ORIGINALRESEARCH published:28February2017 doi:10.3389/fphy.2017.00006 Non-equilibrium Annealed Damage Phenomena: A Path Integral Approach SergeyG.Abaimov* CenterforDesign,Manufacturing,andMaterials,SkolkovoInstituteofScienceandTechnology,Moscow,Russia We investigate the applicability of the path integral of non-equilibrium statistical mechanicstonon-equilibriumdamagephenomena.Asanexample,afiber-bundlemodel with a thermal noise and a fiber-bundle model with a decay of fibers are considered. Initially, we develop an analogy with the Gibbs formalism of non-equilibrium states. Later, we switch from the approach of non-equilibrium states to the approach of non-equilibriumpaths.Behaviorofpathfluctuationsinthesystemisdescribedinterms of effective temperature parameters. An equation of path as an analog of the equation of state and a law of path-balance as an analog of the law of conservation of energy aredeveloped.Also,aformalismofafreeenergypotentialisdeveloped.Forfluctuations ofpathsinthesystem,thestatisticaldistributionisfoundtobeGaussian.Also,wefind the “true” order parameters linearizing the matrix of fluctuations. The last question we Editedby: discuss is the applicability of the phase transition theory to non-equilibrium processes. ZbigniewR.Struzik, UniversityofTokyo,Japan From near-equilibrium processes to stationary processes (dissipative structures), and Reviewedby: then to significantly non-equilibrium processes: Through these steps we generalize the BikasKChakrabarti, conceptofanon-equilibriumphasetransition. SahaInstituteofNuclearPhysics, India Keywords:damage,pathapproach,fluctuations,fiber-bundlemodel,statisticalphysics AlexHansen, PACS. 62.20.M—Structural failure of materials—89.75.-k Complex systems—05. Statistical physics, NorwegianUniversityofScienceand thermodynamics,andnonlineardynamicalsystems. Technology,Norway *Correspondence: INTRODUCTION SergeyG.Abaimov [email protected] Oneofthemostimportantaspectsinthemoderntheoryofdamagephenomenaisthepossibilityto Specialtysection: describedamageoccurrenceonthebaseoftheformalismofstatisticalphysics.Manyapproaches Thisarticlewassubmittedto havebeendeveloped[1–24],andsurveysofdevelopmentscanbefoundin[25–30].However,the InterdisciplinaryPhysics, majorityofthesestudiesaredevotedtoequilibriumornear-equilibriumdamageoccurrence.Inour asectionofthejournal study,weconsideranon-equilibriumdamageprocess(acascade,anavalanche)farfromthepoint FrontiersinPhysics ofaquasi-staticstate. Received:07December2016 Once the analogy between damage mechanics and statistical physics had been discovered, it Accepted:01February2017 immediatelybecameclearthatanydamagephenomenoncanbedescribedasaphasetransition, Published:28February2017 wheredamagegrowthrepresentstheappearanceofanotherphaseandthepointofmaterialfailure Citation: isthepointofthephasetransition.Thebestwaytoillustratethissimilarityistocomparethephase AbaimovSG(2017)Non-equilibrium diagramsinFigure1.Therearedebatesintheliterature[3–13,21,22,24,29–35]aboutwhether AnnealedDamagePhenomena:A thisphasetransitionisofthefirstorderorofthecontinuousorder.Thereasonforthisuncertainty PathIntegralApproach. Front.Phys.5:6. isthatspinodalpointsareverysimilartocriticalpointsinthesensethatintheirvicinityasystem doi:10.3389/fphy.2017.00006 switchesitsbehaviorfromthelawsintrinsiconlytothisparticularsystemtouniversalpower-law FrontiersinPhysics|www.frontiersin.org 1 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach scaling.However,thereisnodoubtthatanystructureunderload isinthemetastablestatesincetherealwaysexistsanucleusofthe criticalsize.Inotherwords,bybreakingenoughofthematerial with external forces, we can always cause material failure. This isaclearindicationofmetastability,becausethereisnocritical nucleus(nomatterhowbigitis)forstablestates. Todiscussourapproach,weinvolvethefiber-bundle model (further,FBM)asthemostillustrativerepresentativeofmodels with damage. However, the applicability of the approach we developliesbeyondtheone-dimensionalformulationoftheFBM andcanbeeasilygeneralizedtotwoorthree-dimensionaldamage phenomena. For all possible approaches, an analogy between statistical physicsanddamagephenomenashouldrelyonmappingofone type of fluctuations on another. In this sense, damage exhibits more complex behavior than classical thermodynamic systems of statistical physics due to the presence of two different forms of disorder: quenched and annealed. Therefore, to perform the aforementionedmapping,thesetwoformsofdisorderareusually studiedapartfromoneanother. The influence of quenched disorder on the model behavior isgenerallyobservedwherethedynamicaltimescaleoffracture is much faster than the timescale of fluctuations. In this case, “frozen” initial disorder in a model plays the crucial role. This type of disorder, employed in the FBM, has been studied by many authors [8, 16, 28, 29, 31–85]. In particular, in our previous studies [21, 22, 30], we introduced quenched disorder bymeansofthefiberstrengthvariabilityandfortheensembleof constantstrainwewereabletomapthefluctuationsobservedon thermodynamicfluctuationsinstatisticalphysics. Another type of behavior is exhibited when a FBM is influenced by the annealed disorder. One of two approaches is generally employed: the disorder is introduced in the model FIGURE1|Phasediagramsof(A)adamagephenomenon,(B)aliquid-gas formulation either by means of a phenomenological rate of system,and(C)amagneticsystem. fiber failures with time [29, 86–109] or by the direct input of a stochastic or thermal noise into the state of stress of a fiber [23,110–118]. Annealed disorder generates fluctuations which are closely the Classical Gibbs Approach of States. In Section Statistical related to their thermodynamic analogs in statistical physics. Mechanics of the Path Integral Approach, we consider a path Indeed,addingstochasticnoisetoafiberloadseemstobequite integral, when instead of non-equilibrium states of the system similartothestatisticalfluctuationsof,forexample,gaspressure we investigate non-equilibrium paths. We map fluctuations of acting on a piston. However, as it was recently discovered damage on the main concepts of statistical mechanics, such [119–123], annealed fluctuations in damage phenomena as temperature, entropy, free energy potential, and the law of exceed thermodynamic fluctuations by orders of magnitude. In conservationofenergy.However,wefindthattheseconceptsare particular, it was found that the effective temperature required nolongerassociatedwiththeenergycharacteristicsofthestates toprovidetheannealedbehaviorobservedinexperimentsis10 ofthesystembutaredefinedoverthespaceofallpossiblepaths. times higher (about 3,000 K) than the real temperature during Also,weinvestigatethebehavioroffluctuationsforthisapproach thetests.Thisdiscrepancyisexplainedbycomplexinteractions andfind“actual”orderparameterswhichdiagonalizethematrix of quenched and annealed types of disorder which are present of the probability distribution. In the conclusions, we discuss simultaneouslyinthemodel[18,113–118,122,124]. howourtheoryleadstothedescriptionofnon-equilibriumphase In our study, we continue our previous research [23, 125, transitions. 126] and investigate a FBM with annealed damage evolution. In Section Model, we introduce models which we will use MODEL to illustrate our approach. In Section The Classical Gibbs ApproachofStates,wedevelopananalogoftheclassicalGibbs Damage is a complex phenomenon. The basic principles of approach for non-equilibrium states. However, this approach damage are often completely disguised by the secondary side has its restrictions which we discuss in Section Restrictions of effects of its appearance. Therefore, to investigate the main FrontiersinPhysics|www.frontiersin.org 2 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach conceptsofnon-equilibriumbehavior,itisreasonabletoconsider of infinite number of fibers N is taken first. Therefore, → +∞ initiallyasimplemodel. althoughonlythesmallfractionoffibersbreaksateachtime-step, Inthispaper,asanillustration,weconsideranannealedFBM. thenumberofbrokenfibersismuchhigherthanunity. WeassumethatthenumberNoffibersinthemodelisconstant Although,weassumethenoisetobeGaussian,thisdoesnot andinfiniteinthethermodynamiclimit.Intactfibersallcarrythe playanyroleinourmodelduetothefactthats,E,andεarethe samestrainε whichisidenticallyequaltothestrainεofthetotal sameforallfibersaswellasforallsystemsintheensembleand f model:ε ε.Thestressofeachintactfiberisassumedtohave aresomeconstants.Therefore,thenoisevalue1σ s Eε f break ≡ = − alinearelasticdependenceonthestrainuntilafiberfailure:σ atwhichfibersbreakisfixedsothatP(s Eε)and1 P(s Eε) f − − − Eε .Inthispaper,weconsiderconstantstrainε constasan aretwoconstantsaswell. f = = externalboundaryconstraintofthemodel.Thecaseofconstant The second modification of the FBM which we consider stress can be considered in a similar way (the redistribution of (following the original terminology, a model with “breaking theloadcomplicatestheanalyticalsolutionbutdoesnotchange kinetics”) is an annealed decay fiber-bundle model [86–109; the applicability of the path integral). The Young modulus E further,DFBM].Thereisnothermalnoiseinthismodel.Instead, hereisassumedtobethesameforallfibers;however,thestress- each(sofarintact)fiberhasaprioriassignedprobabilityp dtof f strain dependence for the total model is non-linear because of failingduringthetimeintervaldt andprobability(1 p dt)of f − fiberfailures.Weintroducean“intact”orderparameterLasthe stayingintact. fraction of intact fibers (unity minus the damage parameter). Here, p is the decay rate which is constant during model f Everywhere further we assume that at the initial time t 0 all evolution. The duration of the time-step dt is considered to be = fibersareintact:L 1. fixedaswell.Therefore,p dtand(1 p dt)aresomeconstants,as t 0 f f Asafirstmodifi=cat=ionoftheFBM,weconsideranannealed wellas1 P(s Eε)andP(s Eε)−respectively,whichrepresent − − − fiber-bundle model with noise [23, 110–118] (further, NFBM). fixedprobabilitiesofafiberbreakingorstayingintactduring a Eachfiberhasthermalfluctuationsofitsenergycharacteristics. time-step.Inthissense,equatingp dtand1 P(s Eε),wecan f − − Fromstatisticalmechanics,weknowthatapiston,whichworks map the DFBM on the NFBM. However, some care should be asagasboundaryandissupportedbyaspring,hasoscillations exercisedwhenwesayp dt 1 P(s Eε).Here,weconsiderdt f = − − duetotheequipartitionofenergy.Inparticular,theelasticenergy tobethedurationofthetimeintervalduringwhichtheGaussian ofthespringfluctuatesasifawhiteGaussiannoisewereaddedto noise loses its correlation with the value of the noise at the the stress of the spring. Similarly, we assume that the stress of previoustime-stepand“switches”toanewvalue.Anillustrative eachfiberσ hasanaddition1σ ofawhiteGaussiannoisewith exampleistheoutputoftheGaussianrandomgeneratorwhich f f zeromeanandstandarddeviation√k T.Asitwasdiscussedin provides a discrete set of uncorrelated i.i.d. numbers. In this B SectionIntroduction,thetemperatureheremaynotcorrespond sense, the duration of dt is not arbitrary but is dictated by the to the temperature during the experiment but may be several physical processes responsible for the correlations in the noise. ordershigher[119–123]. Hence,wecannotsayherethattheconsideredprobabilitiesare Theprobabilitydensityfunctionofthenoiseis proportionaltodt. Theissuecanbeillustratedwiththeaidofextremestatistics. 1 1σ2 Ifweobserveafiberduringalongtimeintervalτconsistingof p(1σf) e−2kBT (1) N >> 1time-stepsdt,theprobabilityofthefiberstayingintact = √2πk T B duringthistimeintervalisPN(s Eε) edτtlnP(s−Eε),whichis − = andthecumulativedistributionfunctionis clearlynotlinearwithtime.Therefore,ourmappingoftheNFBM onto the DFBM is introduced only for one time-step. In this 1σf sense,wewillconsideronlyoneofthesemodels(theDFBMfor 1 x2 P(1σf) e−2kBTdx. (2) theapproachofstatesatSectionsTheClassicalGibbsApproach = √2πk T B Z of States and Restrictions of the Classical Gibbs Approach of −∞ States, which is more convenient for this as it is a model with Eachfiberhasanaprioriassignedstrengththresholdswhichwe a well-known solution; the NFBM for the approach of paths choose to be the same for all fibers and not to change during at Section Statistical Mechanics of the Path Integral Approach the model evolution. A fiber can break only when its stress σ sinceitismoreillustrativeofthegeneralizationtomorecomplex f Eε 1σ exceeds its strength s. We consider discrete time- modelformulations),assuming,however,thatallresultsarevalid f = + stepsdtofthemodelevolution.Ateachtime-step,theprobability fortheothermodelaswell. of a fiber breaking is 1 P(s Eε) and the probability of a Beforegoingfurther,weshouldclarifythatwedonotconsider − − fiberstayingintactisP(s Eε).Hereweassumethatthereare the healing of fibers: Once a fiber has broken, it stays broken − nocorrelationsofnoiseamongadjacentfibers.Also,inspiteof forever. However, our formalism easily allows healing to be thefactthatthetimeintervaldt betweentheconsecutivetime- added to the model formulation. We need only, in addition to steps is assumed to be small, it is supposed to be much larger P(s Eε) and 1 P(s Eε), to introduce the probability of − − − thanthedurationofanynoisecorrelations.Therefore,weassume a broken fiber healing. All formulae can be easily generalized thattherearenocorrelationsofthenoiseintimeeither.Another for this case as well, but due to the complexity of the results assumptionweutilizeisthatalthoughthetimeintervalbetween already present for the simpler model, we leave it for further consecutive time-steps has zero limit, the thermodynamic limit studies. FrontiersinPhysics|www.frontiersin.org 3 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach THE CLASSICAL GIBBS APPROACH OF combinatorialnumberofoutcomeswhenwedistributeNLintact STATES fibersandN(1 L)brokenfibersamongNfibers: − N! The classical Gibbs approach is formulated in terms of states. g . (5) [L] To develop its analog for damage phenomena, we need to give = (NL)!(N(1 L))! − somedefinitions:whatwerefertoasanon-equilibriumstate(or Tosimplifythisexpression,weapplyStirling’sapproximation afluctuation).Inmoredetails,thisapproachisdiscussedin[30], Chapter2. N N LetusimagineaFBMwhereweobserveseparatefibers.Asto N! O Nα (6) ≈ e amicrostate n wewillrefertoaparticularmicroconfiguration (cid:18) (cid:19) { } (cid:0) (cid:1) whenforeachfiberofthemodelweprescribewhetheritisintact and in the limit N >> 1 we discard the slow power-law orbroken.Forexample,forasystemwithN 3fibers,oneof dependenceO(Nα)onNincomparisonwiththefastexponential = themicrostatesis xx wherebytwosymbols“x,”wedenoted multiplier(N/e)N: { |} thatthefirsttwofibersofthemodelarebroken,whilethesymbol “|”meansthatthethirdfiberisstillintact.Foramodelwithonly N N N! . (7) ln three fibers, it is easy to list all the possible microstates: , ≈ e {|||} (cid:18) (cid:19) x , x , x , xx , x x , xx , xxx . {|| } {| |} { ||} {| } { | } { |} { } Thespecialnotationforthelogarithmicaccuracy,“ ,”stands Whatistheprobabilityofobservingaparticularmicrostatein ≈ln heretodenotetheaccuracytotheorderofanarbitrarymultiplier theensembleofidenticalsystems?FortheDFBM,thesolutionis withthepower-lawdependenceonN(fordetailedexamples,see well-knownasthesolutionofradioactivedecay.Theprobability ofafiberfailingattimetduringtimeintervaldtise−pftpfdtand [30],Chapter2). Applying Stirling’s approximation with the logarithmic t theprobabilityofafiberfailingbeforethetimetis e−pftpfdt accuracytoEquation(5),wefind = 0 1−e−pft.Therefore,theprobabilityofobservingamRicrostate{n} g[L]≈lnL−NL(1 − L)−N(1−L). (8) withLintactand(1-L)brokenfibersatthetimetis LetusnowprovethatZensemble(t) 1 e−pft −N isindeedthe w{enn}semble = e−pft NL 1−e−pft N(1−L). (3) fpuanrtcittiioonnifsutnhcetisounmooffthexepeonnseenmtiball=ef.uB(cid:0)nyc−tdioenfisnieti(cid:1)oNnL,/Ttheffe(t)paorvteitrioanll (cid:0) (cid:1) (cid:0) (cid:1) − Weusedherethesuperscript“ensemble”toemphasizethatthis microstates: probabilityisdictatedbytheensembleconsidered. Wecanrewritethedistribution(Equation3)as Zensemble(t)≡ e−NL{n}/Teff(t) (9) n w{enn}semble = Zensem1ble(t)e−NL/Teff(t) (4) GroupingmicrostatesbyfluctuX{at}ions[L],sinceallmicrostatesof thesamefluctuationhaveequalvaluesofL,wecanwrite: wherewedenotedTeff(t) ln−1 epft 1 .Formula(4)could bcaelcloedmeTeiffd(etn)titchael teoffetchteiveG≡itbebmspepr(cid:0)raotbuarbe−.iliTt(cid:1)yhednistZriebnusetmiobnle(ti)f we Zensemble(t)= g[L]e−NL/Teff(t). (10) [L] 1 e−pft −N plays the role of the partition function of th=e X ens−embleaswewillprovelaterinEquations(9–11). Substituting here Teff(t) ln−1 epft 1 and the statistical (cid:0) We see(cid:1) that the probability distribution of microstates is weightfromEquation(5),w≡eeasilyprove−therequiredstatement: (cid:0) (cid:1) Gibbsianwiththeeffectivetemperaturechangingwithtime;this aSlilmowilasrutsotaotchaellrmouordyennasemmicblseysatsembeiwnigth“etffheectliimveiltye”dcsapneocntriucaml. Zensemble(t) = N (NL)!(NN(1! L))! epft1 1 NL1N(1−L) (like an Ising model), the effective temperature can be both NXL=0 − (cid:18) − (cid:19) bpeogsiitninveinganodf ntheegamtivoed,elcheovoosluintigoninbtaucttsswtaittcehsinfogrthfiibsecrshoiincethtoe = epft1 1 +1 N = 1−e−pft −N. (11) (cid:18) − (cid:19) brokenonceatlatertimes. (cid:0) (cid:1) TofindtheprobabilityWensemble(t,L)ofobservingafluctuation As to a fluctuation [L] (a macrostate [L]) we will refer to [L] [L]intheensemble(toobserveasystemwiththegivenfraction the union of all microstates corresponding to the given value Lofintactfibers),wemultiplyprobability(Equations3,4)ofthe of parameter L: [L] n . For example, in the case of a model with three≡fib{ne}r:sLS,{nw}=eLo{bs}erve fluctuation [L 1/3] cmoircrreossptoantedsi:ngmicrostatesbythenumber(Equation8)ofthese = when we observe one of the three microstates: xx , x x , or xx . For the model with arbitrary N, th{e|num}b{er|g[L}] Wensemble g wensemble e−pft NL 1−e−pft N(1−L). of m{ icro|}states, corresponding to the fluctuation [L], is called [L] = [L] {n}∈[L] ≈ln(cid:18) L (cid:19) (cid:18) 1 − L (cid:19) the statistical weight of this fluctuation. It can be found as the (12) FrontiersinPhysics|www.frontiersin.org 4 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach To attribute a microstate n to a particular fluctuation [L], we whichstatesthatthereareabout√Nfluctuationsunderthewidth willwriteeither n [L]o{r}n :L L.Thereisnodifference ofthemaximumofWensemblewiththeprobabilitysimilartothe { }∈ { } {n} = [L] betweenthesetwonotations,andwewillutilizetheonethatwill mostprobablefluctuation[L ].Thereby,theprobabilityof[L ]is 0 0 bemoreconvenientatthemoment. oftheorderof1/√NwhichreturnsustoEquation(18). The dependence Wensemble on NL has a very narrow Let us introduce the partial partition function (of a [L] maximum due to the “wrestling” of two exponentially fast fluctuation) dependences: g and wensemble. The absolute width of this [L] n [L] maximum is of the order{o}f∈√N while the relative width is of Z[L] ≡ e−NL{n}/Teff(t) =g[L]e−NL/Teff(t) (20) the order of 1/√N. The most probable fluctuation (which will n [L] {X}∈ beobservedmostcommonly)isdeterminedbythemaximumof ∂Wensemble whenwesumtheexponentialfunctionsnotoverallmicrostates thisprobability.Equatingthederivativetozero, [L∂]L =0 but only over microstates of the particular fluctuation. Then, (cid:12)0 probability(Equation12)ofobservingthefluctuationisdirectly ∂lnWensemble (cid:12) or [L] 0,wefindtheequationofstate: (cid:12) relatedtothepartialpartitionfunctionofthisfluctuation: ∂L = (cid:12) (cid:12)0 (cid:12) (cid:12)(cid:12) L0 =e−pft (13) W[eLn]semble =g[L]w{enn}s∈em[Lb]le = ZenZs[eLm]ble. (21) TheprobabilitydistributionW[eLn]sembleisnormalizedbyunity: Next,letusconsideraveragingintheensemble: 1=X[L] W[eLn]semble =X[L] g[L]w{enn}s∈em[Lb]le. (14) (cid:10)f(cid:11)ensemble ≡ X{n} f{n}w{enn}semble =X[L] f{n}:L{n}=Lg[L]w{enn}s:eLm{nb}le=L Letusmultiplythisequalitybythepartitionfunction: = f{n}:L{n}=LW[eLn]semble. (22) [L] X Zensemble ≡ e−NL{n}/Teff(t) = g[L]e−NL/Teff(t), (15) Asanexample,weconsideraveragingofL: n [L] X{ } X L ensemble LWensemble. (23) wherewehaveagainsubstitutedthesumovermicrostates n by h i = [L] thesumoverfluctuations[L].Thenumberofnon-zeroter{m}sin X[L] this sum is of the order of √N (the width of the maximum of Wensemble): Here the probability distribution W[eLn]semble is the very fast [L] dependencecomparingtoslowdependenceL.Besides,Wensemble [L] Zensemble = g[L0]e−NL0/Teff(t)O √N . (16) ifsunncotriomnaδli(zLed bLy0)u,naintdy.wTehiemremfoerdei,atitelpylfiaynsd:the role of a delta- (cid:16) (cid:17) − Applying the logarithmic approximation (neglecting slow L ensemble L (24) 0 power-law dependencies and keeping only fast exponential h i ≈ dependencies), we prove the classical result of statistical The definition [L] n of a fluctuation [L] can be mechanicsthatapartitionfunctionequalsitsmaximalterm: presentedintermsof≡non{n-S}e:Lq{{un}i=l}ibLriumprobabilitydistributions Zensemble≈lng[L0]e−NL0/Teff(t). (17) which we developed in details in [30], Chapter 2 as the generalization of the Leontovich approach of non-equilibrium Dividing this equation by Zensemble, we find another important states [127, 128]. We observe a fluctuation [L] when in the relationship: ensembleweobserveasystemwiththegivenfractionLofintact fibers.Itislikeinstatisticalphysicstoobservefluctuationsofgas 1 w{enn}s:eLm{nb}le=L0≈ln g[L0] (18) dalelnististymwohleecnu,lefosrinextahmeplleef,tthhaelfgaosfhitassvgoaltuhmereedwshpiolenttahneeroeuasrlye no molecules in another (right) half of the volume. We call it that the probability of microstates corresponding to the most afluctuation(amacrostate).Buttodeterminethepropertiesof probable fluctuation [L0] is the inversed number of these this state, it is not enough to observe it for a microsecond as a microstates (the principle of the equivalence between the microstate.Tostudythisstate,weisolatethegasinthelefthalf canonicalandmicrocanonicalensembles). ofthevolumebyavirtualboundaryandobserveforsometime Equality (Equation 18) can be found without multiplying beforereleasingit.Butwhatdoesthisadditionalisolationmean? (Equation 14) by Zensemble. Only in this case, we arrive at the Itmeansthatthegasvisitsonlythemicrostatescorrespondingto equation ourfluctuationanddoesnotvisitothermicrostates. The same purpose can be achieved by considering non- 1=g[L0]w{enn}s:eLm{nb}le=L0O √N ≈lng[L0]w{enn}s:eLm{nb}le=L0 (19) equilibrium probability distributions [30]. Distribution (cid:16) (cid:17) FrontiersinPhysics|www.frontiersin.org 5 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach (Equations3,4)isdictatedbytheensembleboundaryconditions We see that the action of the free-energy is directly related to similartothepressureandtemperaturedictatedbytheboundary theprobabilitydistributionoffluctuationsanditsminimization conditions for the gas. However, a system may disregard these is equivalent to the evolution from states of low probability to orders and behave independently. Imagine students who are highlyprobablestates. orderedtoreadonepaperaweekonaverage.Onestudentworks Theensembleactionoffreeenergyis: hardandmayreadfivepapersaweekonaveragewhileanother onedoesnotstudyatall.Similar,asystemmayfluctuateviolating Aensemble NL /Teff Sensemble lnZensemble. (32) 0 ≡ − =− ordersfromtheensemble.Forthegas,itmeansthatthesystem stays in the left half of its volume (have non-zero probabilities Allotherresults,generallydevelopedforthecanonicalensemble, only for those microstates which do not have molecules in the are applicable here as well (for details, see [30], Chapter 2). right half). For our system with damage, it means that only However, our purpose in this manuscript is to suggest possible thosemicrostateshavenon-zeroprobabilitieswhichpossessthe approaches to solve an arbitrary system, not as simple as the correctvalueoftheintactfiberfractionLwhileprobabilitiesof DFBM.Therefore,weprovidedheretheclassicalGibbsapproach othermicrostatesarezero: of states only to the purpose to highlight differences with the 1/g , n :L L proposedbelowpathintegralapproach. w{n} = 0[,L]n{:}Ln{n} =L . (25) (cid:26) { } { } 6= RESTRICTIONS OF THE CLASSICAL Tofindtheentropyofthefluctuation,weapplyGibbsdefinition GIBBS APPROACH OF STATES andarriveatBoltzmann’sentropy: In the previous section, we were lucky that our system had an S w lnw lng . (26) [L] ≡− {n} {n} = [L] analyticalsolution.Thiswastheconsequenceofthefactthatwe X{n} wereabletofindaprioritheprobability te−pftpfdt 1 e−pft Fortheequilibriumprobabilitydistribution,thistransformsinto: = − 0 ofafiberfailingbeforetimet.ForthecasReofanarbitrarysystem, Sensemble ≡ − w{enn}semblelnw{enn}semble wegenerallyknowthedistributionofprobabilitiesonlyforone X{n} time-step.Inotherwords,ifattimeti asystemwithprobability =−X[L] g[L]w{enn}semblelnw{enn}semble twinh{eannatis}emtmhiebclerporissotbainatebailnimtyicorfowtshtiiatshtesNy{snLtei}mwaiintthttahNcetLnfiiebxientrttsaimcistefi-bsteerps,tiw+e1kbneionwg i 1 i 1 Wensemblelnwensemble, (27) { + } + =− [L] {n} where we have substitutedX[tLh]e sum over microstates n by the w{ennis+em1}ble =w{ennis}emble 1−pft NLi+1 pft N(Li−Li+1). (33) { } (cid:0) (cid:1) (cid:0) (cid:1) sumoverfluctuations[L].Intheright-handsideofEquation(27), dwensemble thedependencelnwensembleisaslow,power-lawdependenceon Infact,weutilizedherethegeneralGibbsformula {}dt (t)= NLwhileWensemble i{sn}theexponentiallyfastdependence,acting F wensemble(t′), t′ ≤t thattheevolutionofthedistributionof [L] {} like a delta-function due to its normalization Wensemble prhobabilities of microistates is determined by some functional [L] = dependence on the history of the model. To obtain the [L] 1. Therefore, utilizing Equation (18), we imPmediately find distribution of probabilities of the microstates at time t we Boltzmann’sexpressionfortheequilibriumentropyaswell: have to integrate this equation over all possible paths among allpreviousconfigurations[e.g.,129,130].Wehavetointegrate Sensemble =−lnw{enn}s:eLm{nb}le=L0 =lng[L0]. (28) aarllepionstseirbcloenninetcetremdedbiyatethecontrfiegmuernatdioounss; nthuemsebecronofifgduirffaetiroennst The non-equilibrium free energy potential (the effective paths; each path can have its own probability or be prohibited Helmholtzenergy)ofafluctuationmaybeintroducedas (irreversible damage in the case of the FBM, when a broken fibercannotbecomeintactagain).Therefore,theintegrationof F NL TeffS Teff lnZ . (29) [L] ≡ − [L] =− [L] combinatorialsumsbecomescumbersome. Fortheequilibriumfreeenergywefind This approach corresponds to the classical Gibbs non- equilibrium statistical mechanics, when microstates of a Fensemble NL0 TeffSensemble Teff lnZensemble. (30) system are identified with the system’s microconfigurations ≡ − =− and macrostates of a system are identified with the system’s Insteadoffreeenergy,wecandefinetheactionoffreeenergyon macroconfigurations. However, we know that the probabilities afluctuation(thiswillremovetheasymmetryoftemperature;for forMarkovprocessesareassociatednotwiththestatesbutwith details,see[30,Section2.16]): thepathsamongthesestates(forthefirst-orderMarkovprocess, the probability of a path A-B-C-D is P(DC)P(CB)P(BA)P(A) A[L] ≡NL/Teff −S[L] =−lnZ[L] =−ln ZensembleW[eLn]semble . while to find the probability of the s|ystem|arrivin|g after (cid:16) (31(cid:17)) three steps at state D we should have three sums involved: FrontiersinPhysics|www.frontiersin.org 6 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach which all the broken fibers remain broken. This makes our systemaMarkovprocessoforder1.Forthetotalprocessfrom t0 0 to tν ν dt, we construct all possible chains of = = · microconfigurations. Each such chain, as a possible sequence of particular microconfigurations n0 , n1 ,..., nν , will be { } { } { } referredtoasamicropath n0 n1 ... nν (further, { } → { } → → { } wewillabbreviatethisnotationas n0 nν ).Forexample, { } → { } for a system with N 3 fibers, one of possible micropaths is = x xx xx xxx where symbol “|” {|||} → {| |} → { |} → { |} → { } denotesanintactfiberwhilesymbol“x”denotesabrokenfiber. Letusassumethatforamicropath n0 nν thesequence { }→{ } of the configurations has the evolution of the intact parameter L(t) L0,...,Lν.Thentheprobabilityofthismicropathis ≡ w{enn0se}→mb{lneν} = (1 − P)N|1L1|PN(L0−|1L1|) FIGURE2|ThegraphamongA,B,C,andDstates. ...(1 − P)N|1Lν|PN(Lν−1−|1Lν|) ν 1 P N(1−Lν) N Li 1 − P i 1 − (34) = P P= (cid:18) (cid:19) P(DX3)P(X3 X2)P(X2 X1)P(X1), see Figure2). astheprobabilityofN|∆L|fibersfailingandofNL N(L | | | i i i 1 X1,X2,X3 1L ) fibers staying intact, where P P(s – Eε)≡is cons−tan−t. ThPerefore, it is much easier to find a distribution of the | i| ≡ This probability wensemble is dictated by the ensemble and, probabilities only for a macrogroup of paths than for all the in particular, by th{ne0}→va{lnuνe} of P. This constraint is a model pathsleadingtoamacrostate.IntheGibbsapproach,thestates input and acts similarly to the temperature prescribed in the ofasystemwerechosenasbases,althoughweseethateverything canonical ensemble. In the canonical ensemble, an external points to the fact that as bases we should choose not the states mediumdictatesthedistributionoftheprobabilitiesfordifferent but the paths [125, 126, 131–142]. We will see how we can paths but a system actually can be not in equilibrium with developsuchanapproachinthenextsection.Thebenefitofthis the dictated distribution and realizes some other probability approach will be that its bases will not include integrals of the distribution w for its paths. Therefore, we used here states for the previous evolution of the system but, in contrast, {n0}→{nν} the abbreviation “ensemble” to emphasize that this probability willbethisevolutionitself. distributioncorrespondstotherequirementsstatedregardlessof possiblefluctuations. STATISTICAL MECHANICS OF THE PATH As to the macropath [L0] [L1] ... [Lν] (further, → → → INTEGRAL APPROACH we will abbreviate this notation as [L0] [Lν]) we will refer → toasubsetofallmicropaths n0 nν correspondingtothe { } → { } In the previous sections and in our previous publications [21, specifiedevolutionoftheintactparameterL(t):[L0] [Lν] → = 22, 30], we identified microstates of a system with system’s n0 nν .Forexample,forthesystemwithN 3 mmiaccrrooccoonnfifigguurraattiioonnss.anHdomwaecvreor,stattheissofapapsryosatcehm wwiotrhkesydstewmel’sl {finb0}e→rsS{mn{ν}a:cLr{}noi}p=aLtih→[1]{ }[1/3] [1/3] [0]unitesmicropa=ths → → → onlyforequilibriumstatisticalmechanics.Fornon-equilibrium xx xx xxx , x x x x {|||} → {| } → {| } → { } {|||} → { | } → { | } → statisticalmechanics,weshoulddevelopanotherapproach[126]. xxx ,and xx xx xxx .Infact,macropath { } {|||} → { |} → { |} → { } As an illustration, we consider the case of the NFBM and statesthatthesystempassesduringitsevolutiontheprescribed assumethattheprocessconsistsofν timeintervalsofduration valuesofLbutdoesnotspecifyparticularmicrostates. dt,fromt0 0totν ν dt.Foreachtime-stepti asanorder Theprobabilitiesofmicropathscorrespondingtoamacropath parameterw=echoose=thev·alueoftheintactparameterLi atthis [L0] → [Lν] are all equal one to another and are given by time-step.Forthetotalprocesstheorderparameteristhetotal Equation(34);thenumberofthesemicropathsis historyoftheintactparameterL(t) L0,...,Lν.Weassumethat Lthewpirlolcneostsablewvaayrsiastbalretsinfrothmezeenrsoemda≡bmlea.gWeLe0a=ssu1m,soetthheatqburaonktietny g[L0]→[Lν] = (N 1L1 )!((NN(LL00)! 1L1 ))!. 0 | | −| | fibers cannot become intact again; therefore, increments of the (NL1)! . intactparameter1Li Li Li 1arealwaysnegative. (N 1L2 )!(N(L1 1L2 ))! ≡ − − | | −| | At each time-step ti, a particular system in the ensemble (NLν 1)! has its own value of the intact parameter Li and is in one ...· (N 1Lν )!(N(Lν−1 1Lν ))! (35) of the microconfigurations n corresponding to this intact | | − −| | i { } parameter.Thenextpossiblemicroconfigurationforthissystem asacombinatorialchoiceofN|1L|failedfibersamongNL i i 1 − at the time-step t can only be a microconfiguration in initial fibers. We can cancel (NL)! in the numerators and i 1 i + FrontiersinPhysics|www.frontiersin.org 7 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach (N(L -|1L|))!inthedenominatorstoobtain Asanexample,weagainconsideraveragingofL: i 1 i − ν (NL )! g[L0]→[Lν] = (NLν)!iν1(0N|1Li|)!≈ln(Lν)−NLνiY=1|1Li|−N|1Li|, hLiensemble(t) ≡ {n0}X→{nν}L{n0}→{nν}w{enn0se}→mb{lneν} Q= (36) L0, t t0 = L wensemble , t t (42) wherethenotation“≈ln”meansthatinthethermodynamiclimit ≡({n0}P→{nν} {ni} {n0}→{nν} = i = N allpower-lawmultipliersareneglectedincomparison L , t t 0 0 →+∞ = withtheexponentialdependenceonN. g Lwensemble , t t . Theprobabilityofthesystemhavingamacropath[L0]→[Lν] =([L0]P→[Lν] [L0]→[Lν] i {n0}→{nν} = i (tomoveamongmacroconfigurationswiththespecifiedL(t))is g[L0]→[Lν] Of course, in the thermodynamic limit, hLiensemble(t) coincides Wensemble wensemble g wensemble .(37) withthemostprobablemacropath: L ensemble(t) L(0)(t). [L0]→[Lν] = n 1 {n0}→{nν} = [L0]→[Lν] {n0}→{nν} Probability distribution (Equatihoni 34) is th≈e distribution X= dictated by the ensemble. However, we can also consider Maximization of Wensemble over all the possible macropaths other distributions w which do not obey the givesthemostproba[bLl0e]→m[aLcνr]opathL(0)(t): ensemble requirements.{nN0}e→xt{,nνa}s a fluctuation we consider the Wensemble max Wensemble . (38) cmoarcrerosppoanthdi[nLg0]to→th[iLsνm].aTchroepnauthmbiserggiv[Le0n]→b[Lyν]Eoqfumatiicornop(a3t6h)s. [L(00)]→[L(ν0)] =L1,...,Lν [L0]→[Lν] Again, to observe this macropath, we need to “isolate” our system in this macropath. This is achieved by considering the DependenceL(0)(t)isthedependenceofLwhichwewillobserve probabilitydistributionw suchthatonlythemicropaths generallyintheensemble. {n0}→{nν} The function Wensemble , given by Equation (37), is the belonging to the considered macropath [L0] → [Lν] have product of g [L0]→an[Ldν]wensemble . Both these functions non-zeroprobabilities: containexpon[eLn0]t→ial[lLyν]fastdepen{nd0}e→nc{neνs}onN(Nisinfiniteinthe tfpuhrneorcbmtaiboolndeyWmnaa[emLcn0rsi]eoc→mpbl[aiLlmetνh]it[h)L.a(0sT0)ha]evr→eefroyr[Len,(νa0ir)nr].otwAhesmwspaexaciwmeiluolmfseaelaltbpetahltoehwsm,, ttohhseet w{n0}→{nν} =(cid:26)1/g[L00]→,{n[L0ν}],→{n0{}n→ν}∈/{n[νL}0∈]→[L0[]L→ν] [Lν].(43) absolutewidthofthismaximumisproportionalto√Nwhilethe Here we assumed the equiprobability of all micropaths of the relative width is proportional to 1/√N. Therefore, the number systemisolatedinthemacropath.Actually,thismaynotalways of different macropaths [L0] → [Lν] under the width of this be true but depends solely on how we build fluctuations as an maximumhasapower-lawdependenceonN whilethenumber instrumentofourinvestigation. g[L0]→[Lν] ofmicropaths{n0} → {nν}correspondingtoeachof Theentropyoftheconsideredmacropathis thesemacropaths[L0] [Lν]hastheexponentialdependence → onN.Therefore,forthenormalizationofthefunctionWensemble withthelogarithmicaccuracyweobtain [L0]→[Lν] S w lnw lng . [L0]→[Lν] ≡− {n0}→{nν} {n0}→{nν} = [L0]→[Lν] 1 Wensemble Wensemble {n0}X→{nν} = [L0X]→[Lν] [L0]→[Lν]≈ln [L(00)]→[L(ν0)] (44) g wensemble . (39) ≡ [L(00)]→[L(ν0)] {n0}→{nν}∈[L(00)]→[L(ν0)] We should emphasize here that the entropy introduced is the “dynamical” entropy of the distribution of probabilities for the Thisprovides pathsandmustnotbeconfusedwiththeclassicalGibbsentropy g 1/wensemble . (40) associated with the distribution of probabilities for the states [L(00)]→[L(ν0)]≈ln {n0}→{nν}∈[L(00)]→[L(ν0)] (configurations). In our notations, the classical Gibbs entropy wouldbeS (t) w (t)lnw (t)astheaverageattime In other words, the probability of each micropath of the most {n} ≡ − n {n} {n} probablemacropath[L(00)]→[L(ν0)]hastheorderoftheinverted t over the probabilitieP{s}w{n}(t) of microconfigurations at this numberofthesemicropaths. time. On the contrary, the path entropy is associated with the Next, let us consider averagingover the spaceof all possible probabilities w of micropaths of the total process and {n0}→{nν} paths: cannot be attributed to system’s characteristics at a particular timet. (cid:10)f(cid:11)ensemble(t)≡{n0}X→{nν}w{enn0se}→mb{lneν}f{n0}→{nν}. (41) conIsf,ideinrspteraodbaobiflittyhedisptrroibbuatbiiolinty(Edqiustartiibounti3o4n) dwi{cnt0a}t→ed{nνb}y, twhee FrontiersinPhysics|www.frontiersin.org 8 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach ensemble,fortheensemblepathentropywefind: Since our system has two effective temperatures, instead of a free-energy potential we should define the action of the free- Sensemble wensemble lnwensemble energy potential to avoid the asymmetry of one temperature ≡ −{n0}X→{nν} {n0}→{nν} {n0}→{nν} chosen in favor to another. The action of the free-energy for a =− g[L0]→[Lν]w{enn0se}→mb{lneν}lnw{enn0se}→mb{lneν} = macropath[L0]→[Lν]canbedefinedas: [L0X]→[Lν] ν =−[L0X]→[Lν]W[eLn0s]e→mb[Lleν]lnw{enn0se}→mb{lneν}. (45) A[L0]→[Lν] ≡NLν/Tν +N iX=1Li−1!/TR −νS[L0]→[Lν] = Hdeepreen,dethnece ofunncNtioinncolnmwp{eannr0sie}s→mobn{lneνw}ithhatsheafunsclotiwonspgower-law =−ln(g[L0]→[Lν]exp −NLν/Tν −N i 1Li−1!/T !) aonnd Nw.{enn0sTe}→mheb{rlneeνf}orwe,hicthhehadveepenexdpeonnceentWialelynsemfabslte deapcet[nsL0d]→eans[cLeνas] =−lnZ[L0]→[Lν], X= R (49) delta-function around the most probab[Ll0e]→m[Laνc]ropath, and where Z[L0]→[Lν] is the partial path partition function [23, 125, 126]ofthismacropath: wefind: Sensemble ≈ −lnw{enn0se}→mb{lneν}∈[L(00)]→[L(ν0)] ≈lng[L(00)]→[L(ν0)] Z[L0]→[Lν] ν =S[L(00)]→[L(ν0)]. (46) = exp(−NLν/Tν −N Li−1!/T ) {n0}→{nν}X∈[L0]→[Lν] iX=1 R Inotherwords,theentropyofthetotalensembleofpathsequals ZensembleWensemble . (50) theentropyofthemostprobablemacropath. = [L0]→[Lν] Wecanrewritetheequilibrium distributionofprobabilities, Therefore,fortheactionofthefree-energyweobtain givenbyEquation(34),as w{enn0se}→mb{lneν} = Zens1emble exp −NLν/Tν −N ν Li−1!/T ! A[L0]→[Lν] =−ln(cid:16)ZensembleW[eLn0s]e→mb[Lleν](cid:17). (51) i 1 Hence, the action of the free-energy is directly related to the X= R (47) probability distribution of the fluctuations and the principle of itsminimizationisequivalenttotheevolutionfrompathsoflow N where Zensemble = 1PP , Tν = ln−1 (1−P)/P , probabilitytohighlyprobablepaths. and T ln−1P.(cid:16)It−is(cid:17)easy to see that(cid:8)Zensemble (cid:9)is In Gibbs equilibrium statistical mechanics, an equilibrium = − state is found as a minimum of a free energy potential. For the path partition function of the system: Zensemble R the microcanonical ensemble, the free energy potential is the ν = {pnl0a}yP→t{hnνe}reoxlpes(cid:26)−ofNeffLeνc/tTivνe−teNmp(cid:18)eiPr=at1uLrie−s.1(cid:19)W/eTsRe(cid:27)e.thTatνwahnedn PTiRs ncaengoatnivicealenentrsoepmyb−le,Sth≡e<freelnewn{enr}gy>p≡oteP{nnt}iwal{nis}ltnhwe{Hn}e;lmforhotlhtez close to unity (the probability of a fiber breaking is small), the energy F < Hn > T < lnwn > wn [Hn ≡ { } + { } ≡ n { } { } + temperatureT isinfinitebutpositive. Tlnw ].ForthecaseofgeneralensembleinGibP{b}sequilibrium The definitRion of the temperature Tν is, to some extent, statisti{cna}l mechanics, the principle of the minimization of the ambiguous. Indeed, we can rewrite Equation (34) for free energy potential always works because this potential is probabilitiesas always proportional to the minus logarithm of the probability w{enn0se}→mb{lneν} = (1 − P)N|1L1|PNL1...(1 − P)N|1Lν|PNLν dvaislitdribfourtitohnecoafsefloufcntuoant-ioeqnus.ilWibreiusmeemthecahtatnhiecssaasmweepll,rionncliypnleowis ν =(1 − P)N(1−Lν)PNiP=1Li. (48) wbuethfoavrethteopcaotnhsst.ructthefreeenergypotentialnotforthestates This gives Tν ln−1(1 P) which is different from So, for the path microcanonical ensemble (for the system the expression Tν= ln−1 (−1 P)/P above. However, the isolatedinamacropath),thenegativedynamicalentropy = − differenceisln(P)whichisnegligibleincomparisonwithln(1 P) (cid:8) (cid:9) − S <lnw > w lnw intThheelimtietmPp→era1tu.re T = −ln−1P of the system is − ≡ {n0}→{nν} ≡{n0}X→{nν} {n0}→{nν} {n0}→{nν(}52) complementary to the integral of the evolution of intact R ν tν plays the role of the free energy potential. This immediately parameter:N Li 1 N L(t)dt.Ifatthefinaltime-steptν i 1 − ∝ 0 providesthatallthemicropathscorrespondingtothismacropath thetotalsystemP=fails,L(tν) R0,thenthesingleorderparameter must be equiprobable. If we compare this result with our = tν hypothesis(Equation43),weseethatwewererightconsidering leftistheintegralN L(t)dt. micropathsasequiprobablewithinthemacropath. 0 R FrontiersinPhysics|www.frontiersin.org 9 February2017|Volume5|Article6 Abaimov Non-equilibriumDamagePhenomena:PathApproach For the path canonical ensemble, the role of the free energy at the most probable macropath. These equations could be potentialisplayedbytheactionofthefree-energy used as definitions of the temperatures. As both the entropy of a macropath S lng and the ensemble A ≡ ≡<H{n0}→{nwν}{n>0}→/T{n+ν}[<H{lnn0}w→{n{n0}ν→}/{Tnν+} >lnw{n0}→{nν}] seanmtreopfuynScetniosenmabllede[pL≈0e]n→dS[e[LLnν(0]0c)e]→=o[nL(ν0L)](t)[=La0n]→dln[LLgν(0[]L)((00t))]→re[sLp(ν0e)]cthivaevley,twhee {n0}X→{nν} obtain (53) whereH{n0}→{nν} issupposedtobethedynamicalHamiltonian 1 ∂S[L0]→[Lν] ∂Sensemble,i 1,...,ν 1and Htwwo{ehannmi0scie}th→imνsltba{olnpeonνpw}iaenna=rhstaeiepnmxppdpein(cid:0)es−rtsartHitobu{urcneto0i,}no→nsHi{sn({tνEno}0q/}fu→Ttaw(cid:1){tnio/oνZ}pn/eanT4rst7esm),beo=laefc.thhTechopisrrNroedLbsyνpan/boaTinmlνidtiiicen+asgl: TT1Rν == ∂S[NNL0∂∂]→LLνi[Lν](cid:12)(cid:12)(cid:12)(cid:12)(cid:12)(cid:12)00 ≈≈ ∂SNNen∂∂sLLem(i(ν00b))le. = − (58) (cid:12) N Li 1 /T , and does not correspond to the classical (cid:12) (cid:18)i 1 − (cid:19) This resembles the law of conservation of energy dE/T HamP=iltonian of thRe states because of being determined by the dQ /T dSensemble in the canonical ensemble—an equatio=n ← rulesofthememoryoftheprocess: of“dynam=ical”pathbalance: A ≡ ν {n0}X→{nν} ν N dL(i−0)1!/T +NdL(ν0)/Tν =dSensemble. (59) w{n0}→{nν}"NLν/Tν +N Li−1!/T +lnw{n0}→{nν}# Xi=2 R i 1 X= R (54) Differentials in this equation should be understood as the response of the most probable macropath to the change of the For the most probable macropath [L(00)] → [L(ν0)], the action externalboundaryconstraintP.Thisequationcouldbeobtained of the free-energy is A lnZ which [L(00)]→[L(00)] ≈ − [L(00)]→[L(00)] directly by differentiating (Equation 44) as the logarithm of coincideswiththeensemblepathactionofthefree-energy: Equation(36). ν Aensemble ≡ (NLν/Tν +N Li−1!/T )w{enn0se}→mb{lneν}−Sensemble = {n0}X→{nν} iX=1 R ν =− (−lnw{enn0se}→mb{lneν}−NLν/Tν −N Li−1!/T )w{enn0se}→mb{lneν} =−lnZensemble ≈ {n0}X→{nν} iX=1 R lnZ A . (55) ≈− [L(00)]→[L(00)] = [L(00)]→[L(00)] AtthepointofthemaximumofW[eLn0s]e→mb[Lleν],wehave shoTuoldfifinnddwthheenmtohsetdperroivbaatbivleesmoafctrhoeppartohba[Lb(0i0li)t]yo→fma[Lcr(ν0o)p],atwhes ∂Wensemble ∂lnWensemble (Equation 37) (or of the logarithm of this probability) with [L∂0L]→i [Lν](cid:12) =0or ∂[LL0i]→[Lν](cid:12) =0 (56) respecttoLiequalzero(Equation56).Thisprovides (cid:12)0 (cid:12)0 (cid:12) (cid:12) (where i 1 becau(cid:12)se we assume L to b(cid:12)e non-variable). (cid:12) 0 (cid:12) ≥ For lνnW[eLn0s]e→mb[Lleν] = lng[L0]→[Lν] + Aensemble − NLν/Tν − L(i0) =Pi,|1L(i0)|=(1 − P)Pi−1,i=1,...,ν. (60) N L /T ,wecanwritethat i 1 (cid:18)i 1 − (cid:19) P= R as an analog of the equation of state in equilibrium statistical T1 = ∂lngN[L∂0L]→i [Lν](cid:12)(cid:12)0 = ∂lngN[L∂(00L)](i→0)[L(ν0)],i=1,...,ν−1 meqeuTcahotaioinnnivcoes.sftpiBgaeatihtne.gthceobneshisatveinotr,owffleuschtuoautlidonpsr,owbeasbhlyoucldallfinitdtthhee R (cid:12)(cid:12) (57) secondderivativesoflnW[eLn0s]e→mb[Lleν]: and 1 ∂lng[L0]→[Lν] ∂lng[L(00)]→[L(ν0)] ∂2lnW[eLn0s]e→mb[Lleν] Ki,j, i,j 1,...,ν, (61) Tν = N∂Lν (cid:12)0 = N∂L(ν0) ∂Li∂Lj (cid:12)(cid:12)0 =− = (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) FrontiersinPhysics|www.frontiersin.org 10 February2017|Volume5|Article6
Description: