• Nenhum resultado encontrado

Braz. J. Phys. vol.30 número4

N/A
N/A
Protected

Academic year: 2018

Share "Braz. J. Phys. vol.30 número4"

Copied!
7
0
0

Texto

(1)

Strethed Exponential Relaxation and

Independent Relaxation Modes

R.M.C. de Almeida 1

, N.Lemke 2

,and I.A. Campbell 3;4

1

InstitutodeFsia,UniversidadeFederaldoRioGrandedoSul,

CaixaPostal 15051{91501-970,PortoAlegre,RS,Brazil

2

CentrodeCi^eniasExatasedaTerra {Unisinos

Av. Unisinos,95093022-000{S~aoLeopoldo{RS{Brazil

3

LaboratoiredesVerres,Universite deMontpellierII

++34095MontpellierCedex5,Frane

4

Laboratoire de PhysiquedesSolides,Universite ParisSud,91405 Orsay,Frane

Reeivedon15Otober,2000

WedisusstheoriginofstrethedexponentialrelaxationindisorderedIsingspinsystemsbywriting

themasterequationonthephasespae,andtheevolutionofloalandglobalspinautoorrelation

funtions,intermsofindependentrelaxationmodes, whihareeigenvetors ofthetimeevolution

operator. Inthissenseitisshownthatwhentherelaxationmodesarespatially deloalized, both

loalandglobalautoorrelationfuntionsmaypresentnon-exponentialrelaxation. Wealsoanalyze

resultsforrandomwalksonthedilutehyperube,whihmaybeassoiatedwiththephasespaeofa

disorderedIsingspinsystem. Asexpeted,theresultsshowastrethedexponentialrelaxationnear

theperolationtransition,sineitdealswithrandomwalksonafratalperolatinglusterdened

onalosedsurfae. Wearguethatthesametypeof topologyispresent intheavailableregionof

ongurationspaeinIsingspin-glasssystemsneartheglasstransition,sinethesesystemspresent

verysimilarrelaxationpatternsinthistemperaturerange.

I Introdution

Slowrelaxationnearaphasetransition presentsmany

interestingfeatures thatreetthenature ofthe

tran-sition. This phenomenon may be observed in many

dierent kindsof materials, ranging from the original

observations by Kohlraush for the deay of residual

harge ona Leiden jar [1℄, to suh theoretial models

as the trapmodel [2, 3℄, or random walkers ondilute

hyperubes of high dimensions [4-8℄, and spin glasses

above,orferromagnetsbelowtheirtransition

tempera-tures, struturalglasses, et. [Forareent review,see

Ref. [9℄℄.

Thereisaontroversythatbeginswiththeform of

therelaxation,andintensieswhenthedisussionturns

to thepossibleoriginsforthisbehavior. Twodierent

forms ofnon-exponentialrelaxationofamemory

fun-tion q(t) are often found in the literature, originating

either in theoretial models or invoked in analysis of

experimental data. Therstis

q(t)exp[ (lnt) d

(d 1)

℄; (1)

asproposedbyRanderiaetal. [10℄forshort-rangeIsing

spin-glasses with loal spin dynamis above the

spin-hasneverbeenobserved[11℄. Thepointisthatsystems

shouldrelaxforverylongtimestoreahtheasymptoti

regime,renderingdiÆultaonlusivestatementbased

onnumerialsimulations.

The seond form of non-exponential relaxation of

memoryfuntions istypiallygiven byastrethed

ex-ponentialoftheform

q(t)Cexp[ (t=)

℄ (2)

where isarelaxationtimethat divergesasthe

tran-sitionisapproahedand <1( =1is theaseofa

pureexponentialrelaxation). Thisexpressionhasbeen

foundonsistentwithmanyexperimentsondisordered

systems[9℄andnumerialsimulationsofdierentmodel

systems[2-8,12℄.

In this paper we address this ontroversy by

on-sideringthemasterequationrepresentingtheevolution

ofaspinglassin phasespae. Inthismanner, wean

writetheevolvingdensityfuntionintermsofthe

relax-ationmodesrelatedto thetransition matrix

eigenve-tors. Thetransitionmatrixontainsinformationabout

thedynamisofthesystem,givingtherateforthe

(2)

se-spinsystem anddisuss someof the previous

theoret-ialand numerial resultsfrom this pointof view. In

setion III, we write the master equation for random

walkersonadeimatedhyperube,andusethismodel

asananalogyforthe phasespaeaspinglass. Inthis

waytheglasstransitionisassoiatedwithaperolation

transitionandtheslowrelaxationisexplainedasthe

ex-petedslowingdownofdispersionofrandomwalkerson

perolationlustersembeddedonlosed surfaes. The

results are presented and disussed in setion IV and

nallyonsetionVweonlude.

II The master equation and

re-laxation modes

Two models have been proposed that may yield

strethedexponentialrelaxationintheappropriate

lim-its. The trap model [2℄ onsiders random walkers in

a d-dimensional spae where traps are randomly

dis-tributed. The memory funtion here is given by the

number of living walkers after agiven time. The

ex-petedlifetimeofthewalkers,thatis,theaveragetime

it takesfor awalkerto meetatrap, isdetermined by

the initial distane to the nearesttrap. Consequently

thedistributionofrelaxationtimesontributingtothe

memoryfuntion depends onthedistributionof linear

sizes of trap-free regions. Also, eah walker is

relax-ing in a given region, with adened relaxation time,

thatis,eahwalkerisinanindependentrelaxingmode

of the dynamis of the system. The result of

averag-ing over these independent relaxation modes gives a

strethedexponentialduetothepartiularformofthe

relaxation-timedistributionoriginatingin therandom

loalizationoftraps.

Anothertheoretial modelyielding strethed

expo-nentialrelaxationis thehierarhialmodel [3℄. Inthis

model the spinsare organizedin dierent hierarhial

levels, eah level ontaining a given number of spins.

Thespinsinagivenlevelareallowedtohange states

onlyaftersomeonditionhasbeensatisedbythespins

at the level immediately below. As a onsequene,

the relaxationtimesare alwaysinreasing for

inreas-inglevels,sinesuperior levelsspinsmustwaitforthe

spinsatlowerlevelstorelaxbeforetheyanstart

relax-ing. Eahspinlevelthenanbeviewedasan

indepen-dentmodewithadenedrelaxationtime. Thememory

funtion is the spinautoorrelationfuntion and eah

spinrelaxationisessentiallyruledbyonlyone

indepen-dentmode. Theauthors of this model propose a

spe-i onstraint on the evolutionof upperlevelsyields

strethedexponentialrelaxationfortheglobalmemory

funtion.

Thekeyissueinbothmodelspresentedaboveisthe

fat that the (global) memory funtion is an average

over loal memory funtions that relax exponentially

al memory funtions are measuring the evolution of

independent relaxationmodesof thesystem. We an

verify this statementby onsidering a system with N

Isingspins,inwhihthemonitoredglobalmemory

fun-tionisgivenas

q(t)= 1 N N X i=1 S i (t)S i (0) (3)

and then writing a master equation, dened over the

phasespaeofthesystemas

( ~ S;t) t = X f ~ S 0 g h Q( ~ S; ~ S 0 )( ~ S 0

;t) Q( ~ S 0 ; ~ S)( ~ S;t) i (4) where Q( ~ S 0 ; ~

S) is the transition rate from

ongura-tion ~

S toonguration ~

S 0

and( ~

S;t)istheprobability

of nding the system in the onguration ~

S at time

t. The transition matrix ontains information about

the dynamis of the systemand is dierent for

dier-entmodels. Itisonvenienttoexpressthedynamisin

a Dira-like notation. Writing the aboveequation for

disrete time, the stateof the systemanbeformally

obtainedas

j(t)i=e t

j(t=0)i (5)

where j(t)i represents the probability of nding the

systemin eahpossibleongurationat time t,and

isgivenby

( ~ S; ~ S 0

)=Q( ~ S; ~ S 0 ) Æ( ~ S ~ S 0 ) X f ~ S g Q( ~ S ; ~ S) (6)

Theeigenvalueequationofthetransitionmatrixis

j k i= k j k i (7)

wherek=1;2;:::;2 N

andtheeigenvetorsfj

k igform

a basis for the spae of the solutions of the evolution

equation. Oneanalsobuildanotherbasis,onsidering

thestatesj ~

Sithatrepresentastatewherewehave

prob-abilityoneofndingthesystemin ~

S=(S

1 ;S

2 ;:::;S

N )

andzeroinanyotheronguration. Inthisase,

h ~ Sj ~ S 0 i=Æ(

~ S ~ S 0 ) (8) and ( ~

S;t) = h ~

Sj ~

(t) i. The stationary solution(s) of

Eq.(4) are the eigenstate(s) of Eq.(7) with eigenvalue

zero,normalizedsuhthat

X f ~ Sg h ~

Sj=0i=1 (9)

while for all other eigenstates, with Re() < 0, the

abovesumgiveszero. Itmeansthat someofthe

om-ponentsh ~

Sj6=0iarelessthanzero. Thishastwomain

(3)

rep-the evolution operator provided it ontains a

station-ary omponent suh that all omponents h ~

Sj ~

(t)i of

the physial state are non-negative and (ii) the role

playedbythesenonstationaryomponents,inrelation

to the stationarystates, is to addprobabilityto some

ongurationswhilesubtratingprobabilityfromother

ongurations,onservingtotalprobability. Thesetwo

features are valid due to the probability onservation

guaranteedbytheevolutionequation. Thestateofthe

systemattimet maythenbewrittenas

j(t)i= 2 N X k =1 e k t j k ih k

j(t=0)i (10)

whereitbeomeslearthattheeigenvaluesofthe

tran-sition matrixgivetherelaxationmodesof thesystem.

Letting

k

= 1=

k

, theaboveequationbeomes

j(t)i= 2 N X k =1 e t=k j k ih k

j(t=0)i: (11)

Now, Eq.(11) gives the relaxation of the system as a

wholeintermsofitsindependentmodes. Thequestion

now is to translate this relaxation of the probability

densityof thesystemdened overitsphasespaeinto

therelaxationofthespinautoorrelationfuntionthat

is the ommonly used memory funtion for Ising spin

systems. Forthat wersttake

j(t=0)i=j ~ S 0 i (12) where ~ S 0

istheinitialongurationofthesystem. We

thentaketheprodutofEq.(11)withh ~

Sj,multiplyby

S

i and S

0

i

,that are thevaluesof thei-th spinin

on-gurations ~ S and ~ S 0

,respetively,andaverageoverall

ongurations,toobtaintheloal spinautoorrelation

funtion: q i (t)= 2 N X k =1 e t= k f i ( k ; ~ S 0 ) (13) with f i ( k ; ~ S 0 )= X f ~ Sg S i S 0 i h ~ Sj k ih k j ~ S 0 i (14)

Finally, the global spin autoorrelation funtion

maybewritten as

q(t)= 2 N X k =1 e t= k g( k ; ~ S 0 ) (15) with g( k ; ~ S 0 ) = X f ~ Sg 1 N N X i=1 S i S 0 i ! h ~ Sj k ih k j ~ S 0 i = 1 N N X f i ( k ; ~ S 0 ) (16)

Both equationspresentastationary part,

orrespond-ingto

i

=1(

i

=0)andarelaxationpartrelatedto

smallerrelaxationtimes(

i <0).

Now,onsideringtherelaxingomponentforthe

hi-erarhial model, whih is a spin system so that the

abovememoryfuntion applies,weobservethat

essen-tiallyeah spinrelaxes aordingto its level

indepen-dent mode, that is, the sum over dierent relaxation

modes k splits into sums for dierent levels.

Conse-quently, eah spin is expeted to relax exponentially,

but the sum overindividual spin relaxationfuntions

mayyieldastrethedexponentialformforsuitable

on-straints.

In more general spin systems, it has been argued

that some unfrustrated spin lusters may form, suh

thattherelaxationwithinagivenlusterisexponential,

withrelaxationtimes dened by theluster size. The

sumofthese dierentmodesofexponentialrelaxation

withanappropriatelustersizedistributionouldyield

a global strethed exponential behavior. These

lus-ters would berelated to theGriÆths singularities[13℄,

and hene spatial inhomogeneities would be

responsi-ble for the observed slow relaxation. However, this

modelimpliesthateahspinshouldrelaxexponentially

(eah luster orresponding to an independent

relax-ationmode),andspatialorrelationbetweenloalspin

relaxation modes would signal the appearane of

un-frustratedlusters. Glotzeretal.[12℄monitoredthe

re-laxationoftwoandthreedimensionalIsingspinglasses

andfoundstrethedexponentialrelaxationfor

individ-ualspins,thatis,thesumintheexpressionfortheloal

autoorrelation funtion q

i

(t) in Eq.(13) also yields a

strethedexponential,implyingthatvariousrelaxation

modesontributetotherelaxationofindividualspins.

Infat,itisthestrutureofg(

k ;

~

S 0

)andf

i ( k ; ~ S 0 )that

governs the ontribution of eah relaxation mode for

theglobalandloalspinorrelationfuntions,dening

theiroverallrelaxationbehavior. Forexample,ifthe

re-laxationmodesarenotspatiallyloalized,q(t)ouldin

priniplepresentstrethedexponentialrelaxationwith

agiven and,andsoouldtheloalrelaxation

fun-tionswithdierentvaluesfortheseparameters.

Dier-ent valuesof and haveindeed beenfoundGlotzer

etal. [12℄.

III Ising spin systems and the

hyperube

From Eq.(4) and the evolution equations for the spin

autoorrelationfuntion,itbeomeslearthatitis

in-terestingto investigate the originsof non-exponential

relaxationbyfousingonthephasespaeofthesystem.

Infat,thephasespaeofasystemwithN Isingspins

is the set of the verties of a hyperube of N

(4)

onperolationlustersembeddedonatspaesyielda

sublinear diusion equation hr 2

i t

[14℄, while

ran-dom walkersonlosed surfaes suh as a hypersphere

giveapureexponentialdeayoftheaverage

autoorre-lation funtion, hos((t))i exp( t=), one is led to

theonjetureproposedbyCampbell[4℄thatdispersion

ofrandomwalkersonperolationlustersembeddedin

a losed surfae suh as the hyperube should follow

hr 2

iexp[ (t=)

℄, with !1=3and !1 asthe

perolationthresholdisapproahed.

Inreentsimulationsofaperolationluster

embed-ded inthe surfaeofahypersphere,Jund,Jullien and

Campbell [15℄ show that strethed exponential

relax-ationisdiretlyonnetedtothefratalnatureofthese

lusters. Campbell'sonjeturegoesfurther,by

onsid-eringtheimpliationsoftheseresultstotherelaxation

of suh disorderedsystemsas, for example,Ising spin

glasses. TheideaistoregardtherelaxationofanIsing

spin glass oupled to a heat bath as a random walk

throughtheavailablestatesinphasespae. This

high-dimensional,losedspaehasitslandsapehangedas

thetemperatureapproahesthetransitionfromabove,

sine the number of available states dereases in this

proess.

Nearthetransitionitisonjeturedthat thesetof

available states forms a perolating luster, and that

right at the transition point the giant luster breaks

intomanynitelusters, representingthemany

dier-entfreeenergyvalleystypialofspinglassesbelowthe

glass transition temperature. As temperature is

low-ered towards the transition, the topology of the

lus-ter beomes more labyrinthine, slowing down the

re-laxation towards equilibrium. Observe that this

pi-ture onsiders the relaxation of a density funtion on

the hyperube where eah siteis apossible

ongura-tion of the system. As the dierent verties hange

theiroupation density, theloal and globalspin

au-toorrelationfuntionshangeaordingly,and itmay

bepossiblethat theyaresimultaneouslyinuenedby

manydierentrelaxationmodes.

The simplest ase of the above senario has been

onsideredbyCampbellandollaborators[4-7℄through

MonteCarlosimulationsofarandomwalkeronan

N-dimensional hyperube in whih some of the verties

have been deleted or prohibited to the walker. There

is aritial fration p

of allowedor oupied verties

belowwhihnoperolationlusterisformed. The

sim-ulations,whihwereperformedforoupationfrations

pp

, monitoredtheaverageHammingdistane q(t)

betweentheurrentandinitial positions ofthewalker

asafuntionoftime,whihanbeassoiatedwiththe

global spin autoorrelation funtion. The results are

ompatible with Campbell's onjeture, but the noise

at small values of q(t) for longtimes did not allow a

deisiveonlusion. Analternativeproedureisto

on-sidertheevolutionoftheprobabilitydensityofnding

ter equation of the system[8℄, whih has yielded very

preiseresults.

Startbyimaginingahyperubein(high)dimension

N withafrationpofits2 N

vertiesoupiedat

ran-dom. Theritialperolationonentrationisgivenby

p

=+

3

2

2

+ 15

14

3

+::: (17)

where = 1=(N 1). For p > p

there exists a

gi-ant spanning luster made up of oupied sites

hav-ing one ormore oupied sites as neighbors [16℄. For

p<p

there are onlysmalllusters (withlessthan N

elements).

Consideringeahonguration ~

Sasthebinary

rep-resentationofaninteger,thedensityfuntion( ~

S;t)

beomesthenafuntionoftwointegervariables,(;t),

andthediretiterationofthemasterequationmay

eas-ilybeimplementedonaomputer. Consideringvalues

ofpabovep

,thesystemstartsatagivenvertex

0

be-longingtotheinniteluster. Theprobabilitythatthe

randomwalkerjumpstoaneighboringsiteis1=Nifitis

`oupied'andzerootherwise,inwhihasethewalker

stays where it is. The disrete-time master equation

thenreads

(;t+1)=(;t)+

X

f 0

g

[W(; 0

)( 0

;t) W( 0

;)(;t)℄ (18)

withtransitionmatrixW( 0

;)onvenientlyredened

to desribe the jumping probabilities during disrete

timeintervals.

As theiteration proedure evolveswemonitor the

memory funtion ofthesystem,the spinauto

orrela-tionfuntion,alulatedas

q(t)= 1

N N

X

i=1 X

fg

(;t)S 0

i S

i

; (19)

whereS 0

i andS

i

arethei-thspinsoftheongurations

assoiated,respetively,with

0

and,that is

S 0

i =

(2 i 1

AND

0 )

2 i 2

1

S

i =

(2 i 1

AND )

2 i 2

1: (20)

The advantage of this proedure is that for eah

transition matrix,representingonerealizationofa

hy-perubewithoupationprobabilityp,suessive

iter-ationsof themasterequation providetheaverageover

all possiblerandom walksstarting from the same

ini-tialondition. Averagesoverdierentsamplesarestill

required, butthe aurayissigniantlyenhanedin

omparisonto MonteCarloalulations.

Verynearp

(5)

equa-applied. Inthisase,wetakeadvantageoftheredued

numberofsitesintheperolatinglusterand

diagonal-ize theevolutionoperator,obtain theeigenvaluesand

eigenvetorsandusethemdiretlyinEq.(15).

Astheoupation probabilitypdereasesfrom1.0

down to the perolation threshold, the oupied

ver-ties develop a fratal perolation luster: the vetor

state ~

(t)diuseson afratalaggregate. Inthis ase,

wemayexpetanomalousdiusion. Wethereforestudy

therelaxationoftheautoorrelationfordierentvalues

of p,abovetheperolationthreshold.

IV Results

For eah p studied, we onsidered 100 realizations of

the hyperube. Sine we expliitly solve the master

equation for this system we obtain exat results (to

within numerialroundingerrors)fortheaverageover

allpossiblerandomwalksforeahombinationofa

re-alizationofthehyperubeandonegivenstartingpoint.

Thereforebystudyingevenarelativelysmallnumberof

samplesweanobtainverygood estimatesforthe

av-eragerelaxationfuntion q(t). Themain problemwith

this approahis thatevenwiththeoptimizationused,

putimeandmemoryrestritedusto\low"dimensions

N, sinethenumberofvertiesin thephasespae

in-reases exponentiallywithN (2 N

). Foronveniene

wehosetoarryoutexpliitalulationsindimension

N = 16. 100 samples were studied for eah value of

p : p=0.073, 0.10, 0.24, 0.36 and 0.5, while 500

sam-ples werestudied forp=0:12. Forp=0.073,verynear

the perolation thresholdforN =16, wediagonalized

the evolution operator,while for the other values the

dynamialequationwasdiretlyiterated.

Before disussing the results, it is interesting to

point outthat we expet three relaxation regimes: i)

At short times q(t) will behave as 1 t where is

the average probability per unit time that a stepwill

betaken,i.e.,itisproportionaltothefrationof

ou-piedneighborsofasiteonthegiantluster. Thisisthe

leadingtermintheexponentialfuntionexp( (t)),so

at veryshorttimestherelaxationwillalwaysbe

expo-nential. ii). Atlongertimes,asthesystemexploresthe

labyrinthinegeometryofthegiantluster,aslow

relax-ationregimeappears. iii)Finally,atevenlongertimes,

nite-size eets beome measurable and a rossover

baktoanexponentialrelaxationisexpetedtosetin.

This last rossovertimehasbeenshownto depend on

sample size for systems where relaxation arises from

twoompetingexponentialproesses[17℄.

The results for q(t) are presented in two dierent

ways to verify whether the relaxation is following a

strethedexponentialformand thenobtainthetting

parameters. Fig. 1showsthedata obtainedforq(t)in

alog-logplotfordierentvaluesofp,togetherwiththe

in Fig. 2. The normalization parameter C is always

loseto 1. Itanbeseenimmediatelythatthetsare

ofexellentquality.

Figure1. Deay of the autoorrelation funtionq(t) ona

log-logplotfor dierent valuesof p,as listed intheinset.

Thesolidlinesorrespondtostrethedexponentialts,with

(p)and(p)asindiatedinFig. (2). Theerrorbars

or-respondtoanestimateoftheunertaintyofthepointsdue

tolimitedsampling.

Inamorestringenttestforthestrethed

exponen-tialt,Fig. 3showsthesamedata asbefore but now

in a log(-log(q(t))) versuslog(t) plot. Inthis kind of

plots pure or strethed exponential funtions appears

as straight lines, with the slope giving the exponent

. As expeted, allurvesshowdeviations towardsa

pureexponentialbehaviorforveryshortandverylong

times. At intermediate times, however, the strethed

exponentialbehaviorislearlymarkedbytheextended

straightlineregions. Also,bothFigs. (2)and(3)show

thatasp!p

!1=3and !1.

Figure2. Relaxationtime(p)(irles) andstrethed

(6)

Figure3.Amorestringenttestofthestrethedexponential

behaviorofq(t)isthelog[-log(q(t))℄againstlogtplot.The

dierentvaluesofp arelisted inthe inset. Thesolidlines

orrespond to astrethedexponentialt,with and as

indiatedinFig.(2).

Finally,Fig. 4showsthevalueoftheloalslopeof

q(t)inthelog(-log(q(t)))versuslog(t)plot,thatis,this

gurepresentstheinstantaneousvalueoftheexponent

. The three regimes (exponential - strethed

expo-nential-exponential)maylearlybespotted, withthe

minimumvaluefor theinstantvalueof approahing

1/3asthetransitionisapproahed.

Figure4. Theinstantaneous(t)versuslog(t).Foraperfet

strethedexponentialbehavior the urvesshould be

hori-zontal lines. Theurvesshowthat as weapproahp

the

plateausgetlarger.

In summary, solutions of the master equation

strongly onrm earlier numerial work on random

walkers on the hyperube, but to muh higher

prei-sion[4-7℄.

V Conlusions

Thequestionmayberaisedastowhetherthestrethed

exponential is the true limiting long time relaxation

timerelaxationshouldbedominatedbylarge,ompat,

non-frustrated, isolated lusters of spins for

tempera-turesbelowthe GriÆths transition[10, 18,19℄, but no

numerialevidenehaseverbeenfoundfortheonsetof

thisregime[18,20,21℄. Theprobabilityofenountering

largeunfrustratedlustersinsamplesofthesizes

stud-iednumeriallyanbeestimatedandismirosopially

small;thusanyluster-dominatedregimewould

orre-spondto tinyvaluesofq(t) inhuge samples,and sois

unattainable inpratiefor numerialorexperimental

studies.

Insystems with quenhed-in disordersuh as spin

glasses,simulationsshowthatthereissomespatial

het-erogeneityof relaxationtimes, but that therelaxation

ofindividualspinsisgenerallystronglynon-exponential

[11, 12℄, in agreementwiththe fat that foreah spin

manydierentmodesmaybeontributingtoits

relax-ation, and dierent linear ombinations of relaxation

modesareestablishedfordierentspins,dependingon

initial onditions and on the spatial deloalization of

modes, as shown by the expression for the struture

funtions f

i (

i ;

~

S 0

) in Eq.(14). Inthis sense, loal

re-laxationratesareexpeted tobeheterogeneousandit

may well happen that dierent sites exhibit the most

rapid relaxation at dierent times. In spin-glass

sys-tems, where the probability of large unfrustrated

re-gions is very small, the independent relaxing modes

should notbeexpeted to be loalized, in ontrast to

systems where there are either independent relaxing

lusters or independent hierarhial levels. Hene, in

thepreseneofsomequenhed-instatiinhomogeneity,

heterogeneity an be expeted to be mainly dynami

and the argument for an overall fratal phase spae

topologyleadingtothestrethedexponentialglobal

re-laxationisnotaeted.

In onlusion, using a master equation approah

for random walks on the dilute hyperube, high

pre-ision resultshavebeenobtained ompatible with the

strethed exponentialbeingthe exat funtional form

for thedeayin the limitofinnitely high dimension.

Heterogeneitiesin loal relaxationforms naturally

ap-pear when the autoorrelation funtions are written

in terms of the independent modes of relaxation of

the density funtion dened on the hyperube. The

strethedexponentialappearsasaonsequeneof

fra-tality in a losed spae, both for global and loal

re-laxation funtions: the sum over simple exponentials

yielding astrethed exponential is then also valid for

the loal autoorrelation funtions when the

indepen-dentrelaxationmodesarenotspatially loalized. The

resemblane between the dilute hyperube relaxation

patternandtherelaxationatuallyobservedin

numer-ial studies of spin glasses or experiments on glasses

abovethe freezing temperature strongly suggeststhat

in physial systemsthe independent relaxationmodes

(7)

a fratal struture in this temperature range. For

sums suh as those in Eqs.(13) and (15) to give rise

to strethed exponentials,the weightof eah dierent

relaxationtimeomponentmust havesome

harater-istissuh that,forexample,asteepest-desent

alu-lation ofthesesumsyieldsastrethedexponential. As

astrethedexponentialappearsforthehyperubenear

the perolation threshold, as indeed it isexpeted for

randomwalkersoverperolationlustersembeddedon

losedsurfaes,wearguethatthesametypeoftopology

should bepresent in the available regionof Ising

sys-tems phasespaeneartheglass transition,sinethese

systemsalsopresentverysimilarpatternsofrelaxation

in thistemperaturerange.

Aknowledgements

Thisworkhasbeenpartiallysupported byF

renh-Brazilian Cooperation Program CAPES-COFECUB,

and by Brazilian agenies FAPERGS and CNPq. We

thankD. Stauerforvaluablesuggestions.

Referenes

[1℄ R. Kohlraush, Pogg. Ann. Phys. Chem. 91, 179a

(1854).

[2℄ M.D. Donsker and S.R.S. Varadhan, Commun. Pure

Appl.Math.32,721(1979),P.GrassbergerandI.

Pro-aia,J.Chem.Phys.77,6281(1982),J.C. Rasaiah,

J.Zhu,J.B.Hubbard,andR.J.Rubin,J.Chem.Phys.

93,5768(1990).

[3℄ R.G.Palmer,D.L.Stein, E. Abrahams,and P.W.

An-derson,Phys.Rev.Lett.53,958(1984).

[4℄ I.A. Campbell, J. Phys. (Frane) Lett. 46, L1159

(1985).

[5℄ I. A. Campbell, J.M. Flesselles, R. Jullien, and R.

Botet,J.Phys.C:SolidStatePhys.20,L47(1987).

[6℄ I.A.Campbell,Europhys.Lett.21,959(1993).

[7℄ N. Lemke and I. A. Campbell, Physia A 230, 554

(1996).

[8℄ R. M.C. de Almeida, N. Lemke and I. A. Campbell,

EuropeanPhysialJournalB,inpress.

[9℄ J.C.Phillips,Rep.Prog. Phys.59,1133(1996).

[10℄ M.Randeria,J.P.SethnaandR.G.Palmer,Phys.Rev.

Lett.54,1321(1985).

[11℄ T.Komori,K.HukushimaandH.Takayama,J.Phys.

So.Japan64,4418 (1995).

[12℄ S.C.Glotzer,N.Jan,T.Lookman,A.B.MaIssa and

P.H.Poole,Phys.Rev.E57,7350 (1998).

[13℄ R.B.GriÆths,Phys.Rev.Lett.23,17(1969).

[14℄ S.Alexanderand R.Orbah,J.Phys.Lett.43,L625,

(1982),Y.Gefen,A.AharonyandS.Alexander,Phys.

Rev.Lett.,50,77(1983).

[15℄ P.Jund,R.JullienandI.A.Campbell,tobepublished.

[16℄ P.EordosandJ.Spener,Comput.Math.Appl. 5,33

(1979); K.Weber, Elek. Inf.Verarb. Kybern. 22, 601

(1986).

[17℄ A.Bunde,S.Havlin,J.Klafter,G.Gra,andA.

She-hter,Phys.Rev.Lett.78,3338(1998).

[18℄ A.J.Bray,J.Phys.A22,LL81(1989).

[19℄ F.Cesi,C.MaesandF.Martinelli,Comm.Math.Phys.

188,135(1997)and189,323(1997).

[20℄ A.T.Ogielski,Phys.Rev.B32,7384(1985).

Imagem

Figure 2. Relaxation time  (p) (irles) and strethed ex-
Figure 4. The instantaneous (t) versus log(t). For a perfet

Referências

Documentos relacionados

Neste trabalho o objetivo central foi a ampliação e adequação do procedimento e programa computacional baseado no programa comercial MSC.PATRAN, para a geração automática de modelos

Ousasse apontar algumas hipóteses para a solução desse problema público a partir do exposto dos autores usados como base para fundamentação teórica, da análise dos dados

Despercebido: não visto, não notado, não observado, ignorado.. Não me passou despercebido

Caso utilizado em neonato (recém-nascido), deverá ser utilizado para reconstituição do produto apenas água para injeção e o frasco do diluente não deve ser

i) A condutividade da matriz vítrea diminui com o aumento do tempo de tratamento térmico (Fig.. 241 pequena quantidade de cristais existentes na amostra já provoca um efeito

Peça de mão de alta rotação pneumática com sistema Push Button (botão para remoção de broca), podendo apresentar passagem dupla de ar e acoplamento para engate rápido

didático e resolva as ​listas de exercícios (disponíveis no ​Classroom​) referentes às obras de Carlos Drummond de Andrade, João Guimarães Rosa, Machado de Assis,

Isto é, o multilateralismo, em nível regional, só pode ser construído a partir de uma agenda dos países latino-americanos que leve em consideração os problemas (mas não as