ebook img

Inevitable Power-law Behavior of Isolated Many-Body Quantum Systems and How It Anticipates Thermalization PDF

0.42 MB·
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Inevitable Power-law Behavior of Isolated Many-Body Quantum Systems and How It Anticipates Thermalization

Inevitablepower-law behaviorofisolatedmany-body quantum systemsand how itanticipates thermalization Marco Ta´vora1, E. J. Torres-Herrera2, and Lea F. Santos1 1Department of Physics, Yeshiva University, New York, New York 10016, USA and 2Instituto de F´ısica, Universidad Auto´noma de Puebla, Apartado Postal J-48, Puebla, Puebla, 72570, Mexico (Dated:November3,2016) 6 1 Despite being ubiquitous, out-of-equilibrium quantum systems are much less understood than systems at 0 equilibrium. Progressinthefieldhasbenefitedfromasymbioticrelationshipbetweentheoreticalstudiesand 2 newexperimentsoncoherentdynamics. Thepresentworkstrengthensthisconnectionbyprovidingageneral pictureof the relaxation process of isolatedlatticemany-body quantum systemsthat are routinelystudied in v experimentswithcoldatoms,ionstraps,andnuclearmagneticresonance.Weshownumericallyandanalytically o N thatthelong-timedecayoftheprobabilityforfindingthesysteminitsinitialstatenecessarilyshowsapower-law behavior∝t−γ. Thishappensindependentlyofthedetailsofthesystem,suchasintegrability,levelrepulsion, 2 andthepresenceorabsenceofdisorder. Informationaboutthespectrum,thestructureoftheinitialstate,and thenumberofparticlesthatinteractsimultaneouslyiscontainedinthevalueofγ. Fromit,wecananticipate ] whethertheinitialstatewillorwillnotthermalize. h c e Introduction. A great deal of effort has recently been Ψ(0) with the eigenstates ψ of H, and ρ (E) = m | i | αi 0 putinto improvingourunderstandingof isolatedmany-body C(0) 2δ(E E ) is the energy distribution of Ψ(0) t- quantum systems quenched far from equilibrium. This is in Pweαig|hteαd|bythe−compαonents C(0) 2,theso-calledloca|ldeni- a partmotivatedbythepossibilityofinvestigatingthecoherent | α | sityofstates(LDOS).Thesurvivalprobabilityistheabsolute t s evolution of these systems for long times with different ex- squareoftheFouriertransformoftheLDOS.Allinformation . perimental setups, including those with ultracold atoms [1], t abouttheevolutionofF(t)iscontainedinρ (E). a trapped ions [2, 3], and nuclear magnetic resonance [4, 5]. 0 m We verified that the initial decay of the survival probabil- Alignedwith these efforts, thiswork characterizesand justi- ity is dissociated from the regime (integrable or chaotic) of - fiesthedynamicalbehavioratdifferenttimescalesofexperi- d the Hamiltonian [18–23], but dependson the strength of the mentallyaccessibleintegrableandchaoticlatticemany-body n perturbation. We now show that at long times, regardlessof quantumsystemswithandwithoutdisorder. Fromthisanal- o howfasttheinitialevolutionmaybe,thedynamicsnecessar- c ysis,anewcriterion,basedexclusivelyondynamics,isintro- ily slows down and becomes power-law, F(t) t γ. The [ ducedforidentifyingwhichsystemscanthermalize. ∝ − characterizationofthelong-timedynamicsanditsconnection Thesurvivalprobability(probabilityforfindingthesystem 4 with the viability of thermalization are the central topics of v initsinitialstateattimet)andtheLoschmidtecho(measure thiswork. 7 oftherevivaloftheinitialstateafteratime-reversaloperation) We show that in realistic lattice many-bodyquantum sys- 0 have been extensively considered in the analysis of out-of- 8 equilibriumquantumsystems [6–13]. Severalworkstried to tems with two-body interactions, 0 γ 2. The value of ≤ ≤ 5 establishacorrespondencebetweentheinitialexponentialor the power-law exponentindicates the level of delocalization 0 of the initial state in the energy eigenbasis. When the ini- Gaussian decayswith quantumchaos[12–17] andothersfo- 1. cusedontheonsetofpower-lawdecaysatlongtimes[6–11]. tialstate samplesonlya portionof the Hilbertspace and the 0 In the case of continuous models, the algebraic behavior of LDOS is sparse, γ < 1 and thermalization is not expected. 6 thesurvivalprobabilityhasbeenassociatedwiththepresence Whentheinitialstate ischaotic, sothatitscomponentsCα(0) 1 ofboundsinthespectrum[6–8]whileindisorderednoninter- are uncorrelatedand spread over its entire energyshell [24– : v actingsystemsatthemetal-insulatortransition,thepower-law 27],thermalizationshouldoccur[27–33]. Inparticular,when i exponenthasbeenrelatedwithfractaldimensions[9–11].Ex- theLDOSisergodicallyfilled,thenγ =2.Fromthevaluesof X changesbetweenthesedifferentcommunitieshavebeenvery γ,onecanthusanticipatewhethertheinitialstatewillorwill r a limited. Here,weunifythesemultipleperspectivesintoasin- notthermalize. We also discuss the non-realisticscenarioof gleframeworkanduseittodescribetheevolutionofthesur- fullrandommatrices,wherethepower-lawexponentreaches vivalprobabilityoflatticemany-bodyquantumsystems. theupperboundγ =3. The survival probability (or fidelity) of the initial state is Time scales. The system is initially prepared in an eigen- definedas state of the unperturbed Hamiltonian H , which is abruptly 0 F(t) ≡ hΨ(0)|e−iHt|Ψ(0)i 2 qoufetnhcehepdertiunrtboatHion=V.HA0t+vegrVy,shwohrterteimgesi,sththeedsetcreanygothf (cid:12)(cid:12) 2(cid:12)(cid:12) 2 the survival probability is quadratic [34], as derived from wHh,eCreα(0E)α==ar(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)eψXααthΨe|C(eα0(i0)g)e|2naevr−eailEutheαest(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)oovf=etrhl(cid:12)(cid:12)(cid:12)(cid:12)aZepdssyEosefte−mtihEetHρina0mi(tEiial)lto(cid:12)(cid:12)(cid:12)(cid:12)sntia(a1tne) [EtiPhn0eitαi=ael|xCspPtα(aa0ntα)es|.|i2Co(nEα(0αF)|2−(Et Eα≪0=)2]hσ1Ψ0/−2(10)i)s|H≈th|eΨ1w(0−id)ithσis02ottf2h,ethweenhLeerDrgeOyσSo0fatnh=de h | i 2 Aftertheinitialquadraticbehavior,whetherF(t)switches ables and IPR 3/ [28]. Even though realistic chaotic 0 ∼ D or not to an exponential decay depends on the strength g of many-bodyquantumsystems with two-bodyinteractionsare theperturbation. TheexponentialdecayisvalidintheFermi not described by FRMs, because their Hamiltonian matrices goldenruleregime,wherethetypicalmatrixelementsofgV arebanded,sparse,andrandomelementsmaynotevenexist, are larger than the mean level spacing and the LDOS has a theystillfollowrandommatrixstatisticsandtheirbulkeigen- Lorentzianform.However,forverystrongperturbations,g statesareclosetorandomvectors[28,49,50]. Afterastrong → 1,theLDOSisbroader.Inmany-bodyquantumsystemswith perturbationinto such Hamiltonians, initial states with ener- two-body interactions, where the density of states is Gaus- giesawayfromtheedgesofthespectrumalsogiveveryfilled sian [35–37], the limiting shape of the LDOS is also Gaus- LDOS[24–32]. sian, resulting in the Gaussian decay F(t) = exp( σ2t2) Case 1(i) holds for realistic chaotic many-body quantum − 0 [14, 15, 18–23, 38]. Exponential and Gaussian decays can systems,wheretheLDOSisGaussian,whichleadstoγ =2. thusoccurinbothintegrableandchaoticmodels[18–23].The ForFRM,theLDOSisasemicircle[18],socase1(ii)applies picturebecomesmoresubtleatlongtimes,wherethepower- andγ =3. lawbehavior t γemergesandthefillingoftheLDOSplays Incase2,thepower-lawexponentisobtainedfromthecor- − ∝ akeyrole. relationfunctionC(ω) C(0) 2 C(0) 2δ(E E ω) Causes of the power-law decay. We discuss two distinc- ≡ α,β| β | | α | α− β− presentin F(t) = ∞ dωPeiωtC(ω). A power-lawdecayof tive causes for the long-time algebraic decay of the survival C(ω 0) ωγ R1−,∞with γ < 1, leads to F(t) t γ [9– probability. → ∝ − ∝ − 11, 51–53]. The more correlated the componentsof Ψ(0) , Case1 is related to the unavoidable presence of a lower | i the smaller the exponent γ. This exponent coincides with boundE intheenergyspectrumofanyrealquantumsys- low thefractaldimensionφobtainedfromthescalinganalysisof tem. This point was put forward already in 1958 [39] and IPR φ. ThisrelationwasfoundinstudiesofAnderson in several other early works[40–45]. At long times, the en- 0 ∝D− localization [9–11] and of many-body localization [22, 23]. ergy bound leads to the partial reconstruction of the initial We show thatit holdsalso innoninteractingintegrablemod- state.Thisresultsinthepower-lawdecayofcontinuousmany- els. particle models [6–8] and, as explained here, also of finite This work analyzes how γ depends on the properties of lattice many-body quantum systems with ergodically filled the spectrum, the structure of the initial state, and the num- LDOS. ber of particles that interactsimultaneously. We consider fi- Case2 is induced by the correlations that are present in nite many-body quantum systems described by realistic lat- nonchaotic eigenstates. They are typical of disordered sys- ticemodelswithtwo-bodyinteractionsandbybandedrandom tems undergoinglocalization with [22, 23] or without inter- matrices. All accessible power-law exponents are reached actions[9–11]and,asarguedhere,appearalsoincleaninte- with the disorderedmodels, while with the clean Hamiltoni- grablesystems. The power-lawexponentdueto correlations ans,westudysomespecificvalues. issmallerthanthatresultingfromenergybounds. Realistic many-body quantum systems. We consider one- The exponents of case 1 can be derived from asymptotic dimensional spin-1/2 models with L sites described by the expansions of the integral form of Eq. (1), assuming that followingHamiltonian, ρ (E)isabsolutelyintegrable[46]andthatitsderivativesex- 0 istandarecontinuousin[E , ]. Twoscenariosareidenti- low L ∞ fied[47,48]: H = h Sz +H +λH , (2) ((iii))IfIflimρE(→EE)lodwecρa0y(Es )ab>ru0p,tltyhecnloastelotnogtthimeelsowFe(rt)b∝out−nd2., nX=1 n n NN NNN 0 H =J SxSx +SySy +∆SzSz , such that ρ (E) = (E E )ξη(E) with 0 < ξ < 1 and NN n n+1 n n+1 n n+1 lim 0η(E)>0,th−enFlo(wt) t 2(ξ+1). Xn (cid:0) (cid:1) TEh→esEelorwesultshavebeenobtain∝edf−orcontinuousfunctions. HNNN = J SnxSnx+2+SnySny+2+∆SnzSnz+2 . Yetweshowthattheyremainvalideveninthecaseofdiscrete Xn (cid:0) (cid:1) spectra provided Ψ(0) is chaotic and the LDOS is ergodi- | i It contains nearest-neighbor (NN) and possibly also next- callyfilled. nearest-neighbor(NNN)couplings;~ = 1andSx,y,z arethe To determine if the initial state is chaotic, one performs n spinoperatorsonsiten. h arerandomnumbersfromauni- scaling analysis of the inverse participation ratio (IPR) of n form distribution [ h,h]; the system is clean when h = 0 Ψ(0) writtenintheenergyeigenbasis,IPR C(0) 4. − | i 0 ≡ α| α | and disordered otherwise. J is the coupling strength, ∆ the IPR−01istheeffectivenumberofenergyeigenstatePscontribut- anisotropyparameter,andλ theratiobetweenNNNandNN ingtotheinitialstate. Achaotic Ψ(0) samplesmostenergy couplings. J = 1 sets the energy scale. The Hamiltonian | i eigenbasiswithoutanybias,soIPR0 −1,where isthe conservestotalspininthez direction z. We workwith the dimension of the Hilbert space. Hen∝ceD, as the systDem size largestsubspace z =0ofdimensionS =L!/(L/2)!2. increases,ρ0(E)becomeshomogeneouslyfilledandcloseto TheintegrableSlimitsofH includethDecleannoninteracting an absolutely integrablefunction. An illustrative example is XX (∆,λ,h = 0)and theclean interactingXXZ (∆ = 0, thatofanarbitraryinitialstateprojectedontotheeigenstates λ,h=0)models.Thesystembecomeschaoticasλincre6ases of a full random matrix (FRM). Since these eigenstates are fromzero[54–57]andthelevelspacingdistributionchanges pseudo-random vectors, the overlaps C(0) are random vari- from Poisson [58] to a Wigner-Dysonform [59]. It also be- α 3 caonmdhes<chJao[t6ic0–w6h2e]n. thedisorderstrengthincreasesfromzero F(t) = e−4Nσ022t2 (cid:12)erf(cid:16)E0−E√l2owσ0+iσ02t(cid:17)−erf(cid:16)E0−√Eu2pσ+0iσ02t(cid:17)(cid:12)2, whereerfisthe(cid:12)errorfunctionand isanormalizationco(cid:12)n- The initial states considered are site-basis vectors, where (cid:12) N (cid:12) stant that depends on L through the energy bounds and σ . the spin on each site either points up or down in the z- 0 At long times, after dropping the oscillations from the si- direction. An example is the experimentally [63] accessible nusoidal term cos[t(E + E )], the expression becomes Ne´elstate, ... ,thathasbeenextensivelyusedin up low studiesofth|e↑↓d↑y↓n↑am↓↑i↓csofiintegrablespinsystems. Site-basis F(t ≫ σ0−1) ≃ (2πσ02t2N2)−1 k=up,lowe−(Ek−E0)2/σ02, vectorsevolveunderH (2)afterastrongperturbation,where fromwherethet−2power-lawdecPayisevident. the anisotropyparameter is quenchedfrom ∆ to a fi- When h = 1, we get γ 1. This curve is depicted with nitevalue. TheenvelopeoftheLDOSforthese→in∞itialstates athicklinein Fig.1(a). A∼bovethisline, h > 1andγ < 1. isthereforeGaussian. Anexamplewithγ 1/2isisolatedinFig.1(b). Thisγ is ∼ closetotheexponentφobtainedfromthescalinganalysisof Realistic disordered systems. Figure 1 shows the survival IPR φ [22]. Thisexamplebelongstocase2. probability of site-basis vectors evolving under H (2) with 0 − ∝D Figure 1 (a) demonstrates that with the disordered XXZ ∆ = 1, λ = 0 and various values of h. The initial decay model, we can obtain all power-law exponentsaccessible to is Gaussian, as expectedfromthe Gaussian LDOS. Itagrees verywellwiththeanalyticalexpressionF(t)=exp( σ2t2), realistic lattice many-body quantum systems with two-body − 0 interactions.Byvaryingh,everyγ [0,2]canbereached. as seen for the bottom curve of Fig. 1 (a). Subsequently Bandedrandommatrices. Algebr∈aicdecaysfasterthant 2 the dynamics slows down and becomes a power-law for all − alsosignaltheergodicfillingoftheLDOS.Theyarepossible curves. ifinsteadoftwo-bodyinteractions,many-bodyrandominter- actionsareincluded. Asthenumberofparticlesthatinteract 100 (a) 0.3 (b) simultaneouslygrows,increasingthenumberofuncorrelated γ ∼1/2 nonzero elements in the Hamiltonian matrix, the density of >0.2 10-1 <f statestransitionsfromGaussiantoasemicircle[35]. Thelat- 0.1 ter is typical of FRMs [59]. This transition is reflected also h=1.75 > 0 in the shape of the LDOS [18, 19, 26, 67, 68]. The Fourier <F10-2 0.60 20 40 60 transformof a semicircle gives F(t) = [ (2σ t)]2/(σ2t2), (c) J1 0 0 where is the Bessel function of the first kind [18, 19]. 0.4 J1 10-3 <f>0.2 γ ∼ 2 Tashyemdpetoctaiyceaxtpsahnosirotntirmeveesalissafapsotwerert-hlaawn dGeacuasysiwanithaγnd=th3e, 10-4 1 100 00 5 h=01.20 aFm(tple≫ofσ0−ca1s)e≃1(i[i1),−wshiner(e4σf0otr)]t/h(e2πseσm03ti3c)ir.clTe,hiξs is=an1e/x2-, t t η(E)=(2πσ2) 1(2σ E)1/2,andE = 2σ . 0 − 0− low − 0 FIG.1: Survivalprobability(a)andf(t)(b),(c). In(a): frombot- tomtotop,h = 0.2,0.3,...0.9,h = 0.95,1,1.25,...3,h = 3.5. 0 10 Thicksolidline: h = 1withγ ∼ 1. Circles: analyticalGaussian decayF(t)=exp(−σ02t2).In(b)and(c):Numericalcurve(solid), const−L−1lnt−γ (dashed). Averagesover105dataofdisorderre- -1 10 alizationsandinitialstateswithE0∼0;L=16,closedboundaries. > F < -2 10 When the disorderstrength is small, 0 < h < 1, the sys- tem is chaotic and the LDOS is very filled. This is corrob- oratedfrom the analysisof levelstatistics and by computing 10-3 theinverseparticipationratioaveragedoverinitialstatesand -2 -1 0 1 2 randomrealizations. One findsthat IPR 1. For the 10 10 10 10 10 h 0i ∝ D− t value of h where IPR is maximum, the decay of F(t) at 0 long times is th2, asiillustrated with the bottom curve in ∝ − FIG. 2: Survival probability for basis vectors evolving under Fig.1(a).Forothervaluesofhin(0,1],wehavetheinterme- PBRM with b = 0.1,0.5,1,2,5,10,20,50,100,3000 (solid) diateregion,where1 γ <2. Thesevaluesmayresultfrom from top to bottom. They correspond respectively to the fitted ≤ acompetitionbetweenweakcorrelationsandenergybounds, γ ∼ 0.1,0.5,0.6,0.7,0.9,1.2,1.4,1.9,2.2,2.8 (dashed). Analyt- butthisneedstobefurtherinvestigated. ical F(t) = [J1(2σ0t)]2/(σ02t2) (dotted); D = 3432. Averages The bottom curve of Fig. 1(a) is isolated in Fig. 1(c), over100realizationsand343initialstateswithE0∼0. which shows the rescaled survival probability f(t) = (1/L)lnF(t) [64, 65]. For L 1, this quantity is To illustrate the increase of the value of γ from 2 to the − ≫ independent of L [66]. Figure 1 (c) is a clear exam- upper bound γ = 3, we consider power-law banded ran- ple of the power-law decay caused by energy bounds [case dommatrices(PBRM)[69–71]. DespitethesuccessofFRMs 1(i)]. The Fourier transform of a Gaussian LDOS that has in describing statistically the spectrum of complex systems, lower E and upper E bounds, as in our case, leads to they imply the unphysical scenario of all particles interact- low up 4 ing simultaneously. Banded random matrices were intro- valuesofλand∆, including∆ = 0;andotherinitialstates. duced [67] in an effort to better approach random matrices A t 2 decay has also been speculated for the chaotic Ising − to real systems. We use PBRMs that preserve time rever- modelwithlongitudinalandtransversefields[75]. sal symmetry and whose elements are real random numbers An analyticalexpressionexistsforF(t) for the Ne´elstate frHom2 a G=au1s/s[i1an+di(sntribumtio)n/b[722]]f:orhHnn=nim=. 0T,hheHvna2nluie=of2b, epvaonlsvioinngfuonrdleorngthetimpeersi,odLict X1/X2 mod0e,l g[6iv5e,s7f6]N.e´el(Itt)s ex- dhetnermmiineshowfas|tthe−elemen|tsdecrea6seastheymoveaway L 1ln 2 L 1+2 1Lt 1−/2 , →as indeed confiXrmXed wi→th − − − − fromthediagonal.Whenb ,thePBRMcoincideswitha t−he dashe(cid:2)d line(cid:0)in Fig. 3(d). Su(cid:1)c(cid:3)h small γ indicates that the →D FRM. LDOS is not ergodicallyfilled, as seen in Fig. 3(c) and cor- InFig.2,weshowthesurvivalprobabilityforPBRMswith roboratedbelowbycalculatingIPR . 0 different values of b. As b grows from 50 to and the Amongthetotal = L!/(L/2)!2componentsoftheNe´el ∼ D LDOStransitionsfromcase1(i)tocase1(ii),γincreasesfrom D state,only2L/2arenonzeroandtheyareallequal, C(0) 2 = 2 to 3. In the other direction, as b decreases below 50, the 2 L/2 [76]. ThismeansthatIPR = 2 L/2. Using| thαe|Stir- eigenstatesbecomeless spreadoutandγ decreasesbelow2. − 0 − ling approximationfor large L, we have that ln Lln2. With PBRMs, weobtaina generalpictureofthebehaviorof From lnIPR vs ln , we find that IPR D1/2≃, so φ = thesurvivalprobability,coveringallvaluesofγ,withoutany 0 D 0 ≃ D− 1/2. One sees that, similarly to what is done in disordered restrictiontoaspecificmodel. systems [22, 23], the power-law exponent for the Ne´el state Realistic cleansystems. InFig. 3, we studytheNe´elstate in the XX model, γ = 1/2, can also be extracted from the evolving under a clean chaotic Hamiltonian [Figs. 3(a) and scalinganalysisofIPR . 3(b)] and under the XX Hamiltonian [Figs. 3(c) and 3(d)]. 0 Theenvelopeofthe LDOSisGaussianinbothcases(a)and The nonzero Cα(0) 2 are spread out in energy, result- | | (c),butvisiblysparseinFig.3(c). ing in a very sparse and inhomogeneous LDOS. The non- ergodicity of this state indicates that thermalization should 0.5 not occur. One way to confirm thermalization is by (a) 1 (b) 0.4 verifying the coincidence of the diagonal entropy Sd = LDOS00..23 f0.5 γ∼2 −StPh =α|Clnα(0)|2αlen−|ECαα(/0T)|2−[7(7]aαnEdαthee−tEhαe/rTm)o/d(yTnamαicee−nEtrαo/pTy), 0.1 [29]. HerPe, Sd = (L/2)Pln2 and Sth = lnP[Note that D the Ne´el state has E = 0 and thus infinite temperature T]. 0 0 0 -5 E0 5 0 5 1t0 15 20 Thetwoentropiesdonotcoincideeveninthethermodynamic 0.5 1.5 limit,where(Sth Sd)/L=ln√2. (c) (d) − 0.4 Conclusions. We have shown that the long-time decay of OS0.3 1 the survival probability in isolated lattice many-body quan- D f tum systems is algebraic, F(t) t γ, be the system inte- L0.2 0.5 ∝ − γ∼1/2 grableor chaotic, interactingor noninteracting,clean or dis- 0.1 ordered. The entire range of γ [0,3] can be reached with 0 0 ∈ -5 0 5 0 5 10 15 bandedrandommatrices,whileforrealisticsystemswithtwo- E t body interactions, γ [0,2]. From the value of γ, we infer ∈ howmuchdelocalizedtheinitialstateisintheenergyeigen- FIG. 3: LDOS [(a),(c)] and f(t) [(b),(d)] for the Ne´el state under basis. Thisprovidesawaytoidentifywhethertheinitialstate thechaoticopenH (2)withh = 0,∆ = 1/2,λ = 1[(a),(b)]and undertheclosedXXH[(c),(d)].(a),(c):NumericalLDOS(shaded willthermalizebasedexclusivelyonitsdynamics.Exponents area)andGaussianenvelope(solidline). (b): Numericalresultsfor γ 2signalergodicityandthereforethermalization. Advan- L=22(light),L=24(dark),andconst−L−1lnt−2(dashed).(d): tag≥esofthisapproachtotheproblemofthermalizationinclude L=400(solid)andfNe´el(t)(dashed). the following: any initial state can be considered, numerical XX methodsotherthanexactdiagonalizationareavailableforan- In Fig. 3(b), we observe a power-law decay t 2. The alyzingdynamics,andanaturalconnectionisestablishedwith − agreement between the t 2 decay (dashed line)∝and our nu- experimentsthatroutinelystudythedynamicsofmany-body − merical results (solid lines) suggests that the LDOS must be quantumsystems. ergodically filled and that thermalization should occur. In- Acknowledgments. This work was supported by the NSF deed,theinverseparticipationratiooftheNe´elstateinFig.3 Grant No. DMR-1147430 and Yeshiva University. EJTH (a)givesIPR 1andseveralstudiesforthismodelcon- acknowledges funding from CONACyT, PRODEP-SEP and 0 − ∝D firmthermalization[29,30,73,74]. Wefoundγ =2alsofor VIEP-BUAP,Mexico.WethankAdolfodelCampo,Yevgeny periodic boundary conditions; chaotic models with different BarLev,andMarcosRigolforusefuldiscussions. [1] S. Trotzky, Y.-A. Chen, A. Flesch, I. P. McCulloch, (2012);T.Fukuharaetal.,Nat.Phys.9,235(2013). U. Schollwo¨ck, J. Eisert, and I. Bloch, Nature Phys. 8, 325 5 [2] P. Jurcevic, B. P. Lanyon, P. Hauke, C. Hempel, P. Zoller, S.S.M.Wong,Rev.Mod.Phys53,385(1981). R.Blatt,andC.F.Roos,Nature(London)511,202(2014). [36] V.K.B.Kota,Phys.Rep.347,223(2001). [3] P.Richerme,Z.-X.Gong,A.Lee,C.Senko,J.Smith,M.Foss- [37] P. R. Zangara, A. D. Dente, E. J. Torres-Herrera, H. M. Feig, S. Michalakis, A. V. Gorshkov, and C. Monroe, Nature Pastawski,A.Iucci,andL.F.Santos,Phys.Rev.E88,032913 (London)511,198(2014). (2013). [4] P.Cappellaro,C.Ramanathan,andD.G.Cory,Phys.Rev.Lett. [38] Faster than Gaussian decays emerge in full random matri- 99,250506(2007). ces [18, 19, 21] and real systems with bimodal density of [5] G.Kaur,A.Ajoy,andP.Cappellaro,NewJ.Phys.15,093035 states[21]. (2013). [39] L.A.Khalfin,Zh.Eksp.Teor.Fiz.33,1371(1958)[Sov.Phys. [6] J.G.Muga,A.Ruschhaupt,andA.delCampo,TimeinQuan- JETP6,1053(1958)]. tumMechanics(Springer,London,2009),Vol.2. [40] H.M.Nussenzweig,NuovoCim.X20,694(1961). [7] A.delCampo,Phys.Rev.A84,012113(2011). [41] I. Ersak, Yad. Fiz. 9, 458 (1969) [Sov. J. Nucl. Phys. 9, 263 [8] A.delCampo,NewJ.Phy.18,015014(2016). (1969)]. [9] R.Ketzmerick,G.Petschel,andT.Geisel,Phys.Rev.Lett.69, [42] G.N.Fleming,IlNuovoCimento16,232(1973). 695(1992). [43] P.Knight,Phys.Lett.A61,25 (1977). [10] B. Huckestein and L. Schweitzer, Phys. Rev. Lett. 72, 713 [44] L.Fonda,G.C.Ghirardi,andA.Rimini,Rep.Prog.Phys.,41, (1994). 587(1978). [11] B.HuckesteinandR.Klesse,Phys.Rev.B59,9714(1999). [45] K.M.SluisandE.A.Gislason,Phys.Rev.A43,4581(1991). [12] T. Gorin, T. Prosen, T. H. Seligman, and M. Zˇnidaricˇ, Phys. [46] V.FockandN.Krylov,J.Phys.USSR17,93(1947). Rep.435,33 (2006). [47] A.Erde´lyi,J.Soc.Indust.Appr.Math.4,38(1956). [13] A.Goussev,R.A.Jalabert,H.M.Pastawski,andD.A.Wisni- [48] K.Urbanowski,Eur.Phys.J.D54,25(2009). acki,Scholarpedia7,11687(2012). [49] L.F.SantosandM.Rigol,Phys.Rev.E81,036206(2010). [14] V. V. Flambaum and F. M. Izrailev, Phys. Rev. E 64, 026124 [50] L.F.SantosandM.Rigol,Phys.Rev.E82,031130(2010). (2001). [51] J.T.ChalkerandG.J.Daniell,Phys.Rev.Lett.61,593(1988). [15] F.M. IzrailevandA. Castan˜eda-Mendoza, Phys.Lett.A 350, [52] J.Chalker,PhysicaA167,253 (1990). 355(2006). [53] V.E.Kravtsov,A.Ossipov,andO.M.Yevtushenko,J.Phys.A [16] V.K.B.Kota,LectureNotesinPhysics(Springer,Heidelberg, 44,305003(2011). 2014),Vol.884. [54] T.C.HsuandJ.C.Angle`sd’Auriac, Phys.Rev. B47, 14291 [17] S.K.Haldar,N.D.Chavda,M.Vyas,andV.K.B.Kota,J.Stat. (1993). Mech.2016,043101(2016). [55] K.KudoandT.Deguchi,J.Phys.Soc.Jpn.74,1992(2005). [18] E.J.Torres-HerreraandL.F.Santos,Phys.Rev.A89,043620 [56] L.F.Santos,J.Math.Phys50,095211(2009). (2014). [57] A.GubinandL.F.Santos,Am.J.Phys.80,246(2012). [19] E.J.Torres-Herrera,M.Vyas, andL.F.Santos,NewJ.Phys. [58] Duetodegeneracies,ahighpeakats = 0mayappearforthe 16,063010(2014). XX modelandtheXXZ modelattherootofunit∆ = 1/2 [20] E.J.Torres-HerreraandL.F.Santos,Phys.Rev.E89,062110 [37]. (2014). [59] T.Guhr,A.Mueller-Gro¨eling,andH.A.Weidenmu¨ller,Phys. [21] E.J.Torres-HerreraandL.F.Santos,Phys.Rev.A90,033623 Rep.299,189(1998). (2014). [60] Y. Avishai, J. Richert, and R. Berkovitz, Phys. Rev. B 66, [22] E.J.Torres-HerreraandL.F.Santos,Phys.Rev.B92,014208 0524161(2002). (2015). [61] L.F.Santos,J.Phys.A37,4723(2004). [23] E.J.Torres-Herrera,M.Ta´vora,andL.F.Santos,Braz.J.Phys. [62] F.Dukesz, M. Zilbergerts,andL.F.Santos, NewJ.Phys.11, 46,239(2016). 043026(1(2009). [24] L.F.Santos,F.Borgonovi,andF.M.Izrailev,Phys.Rev.E85, [63] S.Trotzky, P.Cheinet, S.Fo¨lling, M. Feld, U.Schnorrberger, 036209(2012). A. M. Rey, A. Polkovnikov, E. A. Demler, M. D. Lukin, and [25] L.F.Santos,F.Borgonovi,andF.M.Izrailev,Phys.Rev.Lett. I.Bloch,Science319,295(2008). 108,094102(2012). [64] M.Heyl,A.Polkovnikov,andS.Kehrein,Phys.Rev.Lett.110, [26] G.Casati,B.V.Chirikov,I.Guarneri,andF.M.Izrailev,Phys. 135704(2013). Lett.A223,430(1996). [65] F.AndraschkoandJ.Sirker,Phys.Rev.B89,125120(2014). [27] F.Borgonovi,F.Izrailev,L.F.Santos,andV.Zelevinsky,Phys. [66] H.Touchette,Phys.Rep.478,1(2009). Rep.626,1(2016). [67] E.P.Wigner,Ann.Math.62,548(1955). [28] V. Zelevinsky, B. A. Brown, N. Frazier, and M. Horoi, Phys. [68] Y.V.Fyodorov,O.A.Chubykalo,F.M.Izrailev,andG.Casati, Rep.276,85(1996). Phys.Rev.Lett.76,1603(1996). [29] L. F. Santos, A. Polkovnikov, and M. Rigol, Phys. Rev. Lett. [69] A. D. Mirlin, Y. V. Fyodorov, F.-M. Dittes, J. Quezada, and 107,040601(2011). T.H.Seligman,Phys.Rev.E54,3221(1996). [30] E.J.Torres-HerreraandL.F.Santos,Phys.Rev.E88,042121 [70] A.D.MirlinandF.Evers,Phys.Rev.B62,7920(2000). (2013). [71] F.EversandA.D.Mirlin,Rev.Mod.Phys.80,1355(2008). [31] E.J.Torres-Herrera,D.Kollmar,andL.F.Santos,Phys.Scr.T [72] I.Varga,Phys.Rev.B66,094201(2002). 165,014018(2015). [73] M.Rigol,Phys.Rev.Lett.103,100403(2009). [32] K.HeandM.Rigol,Phys.Rev.A85,063609(2012). [74] M.Rigol,Phys.Rev.A80,053607(2009). [33] M.Rigol,Phys.Rev.Lett.116,100601(2016). [75] T.Monnai,J.Phys.Soc.Jpn.83,064001(2014). [34] S.R.Wilkinson,C.F.Bharucha,M.C.Fischer,K.W.Madison, [76] P.P.Mazza,J.-M.Stphan,E.Canovi,V.Alba,M.Brockmann, P.R.Morrow,Q.Niu,B.Sundaram,andM.G.Raizen,Nature andM.Haque,J.Stat.Mech.2016,013104(2016). (London)387,575(1997). [77] A.Polkovnikov,Ann.Phys.(N.Y.)326,486(2011). [35] T.A.Brody,J.Flores,J.B.French,P.A.Mello,A.Pandey,and

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.