ebook img

Correlation decay and large deviations for mixed systems PDF

0.12 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 Correlation decay and large deviations for mixed systems

Correlation decay and large deviations for mixed systems R.Artuso DipartimentoSAT,andCenterforNonlinearandComplexSystems Universita`degliStudidell’Insubria, Como,22100,Italy, IstitutoNazionalediFisicaNucleare,SezionediMilano Milano,20133,Italy E-mail:[email protected] 6 C.Manchein∗andM.Sala 1 DepartamentodeF´ısica,UniversidadedoEstadodeSantaCatarina 0 Joinville,89219-710,Brazil 2 E-mail:[email protected] ∗ n a J 3 1 Abstract ] D We consider low–dimensionaldynamical systems with a mixed phase space and discuss the typical appearance of C slow,polynomialdecayofcorrelations:inparticularweemphasizehowthismixingrateisrelatedtolargedeviations . properties. n i l Keywords: Mixing,largedeviations,correlationfunctions,finite-timeLyapunovexponents. n [ 2 1. Introduction 2. Weakchaos v 1 Hereweprovideanumberofexamplesofthekindof 2 systems our analysis is devoted to: simplest examples 9 involveonedimensionaldynamics. Heretheprototype 2 0 caseisrepresentedbyPomeau-Mannevillemap[9], In this contribution we plan to provide a short re- . 1 view of results concerning mixing properties of de- 0 terministic dynamical systems. In particular our em- xn+1 =TPM(xn)= xn+xznmod1, (1) 6 phasis will be on systems enjoying ergodicity in the (cid:12)(cid:12) 1 whichpresentsanextremelyreachbe(cid:12)haviourasthein- : conventionalsense[1,2], andnotinfiniteergodicity[3], termittency parameter z is varied. At z = 1 it coin- v wherethe veryconceptofmixingturnsouttobe quite i cideswithBernoullimap,thestandardexampleoffully X delicate[4, 5]. The general idea is that when some chaotic(uniformlyhyperbolicdynamics),butassoonas deterministic dynamics presents a mixed phase space, r z > 1the0fixedpointbecomesmarginal,theinvariant a namelycoexistenceofachaoticseawith(evenarbitrar- measuredevelopsasingularityattheorigin,andanoma- ily small) regular structures, sticking of a typical tra- lousfeaturesappear[10,11].Whenz 2thesingularity jectoryclose to regularzonesenhancesseverelycorre- ≥ of the invariantmeasure at the origin becomesnon in- lations, and generically the speed of mixing becomes tegrable,and(1)providesoneofthesimplestexamples polynomial, where instead, for a fully chaotic system of infiniteergodicity[12, 13]. Inthe regime1 < z < 2 we expect exponential decay of correlations. In what thesystemisergodic(theinvariantprobabilitymeasure followswewilldenotebyweaklychaoticsystemsthose willbedenotedbyµ),anddisplayspower-lawdecayof for which a power-law decay of correlations is indeed correlation functions: in particular Hu[14] provedthat present.Weemphasisethat,beyondthesimple“mathe- thereexistLipschitzfunctionsF andGsuchthat matical”exampleswewillmentioninthenextsection, thegeneralobservationsstillapplytonontrivialphysi- dµ(F Tn )G dµF dµG =O(n ξ), (2) calsettings,notablyinfluiddynamics[6–8]. (cid:12)(cid:12)ZI ◦ PM −ZI ZI (cid:12)(cid:12) − (cid:12) (cid:12) Preprintsubmittedtopublication (cid:12)(cid:12) (cid:12)(cid:12)January14,2016 wheretheexponentξ is determinedby theintermit- structure of regular regions, where typical orbits stick PM tencyparameterzthrough for a long time, causing slow correlation decay, is ex- tremelyrich(seeforexample[27–29]). Thereisaclass 1 ξ = 1. (3) of differentarea-preservingmapswith a simpler phase PM z 1 − − spacestructure: In particular one notices that ξ diverges in the Bernoulli limit z 1+, while foPrMz 3/2 correla- L : yn+1 =yn+ f(xn) mod2π, (7) tions are not integr→able, the standard ce≥ntral limit the- (ε,γ) ( xn+1 = xn+yn+1 mod2π, oremdoesnotholdandproperlyrenormalizedBirkhoff where f(x )isdefinedby n averagesconvergetoaLe´vystablelaw[15]. Weremark thatin this examplethe “regular”regionamountsonly f(x )=[x (1 ε)sin(x )]γ. (8) n n n ontheindifferentfixedpointattheorigin. − − Anotherwell knownexample is providedby Pikovsky Suchamapwasintroducedin[30],forε=0andγ=1, mapT [16](see [17]formoredetailedbibliography), andgeneralizedin[31,32],whereatransitiontopoly- P whichisimplicitlydefinedby nomialdecayof correlationswasobservedasε 0+. → Asamatteroffactforε>0themapisfullyhyperbolic 1 [1+T (x)]z, 0< x<1/(2z), (anddisplaysexponentialcorrelationsdecay),while,for P x= T2Pz(x)+ 21z[1−TP(x)]z, 1/(2z)< x<1; (4)iεng=a0rothleeafinxaelodgpoouisnttoatth(e0o,0ri)gbinecinomPoemspeaarua-bMoalincn,epvlailyle- whilefornegativevaluesof x [ 1,0],themapisde- maps (and γ is a sort of intermittency exponent)2. By TfineidsathsaTt,P(w−hxi)le=ret−aiTnPin(xg).indA∈iffre−ermenatrkfiaxbeledfpeoaitnutrse(oaft cinon[3s2id]eirsinwgasthaergduyendatmhaictsthaelocnogrrtehlaetiuonnstdaebclaeymfoarnεifo=ld0, P ispolynomial,with x= 1)withanintermittencyparameterz,theinvariant ± probabilitymeasure µ is the Lebesgue measure (it is a 3(γ+1) ξ = ; (9) simpleexerciseto verifythatbywritingdownthe cor- L 3γ 1 respondingPerron-Frobeniusoperator):thisisobtained − bylettingtheinstabilityunboundedclosetotheorigin. thisinparticularpredictsanexponentξL = 3whenγ = Againwehaveapolynomialdecayrateforcorrelations, 1: forsucha parametervaluethere isa rigorouslower with bound[34]ξL 2. 1 ≥ ξ = ; (5) P z 1 − 3. Indirectapproachtocorrelations:recurrences here any value of z > 1 is allowed, and correlations becomenonintegrable(withageneralizedcentrallimit Direct numerical investigations of correlation func- theorem[17])forz 2. tionsarenotoriouslyhardtoaccomplish,since, evenif ≥ Examplesofsimilarbehaviourinhamiltoniandynamics theergodicinvariantmeasureisknown(likeinthecase include billiard tables in two dimensions, like the sta- of area-preserving maps), a Monte Carlo computation diumorSinaibilliard(seeforinstance[18]),thelimitof ofcorrelationfunctionswithN pointsinvolvesaner- MC kissing discs fordiamondbilliards[19, 20] 1, or mush- rorproportionalto1/√N ,whichcauseshugefluctu- MC roombilliards[21]. ationsaftermoderatetimes. Anotherpopularcontextisthatofarea-preservingmaps, Anindirectapproach,whichhasbeenwidelyusedinthe where the prototype example is the so called standard lastdecades,involvesrecurrencetime(Poincare´)statis- map[22,23]: tics. Such an approach was pioneered in [24, 25, 35], S(K) :( yxnn++11==yxnn−+Kyn+si1n(xn) mmoodd22ππ., (6) abnrideflityaddemsictrsibaeriagosriomupslebavseisr[s3io6n] (osefeita[l2s4o]:[37su])p.poWsee we partition the phase space into two disjoint sub- Despite the simple structure of (6), rigorous analysis sets, labelled 0 and 1: from a long trajectory of the of the standard map is extremely difficult[26], and the system we extract the sequence of residence times n ,n ,...,n ,... (for simplicity we are considering a 1 2 k 1Diamond billiards are in this context particularly interesting, sincebyvarying ageometrical parameter -the radius ofthe bound- ingdiscs-weareabletoturnthecorrelationspeedfromexponential 2Noticethatwhenεbecomesnegativethefixedpointattheorigin topowerlaw. becomeselliptic[33]. 2 discretetimedynamics)onasinglesubset(namelythe butionsoffinitetimeaverages:inparticularmanystud- waitingtimesbeforecrossingtheborder).Thisleadsto ieshaveconcernedfeaturesofdistributionfunctionsof a probabilitydistribution for residence times ℘(n): we finite-timeLyapunovexponents[27,42–46]. Arigorous alsosupposethattheaverageresidencetimeν = n = approachwasproposedin[47](seealso[48,49]):takea h i k ℘(k)isfinite. Nowwemakethe(severe)hypoth- one-dimensionalmapT,thenfinite-timeLyapunovex- · Pesisthatcrossingtheborderleadstoacompletedecor- ponentsaredefinedas relationofthedynamics[38,39]: inthisway[40],ifwe 1 dT(n)(x) consider the autocorrelation function of an observable λ (x )= ln . (13) G: CGG(m)=hG(m0+m)G(m0)i−hGi2, (10) Suchfinitetimenes0timatnesg(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)enerdicxally(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)xd0(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)ependuponthe we have that G(m0 + m)G(m0) = G2 if no cross- initialconditionx0,sotheyarecharacterizedbyaproba- ing takes plache between m0 andim0 +h miand G(m0 + bilitydistributionPn(λn),which,forergodicandchaotic m)G(m ) = G 2 otherwise. IfwedenotebyΠh(m)the systems, collapsestoa delta distributioninthe asymp- 0 probabiliitythhatinocrossingtookplaceinthetimelapse toticlimit [m0,m0+m],then nlim Pn(λn)=δ(λ−λ), (14) →∞ C (m)=Π(m) G2 +(1 Π(m)) G 2 G 2 =Π(m) G2 Gwh2e.reλistheLyapunovexponentofthemapT. Ifwe GG h i − h i −h i h i−h fiix a threshold λ˜ < λ, and compute the sub-threshold ((cid:16)11) (cid:17) Now it is easy to expressΠ(m) in termsof ℘(k), since weights: λ˜ theprobabilitythatapointchosenatrandomisthefirst (n)= dλ P (λ ) (15) pointofaresidencesequenceoflengthkis℘(k)/νwhile Mλ˜ Z n n n −∞ theprobabilitythatapointchosenatrandomisthefirst, wehavethatsuchquantitiesvanishinthelargetimelim- orthesecond,ofak+1residencesequenceis℘(k+1)/ν: its, howevertheirdecayisrelatedtocorrelationdecay: thus: morepreciselytherigorousresultin[47]statesthat,in- dipendentlyofthechoiceofthethresholdλ˜,if Π(m)= 1(℘(m)+2 ℘(m+1)+3 ℘(m+2)+ )= 1 ∞ ∞ ℘(i). 1 ν · · ··· ν (n) , (16) Xj=m Xi=j Mλ˜ ∼ nσ (12) thencorrelationfunctionsofsmoothobservablessatisfy Inthisway,anexponentialdecayof℘(m)yieldsasim- thebound ilarexponentialdecayforcorrelationfunctions,whilea 1 power-lawdecay℘(m) m α correspondstoapolyno- C(m) , (17) mialcorrelationdecay,∼with−ξ = α 2: such a quanti- ≤ mσ−1 − i.e. ξT σ 1. Such a bound cannot be how- tative correspondence has been scrutinized in [20] for ≥ − everoptimal,asforPomeau-Mannevillemapsasimple a number of billiard systems, in full agreement with argument[50]showsthatξ =σ. PM known rigorousresults (see for instance [41]). We re- We notice that an assessment like (16) is a (non- mark again that the crucial approximation involved in exponential) large deviation estimate (see for instance theabovereasoningconsistsinassumingdecorrelation [51]), and as a matter of fact the most general at each crossing: for a different kind of time statistics results[52–54]can bestatedasfollows: if we consider (flight times between collisions in a Lorentz gas with amapT,suchthatξ isthepowerlawmixingrate,than T infinite horizon) such an hypothesis has been investi- Birkhoffsumsofanobservableψsatisfythefollowing gatedin detail[40]: for shorttimes it obviouslymisses estimate: features of real correlations (for instance multiple col- lriespiorondsubceetswteheenanseyimghpbtootuicrirneggdimisec.s),whileitaccurately µx:(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)1nXnj=−01ψ(T(j)(x))−Z dµψ(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)>ǫ≤Cψ,ǫn1ξT(1;8) 4. Indirect approach to correlations: large devia- thatispo(cid:12)lynomiallargedeviations,w(cid:12)ithanexponentin- tions dependentofthethreshold,andcoincidingwiththeone ruling correlations decay3. In one dimension (18) in- The fundamental idea, at a qualitative level, is that cludesthecaseoffinite-timeLyapunovexponents(15), thesamedynamicalmechanismthatslowsdowncorre- lationdecayisresponsibleforanomalous(broad)distri- 3In[52]suchaboundhasbeenshowntobeoptimal. 3 with ψ(x) = ln T (x), while in higher dimensions the phase space, where sticking manifests in slow, poly- ′ | | leadingfinite-time Lyapunovexponentcannotbe writ- nomial mixing rates. Together with the popular use tenasasimpleBirkhoffsum: neverthelessitstillrepre- ofPoincare´ recurrences,weemphasizenewtechniques sentsanaturalindicator. basedonlargedeviationsproperties: we presentin the Such a technique has been used in a variety of last section novel calculations that corroborate the ef- context[55]: fromonedimensionalmapsas(4),toarea fectivenessofsuchamethod,bycomputingpolynomial preservingmaps (7), and it has been also employedto mixingratesofaweaklychaoticarea-preservingmaps corroborate universality claims[56] for correlation de- withveryhighprecision. cay of area preserving maps with mixed phase space. More recently this method was also used in exploring 0.16 n=2×103 (a) mixingpropertiesofcoupledintermittentmaps[57]. n=5×103 0.12 n=1×104 ) 5. Amodelexample n λ To provide an illustration of the technique we de- P(0.08 scribed in the former section, we provide new numer- 1 − ical experiments on the family (7) of area-preserving 00.04 1 maps. Inparticularwewanttoemphasizetwoaspects: the transitionfromchaoticε > 0 to intermittentε = 0 behavior,andhow(18)offersanefficientnumericaltool 0.00 tocomputeexact(polynomial)mixingratesinthelatter 0.80 0.85 0.90 0.95 1.00 1.05 1.10 1.15 λ case. n 1.0 In Fig. (1) we plot Pn(λn) for different cases: the hy- n=2×103 n=1×104 (b) perbolic case (a) has been obtained by 5 106 initial n=5×103 n=5×104 · 0.8 condition, while the intermittent case (b) refers to 107 ) 10−1 initial conditions. One almost obvious feature is that λn0.6 whenwegotoε = 0casethedistributionslookasym- ( 10−3 P metric,duetothefrequentappearanceoflowestimates 1 0.4 10−5 forLyapunovexponentsduetostickingtotheparabolic − 0 fixedpoint4. Thisleadsto a quantitativeanalysisonce 10.2 10−7 10−3 10−2 10−1 100 we estimate how the subthreshold tail of the distribu- tionsshrinkstozero(15):inthehyperboliccase(Fig.(2 0.0 (a)),weobserveapurelyexponentialdecay(noticethat 0.0 0.4 0.8 1.2 the data referring to the case γ = 2 have been shifted λn toavoidoverlapping),asexpectedforafullyhyperbolic case, while in the parabolic ε = 0 case, we observe a powerlawdecay,which,inviewof(18)shouldcoincide Figure1:Distributionoffinite-timemaximalLyapunovexponentsfor themap(7)with(a)ε=1andγ=2and(b)ε=0andγ=2.(Inset: with the mixing rate. A regression analysis yields for distributions of finite-time maximal Lyapunov exponents plotted in aγn=d2fo,rMγλ˜(=n)3∼n−1(.n80)±0.01n,w1.h49ile0.(094,)wprheidleict(s9)ξLpr=ed9i/c5ts, logarithmscale.) ξ =3/2,witneMssiλn˜gho∼wla−rge±deviationanalysisturns L tobeanextremelypowerfultoolinthenumericalanal- Acknowledgments ysisofquantitativemixingpropertiesofdynamicalsys- temswithweakchaoticproperties. R.A. thanksXavier LeonciniandSandroVaientifor many interesting discussions about such topics along 6. Conclusions many years. C.M. thanks CNPq (Brazil), and M.S. thanksCAPES(Brazil)forfinancialsupport. We have reviewed indirect methods for the investi- gations of correlation decay for systems with mixed References 4This motivates[46] systematic investigation of skewness in the distributions,asapossibleindicatorofdeviationsfromhyperbolicbe- [1] V.I.ArnoldandA.Avez, ErgodicProblemsofClassicalme- havior. chanics,(Addison-Wesley,Reading,1989). 4 10−1 preservingintervalmapswithindifferentfixedpoints,Ergod.Th. (a) &Dynam.Sys.20,1519(2000). [13] G.BelandE.Barkai,Weakergodicitybreakingwithdetermin- isticdynamics,Europhys.Lett.74,15(2006). 10−2 [14] H. Hu, Decay of correlations for piecewise smooth maps ) n withindifferentfixedpoints,Ergod.Th.&Dynam.Sys.24,495 (˜λ (2004). M 10−3 [15] S.Goue¨zel,Centrallimittheoremandstablelawsforintermit- tentmaps,ProbabilityTheoryandRelatedFields128,82(2004). [16] A. Pikovsky, Statistical properties of dynamically generated anomalousdiffusion,Phys.Rev.A43,3146(1991). 10−4 γ =2.0 [17] G.Cristadoro, N.Haydn, Ph.MarieandS.Vaienti, Statistical γ =3.0 propertiesofintermittentmapswithunboundedderivative,Non- 10−1 100 linearity,23,1071(2010). 104 n [18] N. Chernov and R. Markarian, Chaotic billiards, (American MathematicalSociety,Providence,1996). [19] J.Machta,Powerlawdecayofcorrelationsinbilliardproblems, 10−2 (b) J.Stat.Phys.32,555(1983). [20] R.Artuso,G.CasatiandI.Guarneri,Numericalexperimentsin billiards,J.Stat.Phys.83,145(1996). 10−3 ) [21] L.A.Bunimovich,Mushroomsandotherbilliardswithdivided n phasespace,Chaos11,802(2001). ( M˜λ10−4 [22] B.V.Chirikov,Auniversalinstabilityofmany-dimensionalos- cillatorsystems,Phys.Rep.52,263(1979). [23] A.J.Lichtenberg andM.A.Lieberman, Regular andchaotic 10−5 dynamics,(Springer-Verlag,NewYork,1992). γ =2.0 [24] C.F.F.Karney,Long-timecorrelationsinthestochasticregime, γ =3.0 PhysicaD8,360(1983). 10−6 [25] B.V.ChirikovandD.L.Shepelyansky, Correlationproperties 103 104 105 indynamicalchaosinhamiltoniansystems,PhysicaD13,395 n (1984). [26] K.BloorandS.Luzzatto,Someremarksonthegeometryofthe standardmap,Int.J.Bif.Chaos19,2213(2009). [27] S.TomsovicandA.Lakshminarayan,Fluctuationsoffinite-time Flinigeusr)efo2r:tDheecmaaypM(7λ)˜(wn)ith(s(yam)bεo=ls)1taongdetγhe=r2waitnhdaγre=gr3eassnidon(bfi)tε(=ful0l stability exponents in the standard map and the detection of smallislands,Phys.Rev.E76,036207(2007). andγ=2andγ=3. [28] M.Sala,C.MancheinandR.Artuso,Estimatinghyperbolicity ofchaoticbidimensionalmaps,Int.J.Bifurc.Chaos22,1250217 (2012). [29] C.MancheinandM.W.Beims,Conservativegeneralizedbifur- [2] I.P.Cornfeld, S.V. FominandYa. G.Sinai, Ergodictheory, cationdiagrams,Phys.Lett.A377,789(2013). (Springer-Verlag,NewYork,1982). [30] J. Lewowicz, Lyapunov functions and topological stability, [3] J.Aaronson,AnIntroductiontoInfiniteErgodicTheory,(Amer- J.Diff.Eq.38,192(1980). icanMathematicalSociety,Providence,1997). [31] R.ArtusoandA.Prampolini,Correlationdecayforanintermit- [4] I. Melbourne and D. Terhesiu, Operator renewal theory and tentarea-preservingmap,Phys.Lett.A246,407(1988). mixingrates fordynamical systemswithinfinite measure, In- [32] R. Artuso, L. Cavallasca and G. Cristadoro, Dynamical and vent.Math.189,61(2012). transportproperties inafamilyofintermittent area-preserving [5] M.Lenci,Oninfinite-volumemixing,Commun.Math.Phys.298, maps,Phys.Rev.E77,046206(2008). 485(2010). [33] C. Liverani, Birth of an elliptic island in a chaotic sea, [6] P.Castiglione,A.Mazzino,P.Muratore-Ginanneschi,A.Vulpi- ani,Onstronganomalousdiffusion,PhysicaD134,75(1999). Math.Phys.El.J.10,1(2004). [34] C. Liverani and M. Martens, Convergence to equilibrium for [7] F.RaynalandP.Carrie`re,Thedistributionof“timeofflight”in intermittent symplectic maps, Commun.Math.Phys. 260, 527 threedimensionalstationarychaoticadvection,Phys.Fluids27, (2005). 043601(2015). [35] S. R. Channon and J. L. Lebowitz, Numerical experiments [8] T.H. Somolon and J.P. Gollub, Chaotic particle transport in instochasticity andheteroclinic oscillation, Ann.N.Y.Acad.Sci. time-dependent Rayleigh-Be´nard convection, Phys.Rev. A 38, 357,108(1980). 6280(1988). [36] L.-S. Young, Recurrence times and rates of mixing, Israel [9] Y.PomeauandP.Manneville, Intermittent transition toturbu- J.Math.110,153(1999). lenceindissipativedynamicalsystems,Commun.Math.Phys.2, [37] V.Baladi,Decayofcorrelations,Proc.SymposiaPureMath.69, 189(1980). 297(2001). [10] P.GaspardandX.-J.Wang,Sporadicity: betweenperiodicand [38] V.Baladi, J.P.EckmannandD.Ruelle, Resonancesforinter- chaoticdynamicalbehaviors,Proc.Natl.Acad.Sci.USA85,4591 mittentsystems,Nonlinearity2,119(1989). (1988). [39] P.Dahlqvist, Approximatezetafunctions fortheSinaibilliard [11] X.-J. Wang, Statistical physics of temporal intermittency, andrelatedsystems,Nonlinearity8,11(1995). Phys.Rev.A40,6647(1989). [40] P.DahlqvistandR.Artuso,OnthedecayofcorrelationsinSinai [12] R. Zweimu¨ller, Ergodic properties of infinite measure- 5 billiardswithinfinitehorizon,Phys.Lett.A219,212(1996). [41] L.A.Bunimovich, Ontherateofdecayofcorrelations indy- namicalsystemswithchaoticbehavior,Sov.Phys.JETP62,842 (1985). [42] M. Falcioni, U. Marini Bettolo Marconi and A. Vulpi- ani, Ergodic properties of high-dimensional symplectic maps, Phys.Rev.A44,2263(1991). [43] H.SchomerusandM.Titov,Statisticsoffinite-timeLyapunov exponents in a random time-dependent potential, Phys.Rev. E 66,066207(2002). [44] C.Anteneodo, Statistics offinite-time Lyapunovexponents in theUlammap,Phys.Rev.E69,016207(2004). [45] M.W.Beims,C.MancheinandJ.M.Rost,Originofchaosin soft interactions and signatures of nonergodicity, Phys.Rev. E 76,056203(2007) [46] C.Manchein,M.W.BeimsandJ.M.Rost,Characterizing the dynamicsofhigherdimensionalnonintegrableconservativesys- tems,Chaos22,033137(2012). [47] J.F.Alves, S.Luzzatto andV.Pinheiro, Lyapunovexponents andratesofmixingforone-dimensionalmaps,Ergod.Th.&Dy- nam.Sys.24,637(2004). [48] J.F.Alves,S.LuzzattoandV.Pinheiro,Markovstructuresand decay ofcorrelations for non-uniformly expanding dynamical systems,Ann.I.H.Poincare´AN22,817(2005). [49] J.F.Alves,J.M.Freitas,S.LuzzattoandS.Vaienti,Fromrates ofmixing torecurrence times via large deviations, Adv.Math. 228,1203(2011). [50] R.ArtusoandC.Manchein,Commenton“Lyapunovstatistics andmixingrates”,Phys.Rev.E87,016901(2013). [51] F. den Hollander, Large deviations, (American Mathematical Society,Providence,2000). [52] I.Melbourne,Largeandmoderatedeviationsforslowlymixing dynamicalsystems,Proc.Am.Math.Soc.137,1735(2009). [53] M. Pollicott and R. Sharp, Large deviations for intermittent maps,Nonlinearity22,2079(2009). [54] I.MelbourneandM.Nicol,Largedeviationsfornonuniformly hyperbolicsystems,Trans.Amer.Math.Soc.360,6661(2008). [55] R. Artuso and C. Manchein, Instability statistics and mixing rates,Phys.Rev.E80,036210(2009). [56] G.CristadoroandR.Ketzmerick,Universalityofalgebraicde- caysinhamiltoniansystems,Phys.Rev.Lett.100,184101(2008). [57] M.Sala, C.MancheinandR.Artuso,Extensivenumericalin- vestigationsontheergodicpropertiesoftwocoupledPomeau- Mannevillemaps,PhysicaA438,40(2015). 6

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.