pulsed - nmr study of the spin - spin relaxation time ( @xmath0 ) and magnetic susceptibility ( @xmath1 ) of fermi fluid phase in the second layer . to remove the large curie susceptibility from the first layer , we replaced the first layer @xmath2he with @xmath3he .
it is also helpful to reduce the substrate heterogeneity effect because the heterogeneous regions with deep potential are preferentially filled with @xmath3he atoms .
the substrate used in this work was grafoil , an exfoliated graphite , with the total surface area of @xmath17 m@xmath18 .
grafoil consists of 10 - 20 nm size micro - crystallites with a mosaic angle spread of @xmath19 @xcite . before adding @xmath2he sample at 2.7 k ,
11.78 nm@xmath5 of @xmath3he was adsorbed at 4.2 k as the first layer .
this density is slightly higher than the highest density of the first layer @xmath2he determined from the neutron scattering ( 11.06 nm@xmath5 ) @xcite and the heat capacity measurement ( 11.4 nm@xmath5 ) @xcite .
therefore , we believe that @xmath2he atoms do nt sink into the first layer . the conventional spin - echo technique with the pulse sequence of @xmath20-@xmath21-@xmath22 were used for measuring the spin - spin relaxation time @xmath0 . in presence of the mosaic angle spread of grafoil ,
the decay of the spin - echo signal shows about 5% of long tail component due to the angular dependence of @xmath0 @xcite .
but , in this work , spin - echo measurements were carried out within @xmath23 where the contribution of the long tail is not important , and @xmath0 was obtained by single exponential fitting .
most of the experiments , except a larmor frequency dependence of @xmath0 , were carried out in a larmor frequency of 5.5 mhz ( @xmath24 mt ) .
the static magnetic field is parallel to the substrate .
the details of this set - up are described in earlier paper , where grafoil is replaced with zyx @xcite .
* a. * temperature dependence of @xmath25 .
the solid curves are fitting lines by eq.([eqn1 ] ) . *
b. * density dependence of the susceptibility at @xmath26 .
the closed circles are the data taken in this work , while the open circles are those in ref.@xcite . *
c. * density dependence of effective temperature @xmath27 . ]
figure 1(a ) shows the temperature dependence of the magnetic susceptibility @xmath6 at @xmath2he density range of @xmath4 nm@xmath5 .
the signals are well described by the phenomenological expression for the susceptibility of a fermi liquid @xcite : @xmath28 down to @xmath8 nm@xmath5 , which is about half of the lowest density in the earlier study @xcite .
now , @xmath29 is the curie constant and @xmath27 is the effective fermi temperature . at @xmath11 nm@xmath5 ,
@xmath1 increases slightly at lowest temperature @xmath30 mk . considering a small amount of paramagnetic solid @xmath2he
, this fraction corresponds to within 2% of the total @xmath2he atoms in experimental accuracy .
we thought that this is trapped @xmath2he atoms in substrate heterogeneities , and it can be removed by introducing a few percent of excess @xmath3he
atoms @xcite . in landau fermi liquid theory ,
the @xmath26 susceptibility of 2d fermi liquid is given by @xmath31 , where @xmath32 is a landau parameter and @xmath33 is the fermi temperature .
since @xmath29 is proportional to @xmath34 , number of @xmath2he atoms , and @xmath33 is proportional to @xmath35 , where @xmath36 is the surface area over which the @xmath2he liquid occupies , we obtain @xmath37 . considering the self - condensation
, @xmath36 would be smaller than the total surface area ( @xmath38 ) .
now , because the @xmath26 susceptibility of an ideal 2d fermi gas is @xmath39 , we obtain @xmath40 .
if @xmath2he atoms are uniformly spread over the surface area ( @xmath41 ) , this function approaches to a constant value ( in ideal gas case , unity ) at dilute limit .
on the other hand , when self - condensation occurs , @xmath36 changes linearly with increasing @xmath34 , resulting in @xmath42 similarly to the density dependence of @xmath43 in heat capacity measurement of the third layer @xcite .
figure 1(b ) shows obtained @xmath44 against @xmath14 .
the value approaches to unity with decreasing @xmath14 , and there is no anomaly down to @xmath8 nm@xmath5 .
therefore , there is no evidence of self - condensation at least at the density above 0.68 nm@xmath5 in the second layer .
the density dependence of @xmath0 at @xmath45 mk and @xmath9 mhz is shown in figure 2 .
it shows a broad maximum of 5.7 ms around @xmath10 nm@xmath5 . at high density side ,
the particle motion is suppressed by the divergence of @xmath15 toward a mott localization @xcite .
therefore , @xmath0 decreases with increasing @xmath14 due to the suppression of motional narrowing . at the lowest density @xmath8 nm@xmath5 ,
the scattering radius @xmath46 is much shorter than a mean free path @xmath47 , and a dilute 2d @xmath2he gas model is applicable . from an analogous approach to a dilute bulk @xmath2he gas @xcite , we obtain : @xmath48 where @xmath49 is magnetic moment of @xmath2he , and the averaged velocity @xmath50 at fermi degenerated gas ( @xmath51 from figure 1(c ) ) . as a consequence
, @xmath0 increases with decreasing @xmath14 by @xmath52 at dilute limit . on the other hand , obtained @xmath0 decreases at @xmath8 nm@xmath5 .
it indicates that there is another relaxation process that becomes important at dilute limit .
cowan and kent measured @xmath0 of very low density @xmath2he films on bare graphite @xcite , and showed that a small amount of @xmath2he ( @xmath53 few percent of monolayer ) would form high - density solid with @xmath54 ms in heterogeneous region of the substrate . by atomic exchanges between this solid and the 2d gas , relaxation in the heterogeneous region dominates the relaxation process of the 2d gas .
the linear @xmath14-dependence of @xmath0 in the `` first layer '' is also shown in the same graph ( fig.[fig2 ] ) for the reference . because the adsorption potential in the second layer is about an order of magnitude smaller than that of the first layer , it is unrealistic to consider such a high - density solid .
however , since the heterogeneous region can exist as a relaxation spot , and very small amount of @xmath2he could also exist as a solid .
this is consistent with the increase of @xmath6 at lowest temperature .
magnetic impurity in substrate may be another causation of the relaxation .
larmor frequency dependence of @xmath0 .
the closed circles are @xmath0 of the second layer fermi fluid at @xmath11 nm@xmath5 and @xmath45 mk .
the open circles are those of the first layer solid at a coverage of 0.76 monolayer and @xmath55 k @xcite . the solid lines are linear fitting to the data . ]
larmor frequency dependence of @xmath0 .
the closed circles are @xmath0 of the second layer fermi fluid at @xmath11 nm@xmath5 and @xmath45 mk .
the open circles are those of the first layer solid at a coverage of 0.76 monolayer and @xmath55 k @xcite .
the solid lines are linear fitting to the data . ] in figure 3 , we show the larmor frequency dependence of @xmath0 in a high - density fermi fluid ( @xmath11 nm@xmath5 ) at @xmath45 mk ( closed circles ) .
we observed a linear relation between @xmath12 and @xmath13 .
this characteristic linear dependence of @xmath0 is also observed in a first layer solid @xmath2he @xcite ( open circles ) .
it is curious since @xmath0 should be independent of @xmath13 as long as @xmath56 , where @xmath21 ( @xmath57 sec ) is the correlation time .
although the explanation for this linearity is not satisfactory established , this could be a microscopic magnetic field inhomogeneity effect caused by the mosaic angle spread and diamagnetism of the graphite substrate .
assuming that the frequency dependence is an extrinsic effect related to the diffusion process in the microscopic magnetic field gradient , the intrinsic @xmath0 is obtained by extrapolating to @xmath58 where the diamagnetic field of graphite disappears .
namely , @xmath59 , where @xmath60 is the intrinsic @xmath0 of the system . from the extrapolation ,
@xmath60 at @xmath11 nm@xmath5 is very long ( @xmath61 s ) although the strict value is not available because @xmath62-intercept is too small .
this is much longer than that of the first layer solid ( @xmath63 ms ) .
nevertheless , the value of @xmath64 , which is a slope of @xmath12 , is almost the same between these two samples .
therefore , it is clear that the @xmath13-dependence of @xmath0 is extrinsic one . to obtain the strict value of @xmath60 , lower frequency experiment or a substrate with larger platelet size and smaller mosaic angle will be required . in conclusion , we measured @xmath6 and @xmath0 of the second layer @xmath2he at the density region of @xmath4 nm@xmath5 on graphite pleplated with a monolayer @xmath3he .
obtained @xmath6 shows fermi fluid behaviour at all densities , and there is no self - condensed state down to @xmath8 nm@xmath5 .
density dependence of @xmath0 at @xmath9 mhz shows a broad maximum of 5.7 msec .
the decrease of @xmath0 at @xmath8 nm@xmath5 could be related to the @xmath2he solid at the heterogeneity in the substrate .
we also observed a @xmath13-linear dependence of @xmath12 , which is similar to that in the earlier study for the first layer solid @xcite .
this @xmath13-dependence of @xmath0 is an extrinsic effect , and the intrinsic @xmath0 in the fermi fluid is much longer than measured value .
9 greywall d s 1990 _ phys .
_ b * 41 * 1842 fukuyama h 2008 _ j. phys .
jpn . _ * 77 * 111013 casey a , patel h , nyki j , cowan b p and saunders j 2003 _ phys . rev .
lett . _ * 90 * 115301 sato d , tsuji d , takayoshi s , obata k , matsui t and fukuyama h 2010 _ j. low temp . phys . _ * 158 * 201 takayoshi s and fukuyama h 2010 _ j. low temp . phys . _
* 158 * 672 lauter h j , godfrin h , frank v l p and schildberg h p 1990 _ physica _ b * 165&166 * 597 takayoshi s , obata k , sato d , matsui t and fukuyama h 2009 _ j. phys . :
* 150 * 032104 murakawa s , niimi y , matsumoto y , buerle c and fukuyama h 2003 _ physica _ b * 329 - 333 * 144 dyugaev a m 1990 _ sov .
_ a _ phys .
_ * 14 * 1 lusher c p , cowan b p and saunders j 1991 _ phys .
* 67 * 2497 collin e , triqueneaux s , harakaly r , roger m , buerle c , bunkov yu m and godfrin h 2001 _ phys . rev .
_ * 86 * 2447 chapman r and richards m g 1974 _ phys .
* 33 * 18 cowan b p and kent a j 1984 _ phys .
lett . _ * 106a * 54 cowan b , el - nasr l a and fardis m 1987 _ jpn . j. app . phys . _
* 26 * suppplement 26 - 3 309 | second layer @xmath2he on graphite ideal two dimensional fermion system controlling adsorption atoms areal density from fermi gas to compressed 2d solid @xcite second layer 4/7 phase gap - less quantum spin liquid state suggested transition between 4/7 mott - hubbard type divergence effective mass @xmath15 @xcite heat study low - density liquid state @xmath16 in third layer not confirmed second layer spin correlation quantum spin liquid self - condensed liquid relaxation process pulsed - nmr preliminary pulsed - nmr study spin relaxation time magnetic susceptibility of fermi fluid phase second layer susceptibility replaced layer @xmath2he with @xmath3he substrate heterogeneity regions filled with @xmath3he atoms substrate grafoil exfoliated graphite surface area @xmath17 m@xmath18 grafoil 10 - 20 nm micro - crystallites mosaic angle spread of @xmath19 @xcite before @xmath2he sample at 2.7 k 11.78 nm@xmath5 of @xmath3he adsorbed at 4.2 k first layerdensity higher than highest first layer @xmath2he determined from neutron scattering 11.06 nm heat capacity measurement 11.4 nm @xmath2he atoms do nt sink into first layer conventional spin - echo technique pulse sequence @xmath20-- used spin relaxation time @xmath0 mosaic angle spread grafoil decay spin - echo signal shows 5% long tail component due to angular dependence @xmath0 spin - echo measurements within @xmath23 not important @xmath0 obtained by single exponential fitting experiments larmor frequency in larmor frequency 5.5 mhz ( @xmath24 mt ) static magnetic field parallel to substrate grafoil replaced with zyx temperature dependence of @xmath25 solid curves fitting lines density dependence of susceptibility at @xmath26 closed circles data open circles density dependence of effective temperature @xmath27 figure 1(a shows temperature dependence magnetic susceptibility @xmath6 at @xmath2he density range nm@xmath5signals described susceptibility fermi liquid @xcite @xmath28 to nm@xmath5 half lowest density earlier study @xmath29 curie constant @xmath27 effective fermi temperature @xmath11 nm@xmath5 @xmath1 increases lowest temperature @xmath30 mk small paramagnetic solid @xmath2he 2% atoms atoms substrate heterogeneities removed introducing excess @xmath3he atoms landau fermi liquid theory @xmath26 susceptibility 2d fermi liquid @xmath31 @xmath32 landau parameter @xmath33 fermi temperature @xmath29 proportional to @xmath34 atoms @xmath33 proportional @xmath35 @xmath36 surface area liquid obtain @xmath37 self - condensation @xmath36 smaller than total surface area @xmath38 @xmath26 susceptibility ideal 2d fermi gas @xmath39 obtain @xmath40 if @xmath2he atoms uniformly spread over surface area @xmath41 function constant value at dilute limitself - condensation @xmath36 changes @xmath34 @xmath42 density dependence @xmath43 heat capacity third layer figure 1(b shows @xmath44 against @xmath14 value unity decreasing @xmath14 no anomaly @xmath8 nm@xmath5 no evidence self - condensation density above 0.68 nm@xmath5 second layer density dependence @xmath0 at @xmath45 mk @xmath9 mhz figure 2 maximum 5.7 ms around @xmath10 nm@xmath5 high density particle motion suppressed divergence @xmath15 mott localization @xmath0 decreases @xmath14 lowest density @xmath8 nm@xmath5 scattering radius @xmath46 shorter than free path @xmath47 dilute 2d @xmath2he gas model applicable @xmath48 @xmath49 magnetic moment @xmath2he averaged velocity @xmath50 at degenerated gas @xmath0 increases decreasing @xmath14 @xmath52 at dilute limit @xmath0 decreases at @xmath8 nm@xmath5 relaxation process at dilute limitcowan kent measured low density films on graphite showed small amount @xmath2he percent monolayer form high - density solid with @xmath54 ms in heterogeneous region substrate atomic exchanges solid 2d gas relaxation heterogeneous region dominates linear @xmath14-dependence of @xmath0 first layer shown in graph fig adsorption potential second layer smaller first layer unrealistic consider high - density solid heterogeneous region relaxation spot small amount @xmath2he could exist as solid consistent with increase @xmath6 at lowest temperature magnetic impurity substrate may relaxation larmor frequency dependence @xmath0 closed circles second layer fermi fluid at @xmath11 nm@xmath5 mk open circles first layer solid coverage 0.76 monolayer @xmath55 k @xcite solid lines linear data frequency dependence closed circles second layer fermi fluid nm mk open circles first layer solid coverage 0.76 monolayer @xmath55 k @xcite solid lines linear datafigure 3 larmor frequency dependence @xmath0 high - density fermi fluid @xmath11 nm@xmath5 at @xmath45 mk closed circles linear relation between @xmath12 @xmath13 dependence observed first layer solid @xmath2he @xmath0 independent @xmath13 @xmath56 @xmath21 @xmath57 sec correlation time microscopic magnetic field inhomogeneity mosaic angle spread diamagnetism graphite substrate frequency dependence extrinsic diffusion process intrinsic @xmath0 obtained extrapolating to @xmath58 where diamagnetic field graphite disappears @xmath59 @xmath60 intrinsic @xmath0 @xmath60 at @xmath11 nm@xmath5 long ( @xmath61 s ) strict value not available @xmath62-intercept too small longer than first layer solid @xmath63 ms ) value @xmath64 slope @xmath12 same between two samples @xmath13-dependence of @xmath0 extrinsic obtain strict value @xmath60 lower frequency experiment substrate larger platelet size smaller mosaic angle requiredmeasured @xmath6 @xmath0 second layer @xmath2he at density region @xmath4 nm@xmath5 on graphite pleplated with monolayer @xmath3he @xmath6 shows fermi fluid behaviour at all densities no self - condensed state to @xmath8 nm@xmath5 density dependence @xmath0 at @xmath9 mhz maximum 5.7 msec decrease @xmath0 at @xmath8 nm@xmath5 related to @xmath2he solid heterogeneity substrate observed @xmath13-linear dependence @xmath12 similar earlier study first layer solid @xcite @xmath13-dependence @xmath0 extrinsic effect @xmath0 in fermi fluid longer than measured value greywall d s 1990 phys. 41 * 1842 fukuyama h 2008 77 * 111013 casey patel nyki cowan saunders j 2003 90 * 115301 sato tsuji takayoshi fukuyama 2010 158 * 201 672 lauter h godfrin frank schildberg 1990 165&166 * 597 takayoshi 2009150 032104 murakawa niimi matsumoto buerle fukuyama h 2003 physica 329 - 333 144 dyugaev 1990 14 lusher cowan b saunders j 1991 67 2497 collin e triqueneaux harakaly roger buerle c bunkov godfrin h 2001 86 2447 chapman r richards m 1974 33 18 cowan b kent a j 1984 106a 54 cowan b el - nasr l fardis m 1987 phys 26 suppplement 26 - 3 309 | 0.49739 | 3,183 | 1,708 | , Saving $0.1 in GPT-4. |
67,847 | a small fraction of tidal disruption events ( tde ) produce relativistic jets , evidenced by their non - thermal x - ray spectra and transient radio emission . here
we present milliarcsecond - resolution imaging results on tde j1644 + 5734 with the european vlbi network ( evn ) .
these provide a strong astrometric constraint on the average apparent jet velocity @xmath0 , that constrains the intrinsic jet velocity for a given viewing angle . | tidal disruption events ( tde ) may help us understand two outstanding questions of black hole astrophysics : how black holes form and grow , and what impact they have on the evolution of their host galaxies ( see a recent review by ( * ? ? ?
* komossa 2015 ) ) .
the first question can be addressed through constraining the low - mass end of the supermassive black hole mass - function ( ( * ? ? ?
* stone & metzger 2016 ) ) .
the idea here is that tdes light - up dormant black holes ( ( * ? ? ?
* giannios & metzger 2011 ) ) that have not shown an active galactic nucleus ( agn ) signature before . as the size of both the black hole and
the disrupted star can be estimated well ( e.g. ( * ? ? ?
* gezari et al . 2012 ) , ( * ? ? ?
* guillochon et al . 2014 ) ) , tdes provide means to weigh black holes down to @xmath1 , the majority of which has remained undetected to date .
the second question can be addressed through observations of newly formed radio jets ( triggered by a tde ) on milliarcsecond scales .
very long baseline interferometry ( vlbi ) is a powerful tool to localise the origin of hi outflows and to reveal the dynamics of atomic gas swept - up by the radio jet , catching agn jet feedback in action ( ( * ? ? ?
* morganti et al . 2013 ) ) ; while vlbi of jetted tdes provide us with a unique new chance of observing and understanding the very early phases of jet formation and evolution in an otherwise pristine environment , without any past agn activity .
future blind radio surveys will potentially detect a large number of tdes , but it has been recognised that a multi - band approach is essential for their identification .
the role of vlbi may be two - fold : very high resolution observations of strong candidates may reveal their nuclear origin , supporting the tde identification at early times . on the other hand ,
securely identified tdes ( from x - ray properties and/or optical spectra ) that occur off - nucleus may point to peculiar systems such as a white dwarf disrupted by an intermediate - mass black hole ( ( * ? ? ?
* shcherbakov et al . 2013 ) ) , a tde from a recoiling black hole ( ( * ? ? ?
* komossa & merritt 2008 ) ) , or it may reveal a secondary galactic nucleus in a merger ( ( * ? ? ? * liu & chen 2013 ) ) .
the latter is potentially interesting since the tde rate is apparently higher in a certain type of post - starburst galaxies that have recently undergone a major merger event ( ( * ? ?
* french , arcavi & zabludoff 2016 ) ) .
it has been shown that jetted tdes will be more efficiently detected in large - scale radio surveys than with current or near - future x - ray missions , therefore tdes will be primary targets of synchrotron transients with the square kilometre array ( ska ) ; the expected numbers however depend strongly on the fraction of jetted events , and their relativistic beaming factors ( ( * ? ? ?
* donnarumma & rossi 2015 ) ) . for the first time
, we obtained a strong constraint on the average apparent jet speed for j1644 + 5734 , briefly described below . more details of this result can be found in @xcite
the transient swift j1644 + 5734 was extremely luminous from radio to the @xmath2 rays , that has been widely interpreted as a non - thermal tde with a relativistic jet ( ( * ? ? ?
* levan et al . 2011 ) ; ( * ? ? ?
* bloom et al . 2011 ) ; ( * ? ? ?
* burrows et al .
2011 ) ; ( * ? ? ?
* zauderer et al . 2011 ) ) .
the variability properties in the different bands indicated that the x - ray and the radio emission were most likely produced in different regions , the latter related to a decelerating external shock in an initially ultrarelativistic outflow ( ( * ? ? ?
* giannios & metzger 2011 ) ) .
similar models have been invoked to explain multi - band properties of @xmath2-ray afterglows ( e.g. ( * ? ? ?
* zauderer et al . 2013 ) ) .
the vlbi technique has the power to directly reveal relativistic ejecta by for example resolving them . in case of j1644
+ 5734 @xcite predicted that marginally resolving the source may be possible at high frequencies with current vlbi instrumentation ; at 22 ghz and above however the emission is expected to fade quickly .
another way , often successfully applied in quasars up to even very high redshifts ( see ( * ? ? ?
* frey et al . 2015 ) ) , is to detect the proper motion of relativistic ejecta .
we followed this alternative route .
we observed swift j1644 + 5734 at six epochs between 12 april 2011 and 9 march 2015 with the evn at 5 ghz .
detection of targets at the few - mjy level requires frequent visits to a bright , compact calibrator , a technique called phase - referencing ( see ( * ? ? ?
* reid & honma 2014 ) ) .
this technique also provides the relative position between the target and the calibrator . for this purpose we used one of the international celestial reference frame ( icrf ) defining sources , j1638 + 5720 as primary calibrator .
the precision of the derived relative position however depends strongly on the separation between these sources ( in our case 55 arcminutes ) .
the larger the angular distance , the larger the systematic errors are .
therefore our strategy was to identify a more nearby compact source in our very first observation , that could be used as a reference point in the following epochs .
we did find a candidate in the first catalogue that was unresolved on milliarcsecond scales , first j1644 + 5736 , with a separation of 2.9 arcminutes .
swift j1644 + 5734 and first j1644 + 5736 were within the primary beam of the smaller telescopes , and only the large dishes had to regularly nod between these sources ( see fig.[fig1 ] for the evn images ) .
using the technique described above , we achieved very high precision relative astrometry between our in - beam reference source and the radio afterglow of swift j1644 + 5734 .
the standard deviation of our five relative positions spanning three years were 13 @xmath3as in right ascension and 11 @xmath3as in declination , just a tiny fraction ( 1/160th ) of the synthesized beam .
we see no detectable proper motion in swift j1644 + 5734 ( relative to first j1644 + 5736 ) .
the 3-@xmath4 upper limit to the average ( over 3 years ) apparent jet velocity is @xmath0 .
the implied limit on the intrinsic jet speed as a function of viewing angle is shown on fig.[fig2 ] . if the radio emission is produced by a forward shock analogous to gamma - ray burst jets with a lorentz - factor @xmath5 ( blue line in fig.[fig2 ] ) or higher ( cf .
* zauderer et al . 2013 ) ) , the implied viewing angle is less than 3 degrees
. a lower average lorentz factor ( e.g. due to a dense interstellar medium ) would allow larger viewing angles .
in this case , the observations would indicate that significant deceleration took place within about three years after relativistic ejecta were triggered in swift j1644 + 5734 , and this may be expected in other non - thermal ( jet - forming ) tdes as well .
ultra - precise vlbi astrometry commenced at a closer epoch to the trigger event may give stronger constraints for other tdes in the future , possibly for systems that are less - beamed ( i.e. have a larger viewing angle , but not completely de - boosted ) . alternatively , the radio jet in swift j1644 + 5734 might have been mildly relativistic ( @xmath6 ) from the beginning , in accordance with what is expected for purely radiatively driven jets ( ( * ? ? ?
* sadowski & narayan 2015 ) ) . while @xcite detected relativistic reverberation in the accretion flow of the system , and argued for a mildly relativistic wind as the source of x - rays .
these models however do not naturally explain the broad - band properties of swift j1644 + 5734 , that imply a jet with an initially high lorentz factor .
relativistic - jet tdes remain in the forefront of transient research with the evn , and the first result claiming a resolved jet in asassn-14li has just been published by @xcite . in the future ,
we intend to launch a ska - vlbi ( ( * ? ? ?
* paragi et al . 2015 ) ) follow - up program for the most promising candidates found by the ska . | tidal disruption events tde ) understand two questions black hole astrophysics : form grow impact on evolution host galaxies review komossa 2015 first question addressed through constraining low - mass end of supermassive black hole mass - function stone metzger 2016 tdes light - up dormant black holes not active galactic nucleus signature size of black hole disrupted star estimated well tdes provide means weigh black holes down to @xmath1 majority undetected second question addressed through observations of newly formed radio jets triggered by tde ) on milliarcsecond scales long baseline interferometry ( vlbi ) localise origin of hi outflows reveal dynamics of atomic gas swept - up by radio jet catching agn jet feedback morganti 2013 vlbi of jetted tdes provide chance observing understanding early phases of jet formation evolution in pristine environment without past agn activity future blind radio surveys detect large number of tdes multi - band approach essential for identification role of vlbi two - fold high resolution observations of strong candidates reveal nuclear origin tde identification earlyidentified tdes from x - ray optical spectra off - nucleus may point to peculiar systems white dwarf disrupted by intermediate - mass black hole shcherbakov et al 2013 tde from recoiling black hole komossa merritt 2008 or may reveal secondary galactic nucleus in merger liu & chen 2013 latter interesting tde rate higher in post - starburst galaxies major merger event french arcavi zabludoff 2016 jetted tdes more efficiently detected in large - scale radio surveys current x - ray missions tdes primary targets of synchrotron transients with square kilometre array expected numbers depend on of jetted events relativistic beaming factors donnarumma & rossi 2015 obtained strong constraint on average apparent jet speed for j1644 + 5734 details in @xcite transient swift j1644 + 5734 extremely luminous from radio to @xmath2 rays interpreted as non - thermal tde with relativistic jet levan et al . 2011 bloom et al. 2011 burrows zauderer 2011variability in bands indicated x - ray radio emission likely produced in different regions related to decelerating external shock in ultrarelativistic outflow giannios metzger 2011 similar models explain multi - band properties of @xmath2-ray afterglows zauderer 2013 vlbi technique relativistic ejecta by resolving them j1644 + 5734 @xcite marginally resolving source possible at high frequencies with instrumentation at 22 ghz and above emission expected fade quickly another way to detect motion of relativistic ejecta followed observed j1644 + 5734 at six epochs between 12 april 2011 and 9 march 2015 with evn at 5 ghz detection of targets at few - mjy level requires frequent visits to bright compact calibrator phase - referencing reid honma 2014 provides relative position between target and calibrator used j1638 + 5720 as primary calibrator precision of relative position depends on separation between sources 55 arcminutes larger angular distance larger systematic errors strategy to identify nearby compact source in first observation as reference point in following epochscandidate first catalogue unresolved first j1644 + 5736 separation 2.9 arcminutes swift j1644 + 5734 within primary beam smaller telescopes large dishes nod between sources fig. achieved high precision relative astrometry between beam reference source radio afterglow swift j1644 + 5734 standard deviation of five relative positions three years 13 @xmath3as in right ascension 11 @xmath3as in declination tiny fraction synthesized beam no detectable motion in swift j1644 + 5734 5736 3-@xmath4 upper limit average over 3 years jet velocity is @xmath0 implied limit intrinsic jet speed function viewing angle shown on fig.[fig2 ] radio emission by forward shock gamma - ray burst jets lorentz - factor @xmath5 or higher implied viewing angle less than 3 degrees lower lorentz factor larger viewing angles observations significant deceleration three years after relativistic ejecta triggered in swift j1644 + 5734 expected in other non - thermal tdes- precise vlbi astrometry closer to trigger event may give constraints for tdes future systems less - beamed larger viewing angle not completely de - boosted radio jet in swift j1644 + 5734 might mildly relativistic @xmath6 beginning radiatively driven jets sadowski narayan 2015 @xcite detected relativistic reverberation accretion flow argued relativistic wind source x - rays models explain broad - band properties swift j1644 + 5734 jet high lorentz factor relativistic - jet tdes forefront transient research first result resolved jet in asassn-14li published by @xcite intend launch ska - vlbi paragi 2015 follow - up program for promising candidates | 0.324763 | 2,024 | 1,032 | , Saving $0.1 in GPT-4. |
67,849 | a new set of mathematical identities is presented for axi - symmetric sessile drops on flat and curved substrates .
the geometrical parameters , including the apex curvature and height , and the contact radius , are related by the identities .
the validity of the identities are checked by various numerical solutions both for flat and curved substrates . * new identities for sessile drops * maryam hajirahimi @xmath0 + fahimeh mokhtari@xmath1 + amir h. fatollahi @xmath1 _ 1 ) physics group , south tehran branch , islamic azad university , + p. o. box 11365 , tehran 4435 , iran , + 2 ) department of physics , alzahra university , + p. o. box 19938 , tehran 91167 , iran _
pacs : 47.55.d- , 47.85.dh , 68.03.cd keywords : drops , hydrostatics , surface tension | the study of drops with the effect of gravity being balanced with the surface tension goes back to more than a century @xcite , followed by renewal updates extending and refining the original treatment @xcite .
more recent efforts have concentrated on presenting approximate analytical solutions or developing more efficient methods for numerical solutions @xcite .
heuristically , the balance between the surface effects and the bulk ones would fix the profile of a drop . while the gravity lowers the center of mass ,
the surface tension ( @xmath2 ) tends to decrease the surface , and the adhesion coefficient ( @xmath3 ) tends to increase the surface of the contact region . for a drop with volume @xmath4 , density @xmath5 and comparable surface effects ( _ i.e. _ @xmath6 )
, the so - called bond number defined by the dimensionless combination @xmath7 would determine whether weight has the dominant contribution or not .
mathematically , at every point of the drop s surface the young - laplace relation holds , @xmath8 where @xmath9 are two principal radii of curvature at the point , and @xmath10 is the pressure jump across the liquid - vapor interface . as the hydrostatic laws express @xmath11 in terms of the surface equation , the young - laplace relation is the differential equation by which , together with the boundary conditions , the drop s profile is determined .
as one of the boundary conditions , the contact angle ( @xmath12 ) is fixed by the young equation @xmath13 the purpose here is to present a set of mathematical identities for sessile drops . in particular , for the cases of sessile drops on flat and curved substrates , by direct integration of young - laplace relation over the entire surface of drop , exact identities are derived .
the geometrical parameters of drop , including the height and curvature at the apex and the contact radius of the drop are related by the identities .
the importance of the mentioned parameters is that , they are initially unknown , and are determined only after the complete solution is available .
the validity of the identities are checked by various numerical solutions both for flat and curved substrates .
using the cylindrical coordinate setup given in fig . 1
the total curvature of the axi - symmetric surface @xmath14 is given by @xmath15 where @xmath16 .
on the other hand , the pressure jump in presence of gravity gets contribution from the weight of the drop s layers as well , leading to @xmath17 in which @xmath18 is the height of the drop s apex , and @xmath19 is a constant representing the pressure jump due to the surface tension .
so , the young - laplace relation reads @xmath20 in which @xmath21 and @xmath22 are denoting the upper and lower parts of the drop , respectively ( see fig .
1b ) , and @xmath23 .
at apex ( @xmath24 ) we have @xmath25 , and so by ( [ 1 ] ) , @xmath26 is simply the curvature at apex .
the main issue with equation ( [ 5 ] ) is that the parameters @xmath18 and @xmath26 are not known at the first place , and would be determined only after the complete solution is available .
so at starting point the main equation is not fully known .
further , the contact radius @xmath27 ( fig . 1 ) , as the limiting value for the variable @xmath28 , is not known at first place . as will be seen shortly , by integrating the young - laplace relation an identity
is obtained which relates the three unknown parameters in a very helpful way .
hereafter , the cases for flat and curved substrates are considered separately . , b ) @xmath29 . ]
the boundary conditions for @xmath30 are : @xmath31 we mention @xmath32 and @xmath33 . in case with @xmath34 ( [ 7 ] ) and ( [ 8 ] ) are valid for @xmath21 . in what follows
we mainly consider the case with @xmath29 . the generalization to case with @xmath35 is rather straightforward .
integrating the young - laplace relation for the upper and lower parts of drop leads to @xmath36 in which we have used @xmath37 for @xmath29 . subtracting ( [ 9 ] ) and ( [ 10 ] ) leads to the identity @xmath38 in which we have used the relation for the volume of drop , @xmath39 it is easy to show that identity ( [ 12 ] ) is valid for the acute contact angle ( @xmath35 ) as well .
it is emphasized in ( [ 12 ] ) no approximation is used , hence it is an exact relation .
it would be useful to check the above identity for the case with absence of gravity , in which , as only the surface effects are present , the drop s surface is part of sphere . by direct insertion it can be seen that the following satisfies the young - laplace relation ( [ 5 ] ) , @xmath40 representing a sphere with radius @xmath41 whose center is located at @xmath42 . following a simple geometrical argument in the sphere ( see fig . 1 ) , we have @xmath43 as on the surface of a sphere the curvature is constant , we have @xmath44 . by setting @xmath45 in the identity (
[ 12 ] ) , it is simply satisfied by the given values . , and definition of the angle @xmath46 ( single - arc ) , as the angle between the drop and horizontal line at the contact point
. the contact angle is represented by double - arc .
other parameters are the same as fig . 1 . ] in the lines similar to the case for the flat case , we can derive the identities for curved substrate as well .
the main differences , as follow , appear in the boundary condition , and in the way that volume of drop comes to the play .
the possible situations are represented in figs . 2 & 3 . here
the angle between the drop s surface at contact point and the horizontal line , denoted by @xmath47 in figs .
2 & 3 , appears as part of the boundary conditions .
first let us consider cases shown in fig . 2 . for cases with obtuse angle ,
the boundary conditions read : @xmath48 again , in case with @xmath34 ( [ 17 ] ) is valid for @xmath21 . in what follows
we mainly consider the case with @xmath29 . the generalization to case with @xmath35
is rather straightforward . integrating the young - laplace relation for the upper and lower parts of drop leads to @xmath49 in which , as mentioned before , @xmath47 is the angle between the drop s surface at contact point and horizontal line ( fig .
2 ) , for which we have @xmath50 subtracting ( [ 18 ] ) and ( [ 19 ] ) leads to the identity @xmath51 in which we have used the following relations for the volumes , @xmath52 in which @xmath53 is the equation of the substrate ( fig . 2 ) .
it is easy to show that identity ( [ 21 ] ) is valid for the acute contact angle as well .
again , it is emphasized in ( [ 21 ] ) no approximation is used , hence it is an exact relation .
now we come to the case shown in fig . 3 , in which @xmath54 makes a dimple .
for this case the young - laplace relation reads : @xmath55 by which after integration one finds @xmath56 or , by the relations for volumes , @xmath57 so , for case in fig .
3 the identity finds a slightly different form than others .
in order to provide the numerical tests of the identities obtained in previous section , here a collection of numerical solutions is presented which covers all the situations for which the identities are claimed to hold .
the outputs of numerical solutions , including the contact radius @xmath27 , apex height @xmath18 and curvature @xmath26 , and in case for curved substrates the angle @xmath47 , are presented in tabs . 1 - 4 , by which the direct tests of identities are made possible .
as illustrations , the plots of all of the numerical solutions are presented in figs .
4 - 7 .
d. j. ryley and b. h. khoshaim , _ a new method of determining the contact angle made by a sessile drop upon a horizontal surface ( sessile drop contact angle ) _ " , j. colloid interface sci . * 59 * ( 1977 ) 243 .
y. rotenberg , l. boruvka & a. w. neumann , _ determination of surface tension and contact angle from the shapes of axisymmetric fluid interfaces _ " , j. colloid interface sci .
* 93 * ( 1983 ) 169 ; p. cheng , d. li , l. boruvka , y. rotenberg & a. w. neumann , _ automation of axisymmetric drop shape analysis for measurements of interfacial tensions and contact angles _ " , colloids surf .
* 43 * ( 1990 ) 151 .
o. i. del rio and a. w. neumann , _ axisymmetric drop shape analysis : computational methods for the measurement of interfacial properties from the shape and dimensions of pendant and sessile drops _ " , j. colloid interface sci .
* 196 * ( 1997 ) 136 .
e. moy , p. cheng , z. policova , s. treppo , d. kwok , d. r. mack , p. m. sherman , and a. w. neumann , _ measurement of contact angles from the maximum diameter of non - wetting drops by means of a modified axisymxnetric drop shape analysis _ " , colloids surf .
* 58 * ( 1991 ) 215 . | study of drops with gravity balanced with surface tension goes back to a century followed by renewal updates original treatment recent efforts concentrated on presenting approximate analytical solutions or efficient methods for numerical solutions balance between surface effects and bulk fix profile drop gravity lowers center of mass surface tension surface adhesion coefficient surface contact region for drop with volume density @xmath5 comparable surface effects bond number defined by dimensionless combination @xmath7 weight dominant contribution mathematically at drop s surface young - laplace relation holds are two principal radii of curvature at point is pressure jump across liquid - vapor interface hydrostatic laws express surface equation young - laplace relation is differential equation boundary conditions drop s profile determined contact angle fixed by young equation purpose to present mathematical identities for sessile drops for on flat and curved substrates by integration of young - laplace relation over surface drop exact identities derived geometrical parameters of drop height curvature contact radius related by identities of mentioned parameters initially unknown determined only after complete solution availablevalidity identities checked by numerical solutions for flat and curved substrates cylindrical coordinate setup fig . 1 total curvature of axi - symmetric surface @xmath14 given by @xmath15 @xmath16 pressure jump in gravity from weight drop s layers leading to @xmath17 @xmath18 is height drop s apex @xmath19 constant pressure jump due surface tension young - laplace relation reads @xmath20 @xmath21 @xmath22 upper lower parts of drop @xmath23 at apex @xmath24 @xmath25 @xmath26 is curvature at apex issue with equation ( 5 ) parameters @xmath18 and @xmath26 not known determined after complete solution main equation not fully known contact radius @xmath27 fig . 1 limiting value for variable @xmath28 not known integrating young - laplace relation identity obtained relates three unknown parameters cases for flat and curved substrates considered separately boundary conditions for @xmath30 are @xmath31 @xmath32 @xmath33 @xmath34 8 valid for @xmath21 consider case with @xmath29generalization case @xmath35 straightforward integrating young - laplace relation upper lower parts drop leads to @xmath36 used @xmath37 for @xmath29 subtracting ( [ 9 ] [ 10 ] ) leads to identity @xmath38 used relation volume drop @xmath39 identity ( [ 12 ] ) valid for acute contact angle @xmath35 ) ( [ 12 ] ) no approximation exact relation check identity case gravity surface effects present drop s surface part sphere direct insertion satisfies young - laplace relation 5 ) @xmath40 sphere radius @xmath41 center at @xmath42 geometrical argument sphere fig 1 @xmath43 curvature constant @xmath44 setting @xmath45 in identity ( [ 12 ] ) satisfied by given values definition angle @xmath46 angle between drop horizontal line contact point contact angle represented by double - arc parameters same as fig . 1 similar flat case derive identities curved substrate main differences in boundary condition volume of drop possible situations represented in figs . 2 & 3 angle between drop s surface contact point horizontal line denoted by @xmath47 figs2 & 3 boundary conditions consider cases fig . 2 obtuse angle boundary conditions read : @xmath48 @xmath34 ( [ 17 ] ) valid for @xmath21 consider case @xmath29 generalization to @xmath35 straightforward integrating young - laplace relation upper lower parts drop leads to @xmath49 @xmath47 is angle between drop s surface at contact point horizontal line 2 @xmath50 subtracting ( [ 18 ] ) and ( [ 19 ] ) leads to identity @xmath51 used relations for volumes , @xmath52 @xmath53 is equation of substrate identity ( [ 21 ] ) valid for acute contact angle 21 no approximation exact relation case fig . 3 @xmath54 makes dimple young - laplace relation reads : @xmath55 after integration finds @xmath56 or @xmath57 identity finds different form numerical tests of identities collection of numerical solutions presented covers situations identities outputs contact radius @xmath27 apex height @xmath18 curvature @xmath26 angle @xmath47 presented in tabs .1 - 4 direct tests identities possible plots numerical solutions figs 4 - 7 j. ryley khoshaim new method determining contact angle sessile drop horizontal surface j colloid interface sci . 59 1977 ) 243 rotenberg boruvka w. neumann determination surface tension contact angle axisymmetric fluid interfaces colloid interface sci 93 1983 ) 169 p. cheng d li boruvka rotenberg neumann automation axisymmetric drop shape analysis interfacial tensions contact angles colloids surf 43 1990 ) 151 del rio a w neumann axisymmetric drop shape analysis computational methods interfacial properties shape dimensions pendant sessile drops j colloid interface sci . 196 1997 ) 136. moy p cheng policova treppo kwok mack sherman neumann measurement contact angles maximum diameter non - wetting drops modified axisymxnetric drop shape analysis colloids surf 58 ( 1991 ) 215 | 0.325849 | 2,175 | 1,093 | , Saving $0.1 in GPT-4. |
67,868 | we present preliminary results and motivation of measurement of the total width of the @xmath0 meson ( @xmath1 ) .
address = institute of physics , jagiellonian university , poland & ikp research centre jlich , germany address = institute of physics , jagiellonian university , poland & ikp research centre jlich , germany | the physics of the @xmath0 meson receives an increasing interest in view of the forthcoming measurements planned e.g. at the cosy , da@xmath2ne-2 and mami - c facilities where the @xmath0 will be produced in hadron - hadron , e@xmath3-e@xmath4 , and @xmath5-hadron reactions , respectively .
experimentally the emphasis will be put on the studies of the @xmath6 meson decays which are of interest on its own account and certainly will provide inputs to the phenomenology of the quantum chromo - dynamics @xcite in the non - perturbative regime .
specifically , precise determinations of the partial widths for the @xmath6 decay channels will be helpful for the development of the chiral perturbation theory .
however , the experimental precision of the partial width for various decay channels where only the branching ratio is known or will be measured is governed by the precision of the knowledge of the total width . in the case of the @xmath0 meson
the branching ratios are typically known with precision better than 1.5% , while the total width is established about 10 times less accurate @xcite .
therefore , we expect that the precise determination of the natural width of the @xmath0 meson will have a strong impact on the physics results which will be derived from measurements carried out by collaborations : wasa - at - cosy @xcite , cball - at - mami @xcite and kloe-2 @xcite .
in the last issue of the review of particle physics only two direct measurements of the natural width of the @xmath0 meson are reported @xcite . in the first experiment
the width was established from the missing mass spectrum of the @xmath7 reaction measured close to the threshold for the production of the @xmath6 meson @xcite .
the experimental mass resolution achieved was equal to 0.75 mev / c@xmath8 ( fwhm ) and the extracted value of @xmath9 amounts to @xmath10 mev / c@xmath8 . in the second experiment
the value of @xmath9 was derived from the threshold excitation function of the @xmath11 reaction @xcite .
the study was performed at 20 different beam momenta , however , the error of the @xmath9 was larger than in the previous measurement due to the large relative monitoring uncertainities . in
this experiment @xmath12 mev / c@xmath8 was determined .
the mean value of the width of the @xmath0 meson from the two direct measurements @xcite amounts to @xcite and differs from the value of determined indirectly from the combinations of partial widths obtained from integrated cross sections and branching ratios @xcite .
the significantly more precise direct determination of the @xmath1 should resolve this discrepancy .
based on many years of experience gained with the cosy-11 apparatus @xcite we expect that combined with the excellent features of the stochastically cooled proton beam of cosy @xcite it will allow to significantly reduce the present uncertainty in the value of the natural width of the @xmath0 meson .
the @xmath1 will be derived directly from the missing mass distribution of the @xmath13 reaction measured near the kinamatical threshold .
the advantage of a study close to the threshold is that the uncertainities of the missing mass determination are much reduced since @xmath14 tends to zero . here by _
mm _ we denoted the missing mass and by _ p _ the momentum of the outgoing protons .
the experiment was conducted for five discret values of beam momentum : 3211 , 3213 , 3214 , 3218 , and 3224
mev / c , where the threshold value amounts to 3208.3 mev / c . in order to improve the experimental resolution of the four - momentum determination of the registered particles and in order to decrease the spread of the momentum of the beam protons reacting with the target two major changes
have been applied to the cosy-11 setup ( fig .
[ jeden ] ) .
namely , the spatial resolution of the measurement of the particle track coordinates in the drift chambers was improved by increasing the supply voltage up to the maximum allowed value and also the dimensions of the target in the direction perpendicular to the cosy beam was decreased from 9 to circa 1 mm .
the changes were possible due to the development of the mnster type cluster - jet target @xcite . due to the high value of the dispersion at the position of the cosy-11 target , the decrease of the size of the reaction volume resulted in a significant reduction of the momentum
spread seen by the target . the effect is visualized in fig .
[ piec]a . as can be clearly implied from the figure ,
the information about target size is crucial for determination of beam momentum spread .
it is also of great importance for the estimation of an error of momentum reconstruction of outgoing protons .
therefore , in order to control the systematical uncertainties each of the crucial parameters ( like target dimesions or beam momentum spread ) were monitored by at least two independent methods .
the systematical error of the extraction of @xmath1 will depend on the accuracy of the determination of the missing mass resolution .
this depends predominantly on the knowledge of the momentum spread of the cosy beam ( fig .
[ piec]a ) and on the accuracy of the four - momentum determination of the registered protons . in the case of the experimental technique used by the cosy-11 collaboration ,
both mentioned factors , depend on the dimensions of the target .
therefore it is crucial to monitor precisely the spatial size of the target perpendicular to the cosy beam . to this end
we exploit two independent methods .
one of the techniques relies on the measurement of the gas pressure in the last chamber of the cluster - jet dump ( fig .
[ piec]b - c ) , and the second method is based on the determination of the momentum distribution of elasticaly scattered protons ( fig .
[ piec]d - e ) .
the gas pressure was measured as a function of the position of wires moving with the constant velocity through the cluster beam .
these wires were rotated around the axis perpendicular to the beam of hydrogen clusters . in case that one or more wires crossed the target the clusters hitting a wire
were stopped causing a decrease of the pressure in the last stage of the cluster dump .
the wire device was located above the reaction point ( fig .
[ piec]c ) allowing to monitor the target dimensions without disturbing the measurement of the @xmath15 reaction .
the knowledge of the wire thickness and geometry enables to calulate the target dimensions from the ditribution of the pressure as a function of the position of the device .
[ piec]b shows an example of the measured pressure as function of time using a constant angular velocity of the device .
the extraction of the target dimensions is in progress . from a preliminary analysis
we expect to achieve an accuracy of about 0.2 mm .
the second technique used for monitoring a target spatial size is based on the measurement of the momentum distribution of the elastically scattered protons .
the momentum reconstruction of registered protons is performed by tracing back trajectories from drift chambers through the dipole magnetic field to the target , which is assumed to be an infinitely thin vertical line . in reality , however , the reactions take place in that region of finite dimensions where beam and target overlap . consequently , assuming in the analysis an infinitesimal target implies a smearing of the momentum vectors and hence a decrease of the resolution of the missing mass signal . therefore , we have developed a method to monitor this overlap by measuring the elastically scattered protons @xcite .
trajectories of the protons scattered in the forward direction are measured by means of two drift chambers ( d1 and d2 ) and a scintillator hodoscope ( s1 ) , whereas the recoil protons are registered in coincidence with the forward ones using a silicon pad detector arrangement ( si ) and a scintillation detector ( s4 ) . the two body kinematics gives an unambiguous relation between the scattering angles of the recoiled and forward flying protons .
therefore , the events of elastically scattered protons can be identified from the correlation line formed between the position in the silicon pad detector si , and the scintillator hodoscope s1 , the latter measured by the two drift chamber stacks . for those protons which are elastically scattered in forward direction and
are deflected in the magnetic field of the dipole the momentum vector at the target point can be determined . according to two
body kinematics , the momentum components parallel and perpendicular to the beam axis form an ellipse .
an example is shown in fig .
[ piec]d - e .
the width of the distributions can be used as a measure of the size of the interaction region . for the appraisal of the effect in fig .
[ piec]d - e we present results obtained with targets of 9 and 1 millimeters .
measurement of the missing mass distributions at five different beam energies will allow for monitoring the systematic uncertainities in the determination of the experimental mass resolution .
this is mainly because the smearing of the missing mass due to the natural width of the @xmath6 remains unaltered when the beam momentum changes , whereas the smearing caused by the experimental uncertainities will narrow with the decreasing beam momentum and at threshold it will reach a constant value directly proportional to the spread of the beam momentum .
the effect is shown in fig .
[ szesc]a .
this figure illustrates also that the change of the target thickness by 8 mm results in a change of the mass resolution by about 0.3 mev . since we expect to control the target thickness with the accuracy better then 0.2 mm , the systematical error due to the determination of the target size should be smaller than 0.01 mev .
moreover , we can also control the influence on the mass resolution caused by a different experimental sources .
for example from the angular distribution of the missing mass spectrum we will be able to estimate contributions to the mass resolution due to the spread of the beam momentum and due to the proton momentum reconstruction .
this is because the resolution of the missing mass due to the spread of the beam momentum is almost independent of the polar emission angle of the @xmath6 meson , whereas the smearing of the missing mass due to the uncertainty of the proton momentum reconstruction does depend on the emission angle of the @xmath6 meson .
an online analysis has revealed a signal originating from the production of the @xmath6 meson at each of the investigated beam momenta ( fig .
[ szesc]b - f ) .
as expected the width of the signal form the @xmath6 meson increases with increasing beam momentum .
the width of the @xmath0 signal determined closest to threshold equals to approximatly 0.4 mev ( fwhm ) .
taking into account that the width of the @xmath6 is around 0.2 mev we may estimate the achieved experimental resolution to be about 0.3 mev just at the same order as the searched signal .
presented spectra were obtained with preliminary calibration of the detection system and they will be corrected for the possible broadening due to the changes of the beam optics which could caused the variations of the beam momentum value at the order of 10@xmath16 . in order to enable such corrections we have monitored various parameters which could influence the beam conditions like current intensity in the cosy dipoles , the temperature of the cooling water of the magnets , air temperature , humidity and barometric pressure inside cosy tunnel . independently , it will be possible to correct the variation of the beam momentum based on the distribution of the elasticaly scattered protons measured simultaneously with the @xmath17 reaction .
thus there is still a room for the improvement of the experimental resolution . at present
the off - line analysis of the data is in progress .
we acknowledge the support of the european community - research infrastructure activity under the fp6 programme ( hadron physics , n4:etamesonnet , rii3-ct-2004 - 506078 ) , the support of the polish ministry of science and higher education under the grants no .
pb1060/p03/2004/26 , 3240/h03/2006/31 and 1202/dfg/2007/03 , and the support of the german research foundation ( dfg ) .
9 b. borasoy and r. niler , these proceedings .
yao _ et al .
_ ( particle data group ) , j. phys .
g * 33 * , 1 ( 2006 ) .
adam _ et al .
_ , e - print archive : nucl - ex/0411038 .
a. thomas , these proceedings .
f. ambrosino _ et al .
_ , e - print archive : hep - ex/0603056 .
c. bloise , these proceedings .
d. m. binnie _ et al .
b * 83 * , 141 ( 1979 ) .
r. wurzinger _
et al . _ ,
b * 374 * , 283 ( 1996 ) .
s. brauksiepe _
a * 376 * , 397 ( 1996 ) .
p. moskal _ et al .
instr . & meth . a * 466 * , 448 ( 2001 ) .
d. prasuhn _ et al .
@xmath18 meth . a * 441 * 167 ( 2000 ) .
h. stockhorst _
et al . _ ,
these proceedings .
h. dombrowski _ et al .
. meth . a * 386 * , 228 ( 1997 ) .
a. tschner _
et al . _ ,
these proceedings .
r. czyykiewicz , phd thesis , jagiellonian university ( 2007 ) .
e. czerwiski , thesis , jagiellonian university ( 2006 ) . | physics @xmath0 meson increasing interest forthcoming measurements at cosy da@xmath2ne-2 mami - c facilities produced in hadron - hadron e@xmath3-e@xmath4 @xmath5-hadron reactions emphasis on @xmath6 meson decays provide inputs to quantum chromo - dynamics non - perturbative regime precise determinations partial widths for @xmath6 decay channels helpful for chiral perturbation theory experimental precision partial width decay channels ratio governed by precision total width @xmath0 meson branching ratios known than 1.5% total width 10 times less accurate precise determination natural width @xmath0 meson physics results derived from measurements wasa - at - cosy cball - at - mami @xcite kloe-2 @xcite last issue review particle physics two direct measurements of natural width @xmath0 meson reported first experiment width established from missing mass spectrum of @xmath7 reaction threshold experimental mass resolution to 0.75 mev / c@xmath8 extracted value of @xmath9 amounts to @xmath10 mev / c@xmath8 second experiment value @xmath9 derived from threshold excitation function @xmath11 reaction @xcite study performed at 20 beam momenta error @xmath9 larger due monitoring uncertainities @xmath12 mev / c@xmath8 determined mean value width @xmath0 meson from two measurements @xcite differs from partial widths cross sections branching ratios @xcite more precise determination @xmath1 resolve discrepancy experience cosy-11 apparatus @xcite stochastically cooled proton beam @xcite reduce uncertainty natural width @xmath0 meson @xmath1 derived from missing mass distribution @xmath13 reaction near kinamatical threshold study close threshold uncertainities missing mass reduced @xmath14 tends to zero _ mm _ missing mass p _ momentum outgoing protons experiment conducted for five values beam momentum 3211 3213 3214 3218 3224 mev / c threshold value 3208.3 mev / c improve resolution - momentum determination decrease spread momentum beam protons reacting target two major changes applied to cosy-11 setup ( fig . [ ] )spatial resolution of particle track coordinates in drift chambers improved by increasing supply voltage to maximum dimensions of target perpendicular to cosy beam decreased from 9 to 1 mm changes due to development mnster type cluster - jet target @xcite due to high value dispersion at position cosy-11 target decrease size reaction volume resulted in significant momentum spread target effect visualized in fig . [ piec]a information about target size crucial for beam momentum spread for estimation error of momentum reconstruction of outgoing protons to control systematical uncertainties crucial parameters target dimesions beam momentum spread monitored by two independent methods systematical error of extraction of @xmath1 on accuracy missing mass resolution depends on knowledge momentum spread of cosy beam and accuracy four - momentum determination of registered protons cosy-11 both factors depend on dimensions target crucial to monitor spatial size of target perpendicular to cosy beam two independent methods on measurement gas pressure in last chamber of cluster - jet dump second momentum distribution of elasticaly scattered protonsgas pressure measured function position of wires moving constant through cluster beam wires rotated around axis perpendicular to beam hydrogen clusters one wires crossed target clusters stopped causing decrease pressure in last stage cluster dump wire device above reaction point fig target dimensions without disturbing measurement reaction knowledge of wire thickness geometry calulate target dimensions from ditribution pressure function position device shows example measured pressure function of time using constant angular velocity extraction of target dimensions in progress expect accuracy of about 0.2 mm second technique for monitoring target spatial size measurement momentum distribution of elastically scattered protons momentum reconstruction protons tracing back trajectories from drift chambers dipole magnetic field to target assumed infinitely thin vertical line reactions take place in region finite dimensions where beam target overlap assuming infinitesimal target implies smearing of momentum vectors decrease resolution missing mass signal developed method to monitor overlap by measuring elastically scattered protons trajectories forward direction measured two drift chambers scintillator hodoscope recoil protons registered forward using silicon pad detector arrangement scintillation detector s4two body kinematics between scattering angles recoiled forward flying protons events of elastically scattered protons identified from correlation line between silicon pad detector scintillator hodoscope s1 measured by two drift chamber stacks protons elastically scattered in forward deflected in magnetic field dipole momentum vector at target point determined momentum components parallel perpendicular to beam axis form ellipse example in fig . [ - e width of distributions measure size interaction region results with targets 9 and 1 millimeters measurement of missing mass distributions at five beam energies systematic uncertainities in experimental mass resolution smearing missing mass due to natural width remains unaltered when beam momentum changes smearing caused by experimental uncertainities with decreasing beam momentum constant proportional to spread beam momentum effect shown in fig . [ szesc]a change target thickness by 8 mm mass resolution by 0.3 mev target thickness 0.2 mm systematical error due to target size should smaller than 0.01 mev control influence mass resolution by different experimental sourcesfrom angular distribution missing mass spectrum estimate contributions to mass resolution due spread beam momentum proton momentum reconstruction resolution missing mass due spread beam momentum independent of polar emission angle @xmath6 meson smearing missing mass due proton momentum reconstruction on emission angle @xmath6 meson online analysis revealed signal from @xmath6 meson at each investigated beam momenta ( fig .]b - f ) width signal @xmath6 meson increases with beam momentum width @xmath0 signal closest to threshold equals 0.4 mev width @xmath6 around 0.2 mev estimate experimental resolution about 0.3 mev same searched signal spectra obtained with preliminary calibration detection system corrected for possible broadening due changes beam optics variations beam momentum value at 10@xmath16 monitored parameters like current intensity temperature cooling water magnets air temperature humidity barometric pressure tunnel to correct variation beam momentum based on distribution elasticaly scattered protons measured with @xmath17 reaction room for improvement experimental resolution off - line analysis data in progresssupport european community research infrastructure fp6 programme hadron physics:etamesonnet rii3-ct-2004 - 506078 polish ministry science higher education grants pb1060/p03/2004/26 3240/h03/2006/31 1202/dfg/2007/03 german research foundation borasoy niler proceedings yao et al particle data group j. phys 33 *, 1 ( 2006 ) adam al e - print archive nucl - ex/0411038 thomas proceedings ambrosino al - print archive hep - ex/0603056 bloise proceedings binnie 83 * , 141 ( 1979 ) wurzinger 374 * 283 ( 1996 ) brauksiepe 376 397 1996 ) moskal meth 466 448 ( 2001 ) prasuhn @xmath18 meth 441 * 167 ( 2000 ) stockhorst al proceedings dombrowski meth 386 * 228 ( 1997 ) tschner proceedings czyykiewicz phd thesis jagiellonian university 2007 czerwiski thesis jagiellonian university 2006 ) | 0.44221 | 3,038 | 1,515 | , Saving $0.1 in GPT-4. |
67,869 | @xmath0fe mssbauer spectroscopy measurements are presented in the underdoped ba(fe@xmath1co@xmath2)@xmath3as@xmath3 series for @xmath4=0.014 ( @xmath5k ) and @xmath4=0.03 and 0.045 ( @xmath6 2 and 12k respectively ) .
the spectral shapes in the so - called spin - density wave ( sdw ) phase are interpreted in terms of incommensurate modulation of the magnetic structure , and allow the shape of the modulation to be determined
. in undoped bafe@xmath3as@xmath3 , the magnetic structure is commensurate , and we find that incommensurability is present at the lowest doping level ( @xmath4=0.014 ) . as co doping increases , the low temperature modulation progressively loses its `` squaredness '' and tends to a sine - wave .
the same trend occurs for a given doping level , as temperature increases .
we find that a magnetic hyperfine component persists far above the sdw transition , its intensity being progressively tranferred to a paramagnetic component on heating . | investigation of the magnetic properties of the newly discovered layered fe - based pnictide supercondutors @xcite with a microscopic probe such as mssbauer spectroscopy is of interest for ( at least ) two reasons : i ) to determine the characteristics of the expected spin - density wave ( sdw ) order and ii ) to give experimental evidence concerning the problem of the interaction and/or local coexistence between superconductivity and magnetic ordering of the fe moments . in the so - called 122 family with ba ,
the parent ( non superconducting ) compound is bafe@xmath3as@xmath3 and doping with co substituted for fe has been shown to induce superconductivity at low doping levels @xcite . in the ba(fe@xmath1co@xmath2)@xmath3as@xmath3 series , the phase diagram as a function of co concentration @xmath4 is `` bell - shaped '' , with @xmath7 presenting a maximum of 26k for @xmath4=0.07 @xcite . below this optimal concentration ,
the material shows apparently the persistence of an antiferromagnetic ( af ) structure in the superconducting phase , the af or spin density wave ( sdw ) transition temperature decreasing rapidly as @xmath4 increases .
neutron diffraction in the undoped compound bafe@xmath3as@xmath3 @xcite has determined the af magnetic structure below @xmath8143k to be _ a priori _ commensurate .
@xmath0fe mssbauer spectroscopy @xcite and @xmath9as nmr @xcite in bafe@xmath3as@xmath3 have confirmed the commensurability of the magnetic structure .
resistivity _ vs. _ temperature for bafe@xmath3as@xmath3 with co dopings @xmath4=0.014 , 0.03 and 0.045 from ref.@xcite .
the data for @xmath4=0.014 have been shifted by 100@xmath10.cm for clarity .
the full arrows mark the inflexion point of @xmath11 , attributed to the onset of magnetic ordering ( @xmath12 ) , and the dash - dotted arrows the superconducting transition temperature.,width=415 ] in this work , we focus on the first problem evoked above : we performed a @xmath0fe mssbauer study of ba(fe@xmath1co@xmath2)@xmath3as@xmath3 in the underdoped region in order to investigate the sdw state . in our samples with co doping @xmath4=0.014 , 0.03 and 0.045 , the resistivity data show an inflexion point @xcite respectively around 115k , 92k and 64k ( see fig.[rhot ] ) .
according to neutron diffraction data in materials with similar co dopings @xcite , this temperature corresponds to the sdw - paramagnetic transition .
we find that the spectra in the magnetic phase have an unusual shape , which we interpret as due to an incommensurate modulation of the af structure , i.e. to an incommensurate ( ic ) spin density wave .
we also find that the magnetic hyperfine subspectrum persists far above @xmath12 , i.e. that af correlations are present up to @xmath13 .
the samples are single crystals prepared by the self - flux method , as described in ref.@xcite .
it was checked that co doping is uniform and that superconductivity occurs in the bulk @xcite . for the mssbauer measurements , the crystals were finely ground .
powder samples made by solid state reaction were also prepared , with @xmath4=0 and 0.03 , for comparison purposes .
the spectra were recorded using a commercial @xmath0co@xmath14:rh @xmath15-ray source mounted on an electromagnetic drive with linear velocity signal .
room temperature spectra ensured that no trace of feas ( with a quadrupole splitting @xmath16=0.69mm / s ) was present . in the high temperature tetragonal phase ( space group @xmath17 ) ,
the @xmath18 fe site has fourfold symmetry ( @xmath19m2 ) . at low temperature , slightly above @xmath12 , a small orthorhombic distortion takes place ( space group @xmath20 ) , and the local symmetry at the @xmath21 fe site is lowered ( 222 ) .
the spectra at 4.2k in the three investigated compounds and in undoped bafe@xmath3as@xmath3 , are represented in fig .
[ spe4k2co ] . in bafe@xmath3as@xmath3
, one observes a single magnetic hyperfine pattern , in agreement with ref.@xcite , with a hyperfine field @xmath22 t .
by contrast , the spectral shapes in the co - doped compounds are rather far from this standard 6-line pattern .
saturation effects can be discarded since the resonant absorption is modest ; for @xmath4=0.03 , where it amounts to 7% , we checked that an absorber with 2% absorption yields exactly the same spectrum .
texture effects are also negligible since a powder absorber ( for @xmath4=0.03 ) yields spectra quasi - identical with those of the single crystal sample . the high statistics reached in these experiments allows us to distinguish fine details of the spectral shape , which is a prerequisite for the fits to be described below . .in
ba(fe@xmath1co@xmath2)@xmath3as@xmath3 : fourier coefficients @xmath23 and maximum value @xmath24 ( in t ) of the hyperfine field modulation at 4.2k . [ cols="^,^,^,^,^,^,^,^",options="header " , ] in intermetallic compounds , it is usually assumed that the fermi contact interaction with @xmath25-electrons is the dominant hyperfine coupling , implying that the fe moment * m * is proportional to the hyperfine field * h@xmath26 * @xcite .
this is true in @xmath27-fe , with a hyperfine constant @xmath28=@xmath29 15t/@xmath30 , and in a number of other intermetallics with similar @xmath28 value . in the 122 family of isostructural pnictides ,
the ratios of hyperfine field and moment are smaller and present rather scattered values ( see table [ tchf ] , where the moment is taken from neutron diffraction data ) .
a strict proportionality between moment and hyperfine field can be however questioned in the pnictides .
indeed , the magnetic susceptibility has been found to be anisotropic ( @xmath31 1.5 - 2 ) both in undoped bafe@xmath3as@xmath3 @xcite and in co - doped compounds @xcite .
this is a hint to the presence of an `` unquenched '' orbital moment at the fe site .
this implies a probably small , but non - zero contribution of the @xmath32-orbitals to the hyperfine field , with a sign opposite to that of the fermi contact field .
then , like in fe@xmath33 compounds @xcite , the hyperfine field is no longer strictly proportional to the moment .
the hyperfine field modulation then would not reflect exactly that of the fe moment .
however , assuming as a first good approximation that proportionality holds in the ba(fe@xmath1co@xmath2)@xmath3as@xmath3 series with the same constant as in bafe@xmath3as@xmath3 , then the maximum fe moment of the modulation at 4.2k would be 0.96 , 0.84 and 0.72@xmath30 for @xmath4=0.014 , 0.03 and @xmath4=0.045 respectively . in the neutron diffraction studies of the series ba(fe@xmath1co@xmath2)@xmath3as@xmath3 @xcite ,
the integrated intensity of the magnetic bragg peaks continuously drops towards zero as temperature is increased towards @xmath12 .
this behaviour is in contrast with the @xmath9as nmr work in bafe@xmath3as@xmath3 @xcite , where the transferred hyperfine field at the as nucleus shows a weak thermal variation and it was concluded that the sdw transition is first order .
a similar discrepancy is observed in the present work in co - doped materials , where the main fourier component @xmath34 of the modulation retains a sizeable value at @xmath12 ( see fig.[pourc3 ] top ) .
however , a correct comparison with the neutron data must take into account the fact that the intensity of a magnetic bragg peak is proportional to the square of the magnetic moment . in case of a modulated structure with fourier coefficients @xmath35 , it should scale with @xmath36 , where the brackets denote an average over a period of the moment modulation : @xmath37 figure [ intnorm ] shows the thermal variations of @xmath38 ( from the @xmath39 values determined by mssbauer spectroscopy ) for @xmath4=0.03 and of the intensity of a magnetic bragg peak for @xmath4=0.04 @xcite .
the @xmath40 value is reduced by half at @xmath12 , in disagreement with the neutron data .
similar behaviours occur for @xmath4=0.014 and 0.045 .
it can thus be concluded that the sdw transition in the ba(fe@xmath1co@xmath2)@xmath3as@xmath3 series , as observed with the mssbauer local probe , also presents a strong first order character .
comparing the moment values obtained from the neutron data @xcite for @xmath4=0.025 and 0.045 with the mssbauer results is more difficult since , in the scaling of the neutron bragg intensity , it was assumed that the magnetic structure is identical to that of bafe@xmath3as@xmath3 , i.e. commensurate with the lattice spacings .
the mean saturated fe moment value @xmath41 obtained from neutron diffraction should be compared with : @xmath42 assuming proportionality between hyperfine field and moment , and using the mssbauer derived values for @xmath43 , one gets : @xmath41=0.70 , 0.59 and 0.45@xmath30 resp . for @xmath4=0.014 , 0.03 and 0.045 .
this must be compared with @xmath41=0.35 and 0.17@xmath30 resp . for @xmath4=0.025 and 0.045
derived from the neutron data in ref.@xcite .
further work is needed to resolve the disagreement between local probe and neutron diffraction data about the fe moment variation with temperature and co doping .
variation of the normalised values of the bragg peak intensity for @xmath4=0.04 from ref.@xcite ( diamonds ) and of the mssbauer derived @xmath38 for @xmath4=0.03 from eqn.[mneutr ] ( squares ) , as a function of the reduced temperature @xmath44 .
the mssbauer data above @xmath12 belong to the short range af regions . ]
@xmath0fe mssbauer absorption spectra have been measured in the ba(fe@xmath1co@xmath2)@xmath3as@xmath3 series for @xmath4=0.014 , 0.03 and 0.045 . in all the samples , unusual magnetic hyperfine lineshapes
are observed from 4.2k to @xmath12 .
they are interpreted in terms of a continuous modulation of hyperfine fields at the fe sites arising from an incommensurate electronic sdw state .
incommensurability is present at the lowest doping level , i.e. for @xmath4=0.014 , while the magnetic structure in bafe@xmath3as@xmath3 is commensurate . above @xmath12
, we do not observe a purely paramagnetic signal , but a coexistence of magnetic hyperfine and paramagnetic subspectra in a large temperature range , up to @xmath45 approximately for all samples .
we believe this is due to the persistence of af short range correlated regions , probably dynamic , well above the sdw transition , which is found to have a strong first order character .
our analysis of the hyperfine field modulations as a function of @xmath4 shows that magnetism is not strongly depleted as doping increases , i.e. the mean fe moment or the maximum of the modulation does not decrease with @xmath4 as rapidly as inferred from neutron diffraction data .
we believe the incommensurability of the magnetic structure has not been evidenced by neutron scattering up to now due to the probable very small value of the incommensurate component(s ) of the sdw wave vector .
this makes mssbauer spectroscopy a valuable tool to investigate these structures : although the propagation vector can not be obtained with this technique , the spectral shape is sensitive to minute incommensurabilities . | magnetic properties layered fe - based pnictide supercondutors @xcite mssbauer spectroscopy spin - density wave order evidence superconductivity magnetic ordering fe moments 122 family ba parent compound bafe@xmath3as@xmath3 doping co fe superconductivity low doping levels ba(fe@xmath1co@xmath3as@xmath3 series phase diagram co concentration @xmath4 bell - shaped @xmath7 maximum 26k for @xmath4=0.07 @xcite below concentration material shows antiferromagnetic af structure superconducting phase wave transition temperature as @xmath4 increases neutron diffraction in undoped compound bafe@xmath3as@xmath3 af magnetic structure below @xmath8143k commensurate mssbauer spectroscopy @xmath9as nmr bafe@xmath3as confirmed commensurability magnetic structure resistivity vs temperature for bafe@xmath3as@xmath3 with co dopings @xmath4=0.014 0.03 0.045 from ref.@xcite data for @xmath4=0.014 shifted by 100@xmath10.cm for clarityfull arrows mark inflexion point @xmath11 onset magnetic ordering @xmath12 dash - dotted arrows superconducting transition temperature focus first problem performed @xmath0fe mssbauer study underdoped region investigate sdw state samples with co doping @xmath4=0.014 0.03 0.045 resistivity data show inflexion point @xcite around 115k 92k 64k neutron diffraction data temperature corresponds sdw - paramagnetic transition spectra magnetic phase unusual shape due incommensurate modulation af structure spin density wave magnetic hyperfine subspectrum persists above @xmath12 af correlations up to @xmath13 samples single crystals self - flux method checked co doping uniform superconductivity occurs bulk @xcite crystals finely ground powder samples solid state reaction prepared @xmath4=0 0.03 comparison spectra recorded commercial @xmath0co@xmath14 @xmath15-ray source electromagnetic drive linear velocity signal room temperature spectra no trace feas quadrupole splitting @xmath16=0.69mm / s )high temperature tetragonal phase @xmath17 @xmath18 site fourfold symmetry @xmath19m2 low temperature above @xmath12 small orthorhombic distortion @xmath20 local symmetry @xmath21 site lowered spectra at 4.2k three compounds undoped bafe@xmath3as@xmath3 represented fig . bafe@xmath3as@xmath3 single magnetic hyperfine pattern.@xcite hyperfine field @xmath22 t spectral shapes co - doped compounds far from standard 6-line pattern saturation effects discarded resonant absorption modest @xmath4=0.03 absorber 2% absorption yields same spectrum texture effects negligible powder absorber yields spectra quasi - identical single crystal sample high statistics distinguish fine details spectral shape prerequisite fits ba@xmath3as@xmath3 fourier coefficients @xmath23 maximum value @xmath24 hyperfine field modulation at 4.2kcols="^ intermetallic compounds fermi contact interaction with @xmath25-electrons dominant hyperfine coupling fe moment proportional to hyperfine field h@xmath26 @xcite true in @xmath27-fe hyperfine constant @xmath28=@xmath29 15t/@xmath30 other intermetallics similar @xmath28 value 122 family isostructural pnictides ratios hyperfine field moment smaller values table moment neutron diffraction data strict proportionality between moment hyperfine field questioned in pnictides magnetic susceptibility anisotropic ( @xmath31 1.5 - 2 ) in undoped bafe@xmath3as@xmath3 @xcite co - doped compounds unquenched orbital moment at fe site implies small non - zero contribution @xmath32-orbitals to hyperfine field sign opposite fermi contact field fe@xmath33 compounds hyperfine field strictly proportional to moment modulation reflect fe moment assuming proportionality holds in ba@xmath3as@xmath3 series same constant maximum fe moment modulation at 4.2k 0.96 0.84 0.@xmath30 @xmath4=0.014 0.03=0.045 neutron diffraction studies series integrated intensity magnetic bragg peaks drops towards zero temperature towards @xmath12 contrast with @xmath9as nmr work @xcite transferred hyperfine field nucleus shows weak thermal variation sdw transition first order similar discrepancy present work co - doped materials main fourier component @xmath34 modulation retains sizeable value at @xmath12 fig comparison neutron data intensity magnetic bragg peak proportional to square magnetic moment modulated structure with fourier coefficients @xmath35 should scale with @xmath36 brackets denote average over period moment modulation @xmath37 figure shows thermal variations of @xmath38 @xmath39 @xmath4=0.03 intensity magnetic bragg peak @xmath4=0.04 @xcite @xmath40 value reduced by half at @xmath12 disagreement with neutron data similar behaviours for @xmath4=0.014 0.045 sdw transition in series presents strong first order charactermoment values neutron data @xmath4=0.025 0.045 with mssbauer results difficult assumed magnetic structure identical bafe@xmath3as mean saturated fe moment value @xmath41 neutron compared with @xmath42 mssbauer values @xmath43 @xmath41=0.70 0.59 0.45 @xmath4=0.014 0.03 0.045 with @xmath41=0.35 0.17 @xmath4=0.025 0.045 neutron data work needed resolve disagreement probe neutron diffraction data fe moment variation with temperature doping variation normalised values bragg peak intensity @xmath4=0.04 from@xcite mssbauer derived @xmath38 @xmath4=0.03 reduced temperature @xmath44 mssbauer data above @xmath12 short range af regions @xmath0fe mssbauer absorption spectra measured in ba@xmath3as series for @xmath4=0.014 0.03 0.045 unusual magnetic hyperfine lineshapes observed from 4.2k to @xmath12 continuous modulation hyperfine fields at fe sites from incommensurate electronic sdw stateincommensurability lowest doping level @xmath4=0.014 magnetic structure bafe@xmath3as@xmath3 commensurate above @xmath12 purely paramagnetic signal coexistence magnetic hyperfine paramagnetic subspectra large temperature range up to @xmath45 due to persistence short range correlated regions above sdw transition strong first order character analysis hyperfine field modulations @xmath4 shows magnetism not depleted as doping increases mean fe moment maximum modulation decrease with @xmath4 rapidly neutron diffraction data incommensurability magnetic structure not evidenced by neutron scattering small value incommensurate component sdw wave vector mssbauer spectroscopy valuable propagation vector spectral shape sensitive to minute incommensurabilities | 0.429737 | 2,893 | 1,537 | , Saving $0.1 in GPT-4. |
67,885 | rich regular frequency patterns were found in the fourier spectra of low - amplitude @xmath0 scuti stars observed by corot satellite ( see papar et al . 2016a , b ) .
the corot observations are , however , influenced by the disturbing effect of the south atlantic anomaly .
the effect is marginal for high amplitude variable stars but it could be critical in the case of low amplitude variables , especially if the frequency range of the intrinsic variation overlaps the interval of the instrumental frequencies .
some tests were carried out both on synthetic and real data for distinguishing technical and stars frequencies . | the photometic space telescope corot @xcite moves on a polar low earth orbit with an orbital period of 1 hour 43 minutes .
the orbital plane is fixed with respect to the center of the earth so ( due to the rotation of the earth ) the satellite flies for each circulation above different geographical lenghts .
this orbital property causes that the satellite passes regularly through the south atlantic anomaly ( saa ) .
the saa is an area where the earth s inner van allen zone comes closest to the surface .
it leads to an increased flux of energetic particles what makes the ccd images to be noisy .
the corot satellite observated continousely even in the saa as well , but the data pipeline signed these data points by a non - zero flag .
if we use data points with zero flag only we get of the noisy data but we lose the continuous sampling which causes alias frequencies in the fourier spectra .
the alias structure needs careful study before we search for regular spacing in the frequency lists of corot @xmath0 scuti stars .
all statement here respect to exo field stars in the 2nd corot long run field located in the galactic anticenter direction ( lra2 ) ( see for the specifications @xcite ) .
we prepared an artificial light curve from 22 frequencies taken from the highest amplitude frequencies of a corot @xmath0 scuti star ( corot 102637079 ) .
the amplitudes were set to be identical with the observed ones and gaussian noise ( @xmath1 mag ) was also added to the sythetic light curve .
( this noise level is typical for a 11 - 13 mag star in the exo field . )
the light curve was sampled with the corot sampling time ( 1 ) either uniformly or ( 2 ) in those data points where corot pipeline yields 0 flag ( viz . without any technical problems ) .
the fourier spectra of these two synthetic time series are given in fig.[fig1 ] .
the equidistantly sampled data has clear spectrum , that is no other peaks then the originally set 22 ones ( top panel in fig.[fig1 ] ) . while the spectrum of flag 0 data shows many low amplitude alias peaks caused by the non - uniform data sampling ( bottom panel in fig.[fig1 ] ) .
the alias phenomenon is well - known for ground - based data where daily and yearly ( seasonal ) alias frequencies are the most frequent .
the spectral window function of flag 0 point of the corot data ( green spectrum in fig.[fig2 ] ) contains frequencies at @xmath2 , where @xmath3 d@xmath4 ( sidereal - day frequency ) , @xmath5 d@xmath4(orbital frequency ) , and @xmath6 are integers .
the amplitude of a given alias frequency is varied form star to star but fig.[fig2 ] shows a typical situation : the highest peak amongst the sideral - day harmonic frequencies is the 2@xmath7 d@xmath4 .
the largest order detectable harmonic is generally the 4th or the 5th one .
the orbital frequency amplitude is lower than the amplitude of its first harmonic .
the amplitudes of the linear combination frequencies @xmath8 are higher than the amplitude of the orbital frequency @xmath9 .
the common pre - whitening processes produce further technical frequencies and increasing noise level @xcite , however , due to the many above mentioned aliases , a pre - whitening method is essential for handling these data .
we used sigspec program @xcite for consecutive pre - whitening process . the sigspec algorithm with its default
spectral significance limit ( @xmath10 ) resulted in the 22 input frequencies for both artificial data sets .
both the residual spectra indicate no significant peaks above the 0.01 mmag average ( noise ) level after the pre - whitening process .
these results are consonant with @xcite diagnose , namely that the higher amplitude frequencies are less affected by the faults of pre - whitening process .
in the same time we found that all of the used input frequencies can be considered as ` higher amplitude frequency ' though they amplitudes are between 1.5 and 0.1 mmag . the red line in fig.[fig2 ] shows the spectrum of corot 102637079 .
this spectrum is very similar to the synthetic one ( in the bottom panel of fig.[fig1 ] ) which is evident because it was generated on the basis of the highest amplitude frequencies of the real star .
more instructive is how heavily overlaps the real star s frequency range with the location of the peaks of the spectral window . despite the previous positive test result
this warns for caution .
the observed data have additional biases from the sythetic case .
the flag 0 data might be still encumbered by the influence of the saa . the strength of this influence differs from star to star . according to our experiences
the alias structure can not be removed totally by a consecutive pre - whitening process .
the probable reason is that the proximity of the saa causes some flux variations ( an increase before the saa and a decrease after that ) which have different time lengths and amplitudes for each passing through .
our conclusion is that the higher amplitude peaks of the fourier spectra can be easily separated from the instrumental ones .
attention has to pay to the increasing noise level is due to many - steps pre - whitening processes @xcite .
any well - established study can be accepted only for the higher amplitude peaks .
they should be referred as intrinsic frequencies of the star .
although , beacuse of the increased noise level , we may miss some of the low amplitude intrinsic frequencies .
when we found regular pattern using the high amplitude frequencies we may extend the pattern with lower amplitude frequencies .
this basic idea using only the highest amplitude peaks was applied in @xcite that in our new investigation on a larger sample of corot @xmath0 scuti stars we extended the regular patterns with lower amplitude frequencies , too .
the results and the details of the analysis will be published soon elsewhere ( papar et al .
2016a , b ) .
scuti star corot 102637079 ( red ) . as a comparison
we show the window function of these data ( green ) .
the two functions are plotted in different scales : fourier amplitudes are in corot flux ( left scale ) , while the window function is represented in the usual normalized scale ( right scale ) . ] | photometic space telescope corot @xcite polar low earth orbit orbital period 1 hour 43 minutes orbital plane fixed center earth rotation satellite flies above different geographical lenghts orbital passes through south atlantic anomaly saa saa earth s inner van allen zone closest to surface increased flux energetic particles ccd images noisy corot satellite observated saa data pipeline signed data points non - zero flag data points zero flag noisy data lose continuous sampling causes alias frequencies fourier spectra alias structure needs study regular spacing frequency lists corot @xmath0 scuti stars 2nd corot long run field galactic anticenter direction prepared artificial light curve from 22 frequencies highest amplitude corot @xmath0 scuti star amplitudes identical observed gaussian noise ( @xmath1 mag ) added curve noise typical for 11 - 13 mag star exo field light curve sampled with corot sampling time ( 1 uniformly or data points corot pipeline yields 0 flag fourier spectra of two time series in fig.[fig1 ] equidistantly sampled data clear spectrum no other peaks then originally set 22 ones[fig1 spectrum flag 0 data shows low amplitude alias peaks non - uniform data sampling alias phenomenon known for ground - based data daily yearly seasonal alias frequencies frequent spectral window function flag 0 point data green spectrum contains frequencies at @xmath2 @xmath3 are integers amplitude alias frequency varied star to star highest peak sideral - day harmonic frequencies 2@xmath7 d@xmath4 largest order detectable harmonic 4th or 5th orbital frequency amplitude lower than first harmonic amplitudes linear combination frequencies @xmath8 higher than orbital frequency @xmath9 pre - whitening processes produce technical frequencies noise level pre - whitening method essential used sigspec program for pre - whitening algorithm default spectral significance limit resulted 22 input frequencies for data sets residual spectra indicate no significant peaks above 0.01 mmag average noise level after pre - whitening consonant with higher amplitude frequencies less affected by faults pre - whiteningfound used input frequencies higher amplitude amplitudes between 1.5 and 0.1 mmag red line in fig.[fig2 shows spectrum of corot 102637079 spectrum similar to synthetic one bottom fig generated on highest amplitude frequencies real star overlaps real star s frequency range with location peaks spectral window positive test result warns for caution observed data have biases from sythetic case flag 0 data might encumbered by influence saa strength differs from star to star alias structure not be removed totally by pre - whitening process proximity of saa causes flux variations different time lengths amplitudes higher amplitude peaks separated from instrumental ones increasing noise level due to - pre - whitening processes study accepted only for higher amplitude peaks referred as intrinsic frequencies of star may miss low amplitude intrinsic frequencies found regular pattern using high amplitude frequencies may extend pattern with lower amplitude frequencies idea using highest amplitude peaks applied in new investigation on larger sample corot stars extended regular patterns with lower amplitude frequencies results details analysis published soon ( papar et al . 2016a , b )star corot 102637079 red comparison window function data green ) two functions plotted different scales fourier amplitudes corot flux left scale window function usual normalized scale right scale | 0.26415 | 1,386 | 662 | , Saving $0.0 in GPT-4. |
67,886 | the longitudinal and transverse beam coupling impedance of the first final totem roman pot unit has been measured in the laboratory with the wire method . for the evaluation of transverse impedance
the wire position has been kept constant , and the insertions of the rp were moved asymmetrically . with the original configuration of the rp , resonances with fairly high q values were observed . in order to mitigate this problem ,
rf - absorbing ferrite plates were mounted in appropriate locations . as a result ,
all resonances were sufficiently damped to meet the stringent lhc beam coupling impedance requirements . | the lhc experiment totem @xcite is designed for measuring the elastic pp scattering cross - section , the total pp cross - section and diffractive processes .
these physics objectives require the detection of leading protons with scattering angles of a few @xmath0rad , which is accomplished using a roman pot ( `` rp '' ) system with stations at 147 m and 220 m from the interaction point 5 where also cms will be located .
each station is composed of two rp units separated by a few metres depending on beam equipment integration constraints .
each rp unit consists of a vacuum chamber equipped with two vertical insertions ( top and bottom ) and a horizontal one ( fig .
[ fig : rp ] ) .
each insertion ( `` pot '' ) contains a package of 10 silicon detectors in a secondary vacuum .
the pots can be moved into the primary vacuum of the machine through vacuum bellows . in order to minimise the distance of the detectors from the beam , and to minimise multiple scattering , the wall thickness of the pot
is locally reduced to a thin window foil .
the low impedance budget of the lhc machine ( broadband longitudinal impedance limit @xmath1 ) imposes a tight limit on the rps beam coupling impedance .
the beam coupling impedance measurement was performed with the wire method like with the first rp prototype in 2004 @xcite . after pulling a 0.3 mm thick wire through the rp along its beam axis
, a vector network analyser was used to measure the complex transmission coefficient @xmath2 between the two ends of the rp ( fig .
[ fig : setup ] ) as a function of the frequency @xmath3 and of the horizontal and vertical pot distances ( @xmath4 , @xmath5 ) from the wire .
[ fig : spectra ] shows the measurement result for all pots in retracted position and compares it with simulations based on two different programs .
while the simulations describe qualitatively all main structures seen in the data , the resonances are shifted in frequency by up to 20% .
this disagreement is attributed to the modelling of the bellows in the simulations .
an exact model requires a very dense mesh of the volume which compromises the simulation convergence in an acceptable cpu time .
substituting the bellows with a longer smooth cylinder of an equivalent total metallic surface provides good agreement at the first mode frequency ( @xmath6500mhz ) , but does not succeed at higher frequencies . improved numerical simulation models will be studied . with all pots in retracted position ( @xmath7 mm ) .
] with all pots in retracted position ( @xmath7 mm ) . ]
the longitudinal impedance @xmath8 was calculated with the `` improved log formula '' @xcite @xmath9 , \end{gathered}\ ] ] where @xmath10 is the characteristic impedance of the unperturbed beam pipe and @xmath11 cm is the length of the perturbation ( i.e. the diameter of the pot insertions ) .
the measurement with all pots in retracted position served as reference measurement @xmath12 after removal of all resonances by interpolation in both modulus and phase .
the approximately gaussian lhc bunch structure with @xmath13ns leads to a gaussian envelope with @xmath14ghz in the frequency distribution of the lhc current with harmonics every 40mhz .
hence the relevant resonances lie well below 1ghz .
[ fig : resonance_vs_d ] shows the evolution of frequency and longitudinal impedance of the first double resonance peak as a function of the vertical pot position @xmath5 ( for this measurement the top and the bottom pot positions were symmetrical w.r.t .
the wire ) .
the impedance value of 700@xmath15 for mode 1b at @xmath16 mm corresponds to @xmath17 , where @xmath18 , or a dissipated power of about 200w , which demonstrates the necessity of mitigating hardware modifications .
the latter were realised in the form of a collar of ferrite tiles around the pot insertions ( fig .
[ fig : rp ] , right ) .
they had the desired effect of smearing out all resonances beyond recognition ( fig .
[ fig : ferrites ] ) . before and after installation of the ferrite tiles .
] the transverse impedance was only measured for the configuration without ferrite tiles where resonances were visible . for technical reasons , neither the two - wire method nor
a movable wire were practicable .
instead , the two vertical pots were moved asymmetrically , keeping their relative distance @xmath19 the jaw width constant . then the longitudinal impedance was measured as a function of the excentricity @xmath20 , i.e. the position of the jaw centre with respect to the wire ( fig .
[ fig : transverse ] , top and middle ) . after a parabolic fit @xmath21 , where ideally @xmath22 by virtue of symmetry
, a combination of vertical transverse impedance @xmath23 and detuning impedance @xmath24 can be obtained from the curvature parameter @xmath25 , like in the moving wire technique @xcite : @xmath26 since there is only one horizontal pot , no analogous measurement of the x - component @xmath27 could be made , which would have enabled the elimination of @xmath24 .
however , based on the calculations in ref .
@xcite for different aperture geometries , the contributions @xmath23 and @xmath24 could be approximately disentangled .
the result for the first group of resonances is shown in fig .
[ fig : transverse ] ( bottom ) as a function of the jaw width @xmath19 .
40 mm and different excentricities @xmath20 of this aperture with respect to the wire ; middle : longitudinal impedance as a function of the excentricity @xmath20 for the fixed aperture width @xmath2840 mm ; bottom : dependence of the transverse impedance on the aperture width @xmath19 . for this entire study ,
the horizontal pot was in retracted position.,title="fig : " ] 40 mm and different excentricities @xmath20 of this aperture with respect to the wire ; middle : longitudinal impedance as a function of the excentricity @xmath20 for the fixed aperture width @xmath2840 mm ; bottom : dependence of the transverse impedance on the aperture width @xmath19 . for this entire study ,
the horizontal pot was in retracted position.,title="fig : " ] 40 mm and different excentricities @xmath20 of this aperture with respect to the wire ; middle : longitudinal impedance as a function of the excentricity @xmath20 for the fixed aperture width @xmath2840 mm ; bottom : dependence of the transverse impedance on the aperture width @xmath19 . for this entire study ,
the horizontal pot was in retracted position.,title="fig : " ] the built - in fourier transformation capability of the network analyser facilitated a transmission study in the time domain .
[ fig : time ] shows the transmission response to an injected gaussian pulse with @xmath290.6ns for two different position configurations of the horizontal and vertical pots .
while without ferrites the resonant behaviour of the insertion cavities leads to oscillations extending beyond 25ns after the main pulse , this `` ringing '' is suppressed by the ferrites after less than 10ns . without ( top ) and with ( bottom )
ferrites for two different combinations of the horizonal ( h ) and vertical ( v ) pot distances from the wire .
] focussing on the main pulse , the loss factor can be calculated according to the formula @xcite @xmath30 dt } } { \left[\int{i_{ref}(t)\ , dt}\right]^2 } \:,\ ] ] where the time integral extends over a range of @xmath31 around the peak .
@xmath32 is the reference pulse measured with all pots in retracted position .
[ fig : lossfactor ] shows , the loss factor does not change strongly with the addition of the ferrites .
we thank j. noel , a.g . martins de oliveira and s. rangod for the assembly of the roman pot . 9 totem , technical design report , cern - lhcc-2004 - 002 . m. deile et al . , tests of a roman pot prototype for the totem experiment , proceedings of pac 2005 , knoxville , tennessee , pp.1701ff .
arxiv : physics/0507080 .
e. jensen , ps - rf note 2000 - 001 , cern 2000 .
t. kroyer , f. caspers , e. gaxiola , longitudinal and transverse wire measurements for the evaluation of impedance reduction measures on the mke extraction kickers , cern - ab - note-2007 - 028 .
h. tsutsui , on single wire technique for transverse coupling impedance measurement , cern - sl - note-2002 - 034 ap .
m. sands , j. rees , slac report pep-95 , aug . | lhc experiment totem @xcite designed measuring elastic pp scattering cross - section total pp cross - section diffractive processes objectives require detection leading protons scattering angles few @xmath0rad accomplished using roman pot rp system stations at 147 m 220 m from interaction point 5 cms located each station two rp units separated few metres beam equipment integration constraints each rp unit vacuum chamber two vertical insertions horizontal one each insertion contains 10 silicon detectors secondary vacuum pots moved into primary vacuum through vacuum bellows minimise distance detectors from beam minimise multiple scattering wall thickness pot reduced to thin window foil low impedance budget lhc machine limit imposes limit rps beam coupling impedance impedance measurement performed with wire method first rp prototype 2004 0.3 mm thick wire through rp vector network analyser complex transmission coefficient between two ends rp function frequency horizontal vertical pot distances from wire shows measurement result for all pots in retracted position compares with simulations two programssimulations describe structures resonances shifted by 20% disagreement attributed to modelling bellows exact model requires dense mesh compromises simulation convergence cpu time substituting bellows with longer smooth cylinder provides agreement at first mode frequency @xmath6500mhz succeed at higher frequencies improved numerical simulation models studied all pots in retracted position mm longitudinal impedance @xmath8 calculated with improved log formula @xmath10 characteristic impedance of unperturbed beam pipe @xmath11 cm length of perturbation diameter pot insertions measurement with pots retracted reference measurement @xmath12 after removal resonances interpolation gaussian lhc bunch structure with @xmath13ns leads to gaussian envelope with @xmath14ghz distribution harmonics every 40mhz relevant resonances below 1ghz fig : resonance_vs_d shows evolution of frequency longitudinal impedance of first double resonance peak vertical pot position @xmath5 top bottom pot positions symmetricalimpedance value 700@xmath15 mode 1b at @xmath16 mm corresponds to @xmath17 @xmath18 dissipated power about 200w necessity mitigating hardware modifications realised collar ferrite tiles around pot insertions fig . smearing resonances installation transverse impedance measured for configuration without ferrite tiles where resonances visible neither two - wire method movable wire practicable two vertical pots moved asymmetrically keeping relative distance jaw width constant longitudinal impedance measured function of excentricity @xmath20 position jaw centre wire fig transverse after parabolic fit @xmath21 @xmath22 combination of vertical transverse impedance @xmath23 and detuning impedance @xmath24 obtained from curvature parameter @xmath25 moving wire technique one horizontal pot no analogous measurement of x - component @xmath27 elimination @xmath24 calculations aperture geometries contributions @xmath23 and @xmath24 could be disentangled result for first group of resonances shown in fig . transverse function of jaw width @xmath19 .40 mm excentricities @xmath20 aperture wire middle longitudinal impedance function excentricity fixed aperture width @xmath2840 mm bottom dependence transverse impedance aperture width @xmath19 horizontal pot retracted position 40 mm excentricities @xmath20 aperture wire middle longitudinal impedance excentricity @xmath20 fixed aperture width @xmath2840 mm bottom transverse impedance aperture width @xmath19 horizontal pot retracted position 40 mm excentricities @xmath20 aperture wire middle longitudinal impedance excentricity @xmath20 fixed aperture width @xmath2840 mm bottom dependence transverse impedance aperture width @xmath19 horizontal pot retracted position built - fourier transformation capability network analyser transmission study time domain fig time transmission response injected gaussian pulse @xmath290.6ns position configurations horizontal vertical pots without ferrites resonant behaviour insertion cavities oscillations beyond 25ns after main pulse suppressed ferrites less than 10ns without bottom ferrites horizonal vertical pot distances from wiremain pulse loss factor calculated formula @xcite @xmath30 time integral extends @xmath31 peak @xmath32 reference pulse pots retracted position lossfactor change addition ferrites j. noel martins de oliveira s rangod assembly roman pot totem technical design report cern - lhcc-2004 - 002 deile tests roman pot prototype totem experiment proceedings pac 2005 knoxville tennessee pp.1701ff arxiv jensen note 2000 - 001 cern 2000 kroyer caspers gaxiola longitudinal transverse wire measurements impedance reduction measures mke extraction kickers cern - ab - note-2007 - 028 tsutsui single wire technique transverse coupling impedance measurement cern sl - note-2002 - 034 sands j. rees slac report pep-95 | 0.317851 | 1,947 | 973 | , Saving $0.1 in GPT-4. |
67,888 | central stars of extra - solar planetary systems are metal - rich .
planet accretion or initial surmetallicity can explain this observationnal fact .
these scenarios can be tested with asteroseismology .
we calibrate two stellar models , one with accretion and one with high initial metallicity , in order to obtain the same external parameters for both of them .
we then compare their internal structures and their oscillation frequencies .
# 1_#1 _ # 1_#1 _ = # 1 1.25 in .125 in .25 in | central stars of planetary systems are known to be overabundant in metals ( see for example santos et al . 2001 ; santos et al . 2003 ) .
two scenarios have been suggested to explain such an excess . in the first case ,
the protostellar gas is metal - rich and so is the star .
the high density of heavy elements , compared to stars without planets , increase the probability of planetesimal formation , and thus of planet formation .
the second explanation assumes that planets are formed out of gas with solar like metallicity . the central stars accrete hydrogene poor - matter during and after planetary formation thereby enhancing their metallicitiesin their outer layer ( murray et al .
we suggest that these two scenarios can be tested by asteroseismology . here
we analyse and compare the oscillation frequencies in stellar models with the same external parameters but different histories : with and without accretion .
we computed 1.1 m@xmath0 stars models with the toulouse - geneva code of stellar evolution .
one of them has a high initial metallicity , from the surface to the core , z=0.0443 .
the other model has initial solar abundances but it experienced accretion at the beginning of its main - sequence evolution .
both models include microscopic diffusion .
we adjusted the mass of accreted materia ( m@xmath1 ) , the mixing - length parameter ( @xmath2 ) and the initial helium abundance ( y@xmath3 ) , which are input parameters in our stellar evolution code , in order to obtain stars with the same effective temperature ( t@xmath4 ) , luminosity ( l ) and metallicity ( z ) .
input and output parameters for these models are given in table 1 and table 2 , evolution tracks are represented in figure 1 .
+ the accreted mass is composed of elements heavier than helium with the same relatives abundances as in the sun .
it is given in units of m@xmath5 ( m@xmath5 = 1.9@xmath610@xmath7 kg ) .
when the convergence of two models is obtained , we compute their oscillation frequencies .
we used the adiabatic pulsations code described by brassard et al .
( 1992 ) .
( m@xmath0 ) & m@xmath1 ( m@xmath5 ) & @xmath2 & x@xmath3 & y@xmath3 & z@xmath3 + 1 & 1.1 & 0.5 & 2.08 & 0.7208 & 0.2600 & 0.0192 + 2 & 1.1 & - & 2.08 & 0.6260 & 0.3297 & 0.0443 + t@xmath4 ( k ) & @xmath8 & @xmath9 & @xmath10 & c.c .
+ 1 & 2.123 & 0.7213 & 0.2389 & 0.0482 & 5850 & 1.128 & 1.0368 & 0.7085 & no + 2 & 2.965 & 0.6541 & 0.3056 & 0.0403 & 5847 & 1.136 & 1.0415 & 0.7035 & yes +
as the physical processes in the stars are not the same , the internal structures of models with the same external parameters present detectable differences . in figure 2a we plotted the relative differences of temperature , pression , density and the square of the sound speed for the point on the hr diagram ( see figure 1 ) at which the stars have similar external parameters . in table
2 @xmath11 is the radius at the bottom of the outer convective zone , r@xmath12 is the star radius .
the last column indicates whether or not the star has a convective core .
we computed the oscillation frequencies of our models for different values of the azimuthal degree _
( from 0 to 3 ) and radial order _
n_. we then compared the oscillation frequencies . the relative difference between oscillation frequencies , as functions of the radial order and the frequency itself , is shown in figure 3 .
sound speed differences , which appear in figure 2a , are worth studying .
it reaches 0.5 % below the convective zone and 1 % near the center . in the first case ,
it reflects the difference in temperature and chemical abundances .
the mean molecular weight , @xmath13 , differs and so does the sound speed .
the peak at @xmath14 0.05 r@xmath12 is due to the presence of a convective core in the supermetallic star .
figure 2b pictures the differences of adiabatic oscillation frequencies linked to the differences in sound speed .
they are of the order of a few tenth of percent which is an observable effect with the means of asteroseismology ( bouchy & carrier 2002 ) . + a further level in analysis consists in computing the small and second differences for both stars , i.e. @xmath15 and @xmath16 .
the former is a good test for stellar interior near the core ( ulrich 1986 ) .
the latter is a good indicator for the region near the convective zone ( gough 1990 ) .
computation of these quantities are under process , they will discussed in a forthcoming paper .
+ + bouchy , f. & carrier , f. 2002 , , 390 , 205 brassard , p. , pelletier , c. , fontaine , g. , & wesemael , f. 1992 , , 80 , 725 gough , d. o. 1990 , lnp vol .
367 : progress of seismology of the sun and stars , 283 murray , n. , chaboyer , b. , arras , p. , hansen , b. , & noyes , r. w. 2001 , , 555 , 801 santos , n. c. , israelian , g. , & mayor , m. 2001 , , 373 , 1019 santos , n. c. , israelian , g. , mayor , m. , rebolo , r. , & udry , s. 2003 , , 398 , 363 ulrich , r. k. 1986 , , 306 , l37 | central stars overabundant in metals santos et 2001 ; 2003 two scenarios suggested excess first protostellar gas metal - rich star high density heavy elements increase probability planetesimal formation planet formation second explanation assumes planets formed gas with solar like metallicity central stars accrete hydrogene matter during after formation enhancing metallicitiesin outer layer murray et al scenarios tested by asteroseismology analyse compare oscillation frequencies in stellar models same different histories with without accretion computed 1.1 m@xmath0 stars models with toulouse - geneva code stellar evolution one high initial metallicity z=0.0443 other initial solar abundances experienced accretion beginning evolution both models include microscopic diffusion adjusted mass of accreted materia mixing - length parameter initial helium abundance stars same effective temperature luminosity metallicity z input output parameters in table 1 table 2 evolution tracks in figure 1 accreted mass elements heavier than helium same abundances as sun given in units of m@xmath5 ( = 1.9@xmath610@xmath7 kg )convergence of models compute oscillation frequencies used adiabatic pulsations code brassard et al 1992 m@xmath0 m@xmath1 m@xmath5 ) @xmath2 x@xmath3 y@xmath3 z@xmath3 1 1.1 0.5 2.08 0.7208 0.2600 0.0192 2 1.1 - 2.08 0.6260 0.3297 0.0443 t@xmath4 @xmath8 @xmath9 @xmath10 1 2.123 0.7213 0.2389 0.0482 5850 1.128 1.0368 0.7085 2 2.965 0.6541 0.3056 0.0403 5847 1.136 1.0415 0.7035 physical processes stars same internal structures of models parameters present detectable differences figure 2a plotted relative differences of temperature pression density sound speed for diagram stars similar parameters table 2 @xmath11 radius bottom outer convective zone r@xmath12 star radius last column indicates star convective core computed oscillation frequencies for azimuthal degree radial order compared relative difference shown in figure 3 sound speed differences in figure 2a 0.5 % below convective zone 1 % near centerfirst case reflects difference temperature chemical abundances mean molecular weight @xmath13 differs sound speed peak at @xmath14 0.05 r@xmath12 due to convective core in supermetallic star figure 2b pictures differences adiabatic oscillation frequencies sound speed few tenth of percent observable asteroseismology ( bouchy carrier 2002 ) further analysis computing small second differences for stars @xmath15 @xmath16 former good test for stellar interior near core ulrich 1986 latter indicator region near convective zone gough 1990 ) computation quantities under process discussed forthcoming paper bouchy carrier 2002 , 390 , 205 brassard pelletier fontaine wesemael 1992 , 80 , 725 gough 1990 vol 367 progress seismology sun stars, 283 murray chaboyer arras hansen noyes 2001 555 , 801 santos israelian mayor 2001 373 , 1019 santos rebolo udry 2003 398 , 363 ulrich 1986 , 306 , l37 | 0.191741 | 1,419 | 772 | , Saving $0.0 in GPT-4. |
67,916 | effective hamiltonian for the kicked double well system was derived using the campbell - baker - hausdorff expansion formula .
asymmetric model for the kicked system was constructed .
analytical description of the quasienergy levels splittings for the low laying doublets was given in the framework of the model .
numerical calculations confirm applicability of the proposed effective asymmetric approach for the double well system with the kick - type perturbation . | the connection between the semiclassical properties of chaotic systems and purely quantum processes such as tunneling is a reach rapidly developing field of research nowadays .
our insight in some novel phenomena in this field was extended in the last decades .
the most intriguing among them are the chaos assisted tunneling ( cat ) and the closely related coherent destruction of tunneling ( cdt ) .
the first one is an enhancement of tunneling in perturbed low - dimensional systems at relatively high external field strengths and high driving frequencies ( in order the singlet - doublet crossings to occur ) @xcite .
this phenomenon takes place when levels of the regular doublet undergo an avoided crossing with the chaotic state @xcite . at the semiclassical level of description one
considers tunneling between kam - tori embedded into the `` chaotic sea '' .
the region of chaotic motion affects tunneling rate because compared to direct tunneling between tori it is easier for the system to penetrate primarily into the chaotic region , to travel then along some classically allowed path and finally to tunnel onto another kam - torus @xcite .
cdt phenomenon is a suppression of tunneling when values of amplitude and frequency of driving force belong to some one - dimensional manifold in the perturbation parameters space @xcite .
this phenomenon occurs due to the exact crossing of two states with different symmetries from the tunneling doublet . in this parameter region tunneling time
diverges which means the total localization of quantum state on the initial torus .
cat phenomenon as well as cdt were experimentally observed in a number of real physical systems .
the cat observation between whispering gallery - type modes of microwave cavity having the form of the annular billiard was reported in the ref .
the same phenomenon for ultracold atoms was experimentally investigated in refs .
the study of the dielectric microcavities provided evidences for cat as well @xcite .
both phenomena were observed in two coupled optical waveguides @xcite .
recently experimental evidence of coherent control of single particle tunneling in strongly driven double well potential was reported in ref .
@xcite .
the most common methods which are used to investigate the cat are numerical methods based on the floquet theory @xcite . among other approaches to cat we would like to mention the scattering approach for billiard systems @xcite and quantum mechanical amplitudes in the complex configuration space @xcite .
there is an analytical approach based on instanton technique , which was proposed in @xcite and independently used in @xcite .
alternative approach based on quantum instantons which are defined using an introduced notion of quantum action was suggested in @xcite .
we will investigate a quasienergy spectrum of paradigmatic model for different physical systems , namely double well potential .
we consider this system with perturbation of the kicked type .
one of the most attractive features of kicked systems is the well investigated simple quantum map which stroboscopically evolves the system from kick @xmath0 to kick @xmath1 and greatly facilitates theoretical analysis .
the main idea of this paper is to study the possibility to construct an effective autonomous model for the non - autonomous perturbed system using campbell - baker - hausdorff expansion formula and to test it in numerical calculations of the quasienergy spectrum .
both cat and cdt are connected with the behavior of the quasienergy spectrum ( avoided or exact crossing of the levels ) .
thus the development of the methods for calculation of this spectrum is important for extending one s knowledge in cat and cdt .
we regard the kick - type perturbation which is proportional to @xmath2 .
this perturbation , in contrast to perturbation proportional to @xmath3 , destroys the spacial symmetry in the system which is important for presence of the cat phenomenon @xcite .
the main role in quantum dynamics of our system is played by the classical asymmetry .
there is no chaos induced processes in it but in our future work we will use this approach to system with cat and cdt . in this paper we propose the effective model for the kicked double well system which gives a possibility to simplify the numerical calculations of the quasienergy spectrum and allows to determine both analytically and numerically the quasienergy splitting dependence on both the perturbation strength and frequency .
now lets construct the effective hamiltonian for the double well system with the perturbation of the kick - type .
hamiltonian of the particle in the double - well potential can be written in the following form : @xmath4 where @xmath5 - mass of the particle , @xmath6 - parameters of the potential .
we consider the perturbation of the kick - type which is proportional to @xmath2 @xmath7 where @xmath8 - perturbation strength , @xmath9 - perturbation period , @xmath10 - time .
full hamiltonian of the system is the following : @xmath11 now we will construct an effective hamiltonian for the system under investigation using the following definition : @xmath12 where rhs is a one - period evolution operator .
we restrict our consideration by sufficiently small values of both the perturbation strength and period . using the campbell - baker - hausdorff expansion formula for the kicked dynamical systems we can rewrite the last expression ( [ def : eff ] ) in the following way @xcite : @xmath13 x,\ ] ] where @xmath14 with the definition of @xmath15 formula ( [ he1 ] )
can be expanded in the following form : @xmath16^k x,\ ] ] where the expression under the integral is a map in the power @xmath17 which for sufficiently small values of both the perturbation strength and period can be rewritten as follows : @xmath18^k x = x - \frac{k t}{m } p + o(\epsilon^2 , \epsilon t , t^2).\ ] ] substituting ( [ map ] ) in expression ( [ he2 ] ) one obtains the following form of the effective hamiltonian : @xmath19 the fouth term in rhs of the last expression has the same order in perturbation parameters as first three main terms .
the fifth term is proportional to small parameter , namely perturbation strength .
we restrict our consideration by terms without small parameters and neglect all terms with order higher than zero in the perturbations parameters . as a result we have the following effective hamiltonian for the kicked double well system :
@xmath20 this is the hamiltonian for the asymmetric double well potential without perturbation .
in contrast to the kicked system it is autonomous . in the next section we will consider the properties for this system and construct the effective asymmetric model for the kicked system . in section [ sec : num ] the correspondence between kicked and asymmetric effective double well systems will be tested numerically .
hamiltonian of the classical particle in the asymmetric double - well potential is the following : @xmath21 where @xmath22 and @xmath23 are parameters of the potential , @xmath24 - asymmetric parameter .
the asymmetric potential with parameters @xmath25 @xmath26 @xmath27 and @xmath28 is shown in the figure [ fig : adwp ] ( thick solid line ) .
eight lowest energy levels are shown in the same figure with thin solid lines .
the form of the last term in the rhs of the hamiltonian ( [ def : hamas ] ) is more handy due to special choice of the asymmetric parameter @xmath24 .
parameter @xmath24 for this form of the hamiltonian is equal to shift between bottoms of the wells ( dashed lines in the figure [ fig : adwp ] ) . in order to construct the effective model for the kicked system we should to introduce an effective value of the shift in the kicked system . comparing obtained effective hamiltonian ( [ def : hameff ] ) and hamiltonian for the asymmetric double well system ( [ def : hamas ] ) we define effective asymmetric parameter in the following way : @xmath29
now we can give definition of the proposed effective model : asymmetric double well system with the effective parameter @xmath30 defining by expression ( [ def : sigma ] ) is an effective model for the kicked system with the perturbation parameters @xmath8 and @xmath31 .
the parameters @xmath22 , @xmath23 and @xmath5 are the same for both systems .
the definition of the effective asymmetric parameter ( [ def : sigma ] ) shows that perturbation strength and frequency appears in it as a product @xmath32 .
this is the first advantage of the effective approach .
we have effectively only one asymmetric parameter ( @xmath24 ) instead of two perturbation parameters ( @xmath8 and @xmath31 ) .
the second advantage is the more simple way of the numerical calculations which will be discussed in the next section .
the third advantage of the proposed approach is that splitting for the doublets laying below the barrier hump in the asymmetric double well can be described analytically .
the asymmetric model can be considered as a pair of shifted harmonic oscillators .
the shift is equal to asymmetric parameter @xmath24 .
it is obvious that in this constructed system non - degenerate energy doublets have splitting @xmath33 . in asymmetric system splitting between levels
remains close to @xmath24 for all doublets lying below the barrier top as for case shown in the figure [ fig : adwp ] .
this correspondence can be used in order to give analytical description of the quasienergy spectrum in the kicked system . using expression ( [ def : sigma ] )
we can write down the formula for low laying quasienergy doublets splittings of the time - dependent system ( [ systemhamiltonian ] ) @xmath34 it worth to mention the linear dependence of the levels splitting on both the perturbation strength and frequency . the applicability of this analytical description will be tested in the numerical calculations in the next section .
for the computational purposes it is convenient to choose the eigenvectors of harmonic oscillator as basis vectors . in this
representation matrix elements of the hamiltonian ( [ eq : h ] ) and the perturbation ( [ eq : v ] ) are real and symmetric .
they have the following forms ( @xmath35 ) : @xmath36 \\ & + \delta_{m + 2 \ ; n } \;\frac{g}{2 } \left(g\ , a_0 ( 2 m + 3 ) - a'_2 \right ) \sqrt{(m + 1)(m + 2)}\\ & + \delta_{m + 4 \ ; n } \frac{a_0 g^2}{4 } \sqrt{(m + 1)(m + 2)(m + 3)(m + 4)},\\ & x_{m\ , n } = \delta_{m + 1 \ ; n } \;\sqrt{\frac g2 } \ ; \sqrt{m + 1},\end{aligned}\ ] ] where @xmath37 and @xmath38 , @xmath39 is planck constant which we put equal to @xmath40 , @xmath41 - frequency of harmonic oscillator which is arbitrary , and so may be adjusted to optimize the computation .
we use the value @xmath42 with parameters @xmath25 @xmath26 @xmath43 .
the matrix size is chosen to be equal to @xmath44 .
calculations with larger matrices give the same results .
system of computer algebra mathematica was used for numerical calculations . in order to obtain quasienergy levels ( @xmath45 ) in the kicked double well system
directly we calculate eigenvalues ( @xmath46 ) of the one - period evolution operator @xmath47 and express quasienergy levels through the definition @xmath48 .
then we get ten levels with the lowest one - period average energy which is calculated using the formula @xmath49 ( @xmath50 are the eigenvectors of the one - period evolution operator ) .
the dependence of the quasienergies of these ten levels on the strength of the perturbation is shown in the figure [ fig : qes ] .
quasienergies of the two doublets with the minimal average energy ( thick lines in the figure [ fig : qes ] ) has a linear dependence on the strength of the perturbation in the considered parameter region .
they are strongly influenced by the perturbation while some of the quasienergy states are not . to obtain the same levels in the framework of the effective model approach we have to calculate the eigenvalues of the hamiltonian of the asymmetric double well potential ( [ def : hamas ] ) with the asymmetric parameter defined by the expression ( [ def : sigma ] ) .
then we shift obtained eigenvalues to zone @xmath51 in order to compare results with ones in the kicked system .
the result of calculations on the base of this procedure is shown in the figure [ fig : qes ] by empty squares for two lowest doublets . comparing results of direct and model calculations we make the conclusion that the levels of the doublets laying below the potential hump are correctly described by the effective model .
performed numerical calculations for the kicked and the effective system give the dependence of the ground quasienergy splitting both on the strength ( fig.[fig : de](a ) ) and the frequency ( fig.[fig : de](b ) ) of the perturbation .
filled circles in figures correspond to kicked double well system , empty squares to numerical results obtained in the framework of the effective model .
there is good agreement between splitting s dependencies for these two systems .
they are linear as it was predicted by expression ( [ an - fomula ] ) .
it should be mentioned that all dependencies for the effective model was obtained from one series of the numerical calculations .
we fix model parameters @xmath52 and calculate numerically one set of numerical points for the dependence on the asymmetric parameter . in kicked system we should to perform one series of numerical calculations for every dependency .
this is the first advantage which was discussed in the previous section .
the second advantage which we should discuss after description of the used numerical methods is a more simple algorithm of the calculations . in the kicked system we should to calculate eigenvalues of the matrix exponents
this is more difficult task than in asymmetric model where we calculate the eigenvalues of the system hamiltonian
. analytical result ( [ an - fomula ] ) which was put forward as third advantage of the method is plotted in the figures [ fig : de ] ( a ) and [ fig : de ] ( b ) by straight solid lines .
numerical points lie close to these lines .
the agreement between numerical calculations and analytical expression ( [ an - fomula ] ) is good ( near @xmath53 ) in the parametric region considered .
effective hamiltonian for the kicked double well system was obtained using the campbell - baker - hausdorff expansion formula .
effective autonomous asymmetric model for this system was constructed .
this model is more convenient in numerical calculations than kicked one .
results of numerical calculations show that model correctly describes quasienergy spectrum of the kicked system for low laying levels .
the analytical formula for the ground quasienergy splitting dependence on both the perturbation strength and frequency was obtained in the framework of the effective asymmetric model .
this formula predicts linear dependence of the ground quasienergy splitting on these parameters for the small perturbation strength and period values .
numerical results for the quasienergy splitting as a function of the perturbation frequency and strength demonstrate linear dependence as well .
they are in a good agreement with the formula ( [ an - fomula ] ) .
proposed approach will be used in future for investigation of the cat and cdt phenomena . | connection between chaotic systems quantum processes tunneling developing field research insight in novel phenomena extended last decades intriguing chaos assisted tunneling cat coherent destruction of tunneling cdt ) first enhancement of tunneling in perturbed low - dimensional systems at high field strengths high driving frequencies singlet - doublet crossings when levels regular doublet crossing with chaotic state tunneling between kam - tori into chaotic sea chaotic motion affects tunneling rate easier penetrate into chaotic region travel tunnel onto another kam - torus cdt suppression of tunneling when amplitude frequency driving force to one - dimensional manifold in perturbation parameters space occurs due to crossing of two states different symmetries tunneling tunneling time diverges localization of quantum state on initial torus cat phenomenon cdt observed in physical systems cat observation between modes microwave cavity annular billiard reported same phenomenon for ultracold atoms investigated in study dielectric microcavities evidences for cat both phenomena observed in two coupled optical waveguidesexperimental evidence of control single particle tunneling in double well potential reported in @xcite common methods cat are numerical methods floquet theory scattering approach for billiard systems quantum mechanical amplitudes in complex configuration space analytical approach on instanton technique proposed in @xcite used alternative approach quantum instantons action suggested in @xcite investigate quasienergy spectrum of paradigmatic model for physical systems double well potential consider system with perturbation of kicked type attractive systems quantum map evolves system from to facilitates theoretical analysis idea study construct effective autonomous model for non - autonomous perturbed system using campbell - baker - hausdorff expansion formula test in numerical calculations of quasienergy spectrum cat and cdt connected with behavior quasienergy spectrum development of methods for calculation important for knowledge in cat cdt kick - type perturbation proportional to @xmath2 destroys spacial symmetry important for cat phenomenon main role in quantum dynamics by classical asymmetry no chaos induced processes future work use this approach to system with cat and cdtpropose model kicked double well system calculations quasienergy spectrum determine quasienergy splitting dependence on perturbation strength frequency construct effective hamiltonian double well system perturbation kick - type hamiltonian particle double - well potential @xmath4 - mass particle - parameters potential perturbation kick - type proportional to @xmath2 @xmath7 perturbation strength period time full hamiltonian system @xmath11 construct effective hamiltonian system investigation definition @xmath12 rhs one - period evolution operator restrict consideration small values perturbation strength period campbell - baker - hausdorff expansion formula systems rewrite last expression def : eff ] @xmath13 x, @xmath14 @xmath15 formula he1 expanded @xmath16^k x, expression integral map power @xmath17 small values perturbation strength period rewritten @xmath18^k x = x - \frac{k t}{m } p + o(\epsilon^2 , t t^2). substituting [ map ] ) in expression [ he2 ] obtains effective hamiltonian : @xmath19 fouth term in rhs last expression same order perturbation parameters as first three main termsfifth term proportional to small parameter perturbation strength restrict consideration by terms without small parameters neglect terms order higher than zero parameters effective hamiltonian for kicked double well system : @xmath20 asymmetric double well potential without perturbation contrast kicked system autonomous next section consider properties system construct effective asymmetric model for kicked system section correspondence between kicked asymmetric double well systems tested numerically hamiltonian classical particle in asymmetric double - well potential @xmath21 @xmath22 @xmath23 parameters potential @xmath24 - asymmetric parameter asymmetric potential with parameters @xmath25 @xmath26 @xmath28 shown in figure eight lowest energy levels shown same figure thin solid lines form last term hamiltonian ( def : hamas ] handy due to special choice asymmetric parameter @xmath24 equal to shift between bottoms wells construct effective model for kicked system introduce effective value of shift effective hamiltonian asymmetric double well system define effective asymmetric parameter @xmath29 definition proposed effective model asymmetric double well system with effective parameter @xmath30 effective model for kicked system with perturbation parameters @xmath8 and @xmath31parameters @xmath22 same for both systems definition effective asymmetric parameter def : sigma shows perturbation strength frequency product @xmath32 first advantage of approach one asymmetric parameter @xmath24 ) instead of two second advantage simple numerical calculations discussed next section third advantage splitting for doublets below barrier hump in asymmetric double well described analytically asymmetric model shifted harmonic oscillators shift equal to asymmetric parameter @xmath24 non - degenerate energy doublets have splitting @xmath33 asymmetric system splitting between levels close to @xmath24 for all doublets below barrier top figure correspondence analytical description of quasienergy spectrum in system expression ( [ def : sigma ] formula for low laying quasienergy doublets splittings of time - dependent system @xmath34 linear dependence of levels splitting on perturbation strength frequency applicability analytical description tested in numerical calculations next section computational choose eigenvectors of harmonic oscillator as basis vectors matrix elements of hamiltonian perturbation ( v ) are real symmetricforms @xmath35 @xmath36\delta_{m + 2\frac{g}{2\left(g_0 ( 2 m + 3 ) - a'_2{(m + 1)(m + 2)\delta_{m + 4 {a_0 g^2}{4{(m + 1) + 2) + 3) + 4)} x_{m =\delta_{m + 1 {\frac g2 }{m + 1}{aligned} @xmath37 @xmath38 @xmath39 planck constant equal @xmath40 @xmath41 frequency harmonic oscillator arbitrary adjusted computation value @xmath42 parameters @xmath25 @xmath26 @xmath43 matrix size equal @xmath44 calculations larger matrices results computer algebra mathematica calculations obtain quasienergy levels @xmath45 ) calculate eigenvalues @xmath46 ) - period evolution operator @xmath47 express quasienergy levels definition @xmath48 ten levels lowest average energy calculated formula @xmath49 @xmath50 eigenvectors evolution operator dependence quasienergies strength perturbation shown figurequasienergies two doublets with minimal average energy figure fig qes linear dependence on strength perturbation in parameter region strongly influenced by perturbation some quasienergy states not obtain same levels effective model calculate eigenvalues of hamiltonian asymmetric double well potential def hamas with asymmetric parameter expression def : sigma ] ) shift eigenvalues to zone @xmath51 compare results with kicked system result calculations shown in figure fig qes by empty squares for two lowest doublets levels of doublets below potential hump correctly described by effective model numerical calculations for system give dependence of ground quasienergy splitting on strength and frequency perturbation filled circles figures correspond to kicked double well system empty squares to numerical results effective model good agreement between splitting s dependencies for systems linear predicted by expression an - fomula all dependencies for effective model obtained from one series numerical calculations fix model parameters @xmath52 calculate one set numerical points for dependence on asymmetric parameter in kicked system perform one series numerical calculations for every dependency first advantage discussed second advantage simple algorithm of calculationskicked system calculate eigenvalues matrix exponents difficult than asymmetric model analytical result an - fomula ] third advantage method plotted figures by straight lines numerical points close lines agreement between numerical calculations analytical expression - fomula good near @xmath53 parametric region effective hamiltonian kicked double well system obtained campbell - baker - hausdorff expansion formula effective autonomous asymmetric model constructed convenient numerical calculations kicked calculations describes quasienergy spectrum kicked system low laying levels analytical formula ground quasienergy splitting dependence on perturbation strength frequency obtained effective asymmetric model formula predicts linear dependence ground quasienergy splitting parameters small perturbation strength period values numerical results quasienergy splitting perturbation frequency strength demonstrate linear dependence good agreement with formula an - fomula ] proposed approach investigation cat cdt phenomena | 0.508909 | 3,310 | 1,630 | , Saving $0.1 in GPT-4. |
67,919 | according to a 2007 paper there was no evidence for a neutrino burst of two or more events in super - kamiokande ( sk ) during the entire period of data - taking from 1996 to 2005 from andromeda or anywhere else .
there is , however a scenario under which a detectable signal could have been missed given the search method employed by the analysis , and it would have been found using an alternate method .
the alternate method depends on the hypothesis that two of the neutrino mass eigenstates have masses 4.0 ev and 21.4 ev which was inferred from an analysis of the sn 1987a data .
although one might argue that the hypothesis of such large neutrino masses is remote , there is a way they could be compatible with observed constraints on neutrino masses involving a third tachyonic ( @xmath0 ) eigenstate , plus three sterile neutrinos .
given the importance of a positive supernova search result , and the ease of conducting it using existing sk data , there would seem to be little reason not to do it . | to date the only supernova giving rise to an observed neutrino burst is sn 1987a for which 24 events were seen in 3 detectors in a burst lasting about 15 s @xcite . given the probable significant delay before future mton neutrino detectors are completed , there is considerable interest in maximizing the efficiency for supernova detection using the best detectors currently available , especially the 50 kton super - kamiokande ( sk ) .
two important ways of enhancing the sensitivity involve modifying sk in some way , or alternatively selecting an optimum energy threshold cut @xmath1 to impose on the data that is based on the very different dependencies of the expected signal and background events @xcite .
of course , a supernova from a core collapse within our galaxy would be completely obvious in sk without needing to maximize its detection efficiency , and one well beyond andromeda likely would not yield a single event in current detectors .
this makes the possibility of seeing a supernova burst from andromeda ( expected signal 2 events in sk ) one of great interest .
thus , given the spatial distribution of distances of potential sources , andromeda plays a very special role in extragalactic supernova searches with sk .
in fact , since the expected frequency of supernovae closer than a distance d does not increase significantly between the edge of our galaxy and andromeda , and also between andromeda ( d = 778 kpc ) and the next rich source at a distance of 2.7 mpc , at which point the expected flux would have dropped by a further factor of 10 by virtue of the inverse square law @xcite .
one may therefore conclude that if one should observe genuine signal of 2 - 3 events in sk the source is very likely to be andromeda a fact that could be checked based on directional information on those events .
a 2007 paper analyzed data from sk for the period 1996 - 2005 , and it reported no signal from andromeda or anywhere else @xcite .
however , for the pre-2002 period sk had uncertainties in its energy calibration and it was not used in the 2007 analysis . while the pre-2002 sk could easily show a supernova in our galaxy , it was not sensitive enough to look for a supernova in andromeda . thus , the search for a signal from andromeda , which required making an energy cut on the data in order to enhance the signal to background ratio , was based only on 3 years of data : 2005 - 2007 . as of 2013 there now exists 7 more years of data to examine for sk , so it might be timely to redo the analysis to look for a low event rate signal from andromeda .
however , far more important than the additional 7 years of data since 2005 is the possibility that the earlier analysis left open a way that a detectable signal might have escaped notice .
given that the signal from sn 1987a lasted around 15 seconds , the search reported in 2007 used a sliding 20 second time window , and it required the presence of at least n = 2 candidate events ( a `` cluster '' ) in that window , nearly all of which would be expected to be background . in order to reduce background a threshold energy cut @xmath2 was imposed which drastically reduced background to about one event per day , while simultaneously reducing any real supernova signal by about @xmath3 , based on the expected supernova spectrum thereby reducing the expected counts from a supernova in andromeda from 2.0 to 1.4 @xcite .
the background was reduced further by considering the spatial positions of event vertices within the detector , since background events tend to cluster spatially , while real neutrino events would not .
therefore a cut imposed on the average spacing of the vertices for the clusters in any 20 second time window was found to reduce background almost to zero , while having a negligible effect on eliminating any real n event cluster @xcite .
in fact , the combined cuts on energy and vertex proximity left only 3 remaining candidate @xmath4 event clusters for the entire data - taking period . unfortunately , each of those 3 clusters occurred within 0 to 6 seconds after blastings in the mine that housed sk , and they had to be excluded on that basis , thereby reducing the final number of candidates to zero .
most analyses of sn 1987a assume that the 24 events can only set upper limits on the mass of the electron neutrino @xcite .
however , a recent analysis of sn 1987a neutrinos suggests that the ( anti)electron neutrinos detected consist of two neutrino mass eigenstates having masses @xmath5 and @xmath6 @xcite , using an idea that was explored much earlier@xcite .
although the suggested masses seem preposterously large , the 2012 - 2013 papers suggest a way they could be consistent with much smaller empirical upper limits on the neutrino masses @xcite . for example , the upper limit on the electron neutrino mass , and the sum of the flavor state masses from cosmology can all be addressed if the third mass eigenstate is tachyonic ( @xmath0 ) which would permit all three flavor states to have masses very close to zero .
it is also necessary to have 3 sterile neutrinos nearly degenerate with the three active ones to yield the well established @xmath7 and @xmath8 values .
the claim of 4.0 ev and 21.4 ev mass eigenstates rests on the idea that most of the 24 neutrinos observed were emitted during a time much less than the 15 second burst , possibly in as short a time as 1 second .
core collapse modellers differ on the time distribution in neutrino luminosity , and only some of them suggest a very large fraction in the first second .
for example , while totani et al .
show @xmath9 emitted in the first second and @xmath10 in the first 2 s ( based on a numerical integration of their fig .
1),@xcite other sources suggest significantly less early luminosity @xcite . thus ,
given the uncertainties , one can not rule out the possibility that a large majority of the observed neutrinos from sn 1987a were emitted during an interval perhaps as short as 1 s. if one assumes near - simultaneous neutrino emissions then the neutrino arrival times at the detectors depend primarily on their mass @xmath11 and energy @xmath12 , based on relativistic kinematics according to : @xmath13 where @xmath14 is the time a given neutrino is detected , @xmath15 is the light travel time from the supernova , and @xmath16 is some fixed time for all neutrinos associated with the supernova burst , i.e. , the time @xmath17 signifies when neutrinos of infinite energy e would have arrived .
note that equation 1 implies that on a plot of @xmath18 versus @xmath14 all neutrinos having a fixed mass @xmath11 lie on a straight line of slope @xmath19 .
thus , if neutrinos having two distinct masses were present among the 24 they would all lie on one of two straight lines having slopes @xmath20 and @xmath21 that intersect at a point on the @xmath14 axis .
remarkably , as shown in @xcite , this pattern is exactly what the data show ( within the uncertainties in @xmath18 ) for every one of the 24 observed neutrinos .
finally , we note that the goodness of the two straight line fit ( from which the claim was inferred ) is not significantly impaired if there is a limited spread in emission times . in a monte carlo simulation in which 24 events are generated , assuming an emission time distribution and energy spectrum following totanti et al.@xcite , and assuming the masses of individual neutrinos are either 4.0 ev or 21.4 ev , the agreement of the fit to two straight lines is usually just as good as for the actual data .
finally , it should be noted that the seemingly implausible claim of two such large mass eigenstates has recently received additional support from a sterile neutrino fit to the dark matter halos in the milky way and in galaxy clusters , using that same pair of masses@xcite .
in the @xmath18 versus @xmath14 plane the search window used in the 2007 search is a rectangle of height @xmath22 and width 20 s which slides along the @xmath14 axis .
this search window is appropriate if we expect a neutrino burst from a source in andromeda to have a similar width on arrival to that of sn 1987a , i.e. , around 15 s. however , if the arrival times are primarily due to different travel times for neutrinos having different energies and masses , and if the masses are as large as 4.0 ev and 21.4 ev the arriving pulse will broaden the more remote the source , and _ individual _ neutrino arrival times will correlate with their energy according to eq .
since andromeda is about 14 times as far as sn 1987a , a 15 s burst from andromeda would broaden to about @xmath23 on reaching earth . the alternate search method proposed here , therefore ,
replaces the fixed width rectangular search window by one consisting of two triangular regions see figure 1 which bracket ( within expected neutrino energy uncertainties ) the two dashed lines having slopes corresponding to the masses 4.0 ev and 21.4 ev .
thus , the dashed line down the center of the larger triangle has a slope @xmath19 where @xmath24 is the light travel time from andromeda , and where @xmath25 similarly , the dashed line bisecting the skinnier triangle has a slope given by the same equation but with @xmath26 the sides of the 2 triangles are based on a @xmath27 @xmath28 uncertainty in @xmath12 , and hence a @xmath29 uncertainty in @xmath18 @xcite . as the two - triangle search window slides smoothly along the @xmath14 axis both triangular regions must of course move together to keep their bottom vertices in contact and on the t - axis .
note that unlike the fixed ( 20 s ) width rectangular search window , the width of the two - triangle search window , @xmath30 along the time axis is proportional to @xmath31 so that on a plot like fig .
1 , where the scale of the two axes is @xmath32 and @xmath30 , it is the rectangular search window that appears to change its width for two different values of @xmath33 for example , a choice @xmath34 yields a width @xmath35 whereas the choice @xmath36 yields a width @xmath37 for the two - triangle window .
it is clear that the proposed two - triangle search window might indeed produce candidate events that would be missed by the standard search window especially when a threshold @xmath36 were used instead of @xmath34 and the actual area of the search window becomes nine times larger . as an example , consider the 3 simulated events shown in figure 1 as black circles and assume that their locations in time are appropriate to the choice @xmath38 they clearly are all inside the two - triangle search window with @xmath39 but only one of the 3 would be picked up as a candidate either for a rectangular search window or for a two - triangle search window with @xmath40 hence no @xmath41 event cluster would register in this case using either of these latter two search windows the standard rectangular one or the two - triangle one with too high an energy threshold .
one might argue justifiably that using an energy threshold as low as @xmath36 could generate a number of false signals .
however , while @xmath34 may be the optimum choice in terms of enhancing the signal to background rates , it may not be an optimum choice if very few real events are expected as in the case of a signal from andromeda .
if we expect only two events in sk from a supernova in andromeda , then according to poisson statistics there is a @xmath42 chance of getting 2 events and a @xmath3 chance of getting 3 or more events .
in contrast , if a cut @xmath34 is applied the expected number of events would drop to 1.4 instead of 2.0 , and the previous numbers become : a @xmath43 chance of getting 2 events and a @xmath44 chance of getting 3 or more events .
this means that the chances of finding a very persuasive signal ( @xmath41 events ) essentially is reduced in half by virtue of the @xmath34 cut .
furthermore , given that zero events were seen in the original search , the possibility of a few background events masquerading as a signal is less worrisome when one might easily eliminate them if they do not point back to andromeda .
the authors of the original search did in fact try lowering @xmath1 after finding zero events with the original choice , and they still found no signal . the main argument being advanced in the present paper ,
however , is not the that a signal could have been missed due to too low an @xmath34 cut , but rather due to the rectangular shape of the search window .
the frequency of supernovae in a galaxy is very uncertain , since perhaps as many as @xmath45 are hidden by dust @xcite .
andromeda is a galaxy of comparable size to the milky way , and the expected frequency of supernovae is also comparable perhaps 5 per century , or one every 20 years which suggests there could be a 50% chance that a supernova occurred during observing period without any visible indication of it .
if no signal is found by examining the 2002 - 2012 record for sk using the alternate search method , it could simply mean that there were no supernovae in andromeda during the 10 years since 2002 . if 2 to 3 events are observed in sk
, one would need to see how well they point back to andromeda . at an energy of @xmath46
the directional uncertainty of arriving neutrinos is about @xmath47 @xcite .
a directional uncertainty in the source of about @xmath48 amounts to about @xmath49 of all possible solid angles , so the chances of finding a cluster of n events all pointing back to andromeda within @xmath48 would be @xmath50 which would be highly convincing for @xmath51 were such a result found and the events not picked up within the earlier 20 s wide rectangular search window , it would constitute evidence for a pulse broadening with distance , in conformity with the existence of 4.0 ev and 21.4 ev neutrinos .
hence , the significance of such a positive result would be far greater than being the first extragalactic observation of a supernova burst of neutrinos , since it would also give support to the hypothesis of two identified mass eigenstates , and all that follows from that hypothesis , including the existence of tachyonic neutrinos .
such a hypothetical positive search result would constitute a third piece of evidence in support of the two claimed mass eigenstates in addition to the sn 1987a analysis@xcite and the sterile neutrino dark matter fits.@xcite .
e. n. alekseev , l. n. alekseeva , i. v. krivosheina , v. i. and volchenko , `` detection of the neutrino signal from sn 1987a using the inr baksan underground scintillation telescope , '' eso workshop on the sn 1987a , garching , federal republic of germany , july 6 - 8 , 1987 , proceedings ( a88 - 35301 14 - 90 ) . garching , federal republic of germany , european southern observatory , 237 - 247 ( 1987 ) | only supernova neutrino burst is sn 1987a 24 events seen in 3 detectors burst 15 s probable delay before future mton neutrino detectors interest in maximizing efficiency for supernova detection using best detectors especially 50 kton super - kamiokande ( sk ) enhancing sensitivity involve modifying sk or selecting optimum energy threshold cut signal supernova from core collapse within galaxy obvious in sk without detection efficiency one beyond andromeda yield event in current detectors possibility supernova burst from andromeda ( expected signal 2 events in sk interest andromeda special role in extragalactic supernova searches with sk expected frequency of supernovae closer than distance d increase between edge galaxy and andromeda between andromeda and next source at distance 2.7 mpc expected flux dropped by factor 10 inverse square law if genuine signal of 2 - 3 events in sk source likely to be andromeda directional information 2007 paper analyzed data from sk 1996 - 2005 reported no signal from andromeda or pre-2002 period sk had uncertainties in energy calibration not used in 2007 analysispre-2002 sk could show supernova in our galaxy not sensitive enough in andromeda search for signal from andromeda required energy cut to enhance signal to background ratio based on 3 years data 2005 - 2007 2013 7 more years data for sk might to redo analysis for low event rate signal from andromeda important possibility earlier analysis detectable signal might have escaped notice signal from sn 1987a lasted 15 seconds search in 2007 used sliding 20 second time window required n = 2 candidate events all expected be background to reduce background threshold energy cut @xmath2 imposed reduced background to one event per day real supernova signal by @xmath3 expected counts from supernova in andromeda from 2.0 to 1.4 @xcite background reduced by spatial positions of event vertices detector cut on average spacing vertices for clusters in 20 second window background almost to zero negligible effect on eliminating real n event cluster combined cuts on energy and vertex proximity left 3 remaining candidate @xmath4 event clusters for data period 3 clusters occurred within 0 to 6 seconds after blastings in mine sk excluded final number of candidates to zeroanalyses sn 1987a assume 24 events upper limits mass electron neutrino @xcite recent analysis suggests anti)electron neutrinos two neutrino mass eigenstates masses @xmath5 and @xmath6 @xcite idea explored earlier suggested masses large 2012 - 2013 papers suggest with smaller empirical upper limits neutrino masses @xcite upper limit electron neutrino mass flavor state masses addressed if third mass eigenstate tachyonic ( @xmath0 ) three flavor states masses close to zero necessary 3 sterile neutrinos degenerate with three active ones yield @xmath7 and @xmath8 values claim of 4.0 ev 21.4 ev mass eigenstates most 24 neutrinos emitted less than 15 second burst possibly short 1 second core collapse modellers differ time distribution neutrino luminosity some suggest large fraction first second totani et al . show @xmath9 first second @xmath10 first 2 s other sources suggest less early luminosity majority neutrinos sn 1987a emitted during interval short 1 sassumes near - simultaneous neutrino emissions arrival times at detectors depend on mass @xmath11 energy @xmath12 relativistic kinematics @xmath13 @xmath14 time neutrino detected @xmath15 light travel time from supernova @xmath16 fixed time for neutrinos supernova burst time @xmath17 signifies neutrinos infinite energy arrived equation 1 implies plot @xmath18 versus @xmath14 neutrinos fixed mass @xmath11 lie on straight line of slope @xmath19 if neutrinos two distinct masses present lie on straight lines slopes @xmath20 @xmath21 intersect @xmath14 axis pattern data show for 24 observed neutrinos goodness of two straight line fit not impaired limited spread in emission times monte carlo simulation 24 events generated emission time distribution energy spectrum masses neutrinos 4.0 ev or 21.4 ev agreement of fit to two straight lines usually good actual data implausible claim of two large mass eigenstates support from sterile neutrino fit to dark matter halos in milky way galaxy clusters@xmath18 versus @xmath14 plane search window 2007 search rectangle height @xmath22 width 20 s slides along @xmath14 axis neutrino burst from andromeda similar width arrival sn 1987a around 15 s arrival times due to different travel times neutrinos different energies masses masses large 4.0 21.4 ev arriving pulse remote source neutrino arrival times correlate with energy andromeda 14 times far as sn 1987a 15 s burst to @xmath23 reaching earth alternate search method replaces fixed width rectangular search window by two triangular regions figure 1 lines slopes masses 4.0 21.4 ev line center larger triangle slope @xmath19 @xmath24 light travel time from andromeda @xmath25 line bisecting triangle slope @xmath26 sides triangles based on @xmath27 @xmath28 uncertainty in @xmath12 @xmath29 uncertainty in @xmath18 two - triangle search window slides along @xmath14 axis triangular regions move together keep bottom vertices contact t - axisunlike fixed width rectangular search window width two - triangle search window @xmath30 proportional to @xmath31 on plot like fig . 1 scale @xmath32 rectangular search window width for two values of @xmath33 choice @xmath34 yields width @xmath35 @xmath36 @xmath37 for proposed two - triangle search window might produce candidate events missed by standard search window especially when threshold @xmath36 used instead of @xmath34 area search window becomes nine times larger 3 simulated events in figure 1 black circles locations appropriate to choice @xmath38 all inside two - triangle search window with @xmath39 only one of picked up candidate for rectangular or two - triangle with @xmath40 no @xmath41 event cluster register using search windows with high energy threshold energy threshold low as @xmath36 could generate false signals @xmath34 may be optimum choice signal to background rates not optimum if few real events expected signal from andromeda two events from supernova in andromeda @xmath42 chance 2 events and @xmath3 3 or more eventsif cut @xmath34 applied expected events to 1.4 instead 2.0 previous numbers @xmath43 2 @xmath44 3 or more chances finding persuasive signal ( @xmath41 events reduced half @xmath34 cut zero events seen in original search possibility of background events masquerading as signal less worrisome if point back to andromeda authors lowering @xmath1 after zero events found no signal main argument not signal missed due to low @xmath34 cut but to rectangular shape of search window frequency of supernovae in galaxy uncertain many @xmath45 hidden by dust andromeda comparable size to milky way expected frequency supernovae perhaps 5 per century or one every 20 years 50% chance supernova without if no signal found 2002 - 2012 record sk could no supernovae in andromeda during 10 years since 2002 if 2 to 3 events observed in sk need to see point back to andromeda at energy of @xmath46 directional uncertainty of arriving neutrinos about @xmath47 @xcitedirectional uncertainty source @xmath48 amounts to @xmath49 solid angles chances finding n events pointing back to andromeda within @xmath48 @xmath50 convincing for @xmath51 result found events not 20 s search window evidence for pulse broadening with distance 4.0 and 21.4 ev neutrinos significance positive result greater than first extragalactic observation supernova burst neutrinos hypothesis two mass eigenstates tachyonic neutrinos hypothetical positive search result third evidence two mass eigenstates sn 1987a analysis@xcite sterile neutrino dark matter fits alekseev alekseeva krivosheina volchenko detection neutrino signal from sn 1987a baksan underground scintillation telescope workshop on sn 1987a garching july 6 - 8 1987 proceedings ( a88 - 35301 14 - 90 ) european southern observatory 237 - 247 ( 1987 ) | 0.509661 | 3,392 | 1,788 | , Saving $0.1 in GPT-4. |
67,924 | we employ renormalization group ( rg ) summation techniques to obtain portions of laplace qcd sum rules for scalar gluon currents beyond the order to which they have been explicitly calculated .
the first two of these sum rules are considered in some detail , and it is shown that they have significantly less dependence on the renormalization scale parameter @xmath0 once the rg summation is used to extend the perturbative results . using the sum rules , we then compute the bound on the scalar glueball mass and demonstrate that the 3 and 4-loop perturbative results form lower and upper bounds to their rg summed counterparts .
we further demonstrate improved convergence of the rg summed expressions with respect to perturbative results . 10.0 in 9.0 in -0.60 in key words : sum rules , renormalization group , scale dependence + pacs no .
: 12.38cy | when computing the radiative corrections to physical processes , it is necessary to introduce a scale parameter @xmath0 in order to remove divergences through the renormalization procedure .
exploiting the fact that any explicit dependence on @xmath0 must be cancelled by implicit dependence of physical parameters on @xmath0 , one obtains the renormalization group ( rg ) equation .
this equation has been used to fix portions of radiative effects beyond those determined by direct perturbative calculation .
( for example , see refs . [ 1,2,3,4,5,6 ] . ) in this letter we apply this approach to laplace qcd sum rules for scalar gluonic currents .
we find that within rg summation , the dependence on the scale parameter @xmath0 is significantly diminished ; this is expected as any exact solution of the rg equation would necessarily have no dependence on @xmath0 .
to illlustrate the physical significance of our approach , we do a full qcd sum rule calculation and apply it to determine the mass bound for the scalar glueball .
the rg summation approach provided better results with improved convergence and lesser scheme dependence than those obtained via using purely perturbative inputs .
this indicates that the rg summation approach can potentially be beneficial to other qcd sum rule applications as well .
the scalar gluonic correlation function is expressed as @xmath1 where @xmath2 , @xmath3 and @xmath4 is the qcd @xmath5-function defined for the evolution of the qcd strong coupling constant , @xmath6 .
the perturbative laplace sum rule @xmath7 [ 7 , 8 , 9 , 10 ] is given by @xmath8 where @xmath9 is the inverse square of the borel mass .
the imaginary part of the perturbative scalar gluonic correlator at centre of mass energy @xmath10 , can be extracted from @xmath11 , which has been computed to @xmath12 in [ 11 ] and @xmath13 in the qcd coupling @xmath6 [ 12 ] .
one can extract @xmath14 in the following way using the expression [ 13 ] @xmath15\ ] ] where @xmath16 . in ref .
[ 13 ] the results to order @xmath12 appear ; here we make use of the following results to order @xmath13 with 3 active quark flavours . @xmath17 together eqs .
( 2,3 ) lead to consideration of integrals of the form @xmath18 which satisfy @xmath19 in particular we find that @xmath20\nonumber \\ & + \log^2(\tau \mu^2 ) \left [ 9 - 6\gamma_e \right ] - 2 \log^3(\tau \mu^2)\nonumber\end{aligned}\ ] ] where @xmath21 = euler s constant @xmath22 . together ,
( 4,7a - d ) result in @xmath23 \quad ( l^\prime \equiv \log ( \tau \mu^2))\end{aligned}\ ] ] where @xmath24 in a similar fashion we find that @xmath25 eqs .
( 4,10a - d ) together lead to @xmath26\end{aligned}\ ] ] where @xmath27
we now define @xmath28 so that @xmath29 by eq . ( 2 ) .
so also by eq .
( 2 ) @xmath30 and so by eqs .
( 13 - 15 ) @xmath31 showing that @xmath32 is fixed by @xmath33 . regrouping terms in the sum in eq .
( 13 ) , we can write @xmath34 where @xmath35 where @xmath36 . @xmath37 is the leading - log ( ll ) contribution to @xmath7 , @xmath38 the next - to - leading - log ( nll ) contribution @xmath39 @xmath40 the @xmath41 contribution .
since the explicit and implicit dependence of @xmath7 on the unphysical parameter must cancel , we have the rg equation @xmath42 which by eq .
( 14 ) becomes @xmath43 s^{(k ) } = 0\ ] ] where we have the qcd @xmath5-function @xmath44 where @xmath45 , @xmath46 , @xmath47 and @xmath48 for 3 active flavours .
( note : the anomalous dimension @xmath49 for scalar gluonic currents . ) + order - by - order in powers of @xmath50 , eqs .
( 17,20 ) lead to @xmath51 the boundary conditions for these nested equations are @xmath52 solving eqs .
( 22a - c ) in turn , we obtain @xmath53.\end{aligned}\ ] ] ( for @xmath54 , see the appendix . )
+ to obtain @xmath55 exactly , one needs @xmath56 and @xmath57 , neither of which have been computed as this involves five loop calculations .
however one could in the approximation @xmath58 solve for @xmath59 . using explicit numerical values of the parameters occurring in eqs .
( 24 , a.3 ) we find that with three quark flavours @xmath60 @xmath61.\nonumber\end{aligned}\ ] ] an explicit four loop calculation with three quark flavours and taking @xmath62leads to [ 14 ] @xmath63 where @xmath64 .
+ we plot the purely perturbative @xmath65 and @xmath66 of eqs .
( 8 , 11 ) with the rg improved expressions following from eqs .
( 14,17,26,27 ) in figs . 1 , 2 , 3 and 4 . for parametrizing @xmath67 dependence ,
we define @xmath68 , and plot perturbative and rg - summed expressions for phenomenologically relevant values of @xmath69 0.8 , 1 and 1.2 respectively . in both sum rules
, we note that the rg summed values are remarkably less renormalization scale dependent than the fixed order perturbative results . to demonstrate the usefulness of our approach , we compute the mass of the scalar glueball using both purely perturbative and rg summation results . utilizing a standard qcd sum rule approach as in ref .
[ 10 ] , we incorprate non - perturbative parts which include condensate and instanton contributions to the laplace sum rules . these pieces are combined as follows , @xmath70 where @xmath71 and the second and third term are condensate and instanton contributions respectively .
we use the provided expressions for @xmath72 and @xmath72 in [ 10 ] and use the same set of qcd input parameters .
the sum - rules provide a robust upper bound on the scalar glueball mass @xmath73 @xmath74 in figure 5 , we plot the mass bound computed from both perturbative and rg summed laplace sum rules .
we not only find reduced scale dependence for the rg summed expressions , but also note that the purely 3-loop and 4-loop estimates are upper bounds to the rg summed mass estimates .
this amply demonstrates ( using a full qcd sum rule calculation ) the benefit of using rg - summed expressions , as compared to using the purely perturbative results . +
towards demonstrating the convergence properties , we plot the 3-loop and 4-loop mass estimates separately , both for perturbative and rg - summed results .
figures 6 and 7 indicate better convergence properties of the rg summed results .
finally , we also propose an alternate rearrangement of the sum in eq .
( 13 ) , so that in place of eq .
( 17 ) we have @xmath75 where @xmath76 substitution of eq .
( 30 ) into eq .
( 20 ) shows that the rg equation is satisfied at each order in @xmath77 provided @xmath78 if now @xmath79 b_n^{(k ) } ( x)\ ] ] and @xmath80 then @xmath81 together , eqs .
( 30 - 35 ) show that @xmath82 \exp \left ( -2 \int^x \frac{d\tilde{x}}{\tilde{x}}\right)\nonumber \\ & = a_0^{(k ) } x ( \eta - l).\end{aligned}\ ] ] ( changes in the boundary condition of eq .
( 34 ) can be compensated by changes in @xmath0 in @xmath77 . ) eq .
( 36 ) is not unexpected ; it shows how all log - dependent contributions to @xmath33 are fixed by the rg equation to be given in terms of the log - independent contribution to @xmath33 ( i.e. , @xmath83 ) .
using the four loop @xmath5-function in qcd as well as the four loop contribution to the scalar gluonic correlation function , we have explicitly summed the @xmath84 contribution to the corresponding laplace qcd sum rules . by having incorporated these contributions , the sum rules @xmath65 and @xmath66 have a considerably reduced dependence on the non - physical renormalization scale @xmath0 .
it is also possible to use the rg equation to show how all log - dependent contributions to the laplace sum rules are fixed by the log - independent contributions .
fc would like to thank the support and generous hospitality of the cern theory group , where work on this paper was conducted .
tgs would like to thank natural science and engineering research council ( nserc ) for their support .
r. mcleod provided useful discussions .
the equation for @xmath85 that follows from eq .
( 17,20,21 ) is @xmath86 s_3^{(k ) } - \beta_1 \left [ 4 + u \frac{d}{du}\right ] s_2^{(k ) } - \beta_2 \left [ 3 + u \frac{d}{du}\right ] s_1^{(k ) } \nonumber\ ] ] @xmath87 s_0^{(k)}= 0 .
\eqno(a.1)\ ] ] writing eqs .
( 24a , b ) as @xmath88 it is easily shown that the solution to eq .
( a.1 ) is @xmath89 @xmath90 @xmath91.\nonumber\ ] ] 99 m.r .
ahmady , f.a .
chishtie , v. elias , a.h .
fariborz , n. fattahi , d.g.c .
mckeon , t.n .
sherry and t.g .
steele , _ phys .
rev . _ * d66 * , 014010 ( 2002 ) .
chishtie , t. hanif , j. jia , r.b .
mann , d.g.c .
mckeon , t.n .
sherry and t.g .
steele , _ phys .
_ * d83 * , 105009 ( 2011 ) .
maxwell , _ nucl .
suppl . _ * 86 * , 74 ( 2000 ) .
maxwell and a. mirjalili , _ nucl .
phys . _ * b577 * , 209 ( 2000 ) ; _ nucl .
phys . _ * b611 * , 423 ( 2001 ) . g. abbas , b. ananthanarayan and i. caprini , _ phys . rev . _ * d85 * , 094018 ( 2012 ) .
g. abbas , b. ananthanarayan , i. caprini and j. fischer , _ phys .
rev . _ * d87 * , 014008 ( 2013 ) .
m. a. shifman , a. i. vainshtein and v. i. zakharov , nucl . phys .
b 147 ( 1979 ) 385 .
m. a. shifman , a. i. vainshtein and v. i. zakharov , nucl .
b 147 ( 1979 ) 448 . l. j. reinders , h. rubinstein and s. yazaki , phys .
( 1985 ) 1 .
d. harnett , t.g .
steele and v. elias , _ nucl .
phys . _ * a686 * , 393 ( 2001 ) .
k. g. chetyrkin , b. a. kniehl , and m. steinhauser , _ phys .
_ * 79 * , 353 ( 1997 ) .
baikov and k.g .
chetyrkin , _ phys .
lett . _ * 97 * , 061803 ( 2006 ) .
chishtie , v. elias and t.g .
steele , _ phys .
rev . _ * d59 * , 105013 ( 1999 ) .
chetyrkin , p.a .
kniehl and m. steinhauser , _ nucl .
phys . _ * b510 * , 61 ( 1998 ) . | computing radiative corrections to physical processes introduce scale parameter @xmath0 remove divergences renormalization procedure explicit dependence on cancelled by implicit dependence physical parameters obtains renormalization group rg ) equation used radiative effects beyond direct perturbative calculation apply approach to laplace qcd sum rules for scalar gluonic currents dependence on scale parameter @xmath0 diminished exact solution no dependence on @xmath0 full qcd sum rule calculation determine mass bound for scalar glueball rg summation approach better results improved convergence lesser scheme dependence perturbative inputs beneficial to other qcd sum rule applications scalar gluonic correlation function expressed as @xmath1 @xmath2 qcd @xmath5-function for evolution of qcd strong coupling constant @xmath6 perturbative laplace sum rule @xmath7 given by @xmath8 @xmath9 inverse square of borel mass imaginary part perturbative scalar gluonic correlator centre mass energy @xmath10 extracted from @xmath11 computed to @xmath12 11 @xmath13 in qcd coupling @xmath6 [ 12 ]extract @xmath14 expression [ 13 @xmath15 @xmath16 13 results order @xmath12 appear results order @xmath13 3 quark flavours @xmath17 eqs 2,3 ) integrals @xmath18 satisfy @xmath19 @xmath20\nonumber \log^2(\tau \mu^2 ) 9 - 6\gamma - 2 \log^3(\tau^2)\nonumber\end{aligned} @xmath21 = euler s constant @xmath22 ( 4,7a - d ) result @xmath23 \equiv \log^2)\end{aligned} @xmath24 @xmath25 eqs 4,10a - d ) lead @xmath26\end{aligned @xmath27 define @xmath28 @xmath29 eq 2 eq @xmath30 eqs 13 - 15 @xmath31 @xmath32 fixed by @xmath33 regrouping terms sum 13 write @xmath34 @xmath35 @xmath36 @xmath37 leading - contribution @xmath7 @xmath38 next contribution @xmath39 @xmath40 @xmath41 contribution dependence @xmath7 unphysical parameter rg equation @xmath42 eq 14 ) becomes @xmath43 s^{(k ) } = 0 qcd @xmath5-function @xmath44 @xmath45 3 active flavours anomalous dimension @xmath49 scalar gluonic currents powers @xmath50 eqs lead @xmath51 boundary conditions @xmath52 solving eqs obtain @xmath53.\end{aligned} @xmath54 see appendix obtain @xmath55 needs @xmath56 @xmath57 computed five loop calculations @xmath58 solve @xmath59 numerical values parameters eqs three quark flavours @xmath60 @xmath61.\nonumber\end{aligned} four loop calculation three quark flavours @xmath62leads 14 @xmath63 @xmath64 plot perturbative @xmath65 @xmath66 eqs 11 rg improved expressions eqs 14,17,26,27 ) figs . 1 2 3 4 parametrizing @xmath67 dependence define @xmath68 plot perturbative rg - summed expressions relevant values @xmath69 0.8 1 1.2 rg summed values less renormalization scale dependent fixed order perturbative resultscompute mass scalar glueball using perturbative rg summation results standard qcd sum rule approach incorprate non - perturbative parts condensate instanton contributions laplace sum rules pieces combined @xmath70 @xmath71 second third term condensate instanton contributions use expressions @xmath72 @xmath72 10 same qcd input parameters sum - rules provide upper bound scalar glueball mass @xmath73 @xmath74 figure 5 plot mass bound perturbative rg summed laplace sum rules reduced scale dependence rg summed expressions 3-loop 4-loop estimates upper bounds rg summed mass estimates demonstrates benefit rg - summed expressions perturbative plot 3-loop 4-loop mass estimates separately perturbative rg - summed results figures 6 7 indicate better convergence properties rg summed results propose alternate rearrangement sum in eq . ( 13 ) 17 @xmath75 @xmath76 substitution eq 30 ) into eq ( 20 ) shows rg equation satisfied each order @xmath77 @xmath78 @xmath79 @xmath80 @xmath81 eqs .30 - 35 ) show @xmath82 -2^x\tilde{x}}\nonumber = a_0^{(k ) } x -{aligned changes boundary condition eq 34 ) compensated changes @xmath0 @xmath77 eq 36 ) unexpected shows log - dependent contributions @xmath33 fixed rg equation log - independent contribution @xmath33 @xmath83 four loop @xmath5-function qcd contribution scalar gluonic correlation function summed @xmath84 contribution laplace qcd sum rules incorporated contributions sum rules @xmath65 @xmath66 reduced dependence non - physical renormalization scale @xmath0 use rg equation show log - dependent contributions laplace sum rules fixed log - independent contributions support cern theory group natural science engineering research council ) mcleod provided discussions equation @xmath85 eq 17,20,21 ) @xmath86 s_3^{(k ) } - \beta_1 4 + s_2^{(k ) -_2 3 + u s_1^{(k )\nonumber @xmath87 s_0^{(k)}= 0 (a.1) writing eqs 24a b ) @xmath88 solution to eq a.1 ) @xmath89 @xmath90 @xmath91 99 m ahmady chishtie v elias fariborz fattahi mckeon sherry steele . rev d66 * 014010 ( 2002 ) chishtie hanif j. jia mann mckeon sherry steele d83 * 105009 ( 2011 ) maxwell nucl suppl 86 * 74 ( 2000 ) maxwell mirjalili b577 * 209 ( 2000 ) b611 * 423 ( 2001 ) g. abbas b ananthanarayan i. caprini rev d85 094018 ( 2012 ) abbas ananthanarayan caprini j. fischer phys rev d87 * 014008 ( 2013 ) shifman a i. vainshtein v zakharov phys 147 ( 1979 ) 385 shifman vainshtein zakharov 147 ( 1979 ) 448 l j. reinders hrubinstein yazaki phys . 1985 ) harnett t steele v. elias nucl phys . a686 393 2001 ) chetyrkin b a kniehl m steinhauser . 79 353 1997 ) baikov chetyrkin phys . 97 061803 2006 ) chishtie v. elias steele . rev d59 105013 1999 ) chetyrkin kniehl steinhauser nucl phys . b510 * 61 1998 ) | 0.432913 | 2,929 | 1,581 | , Saving $0.1 in GPT-4. |
67,927 | we consider chains consisting of several identical subsystems weakly coupled by various types of next neighbor interactions . at both
ends the chain is coupled to a respective heat bath with different temperature modeled by a lindblad formalism .
the temperature gradient introduced by this environment is then treated as an external perturbation .
we propose a method to calculate the heat current and the local temperature profile of the resulting stationary state as well as the heat conductivity in such systems .
this method is similar to kubo techniques used e.g. for electrical transport but extended here to the liouville space . , and quantum transport , quantum statistical mechanics , nonequilibrium and irreversible thermodynamics 05.60.gg , 05.30.-d , 05.70.ln | as a specific topic of non - equilibrium thermodynamics , heat conduction has long since been of central interest . instead of reaching a complete equilibrium state , the composite system under some appropriate perturbation enters a local equilibrium state small parts of the system approach equilibrium but not the whole system . within non - equilibrium statistical mechanics
the theory of linear reponse , originally developed to account for electric conductivity , is a very import method to investigate dynamical as well as static properties of materials @xcite . in this context
the famous kubo - formulas @xcite have led to a rapid developement in the theoretical understanding of processes induced by an external perturbation of the system .
however , a direct mapping of these ideas on pure thermal transport phenomena ( perturbations due to thermal gradients @xcite ) faces serious problems : contrary to the case of external perturbations by an electric field , thermal perturbations can not directly be described by a potential term in the hamiltonian of the system .
rather , the thermal perturbation is introduced by heat baths with different temperatures coupled to the system , thus calling for a more detailed description than is needed for electric transport . nevertheless , those methods are often used , eventually because of their immediate success in describing non - equilibrium processes @xcite . recently , the main focus of considerations on heat conduction and fourier s law has shifted towards small ( one dimensional ) quantum systems @xcite .
typically , these systems are chains of identical subsystems weakly coupled by some next neighbor interaction . based on the lindblad formalism @xcite or on techniques of quantum master equations @xcite ,
heat baths are then weakly coupled to the chain at both ends .
it has been found that in such systems the appearance of a normal heat conduction depends on the type of the interaction between these elementary subsystems @xcite .
most quantum mechanical interaction types show a normal heat conduction behavior ( constant non - vanishing local temperature gradients ) , whereas for some special coupling the local gradient within the chain vanishes ( divergence of the conductivity , non - normal scenario ) .
let us now introduce the model system , which we are going to investigate in the following based on the full numerical integration of liouville - von neumann equation as well as on a perturbation theory in liouville space .
the dynamics of the quantum model for heat transport is given by the liouville - von - neumann equation ( lvn ) for open systems @xmath0 thus we have to consider super operators here @xmath1 acting on operators in hilbert space , e.g. the density operator of the system ( see @xcite ) .
the complete liouville operator of the open system under consideration is given by @xmath2 the first term controls the coherent evolution of the quantum system defined by the hamiltonian @xmath3 : it is defined by its action on the density operator @xmath4 according to @xmath5\;,\ ] ] like for closed quantum systems .
the system @xmath3 is here a chain of @xmath6 identical subunits with @xmath7 levels each , coupled weakly by a next neighbor interaction , thus living in a liouville space of dimension @xmath8 .
one could think of several concrete model systems , for example spin models ( @xmath9 ) , for which the hamiltonian would read @xmath10 the first term is the local part of the hamiltonian , whereas the second defines the interaction between the subsystems ( @xmath11 denote the pauli operators of the @xmath12th spin ) .
choosing @xmath13 we get the heisenberg interaction and for @xmath14 , @xmath15 an energy transfer coupling only ( xy model ) .
furthermore , to avoid any bias we will often use a random next neighbor interaction but without disorder ( the same random interaction between different subsystems ) .
the chain is weakly coupled to two heat baths , one at each end of the system , given by the super operators @xmath16 and @xmath17 in ( [ eq:1 ] )
. this bath coupling could be realized by standard lindblad operators @xcite , well known from the theory of open systems in quantum optics . another equivalent possibility is to derive a quantum master equation for the model system leading to a special bath coupling like e.g. in @xcite .
we are interrested in the stationary state of the above described system , namely of the lvn equation ( [ eq:8 ] ) in heat conducting scenario .
this stationary state - a local equilibrium state defines a temperture profile and a heat current . as a measure for the temperature @xmath18 of a single subsystem
we use here the local energy of the respective system so that @xmath19 in units of the local level spacing , c.f .
this should be appropriate for weakly coupled subsystems within the chain ( only a very small amount of energy is within the interaction ) . the operator @xmath20 measures the local energy difference between two adjacent subsystems @xmath12 and @xmath21 ( @xmath22 ) .
the energy current operator @xmath23 can be derived from a discretized version of the continuity equation ( formulating a current into and out of the site @xmath12 , respectively ) @xmath24 = \hat{j}^{(\mu-1,\mu ) } - \hat{j}^{(\mu,\mu+1)}\;.\ ] ] both operators @xmath25 , @xmath23 , will be used now to investigate the stationary state .
beside a complete numerical investigation of the mentioned stationary state , let us try to get the local equilibrium state of the system out of a perturbation theory .
since the system of the last section is living in liouville space we have to work in this respective space , too .
firstly , we consider the unperturbed system as our chain with both bath systems at the same temperature @xmath26 . since the two baths have exactly the same temperature we expect the whole system to settle in a thermal stationary state @xmath27 : this state should support neither a heat current nor temperature gradients it is a global equilibrium state with temperature @xmath28 .
the eigenvalues and eigenvectors of the unperturbed system are given by the eigen equation @xmath29 where the ket - vectors in liouville space have been denoted as @xmath30 .
a scalar product of vectors in liouville space can be defined by @xmath31 the ( unique ) stationary state @xmath27 is also an eigenvector of the system with eigenvalue zero , @xmath32 , whereas all other eigenvalues have a negative real part .
this is due to the fact that asymptotically the system should enter the unique equilibrium state @xmath33 regardless of which state the system was at the beginning .
no other eigenvector is able to contribute to the equilibrium state , i.e. all other eigenvectors must be unstable .
since the liouville operator @xmath34 is not hermitian , the eigenvectors do not form an orthogonal basis , i.e.@xmath35 is in general not the unit operator in liouville space ( c.f .
but with the help of the super operator @xmath36 it is possible to find a dual basis @xmath37 with the property @xmath38 the system will be perturbed now by applying a small temperature gradient @xmath39 .
we are interested in the properties of the stationary _ local _ equilibrium state of the system reached because of this perturbation .
the liouville operator of the perturbation is thus given by @xmath40 the two environment operators are the same as before but now with two different temperatures . in the case of a chain of two level systems
each of the two super operators @xmath41 of the bath coupling at both ends of the system consists of two transition processes ( in case of finite temperatures ) @xmath42 with the two rates @xmath43 and @xmath44 ( @xmath45 is the coupling strength of the environment , @xmath28 its temperature ) and @xmath46 , @xmath47 are transition operators . with this definition the perturbed super operator ( [ eq:7 ] ) can be rewritten as @xmath48 with @xmath49 .
the first two terms are just bath systems at temperature @xmath28 like in the unperturbed case , we therefore neglect these terms in the following . like in standard perturbation theory in hilbert space we can calculate the state correction @xmath50 in first order perturbation theory with respect to the stationary state of the unperturbed system @xmath27 ( keeping in mind that the eigensystem is not orthogonal ) as @xmath51 for a more detailed derivation of this formula see @xcite .
since the stationary density operator of the system is now given by @xmath52 and since we know that @xmath27 does not give rise to any local temperature difference and current , the expectation values of the operators defined above are determined only by @xmath50 from ( [ eq:27 ] ) .
we thus find for the local internal temperature gradient @xmath53 and for the local current within the system @xmath54 the current as well as the local temperature gradient depend linearly on the global temperature difference of the bath systems . under stationary conditions
the current must be independent of @xmath12 , @xmath55 , so that ( [ eq:33 ] ) can be rewritten as @xmath56 eigenstates and eigenvalues entering here the _
conductivity @xmath57 depend only on the mean temperature of the unperturbed system , not on @xmath39 .
based on this @xmath57 as a global property of the system , including its contact properties to the environments , let us call ( [ eq:34 ] ) `` external fourier s law '' . furthermore , combining ( [ eq:23 ] ) and ( [ eq:33 ] )
, we can define a _
conductivity within the system @xmath58 implying also @xmath59 to be independent of the external gradient @xmath39 .
we compare these result with the complete numerical solution of the lvn equation ( [ eq:8 ] ) , here for a heisenberg spin chain with four spins . in fig .
[ fig:1 ] we show the local conductivity of the two central spin systems @xmath60 as a function of the external gradient @xmath39 .
indeed , we find numerically that @xmath60 according to ( [ eq:21 ] ) does not depend on @xmath39 ( dashed line ) .
the exact numerical solution of the lvn equation shows a weak @xmath39 dependence ( solid line ) .
as expected , linear transport and our perturbational theory apply for not too large external gradients @xmath39 only .
note that the full range of temperatures @xmath61 has been mapped here onto the interval @xmath62 $ ] ( @xmath63 is thus already a large gradient ) . in a heisenberg spin chain of 4 spins as a function of the external perturbation @xmath39 ; the solid line refers to the solution of the full lvn equation , the dashed line shows ( [ eq:21]).,width=188 ] we address now some remarkable features of very small heat conducting model systems , based on the numerical solution of the complete lvn equation ( compare @xcite ) .
`` normal '' heat conduction ( fourier s law ) is associated with a constant but non - zero local temperature gradient and thus a finite conductivity @xmath64 independent of site @xmath12 everywhere in the system ( see @xcite ) . in fig .
[ fig:2 ] we show the temperature profile of the heisenberg spin chain for different coupling types within the chain .
the data are obtained from the full solution of ( [ eq:8 ] ) .
the majority of coupling types , the heisenberg coupling and the random next neighbor interaction , indeed show the normal behavior in the weak coupling limit as demonstrated in fig .
[ fig:2 ] ( @xcite ) .
but this `` normal '' transport type does not always show up : a spin - spin interaction consisting of an energy transfer coupling only ( xy model ) leads to a vanishing temperature gradient off the contact regions ( see fig .
[ fig:3 ] ) .
the vanishing gradient in fig .
[ fig:3 ] implies a divergent conductivity within the chain @xmath59 .
nevertheless the current remains finite because of the resistance at the contacts , therefore the global conductivity @xmath65 defined in ( [ eq:34 ] ) also remains finite for this special coupling type .
therefore we could state that the `` external fourier s law '' is valid even if fourier s law proper does not apply . for the current , in case of a normal transport behavior , in dependence of the internal gradient
we find a linear behavior , like proposed by fourier s law shown for three different mean temperatures of the bath systems in fig .
[ fig:4 ] . obviously , the _ local _ conductivity depends on the mean temperature of the bath systems .
furthermore , we show the dependence of the conductivity on mean temperature ( see fig .
[ fig:5 ] ) . by comparing these solutions of the full lvn equation with the results of the perturbation theory we find a very good accordance for not to large gradients , as expected .
the perturbation theory reflects the vanishing gradient in energy exchange coupling as well as the normal linear gradient for heisenberg chains .
also the conductivities are the same as found in the full solution of lvn equation .
we have considered heat conduction in small quantum systems built up from identical subsystems weakly coupled by some next neighbor interaction . by a perturbation theory similar to
that introduced by kubo but extended to the full liouville space of the system , we have been able to derive a quantitative equation for the temperature gradient and heat currents within such systems .
these equations depend only on properties of the unperturbed system and linearly on the strength of the perturbation @xmath39 .
the most remarkable point of the equation for the heat current and the temperature profile is the fact that the global temperature difference of the external bath systems shows up only as a parameter .
this is not only a numerical advantage , since a diagonalization for different global gradients is no longer necessary , but also an interesting physical fact : the heat conductivity is independent of the external gradient @xmath39 , and the `` external fourier s law '' is always fulfilled , even if the internal gradient of the system is not constant , as long as the perturbation theory applies .
our approach does not have the problem of introducing a potential term into the hamiltonian of the system , like in standard kubo formulas for heat conduction .
the bath systems , modeled by a linblad formalism , directly define the perturbation in liouville space . like in standard perturbation theory in hilbert space , the first order correction to the stationary state of the system is expressed in terms of transition matrix elements of the perturbation operator and the eigenstates and eigenvalues of the unperturbed system .
only the non - orthogonality of the eigensystem of the unperturbed system needs a more careful treatment , formally the equations are very similar
. investigations on further aspects of the derived formalism are in progress .
we hope to clarify the question of the different transport behavior ( non - vanishing and vanishing local gradients ) under different coupling types .
we thank m. hartmann , m. henrich , ch .
kostoglou , h. michel , h. schmidt , m. stollsteimer and f. tonner for fruitful discussions . financial support by the deutsche forschungsgesellschaft is gratefully acknowledged . | specific topic non - equilibrium thermodynamics heat conduction central interest complete equilibrium composite system under perturbation enters local equilibrium state small parts approach equilibrium not whole non - equilibrium statistical mechanics theory of linear reponse for electric conductivity import method dynamical static properties of materials kubo - formulas led to rapid theoretical understanding of processes induced by external perturbation direct mapping on thermal transport phenomena perturbations due to thermal gradients faces problems external perturbations electric field thermal perturbations not directly described by potential term in hamiltonian system thermal perturbation introduced by heat baths with different temperatures more detailed description for electric transport methods often used success describing non - equilibrium processes focus on heat conduction fourier s law shifted towards small one dimensional ) quantum systems systems chains of identical subsystems weakly coupled by next neighbor interaction heat baths weakly coupled to chain at both ends normal heat conduction depends on interaction between elementary subsystemsquantum mechanical interaction types show normal heat conduction behavior constant non - vanishing local temperature gradients special coupling local gradient vanishes divergence conductivity non - normal introduce model system based on numerical integration liouville - von neumann equation perturbation theory in liouville space dynamics quantum model for heat transport by liouville - von - neumann equation for open systems @xmath0 consider super operators @xmath1 acting on operators hilbert space density operator complete liouville operator by @xmath2 first controls evolution quantum system defined by hamiltonian @xmath3 defined by action on density operator @xmath4 systems system @xmath3 chain of @xmath6 identical subunits with @xmath7 levels coupled by next neighbor interaction living in liouville space of dimension @xmath8 concrete model systems spin models ( @xmath9 hamiltonian @xmath10 first term local part hamiltonian second defines interaction between subsystems choosing @xmath13 heisenberg interaction @xmath14 @xmath15 energy transfer coupling avoid bias use random next neighbor interaction without disorderchain coupled to two heat baths one each end by super operators @xmath16 @xmath17 bath coupling realized by standard lindblad operators @xcite from theory open systems quantum optics derive quantum master equation for model system to special bath coupling like interrested in stationary state system lvn equation ( [ eq:8 ] ) in heat conducting scenario stationary state local defines temperture profile heat current measure for temperature @xmath18 single subsystem local energy system @xmath19 local level spacing appropriate for weakly coupled subsystems small energy interaction operator @xmath20 measures local energy difference between adjacent subsystems @xmath12 @xmath21 @xmath22 energy current operator @xmath23 derived from discretized continuity equation current into out site @xmath12 @xmath24 = \hat{j^{(\mu-1, - \hat{(\mu,\mu+1)} operators @xmath25 @xmath23 used to investigate stationary state local equilibrium state out of perturbation theory system in liouville space work in consider unperturbed system as chain with both bath systems same temperature @xmath26two baths same temperature expect system in thermal stationary state @xmath27 support heat current temperature gradients global equilibrium state with temperature @xmath28 eigenvalues eigenvectors unperturbed system given by equation @xmath29 ket - vectors liouville space denoted as @xmath30 scalar product of vectors defined by @xmath31 unique stationary state @xmath27 eigenvector eigenvalue zero @xmath32 other eigenvalues negative part due system enter unique equilibrium state @xmath33 no other eigenvector equilibrium other eigenvectors unstable liouville operator @xmath34 not hermitian eigenvectors form orthogonal basis@xmath35 not unit operator liouville super operator @xmath36 possible find dual basis @xmath37 property @xmath38 system perturbed small temperature gradient @xmath39 interested in properties stationary _ local _ equilibrium state reached perturbation liouville operator perturbation given by @xmath40 two environment operators same two different temperaturestwo level systems two super operators @xmath41 bath coupling two transition processes finite temperatures ) @xmath42 two rates @xmath43 and @xmath44 @xmath45 coupling strength environment temperature @xmath46 @xmath47 are transition operators perturbed super operator [ eq:7 ] rewritten as @xmath48 with @xmath49 first two terms bath systems at temperature @xmath28 neglect terms calculate state correction @xmath50 stationary state unperturbed system @xmath27 eigensystem not orthogonal @xmath51 see @xcite stationary density operator by @xmath52 @xmath27 local temperature difference current expectation values operators determined by @xmath50 [ eq:27 ] ) local internal temperature gradient @xmath53 local current system @xmath54 temperature gradient depend on global temperature difference bath systems stationary conditions current independent of @xmath12 , @xmath55 ( [ eq:33 ] ) rewritten as @xmath56 eigenstates eigenvalues conductivity @xmath57 depend on mean temperature unperturbed system not @xmath39 @xmath57 global property system contact properties call ( [ eq:34 ] ) `` external fourier s law ''combining eq:23 eq:33 ] conductivity system @xmath58 @xmath59 independent of external gradient @xmath39 compare with solution lvn equation eq:8 ] for heisenberg spin chain with four spins fig fig:1 local conductivity of central spin systems @xmath60 function of external gradient @xmath39 @xmath60 depend on @xmath39 shows weak @xmath39 dependence linear transport perturbational theory apply for large external gradients @xmath39 full range of temperatures @xmath61 mapped onto interval @xmath62 @xmath63 large gradient heisenberg spin chain of 4 spins function of external perturbation @xmath39 solid line refers to solution full lvn equation dashed line shows [ eq:21]). address features of small heat conducting model systems based on solution lvn equation normal heat conduction with constant non - zero local temperature gradient finite conductivity @xmath64 independent of site @xmath12 fig :2 ] temperature profile of heisenberg spin chain for different coupling types data from full solution of ( [ eq:8 ] ) majority coupling types heisenberg coupling random next neighbor interaction show normal behavior in weak coupling limit in fig fig:2 ] normal transport type always show spin - spin interaction energy transfer coupling leads to vanishing temperature gradient off contact regions fig:3 ] vanishing gradient implies divergent conductivity chain @xmath59 current remains finite resistance at contacts global conductivity @xmath65 remains finite for special coupling type external fourier s law '' valid even if current normal transport behavior linear behavior fourier s law for three mean temperatures bath systems in fig:4 ] local conductivity depends on mean temperature bath systems dependence conductivity on mean temperature fig:5 ] comparing solutions full lvn equation with perturbation theory good accordance for not large gradients perturbation theory reflects vanishing gradient in energy exchange coupling normal linear gradient for heisenberg chains conductivities same as in full solution lvn equation considered heat conduction in small quantum systems from identical subsystems weakly coupled by next neighbor interactionperturbation theory similar to kubo extended to full liouville space quantitative equation for temperature gradient heat currents equations depend on properties unperturbed system strength perturbation @xmath39 global temperature difference of external bath systems shows as parameter numerical advantage diagonalization necessary physical heat conductivity independent of external gradient @xmath39 external fourier s law '' fulfilled even if internal gradient not constant approach introducing potential term into hamiltonian system standard kubo formulas bath systems modeled by linblad formalism define perturbation in liouville space first order correction to stationary state expressed in transition matrix elements of perturbation operator eigenstates eigenvalues of unperturbed system non - orthogonality of eigensystem unperturbed system needs careful treatment equations similar investigations on further aspects formalism in progress hope to clarify different transport behavior non - vanishing vanishing local gradients under different coupling types thank m. hartmann m. henrich kostoglou h. michel h. schmidt m. stollsteimer f. tonner for discussions financial support by deutsche forschungsgesellschaft acknowledged | 0.509393 | 3,349 | 1,703 | , Saving $0.1 in GPT-4. |
67,930 | we searched for @xmath0 6(1,6)-5(2,3 ) maser emission at 22.235 ghz from several saturnian satellites with the nobeyama 45 m radio telescope in may 2009 .
observations were made for titan , hyperion , enceladus and atlas , for which pogrebenko _ et al . _
( 2009 ) had reported detections of water masers at 22.235 ghz , and in addition for iapetus and other inner satellites .
we detected no emission of the water maser line for all the satellites observed , although sensitivities of our observations were comparable or even better than those of pogrebenko _
et al._. we infer that the water maser emission from the saturnian system is extremely weak , or sporadic in nature .
monitoring over a long period and obtaining statistical results must be made for the further understanding of the water maser emission in the saturnian system . | maser emissions are widely found in celestial objects such as dense cores of molecular clouds and circumstellar envelopes of late - type stars @xcite .
masers have been used as probes of gas with the h@xmath1 number density of typically @xmath2@xmath3 @xmath4 . for solar system objects , several maser and laser phenomena
have been found ; @xmath5 , @xmath6 ( venus and mars : mumma 1992 ) and @xmath7 ( for many comets : _
e.g. _ crovisier _ et al .
each phenomenon would be induced by different physical processes . while a thermal 22.235 ghz water line was possibly detected for comet hale - bopp ( bird _ et al .
_ 1997 ) , the first detection of @xmath0 maser in the solar system was reported at the catastrophic impact of comet shoemaker - levy9 and jupiter ( cosmovici _ et al .
this report suggests that such an incident can induce collisional pumping for water masers .
recently , pogrebenko _ et al . _
( 2009 ) ( abbreviated as pog hereafter ) have reported the detections of @xmath0 masers from the saturnian satellites ( titan , hyperion , enceladus and atlas ) with the medicina 32 m and mets@xmath8hovi 14 m telescopes .
this is interesting because , unlike a temporal phenomenon such as the break - up and disruption of a comet , we can perform long period monitoring of @xmath0 emission using ground - based telescopes , space telescopes ( _ e.g. _ herschel space telescope ) and spacecrafts .
so far , we do not have much knowledge about the maser mechanism in the solar system , although a lot of water maser phenomena are observationally studied for extra - solar objects .
the combination of ground , space and in - situ observations would contribute to understand the nature of water maser emission if the presence in the saturnian system is verified .
therefore , we must accumulate more data of the water maser lines in the saturnian system . in this letter
, we report our trial of detecting water maser satellites with the 45 m radio telescope at nobeyama radio observatory ( nro ) .
we observed the major saturnian satellites for which pog reported the detections , and in addition , we observed a few other inner satellites .
we observed the saturnian satellites with the nro 45 m telescope at the water maser frequency 22.23508 ghz on 15 - 17 , 24 , 27 , 28 may in 2009 .
we used a cooled hemt amplifier as the receiver front ends and an acousto - optical spectrometer ( aos - h ) as back ends .
the total band width of an acousto - optical spectrometer aos - h was 40 mhz , and frequency resolution was 37 khz , which corresponds to velocity resolution of @xmath9 km s@xmath10 .
all the observations were made using the position switching method .
we pointed the telescope toward the center of titan and saturn .
a typical integration time of one scan was 20 sec .
as the half - power beam width ( hpbw ) of the telescope ( @xmath11 at 23 ghz ) was larger than the angular diameter of the saturn ( @xmath12 including the ring ) , we observed several satellites simultaneously .
for example , we could observe hyperion when we observed titan on 16 , 17 , 18 and 28 may , and enceladus and atlas when we observed the saturn throughout the observational days , though a slight offset of pointing resulted in the worse detection upper limit .
furthermore , we select iapetus as the off points of the position switching .
we segregated these satellites using the differences of doppler shift frequency in the data analysis .
telescope pointing was checked with nearby strong h@xmath1o maser stars .
the antenna temperature , @xmath13 , was obtained by the chopper wheel method correction for atmospheric and antenna ohmic losses .
a typical range of the system temperature was 120300k , which depended on weather condition and airmass at the observations .
table 1 gives a summary of the observations .
we performed base line fitting with the 3rd order polynomial to eliminate the continuum emission from the sky and saturn .
then , we co - added and binned these spectra with taking the doppler shift for each object into consideration .
radial velocities of the saturnian satellites were calculated using the jpl horizons on - line ephemeris system ( giorgini _ et al .
the resultant velocity resolution was about @xmath14 km s@xmath10 .
these processes were done on the software package newstar , and procedures to obtain the final spectra after the doppler correction were carried out on the software which had been developed for the solar system objects .
figure 1 shows the obtained spectra ( upper two and lower - left panels ) and the illustration of the satellite positions ( lower - left panel ) at the time of observations for titan , hyperion and iapetus .
figure 2 also shows the spectra ( upper panels ) and illustrations of positions ( lower panels ) for enceladus and atlas .
the position of a satellite @xmath15 is defined as an angle between the line of sight towards saturn and the line of a satellite and saturn ( @xmath16 @xmath17 when the object transits on the saturn seen from the earth ) .
table 2 shows a typical 1@xmath18 upper limit of the signal considering the factor of a satellite position inside the telescope beam .
the conversion efficiency of flux density to antenna temperature ( @xmath13 ) , 2.8 jy @xmath19 , was used for the calculations .
the observational results show the data combined on each day and the entire days during the observation period .
pog reported that the water line detections of titan and hyperion were @xmath2030 mk ( @xmath21 for 8 hour integration ) and @xmath2050 mk ( @xmath22 for 6 hours integration ) , respectively , which correspond to 300 mjy and 500 mjy , respectively .
they found that these emissions were seen on almost the same satellite positions for both objects , and based on these results , they inferred a common mechanism of the emission such as saturnian magnetosphere bow shock . on 16 - 18 and 28 may 2009 , titan and hyperion were near conjunction so that we could observe them simultaneously in the telescope beam . during 16 - 18
may , especially , these objects were at almost the same position .
therefore , if the maser emissions had been caused by the common mechanism , we would have detected them simultaneously for these two objects .
however , our observations this time could not detect the signals of the emissions , and all the data combined also showed no symptom .
we estimated typical values of the @xmath23 upper limit for each daily data .
we also obtained @xmath23 upper limits during the observations using all the daily data combined .
the typical(all ) @xmath23 upper limits were 200(60 ) mjy for titan and 240(80 ) mjy for hyperion , which means that we could certainly detect the line emissions if the levels of flux densities were as high as those of pog .
although pog did not report the water maser emission on iapetus , we monitored this satellite during our observation period aiming this object at the off - point of the position switching observations .
we could not find any signal stronger than @xmath20210(75 ) mjy ( @xmath23 upper limit ) .
the @xmath0 ice and vapor plume on enceladus has been reported by the cassini ultraviolet imaging spectrometer ( hansen _ et al .
_ 2006 and hansen _ et al .
_ 2008 ) , and the hypothesis that there exists liquid water in the crust has been proposed ( porco _ et al .
_ 2006 ) . as for the water maser emission ,
the mass ratio of water vapor to ice in the plume is important because a majority of water molecules must be in the form of vapor in order to have maser emission originated from the enceladus plume .
a recent theoretical study indicates that the plume would be dominated by vapor from the thermodynamics perspective ( kieffer _ et al .
the maser line intensity reported by pog was 500 mjy ( 4.2@xmath18 ) , and they estimated the column density of water vapor from the observed maser intensity , which agrees with the column density of the water vapor plume observed in uv ( @xmath24 : hansen _ et al .
nevertheless , we did not find appreciable maser emission in our data .
the @xmath23 upper limits were 780(540 ) mjy .
our data were worse compared with those of pog because enceladus was located around the edge of the telescope beam much of the observational time .
however , these results might indicate that the flux of the plume is varying , or the plume is sporadic ; like a geyser .
pog reported the most certain detections for atlas .
the averaged spectrum showed a peak with 32 mk and s / n=7.0 . from the satellite positions
, they found that the maser emissions occurred on the trailing side which was several thousand km away from atlas , and suggested that the disturbance of the atlas s motion had caused the emission in the edge regions of saturnian rings a and f. we attempted to verify this subject in our data .
we divided the positions of atlas into 3 parts ; @xmath16 , position 1 and 3 : positions passing before the geometrical edges of the saturn s rings seen from the earth , and position 2 : those after the rings ( see the positions of atlas in figure 2 ) . for each position
, we combined the data , and furthermore , we tested the data on 19 may because atlas passed by the geometrical edges of the rings ( @xmath25 ) during the observations .
however , we could not have positive results for each case .
all the combined data did not show any prominent feature of the emission .
the 3@xmath18 detection limits for each data set were in the range of 430520(400 ) mjy .
we also checked several inner satellites which have a diameter larger than @xmath2010 km ; mimas , janus , epimeteus , prometheus , pandora and pan .
for all the satellites , the acquired data did not show indicative of the maser emission .
the typical value of @xmath23 upper limits were 500(210 ) mjy .
we had an opportunity to observe the saturnian water maser line at the time of the ring s disappearance ; that is , we see the saturn ring almost in the edge - on view , and we obtained the data which had comparable or even better sensitivities than those by pog for almost of the satellites , though our observations were made for a limited period in 2009 may . if maser emission is stationary in intensity at the levels as those reported by pog , we could detect them .
however the results were negative for all of the observed saturnian satellites . from these results ,
we conclude that the water maser in the saturnian system may be sporadic in nature and it is strongly restricted to the time and position of satellites .
we have to monitor the satellites for longer periods and to obtain statistical results .
these studies would be useful to figure out the water maser emission reported in the saturnian system .
in addition , we should also perform monitoring observations for other icy bodies ; jovian satellites , comets , outer asteroids and kuiper belt objects would be inside the scope . as suggested by cosmovici _
( 1996 ) , the maser emission may be induced by catastrophic events .
such events like a disruption or eruption can occasionally be found among the solar system objects ; ( @xmath5 c/1999 s4 linear : disruption and dissipation , 29p / schwassmann - wachmann : outburst , 7968 elst - pizarro : impact or cometary activity ( toth 2000 and hsieh et al . 2004 ) , etc . ) .
we should not miss events which will occasionally happen and carry out the observations to accumulate the data .
* a search for water maser emission ( 22.22351 ghz ) from several saturnian satellites ; titan , hyperion , enceladus atlas , iapetus , and other inner satellites with a diameter larger than 10 km ( mimas , janus , epimeteus , prometheus , pandora and pan ) were carried out with the 45 m radio telescope at nobeyama radio observatory .
* we could not confirm any emission line for all the satellites .
the typical @xmath23 upper limits of daily(average ) data for titan , hyperion , enceladus and atlas were 200(60 ) , 240(80 ) 780(540 ) and 430 - 520(400 ) mjy , respectively , and for the other inner satellites , the s / n obtained were almost the same . * the sensitivities were comparable or even better than those of pog for most of the satellites . * from our observations
, we infer that water maser emission in the saturnian system is sporadic in nature .
we are grateful to dr .
j. crovisier , the referee of this paper for suggestions to improve the manuscript , and s. takano and j. maekawa for advice and discussions .
this research made use of ephemeris given by nasa jet propulsion laboratory , california institute of technology . | maser emissions found in celestial objects dense cores molecular clouds circumstellar envelopes late - type stars masers used as probes gas@xmath1 density @xmath2@xmath3 @xmath4 solar system objects several maser laser phenomena found @xmath5 @xmath6 @xmath7 comets induced by different physical processes thermal 22.235 ghz water line detected for comet hale - bopp first detection of @xmath0 maser reported at catastrophic impact of comet shoemaker - levy9 jupiter suggests incident can induce collisional pumping for water masers pogrebenko reported detections @xmath0 masers from saturnian satellites with medicina 32 m mets@xmath8hovi 14 m telescopes long period monitoring of @xmath0 emission using ground - based telescopes space telescopes spacecrafts much knowledge about maser mechanism in solar system water maser phenomena studied for extra - solar objects combination ground space in - situ observations contribute to understand nature water maser emission if presence in saturnian system verifiedaccumulate more data water maser lines saturnian system report trial detecting water maser satellites with 45 m radio telescope at nobeyama radio observatory observed major saturnian satellites pog other inner satellites observed saturnian satellites 45 m telescope at water maser frequency 22.23508 ghz 15 - 17 24 27 28 may 2009 used cooled hemt amplifier acousto - optical spectrometer back total band width 40 mhz frequency resolution 37 khz velocity resolution @xmath9 km s@xmath10 observations using position switching method pointed telescope toward center titan and saturn typical integration time scan 20 sec half - power beam width telescope @xmath11 23 ghz larger than angular diameter saturn @xmath12 observed several satellites simultaneously hyperion enceladus atlas offset pointing detection iapetus as off points position switching segregated satellites using differences doppler shift frequency telescope pointing checked with nearby strong h@xmath1o maser stars antenna temperature , @xmath13 obtained by chopper wheel method correction for atmospheric antenna ohmic lossestypical system temperature was 120300k depended on weather airmass at observations table 1 summary observations performed base line fitting with 3rd order polynomial to eliminate continuum emission from sky saturn co - added binned spectra doppler shift for each object radial velocities saturnian satellites calculated using jpl horizons on - line ephemeris system resultant velocity resolution about @xmath14 km s@xmath10 processes on software newstar final spectra after doppler correction on software for solar system objects figure 1 shows obtained spectra satellite positions for titan hyperion iapetus figure 2 spectra for enceladus atlas position of satellite @xmath15 defined as angle between line of sight towards saturn table 2 shows typical 1@xmath18 upper limit signal satellite position telescope beam conversion efficiency of flux density to antenna temperature ( @xmath13 2.8 jy @xmath19 used for calculations observational results show data combined on each day days during observation period water line detections of titan hyperion were @xmath2030 mk mk correspond to 300 mjy and 500 mjyfound emissions seen same satellite positions for both objects inferred common mechanism saturnian magnetosphere bow shock 16 - 18 28 may 2009 titan hyperion near conjunction simultaneously telescope 16 - 18 may objects same position if maser emissions caused by common mechanism detected simultaneously objects observations detect signals emissions data combined showed no symptom estimated typical values @xmath23 upper limit for each daily data obtained upper limits data typical upper limits 200(60 ) mjy for titan 240(80 ) mjy for hyperion could detect line emissions if levels flux densities high as pog pog report water maser emission on iapetus monitored satellite aiming object off - point position switching observations signal stronger than @xmath20210(75 ) mjy ( @xmath23 upper limit ) @xmath0 ice and vapor plume on enceladus reported by cassini ultraviolet imaging spectrometer hypothesis liquid water in crust proposed mass ratio of water vapor to ice in plume important majority of water molecules must form vapor maser emission enceladus plumestudy indicates plume dominated by vapor kieffer maser line intensity reported by pog was 500 mjy ( 4.2@xmath18 ) estimated column density of water vapor from agrees with water vapor plume in uv @xmath24 hansen _ al appreciable maser emission in our data @xmath23 upper limits were 780(540 ) mjy data worse compared with pog because enceladus located around edge telescope beam results might indicate flux plume varying or sporadic pog reported certain detections for atlas averaged spectrum showed peak with 32 mk s / n=7.0 maser emissions occurred on trailing side several thousand km away from atlas disturbance of atlas s motion caused emission in edge regions of saturnian rings a and f attempted to verify in data divided positions of atlas into 3 parts 1 before 2 after rings figure 2 combined data tested data on 19 may because atlas passed by edges rings positive results for each case combined data show prominent feature emission 3@xmath18 detection limits for 430520(400 ) mjy .checked inner satellites diameter larger than @xmath2010 km ; mimas janus epimeteus prometheus pandora pan data show maser emission typical value of @xmath23 upper limits were 500(210 ) mjy saturnian water maser line at ring s disappearance saturn ring in edge - on view obtained data comparable or better sensitivities than pog for satellites observations limited period in 2009 may if maser emission stationary at levels pog could detect results negative for all observed saturnian satellites water maser in saturnian system sporadic restricted to time position satellites monitor satellites for longer periods obtain statistical results studies useful figure out water maser emission perform monitoring observations for other icy bodies ; jovian satellites comets outer asteroids kuiper belt objects maser emission induced by catastrophic events events like disruption eruption among solar system objects @xmath5 c/1999 s4 linear disruption 29p / schwassmann - wachmann outburst 7968 elst - pizarro impact cometary activity miss events carry out observations to accumulate datasearch for water maser emission ( 22.22351 ghz from saturnian satellites titan hyperion enceladus atlas iapetus diameter than 10 km with 45 m radio telescope at nobeyama radio observatory confirm emission line all satellites typical upper limits daily data for titan hyperion enceladus atlas 200(60 ) 240(80 ) 780(540 ) 430 - 520(400 ) mjy other satellites almost same comparable or better than pog for most satellites water maser emission in saturnian system sporadic grateful to dr . j. crovisier referee s. takano j. maekawa for advice research ephemeris nasa jet propulsion laboratory california institute of technology | 0.440191 | 2,929 | 1,470 | , Saving $0.1 in GPT-4. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.