• Nenhum resultado encontrado

Braz. J. Phys. vol.30 número1

N/A
N/A
Protected

Academic year: 2018

Share "Braz. J. Phys. vol.30 número1"

Copied!
15
0
0

Texto

(1)

Nonequilibrium Kineti Ising Models:

Phase Transitions and Universality

Classes in One Dimension

Nora Menyhard

and Geza

Odor

ResearhInstitute forSolidState PhysisandOptis,

H-1525Budapest,P.O.Box49,Hungary

Researh InstituteforTehnialPhysisandMaterialsSiene,

H-1525 Budapest,P.O.Box49, Hungary

Reeived1Deember1999

NonequilibriumkinetiIsingmodelsevolving undertheompetingeetofspinipsatzero

tem-peratureandKawasaki-typespin-exhangekinetisatinnitetemperatureT areinvestigatedhere

inonedimensionfromthepointofviewofphasetransitionandritialbehaviour. Branhing

anni-hilatingrandomwalkswithanevennumberofospring(onthepart oftheferromagneti domain

boundaries),isadeisiveproessinformingthesteadystateofthesystemforarangeofparameters,

inthefamilyofmodelsonsidered. Awidevarietyofquantitiesharaterizetheritialbehaviour

ofthesystem.Resultsofomputersimulationsandofageneralizedmeaneldtheoryarepresented

anddisussed.

I Introdution

The Ising model is a well known stati, equilibrium

model. Its dynamial generalizations, the kineti

Ising models, were originally intended to study

relax-ationalproessesnearequilibriumstates[1,2℄. Glauber

introdued the single spin-ip kineti Ising model,

while Kawasaki onstruted a spin-exhange version

for studying the ase of onserved magnetization. To

hek ideas of dynami ritial phenomena [3℄ kineti

Ising models were simple enough for produing

ana-lytial and numerial results,espeially in one

dimen-sion,wherealsoexatsolutionsouldbeobtained,and

thushaveproventobeausefulgroundfortestingideas

and theories [4℄. On the other hand, when attention

turnedtonon-equilibriumproesses,steadystates,and

phasetransitions,kinetiIsingmodelswereagainnear

athand. NonequilibriumkinetiIsingmodels,inwhih

the steady state is produed by kineti proesses in

onnetion with heat baths at dierent temperatures,

have been widely investigated [5-8℄, and results have

shown that various phasetransitions are possible

un-der nonequilibrium onditions, even in one dimension

(1d) (forareviewseetheartilebyRazinRef. [9℄.

Alsointheeldoforderingkinetis,universalityis

strongly inuenedbythe onservedornon-onserved

harater of the order parameter. The saling

be-areofentral importane. Kineti Isingmodelsoera

usefullaboratoryagain toexplore thedierentfators

whihinuene theseproperties[10℄.

Mostofthesestudies,however,havebeenonerned

with the eets the nonequilibrium nature of the

dy-namismightexertonphasetransitionsdrivenby

tem-perature. In some examples nonloal dynamis an

generatelong-rangeeetiveinterationswhihleadto

mean-eld(MF)-type phasetransitions(see, e.g.,Ref.

[9℄).

A dierent line of investigating nonequilibrium

phasetransitions has been via branhing annihilating

randomwalk(BARW) proesses. Here partiles

ho-sen at random arry out random walks (with

proba-bility p) and annihilate pairwise on enounter. The

inrease of partiles is ensured throughprodution of

n ospring, with probability 1 p. Numerial

stud-ieshaveledto theonlusionthat ageneralfeature of

BARW is a transition as a funtion of p between an

ative stationary state with non-zero partile density

and an absorbing, inative one in whih all partiles

areextint. Theparityonservation ofpartilesis

de-isiveindeterminingtheuniversalitylassofthephase

transition. In the odd-n ase the universalitylass is

thatofdiretedperolation(DP)[11℄,whiletheritial

(2)

numer-andTretyakov[13℄. Thisnewuniversalitylassisoften

alled parity onserving(PC)(however,other authors

allittheDI[14℄orBAWelass[15℄);wewillalsorefer

to itbythisname. A oherentpitureofthis senario

isprovidedfromarenormalizationpointofviewin[16℄.

TherstexampleofthePCtransitionwasreported

byGrassbergeret al.[17,18℄,whostudied probabilisti

ellular automata. These 1d models involve the

pro-esses k ! 3k and 2k ! 0 (k stands for kink), very

similar to BARW with n = 2. The two-omponent

interating monomer-dimer model introdued by Kim

andPark[14℄representsamoreomplexsystemwitha

PCtypetransition. Otherrepresentativesof thislass

arethethree-speiesmonomer-monomermodelsof[15℄,

andageneralizedDomany-KinzelSCA[19℄,whih has

twoabsorbingphasesandanativeone.

AlassofgeneralnonequilibriumkinetiIsing

mod-els(NEKIM)withombinedspinipdynamisatT =

0andKawasakispinexhangedynamisatT =1has

beenproposedbyoneoftheauthors[20℄inwhih,fora

rangeofparametersofthemodel,aPC-typetransition

takesplae. Thismodelhasturnedouttobeveryrih

inseveralrespets. AsomparedtotheGrassbergerCA

models, in NEKIM the rates of random walk,

annihi-lation and kink-produtionanbeontrolled

indepen-dently. Italsooersthesimplestpossibilityofstudying

the eet of a transition whih ours on the level of

kinks upon the behaviour of the underlying spin

sys-tem. Thepresentreviewisintendedtogiveasummary

oftheresultsobtainedviaomputersimulationsofthe

ritialpropertiesof NEKIM [21-25℄. Dynamial

sal-ingandageneralizedmeaneldapproximationsheme

havebeenappliedintheinterpretation. Somenew

vari-ants ofNEKIM arealso presentedhere, togetherwith

new high-preisiondata for some of the ritial

expo-nents.

II Non- equilibrium kineti

Ising model (NEKIM)

A. 1d Ising model

BeforegoingintothedetailsofNEKIM,letusreall

somewellknownfatsabouttheone-dimensionalIsing

model. Itisdened onahainoflengthL;the

Hamil-tonianinthepreseneofamagnetieldH isgivenby

H=3D J P

i s

i s

i+1 H

P

i s

i ,(s

i

=1). Themodel

doesnothaveaphasetransitionintheusualsense,but

whenthetemperatureT approaheszeroaritial

be-haviour is shown. Speially, p

T = e

4J

k T

plays the

roleof T T

T

in1d,andintheviinityofT =0ritial

exponentsanbe dened aspowersofp

T

. Thus, e.g.,

theritialexponent oftheoherenelengthisgiven

through /p

T

. Conerning statis,from exat

so-lutions,keepingtheleading-ordertermsforT !0,the

ritialexponentsof(k

B

T times)thesuseptibility,the

oherene length and the magnetisationare known to

be

s

= = 1=2, and

s

= 0, respetively. We an

say thatthe transitionis of rstorder. Fisher's stati

salinglaw

s

=d 2

s

isvalid.

The Ising model possesses no intrinsi dynamis;

with the simplest versionof Glauberkinetis (see

be-low)itis exatlysolvableandthedynami ritial

ex-ponentZ denedviatherelaxationtimeofthe

magne-tizationas =

0

Z

hasthevalueZ=2. (Here

0 isa

harateristitimeoforderunity.)

Turningbaktostatis,ifthemagnetielddiers

from zero themagnetization is alsoknownexatly; at

T =0thesolutiongives:

m(T =0;H)=sgn(H): (1)

Moreover,for 1andH=kT 1theexatsolution

reduesto

m2h; h=H=k

B

T: (2)

Insalingformonewrites:

m

s

g(h

) (3)

whereisthestatimagnetiritialexponent.

Com-parisonofeqs. (2)and(3)resultsin

s

=0and=.

ItislearthatthetransitionisdisontinuousatH =0

aswell[i.e.,uponhangingthesignofH inEq. (1)℄. In

the following theorder of limits will always bemeant

as: (1)H !0,andthen(2)T !0.

B. Denition of the model

A general form of the Glauber spin-ip transition

ratein one-dimension forspin s

i

sittingat sitei is [1℄

(s

i =1):

W

i =

2 (1+Æs

i 1 s

i+1 )

1

2 s

i (s

i 1 +s

i+1 )

(4)

where =tanh2J=kT , withJ denotingthe oupling

onstantintheferromagnetiIsingHamiltonian, and

Æarefurtherparameterswhihan,ingeneral,also

de-pend on temperature. (Usually the Glauber model is

understood asthespeialaseÆ=0, =1,while the

Metropolismodel[26℄isobtainedbyhoosing =3=2,

Æ= 1=3). Therearethreeindependentrates:

w

same =

2

(1+Æ)(1 ) (5)

w

oppo =

2

(1+Æ)(1+) (6)

w

indif

(3)

wherethesubsriptssame,et.,indiatethethree

pos-sible neighborhoods ofagivenspin(""";#"#and ""#,

respetively). InthefollowingT =0willbetaken,thus

=1,w

same

=0and ,Æareparameterstobevaried.

The Kawasaki spin-exhange transition rate of

neighbouringspinsis:

w

i;i+1 (s

i ;s

i+1 )=

p

ex

2 (1 s

i s

i+1 )[1

2 (s

i 1 s

i +s

i+1 s

i+2

)℄: (8)

d

AtT =1(=0)theaboveexhangeissimplyan

unonditionalnearestneighborexhange:

w

i;i+1 =

1

2 p

ex [1 s

i s

i+1

℄ (9)

where p

ex

istheprobabilityofspinexhange.

The transition probabilities in Eqs. (4) and (9)

are responsible for the basi elementary proesses of

kinks. Kinksseparatingtwoferromagnetially ordered

domainsanarryoutrandomwalkswithprobability

p

rw /2w

indif

= (1 Æ) (10)

while twokinks at neighbouringpositions will

annihi-latewith probability

p

an /w

oppo

= (1+Æ) (11)

(w

same

isresponsibleforreationofkinkpairsinsideof

ordereddomainsat T 6=0).

In aseof thespin exhanges, whih also at only

at domainboundaries,theproessofmainimportane

here is that a kink an produe two ospring at the

nexttimestepwithprobability

p

k !3k /p

ex

: (12)

The above-mentionedthree proesses ompete, and it

depends onthe valuesof the parameters , Æ and p

ex

whattheresultofthisompetitionwillbe. Itis

impor-tanttorealizethattheproessk!3kandevelopinto

propagatingospringprodutiononlyifp

rw >p

an ,i.e.,

the new kinks are able to travelon the average some

lattie points away from their plae of birth, and an

thus avoidimmediate annihilation. Itisseenfrom the

abovedenitionsthatÆ<0isneessaryforthisto

hap-pen. Intheoppositeasetheonlyeet ofthek!3k

proessontheusual Isingkinetisis tosoftendomain

walls. This heuristi argumentis supported by

simu-lations aswellastheoretialonsiderations, aswill be

C. Phase boundary of the PC transition

in NEKIM

Inallofourinvestigationswehaveonsidereda

sim-pliedversionoftheabovemodel,keepingonlytwo

pa-rametersinsteadofthreebyimposingthenormalization

onditionp

rw +p

an +p

k !3k

=1,i.e.,

2 +p

ex

=1: (13)

Intheplaneofparametersp

ex

and Æ=

pr w=pan 1

p

r w =p

an +1

the

phasediagram,obtainedbyomputersimulation,isas

shownin Fig. 1. There isalineofseond-orderphase

transitions, the order parameter being the density of

kinks. The phaseboundary was obtainedby

measur-ing(t), the density of kinks, startingfrom arandom

initial distribution, and loating the phase transition

pointsvia (t)/t

with =:28:01. The lattie

was typially L = 2000 sites; we generally averaged

over 2000 independent runs. The dotted vertialline

in Fig. 1 indiates the ritial point that we

investi-gated in greater detail; further ritial harateristis

weredeterminedalongthisline.

Figure1. Phase diagramofthetwo-parametermodel. The

ative phase lies above the boundary. The dotted

verti-alline indiatesthe ritial point that we investigatedin

greaterdetail.

Theline ofphasetransitionsseparatestwokindsof

steadystates reahable bythe systemfor large times:

(4)

in one of thepossibleferromagnetistatesof all spins

up or all spins down, while the ative phase is

disor-dered, from thepointof viewof theunderlying spins,

duetothesteadilygrowingnumberofkinks.

For determining the phase boundary, as well as

other exponents disussed below, we used the

sal-ingonsiderationsof Grassberger[17,18℄. The density

(x;t;)ofkinkswasassumedtofollowthesalingform

(x;t;)/t

(x 1=?

;t 1=

k

) (14)

Here measuresthe deviationfrom theritial

proba-bilityatwhihthebranhingtransitionours,

? and

k

areexponentsofoherenelengthsinspaeandtime

diretions,respetively. Bydenition,

k =

?

Z,where

Z is thedynami ritialexponent. (a;b) isanalyti

neara=0andb=0. UsingEq. (14)thefollowing

re-lationsanbededued. Whenstartingfromarandom

initialstatetheexponent haraterizesthegrowthof

theaveragekink densityintheativephase:

()= lim

t!1

(x;t;)/

(15)

while the derease of density at the ritial point is

givenby

(t)/t

: (16)

Theexponentsareonnetedbythesalinglaw:

=

k

=

?

Z (17)

The phase boundary shown in Fig. 1 was

identi-edusing Eq. (16)[20℄. Theinitialstatewasrandom,

withzeroaveragemagnetization. At theritialpoint

marked on the phase boundary we reently obtained

thevalue =:280(5). This point washosen in a

re-gionoftheparameters(p

ex

;Æ),wheretheeetof

tran-sientsin time-dependent simulationsis small. Critial

exponentswere measuredaroundthepoint: =0:35,

Æ

= 0:395:001and p

ex

=0:3. Thedeviation from

the ritial point was taken in the Æ-diretion. It

is worth noting that in all of our simulations of the

NEKIM rule, spin-ips and spin-exhanges alternated

at eah time step. Spin-ips were implemented

us-ingtwo-sublattieupdating,whileLexhangeattempts

were ountedas one time-stepfor exhangeupdating.

The numerial value of the phase transition point is

sensitive to the manner of updating (thus, e.g., with

random-sequential updating in both proesses, j Æ

j

dereasesbymorethantenperentatthesamevalues

of andp

ex ).

Conerning, wearriedoutquitereentlya

high-preision numerial measurement; the result is shown

determinedtheeetiveexponent

eff

(),whihis

de-nedastheloalslopeof()inalog-logrepresentation

betweenthedatapoints(i 1;i)

eff (

i )=

ln(

i

) ln(

i 1 )

ln

i ln

i 1

; (18)

where

i =Æ

Æ

i

,providinganestimate

=lim

!0

eff

(): (19)

As shown in Fig. 2, the eetive exponent inreases

slowly and tends towards the expeted PC value as

! 0. A simple linear extrapolation yields the

esti-mate =0:95(2).

Figure 2. Eetive exponent

eff

near the ritial point.

Simulations were performed on a 1d NEKIM ring of size

L=24000. Averagingwasdoneafterthe steadystatewas

reahedforonethousanddatapointsand10 4

independent

samples.

Thetransientsmentionedabovein onnetionwith

thetime-dependentsimulationsshowupforsmalltimes

and arepartiularly disernibleat thetwoendsof the

phaseboundary,wherethespin-ip andspin-exhange

proesseshaveverydierenttimesales. For this

rea-son, near p

ex

= 1:0, where we get loseto Æ =0 (the

Glauberase),itisverydiÆulttogetreliable

numer-ialestimates,evenfor;theexponenthasbeenfound

to grow slowly with time, from a value near zero at

smalltimes. Itisimportanttonotiethataordingto

Eq. (13),p

ex

=1is approahedtogether with !0;

thus pex

!1,asotherwiseÆ

=0annotbereahed.

Severalrunswithdierentparametervalues[inluding

aseswithouttherestrition,Eq. (13)℄,havebeen

per-formed in this Glauber limit; all show that the Æ =0

aseremains Ising-likeforallvaluesof andp

ex : the

exponent tends to 0:5 for late times. Wetherefore

onjetured[20℄thattheasymptoteofthephase

bound-aryisÆ=0forp

ex

=1:0,andthusthatp

rw >p

(5)

In a reent paper, using exat methods,

Mussaw-isade,SantosandShutz[27℄onrmedtheabove

on-jeture. Theseauthorsstudyaone-dimensionalBARW

model with an even number of nearest-neighbour

o-spring. Theystartwith spinkinetis;intheirnotation

D, and are the rates of spin diusion,

annihila-tion andspin-exhange, respetively, andtransformto

a orresponding partile (kink) kinetis to arrive at a

BARW model. (In theirnotation D = p

RW

, =p

an

and=2p

ex .)

Usingstandardeld-theoretitehniques,

Mussaw-isadeetal. deriveddualityrelations

~

=; (20)

~

D=+; (21)

~

=D (22)

whihmapthephasediagramontoitselfinanontrivial

way and divide the parameter spae into two distint

regionsseparatedbytheself-dualline :

D=+: (23)

Theregionsmappedontoeahotherhave,ofourse,the

samephysial properties. Inpartiular,theline =0

mapsontothelineD=andthefast-diusionlimitto

thelimit!1. Inthefastdiusionlimittheauthors

ndthatthesystemundergoesamean-eldtransition,

butwithpartile-numberutuationsthatdeviatefrom

those givenby the MF approximation. Sothe nature

ofthephasetransitioninthislimitisnotPCbutMF.

The observation that for largeD, any small brings

the system into the ativephase, together with

dual-ity, preditsaphase transition at D =, in the limit

!1. The exatresultof [27℄ onrms the

onje-ture-stemmingfromoursimulations-ontheloation

of the phasetransition point in thephase diagram of

NEKIMin thelimitp

ex

!1,whihisindeed atÆ=0.

TheduallimitD!1overstheneighbourhoodofthe

pointÆ= 1,p

ex

=0. Thisresultpreditsaninnite

slopeof thephaseboundary in this representation,at

this point,asisalsoapparentin Fig.1. Itisimportant

tonotiethatthedualitytransformationdoesnot

pre-serve the normalization ondition of NEKIM, eq(13),

used in allofournumerialstudies. Theself-dualline

of[27℄D=+,whihintermsofourparametrization

reads

Æ= 2p

ex

1 p

ex

(24)

doesnot onserve the above-mentioned normalization

ondition either. In the phase diagram of Fig. 1 it

orrespondstoalinewhihstartsat(0;0)andendsat

(1=3; 1)in the(p ;Æ)plane.

Finally, in ref. [27℄ a proportionality relation

be-tween the (time-dependent) kink density for a

half-lledrandominitialstate,andthesurvivalprobability

of twosingle partiles in an initially otherwise empty

systemhasalso beenderived. Conerning the

respe-tiveritialexponentsandÆ(thelatterbeingdened

through the ritial behaviour of the survival

proba-bility P(t) / t Æ

), they havebeen shown to oinide

withintheerrorofnumerialsimulationsin[24℄,as

re-quiredbysaling.

D. Long-range initial onditions

Thesymmetrybetweenthetwoextremeinitalases

in time dependent simulations (single seed versus

ho-mogeneousrandomstate) onerning thekink density

deay inspiredusto investigateinitial onditionswith

long-rangetwo-pointorrelationsintheNEKIMmodel.

Thisproedurewasshowntoauseontinuously

hang-ingdensity deayexponentsin aseof DP transitions

[28℄. In this way it has been possible to interpolate

ontinuouslybetweenthetwoextremeinitialases.

In the present ase we have started from initial

states ontaining an even number of kinks to ensure

kinknumberonservation mod 2and possessing

kink-kink orrelations of the form < k

i k

i+x

> / x (1 )

with lying in the interval (0;1). Here k

i

denotes a

kink at site i. These states have been generated by

thesameserialalgorithmasdesribedandnumerially

testedinref. [28℄.

Thekinkdensityinsurvivingsampleswasmeasured

in systemsofL=12000sites, forupto t

max

=80000

time steps; we observe good quality of saling overa

timeintervalof three deades (80;80000)(see Fig. 3)

[32℄.

Asone ansee, thekinkdensityinreaseswith

ex-ponent0:28for =0,where inprinipleonlyone

pairof kinksisplaedonthelattieowingto thevery

shortrangeorrelations. Withouttherestritionto

sur-vivingsampleswewouldhaveexpetedaonstant,the

above exponent orresponds to dividingthe kink

den-sity by the survival probability, P(t) / t Æ

and the

abovevalueisinagreementwithourformersimulation

resultsforÆ[24℄.

In theother extreme ase for =1 thekink

den-sitydeayswithwithanexponent 0:28,againin

agreement with our expetations in ase of a random

initial state [20℄. In between the exponent hanges

linearly as a funtion of and hanges its sign at

= 0:5. This means that the state generated with

= 1=2 is very near to the situation whih an be

(6)

Figure3.log(

k ink

(t))versuslog(t)inNEKIMsimulations

for initialonditions with=0;0:1;0:2:::;1(from bottom

totop).

Wementionedattheendofthelastsetionthatthe

dualitytransformation of ref.[27℄preditstheequality

oftheabovetwoextremalexponents(thoughitalso

fol-lowsfromsaling). Oneanmaketheonjeturethatin

greatergeneralitythe duality transformation may

on-net also ases with initial two-point orrelation

expo-nents: $(1 ).

E. Variants of the NEKIM model

1. Varying exhange-range model

NowwegeneralizetheoriginalNEKIMmodelby

al-lowingtherangeofthespin-exhangetovary. Namely,

Eq. (9)isreplaedby

w

i;i+k =

1

2 p

ex [1 s

i s

i+k

℄; (25)

where i isarandomly hosensiteand s

i

isallowedto

exhangewiths

i+k

withprobabilityp

ex

. Sitekisagain

randomly hosenin theinterval1R , R beingthus

therangeofexhange. Thespin-ippartofthemodel

willbeunhanged.Wehavearriedoutnumerial

stud-ies with this generalizedmodel [21℄ in order to loate

the lines of Ising-to-ative PC-type phase transitions.

The method of updating was as before. It is worth

mentioning, that besidesk !3k, the proessk !5k

an also our forR 3, and the newkink pairsare

notneessarily neighbors. The harater of the phase

transition line at R >1 is similar to that for R = 1,

exept that the ative phase extends, asymptotially,

downto Æ =0.

Figure 4. Phase diagram of NEKIM for Æ

(R ;p

ex ) and

a) R = 1, and b) R = 3, as a funtion of p

ex (with

2 =1 p

ex

),andfor) =0:35(p

ex

=0:3)asafuntion

of(R =1+R) 4

(fullline).

This isillustrated in Fig. 4, where besides R =1,

the ase R = 3is also depited: the ritial value of

Æ

is shown as a funtion of p

ex

[urves a) and b)℄.

Moreover, Æ

asafuntionofR isalsoshownat

on-stant =:35 . Theabsissahasbeensuitablyhosen

to squeeze the whole (innite) range of R between 0

and 1 and for getting phase lines of omparable size

(hene the power4 of R =(1+R ) in aseof urve)).

Thephaseboundarieswereobtainedbymeasuring(t),

thedensityofkinks,startingfromarandominitial

dis-tribution and loating the phase transition points by

(t) / t

with =0:27:04. Wetypially used a

system ofL =2000 sites and averagedover500

inde-pendentruns.

Besides the ritial exponents we have also

deter-minedthehangeof Æ

with1=Rnumeriallyatxed

=:35;p

ex

= :3. Overthe deadeof R = 4 40 we

havefound

Æ

2:0(1=R ) 2

; (26)

whih is reminisent of a rossovertype behaviour of

equilibrium and non-equilibrium phasetransitions [4℄,

herewith rossoverexponent1=2. It should be noted

here,thattogetlosertotheexpetedÆ

MF

=0,longer

hains (weused lvaluesupto 20000)and longerruns

(here up to t = 510 4

) would have been neessary.

The former to ensure l R [29℄ and the latter to

overome the long transients present in the rst few

deades of time steps. Inwhat follows wewill always

refertotheMF limitin onnetionwithp

ex

!1(i.e.,

p

ex

= !1), forthesakeofonreteness, butkeepin

(7)

2. Cellular automatonversionof NEKIM

Another generalization is the probabilisti

ellu-lar automaton version of NEKIM, whih onsists in

keepingtheratesgivenin eqs.(7)and presribing

syn-hronousupdating. Vihnia[30℄investigatedthe

ques-tion how well ellular automata simulate the Ising

model with the onlusion that the unwanted

feed-bak eets present at nite temperatures are absent

atT =0andthegrowthofdomainsisenhaned;

equi-librium isreahedquikly. Inthelanguageof Glauber

kinetis,thisreferstotheÆ0ase.

With synhronous updating, it turns out that the

T =0Glauberspin-ip partofNEKIM itselfleadsto

proesses of the type k ! 3k, and for ertain values

of parameter-pairs( ;Æ) with Æ < 0,a PC-type

tran-sition takesplae. For =:5the ritialvalueof Æis

Æ

= :425(5), while e.g., for =:35;Æ

= :535(5).

For j Æj >j Æ

j the phase is the ative one,while in

the opposite aseit is absorbing. All the

harateris-tiexponentsanbehekedwithamuhhigherspeed

than for theoriginal NEKIM. The phaseboundaryof

theNEKIM-CA in the( ; Æ)plane issimilar tothat

in Fig. 1exeptfor thehighestvalueof =1,Æ

=0

annot be reahed; the limiting value is Æ

= :065.

For the limit Æ = 0 Vihnia's onsiderations apply,

i.e., itisanIsingphase. Thelimitingase !0with

Æ

! 1is veryhard from the pointof viewof

simu-lations, moreoverthe ase = 0 is pathologial: the

initial spin-distribution freezes in. Theorresponding

phasediagramisshownin Fig. 5.

Figure5. PhasediagramoftheCAversionofNEKIM.The

phase transition pointshave been determinedinthesame

wayasfor Fig. 1.

3. NEKIM in a magneti eld

It has been shown earlier, in the framework of

a model dierent from NEKIM exhibiting a PC-type

phase transition[31℄, that upon breaking theZ

2

sym-ered. Inaseof theGlaubermodel asimilarsituation

ariseswiththeintrodutionofamagnetield. Inthe

preseneofamagnetieldtheGlaubertransitionrates

giveninSe. 2aremodiedso:

w h

indif

= w

indif (1 hs

i

); (27)

w h

oppo

= w

oppo (1 hs

i

); (28)

h = tanh( H

k

B T

) : (29)

Fig. 6showsthephase diagram of NEKIM in the

(h;Æ)plane[22℄,startingatthereferenePCpointfor

h=0usedinthispaper(Æ

= :395).

We haveapplied only random initial state

simula-tionsto ndthepointsof thelineof phasetransitions

(ritialexponents:=:17(2), =:26(2),

orrespond-ingtotheDPuniversalitylass,asexpeted.) Itisseen,

thatwith inreasing eld strengththe ritialpoint is

shiftedtomoreandmorenegativevaluesofÆ.

Finally, it is worth mentioning a further possible

varint of NEKIM - again without a magneti eld

-in whih the Kawasaki rate Eq. (8) is onsidered at

somenitetemperature,insteadofT =1,butkeeping

T =0in theGlauberdynamis. Asloweringthe

tem-peratureofthespin-exhangeproessatsagainstkink

prodution,theresultisthattheativephaseshrinks.

Forfurther detailsseeRef. [32℄.

Figure6. Phase diagram ofNEKIM inthe (h;Æ) planein

the preseneof anexternal magnetield. Parameters of

the transition probabilities: = :35;p

ex

=:3. Naturally,

thephasediagramanbedrawnsymmetriallyfornegative

(8)

III Generalized mean- eld

the-ory and Coherent anomaly

extrapolation

UsingtheCAversionofNEKIMdesribedinthe

previ-oussetionweperformedn-sitelustermean-eld

al-ulations(GMF)upto n=6,andaoherentanomaly

extrapolation(CAM), resultingin numerialestimates

for ritial exponents in fairly good agreement with

those of other methods. With this method we ould

also handle a symmetry-breakingeld that favors the

spin-ips inonediretion. ForthisSCAmodelwe

ap-plied the generalized mean-eld tehnique introdued

byGutowitz[33℄ andDikman[34℄ fornonequilibrium

statistial systems. Thisisbasedonthealulationof

transitionsofn-siteluster probabilitiesP

n (fs i g) as P t+1 n (s 1 ;:::;s

n

)=F(fP t

m (s

1 ;:::;s

m )g)=P

t n (s 1 ;:::s n ) (30)

where F is a funtion depending on the update rules

andthelastequationreferstooursolutioninthesteady

statelimit. Atthek-pointlevelofapproximation,

or-relationsarenegletedforn>k,that is,P

n (s

1 ;:::;s

n )

is expressed by using the Bayesian extension proess

[33,35℄.

P

n (s

1 ;:::;s

n )= Q j=n k j=0 P k (s 1+j ;:::;s

k +j ) Q j=n k j=1 P k 1 (s 1+j ;:::;s

k 1+j )

: (31)

Thenumberofindependentvariablesgrowsmoreslowly

than 2 n

1, owingto internal symmetries(reetion,

translation)oftheupdateruleandthe`blok

probabil-ityonsisteny'relations:

P

n (s

1 ;:::;s

n ) = X sn+1 P n+1 (s 1 ;:::;s

n ;s n+1 ); P n (s 1 ;:::;s

n ) = X s0 P n+1 (s 0 ;s 1 ;:::;s

n ):

The series of GMF approximations an now

pro-videabasisforanextrapolationtehnique. Onesuha

methodistheoherentanomalymethod(CAM)

intro-dued by Suzuki [36℄. Aording to the CAM (based

onsaling)theGMFsolutionforaphysialvariableA

(example the kink density)at agiven level (n)of

ap-proximation{in theviinity oftheritialpoint, Æ

{

anbeexpressedastheprodutofthemean-eld

sal-inglawmultipliedbytheanomalyfatora

n : A n =a n (Æ=Æ n 1) MF ; (32)

andthen-thapproximationofritialexponentsofthe

true singular behavior (

n

) an be obtained via the

salingbehaviorofanomaly fators:

a n n MF n (33)

wherewehaveusedthe

n =(Æ =Æ n Æ n =Æ )invariant

variable insteadof , thatwasintrodued tomakethe

CAMresults independent ofusing Æ

or1=Æ

oupling

([37℄). HereÆ

istheritialpointestimatedby

extrap-olationfromthesequeneapproximationsÆ n

. Sinewe

ansolvetheGMFequationsforn6,wehavetaken

intoaountorretionstosaling,anddeterminedthe

trueexponentswiththenonlinearttingform :

a n =b n MF n +

n MF+1

n

(34)

whereb andareoeÆientstobevaried.

Asdisussedin aprevioussetion,byduality

sym-metry thep

ex

=0axisis mapped into theÆ =0axis

andtheneighborhoodofthe(p

ex

=1;Æ=0)triritial

pointwhihanbewelldesribedbythemean-eld

ap-proximationstothe(p

ex

=0;Æ= 1)point. Therefore

this pointis also atriritial pointand themean-eld

resultsareappliablein theneighborhood.

A. Z

2

symmetriase

FirstweshowtheresultsoftheGMF+CAM

alu-lations withoutexternal eld. Forn=1,we havethe

traditionalmean-eldequationforthemagnetization:

dm=dt= Æ m(m 2

1) (35)

andforthekinkdensity:

d=dt=2Æ (1 3+2 2

) (36)

The stable, analyti solution of the kink density

ex-hibitsajump atÆ=0.

1 (1)=

1=2 : Æ<0

0 : Æ0

Then=2approximationgivesforthedensityofkinks,

(1), thefollowingexpression:

(9)

forÆ<0. ForÆ>0(1)=0,i.e., GMFstill preditsarst-ordertransition forÆ=0;thejump sizein (1) at

Æ=0,however,dereasesmonotoniallywithdereasing ,aordingto eq. (37).

Figure7. (1)asafuntionofÆasobtainedfromGMF forn=3;::;6inaseofa: =:35,b: =:05

The GMF equations an only be solved

numeri-ally for n > 2. We determined the solutions of the

n=3;4;5;6approximationsforthekinkdensityati).

=0:35(Fig. 7a)andofthen=3;4;5approximations

at ii). =0:05(Fig. 7b).

Asweansee,thetransitionurvesbeome

ontinu-ous,withnegativevaluesforÆ n

n

denotesthevalueof

Æ inthen-th approximationfor whihthe

orrespond-ing (1) beomes zero). Moreover, j Æ n

j inreases

with growing n values. As inreasing n orresponds

todereasingmixing,i.e.,dereasingp

ex

,thetendeny

shownbytheaboveresultsisorret.

Figs. 8a) and 8b) show a quantitative - though

onlytentative-omparisonbetweentheresultsofGMF

and the simulated NEKIM phase diagrams. The

ob-tainedGMFdataforÆ n

,orrespondington=3;4;:::;6

( = 0:35) are depited in Fig. 8a) as a funtion of

1=(n 3), together with results of simulations. The

orrespondene between n and p

ex

washosen as the

simplestoneivableone. (NotethatÆ

6=0isobtained

rstforn=4.) Thesimulatedphasediagramwas

ob-tained without requiring the normalization ondition,

Eq. (13), at xed =0:35. In this ase, the Æ

=0

limit, of ourse, is not reahed and a purely

seond-order phase transition line an be ompared with the

GMFresults(fornvalueswherethelatteralsopredits

aseond-ordertransition).

Figure 8. Comparison between simulation results and GMF preditions. GMF results for Æ

are plotted as afuntion of

1=(n 3),whilesimulationresultsfor Æ

aredepitedasafuntionof p

ex

atonstant ,withR=3, =0:35inase(a),

(10)

Simulationsfor R = 3were found to lead to j Æ

j

valueslowenoughtotGMF data. The(polynomial)

extrapolation of GMFdata to n! 1(orresponding

to p

ex

=0, i.e.,plain spin-ip), alsoshownin Fig. 8,

ould be expeted to approah Æ = 1. That this is

not ase may be due to the irumstanes that upon

inreasingni) GMFstartsherefrom arst-orderMF

phase transition,whih ii) beomesseond-order, and

iii) GMFshould endupat thedualpointdisussed in

Se. II.C, while this symmetry is not inluded in the

appliedapproximationproedure. Fig. 8(b)showsthe

n = 3;4;5 results for = 0:05, whih are ompared

nowwiththeR=1simulationdata.

TheresultsofourGMFapproximationusefulfora

CAManalysisarethen=4;5;6data,whilethen=3

result an be taken to represent the lowest-order MF

approximation(withÆ

MF

=0)foraontinuous

tran-sition (no jump in

3

). For Æ

we use a polynomial

extrapolation.

In the n = 3 approximation the exponent =

1:0064,thus

MF

1. Consequently, asthetable

be-low shows, the anomaly fator doesnot depend on n.

This means,aordingto Eq. (33), that theexponent

isestimatedtobeequaltotheMFvalue '

MF =1.

TableI.CAMalulationresults

n

n

n

4 2:49043 0:01083

5 1:81022 0:01074

6 1:45766 0:01079

B. Broken Z

2

symmetryase

ThetransitionprobabilitiesofNEKIM,inthe

pres-eneofenexternalmagnetieldHhavebeengivenin

eqs.(27)-(29), in the previoussetion. Now we extend

the method for the determination of the exponent of

theorder-parameterutuationaswell:

()=L(< 2

> <> 2

) n

: (38)

The traditional mean-eld solution (n = 1) results in

stable solutionsforthemagnetization :

m

1 =

h

Æ

; if Æ<0 and h 2

=Æ 2

<1 (39)

m

1

= sgn(h); otherwise: (40)

andforthekink-onentration:

1 (1)=

1

2 (1 (

h

Æ )

2

); if Æ<0andh 2

=Æ 2

<1 (41)

1

(1)=0; otherwise: (42)

Forn>1thesolutionsanagainonlybefound

nu-merially. Inreasing the order of approximation, the

ritialpointestimatesÆ n

shifttomorenegativevalues

similarly totheH =0ase. Thelim

n!1 Æ

n

(h) values

havebeendeterminedwithquadratiextrapolationfor

h =0:01;0:05;0:08,and 0.1. The resulting urvesfor

n

(Æ) and

n

(Æ)are shown in Figs. 9(a) and (b),

re-spetively,for theaseofh=0:1 in dierentordersn

oftheGMFapproximation.

Naturally, these urves exhibit a mean-eld type

singularityattheritialpoint:

n

n (Æ=Æ

n

1)

MF

(43)

n

n (Æ=Æ

n

1)

MF

; (44)

with

MF

= 1 and

MF

= 1. The results of the

CAM extrapolation for various h-values are shown in

thetablebelow:

As to the aseh =0, weould not determine the

exponent

n

beause the low level GMF alulations

resulted in disontinuous phase transition solutions {

whih weannot usein theCAMextrapolation{and

sowehadtoofewdatapointstoahieveastable

non-lineartting. Higher-orderGMFsolutionswouldhelp,

butthatrequiresthesolutionofasetofnonlinear

equa-tions with morethan 72 independent variables. This

problemdoesnotourforh6=0;theaboveresults-

be-ingbasedonthefullsetofapproximations(n=1;:::;6)

-arefairlystable.

Table2. CAMalulationresults

h 0.0 0.01 0.05 0.08 0.1 DP PC

1.0 0.281 0.270 0.258 0.285 0.2767(4) 0.94(1)

(11)

Figure9. Thekinkdensity(a)and(b)theseondmomentofthekinkdensityintheneighbourhoodoftheritialpointÆ

(h)

forh=0:1. Theurvesfromrighttoleftorrespondton=1;:::;6(levelofGMFalulation). Thepointsweredetermined

witharesolutionof10 5

inÆ,inordertobeable toextratCAManomalyoeÆients.

IV Domain growth versus

luster growth properties;

Hypersaling

Intheeld ofdomain-growthkinetisithaslongbeen

aeptedthatthesalingexponentofL(t),the

hara-teristidomainsize,isequalto1=2iftheorder

parame-terisnotonserved. Intheaseofa1dIsingspinhain

of lengthL, the struture fator at the ferromagneti

Bragg peak is dened as: S(0;t) = L[< m 2

> <

m > 2

℄; m = 1

L P

i s

i , (s

i

=1). If the onditionsof

validityofsalingarefullled [38℄then

S(0;t))/[L(t)℄ d

; L(t)/t x

(45)

where now d = 1 and x = 1=2. Another quantity

usually onsidered [38℄ is the exess energy E(t) =

E(t) E

T (E

T

istheinternalenergyofamonodomain

sampleatthetemperatureofquenh),whihinourase

isproportionalto thekink density

(t)= 1

L <

X

i 1

2 (1 s

i s

i+1 )>;

whihfollows

(t)/ 1

L(t) /t

y

(46)

with y=1=2intheGlauber-Isingase, expressingthe

well known time dependene of annihilating random

walks (y = in the notation of setion 2.). Wealso

measured S(t) at the PC point and found power-law

behaviourwithanexponentx=0:575(5)[22℄.

Theseond harateristigrowthlengthatthe PC

pointisthelustersize,denedthroughthesquare-root

of <R 2

(t) >/ t z

. The latteris obtainedby starting

e.g.[12℄), or from a single kink [18,20℄. Both

length-exponentsare,howeveronnetedwithZ,thedynami

ritial exponent, sine at the PC transition the only

dominant length is the (time-dependent) orrelation

length. Itwasshownin[18℄and[39℄that z

2 =

1

Z

follows

fromsalingforone-kinkandtwo-kinkinitialstates,

re-spetively. In our simulationswealso found [22℄ that

x=1=Z,within numerialerror,atthePCpoint,and

thus

x=z=2 (47)

follows. Exponentsof lustergrowthareonnetedby

ahypersalingrelationrstestablishedbyGrassberger

anddelaTorre[40℄forthediretedperolation

transi-tion. ItdoesnotapplytothePCtransitioninthesame

form [12℄, where the dependene on the initial state

(oneortwokinks)manifestsitselfintwoluster-growth

quantities: thekink-numberN(t;)/t

f(t 1

? Z

) and

the survival probability P(t;) / t Æ

g(t 1

? Z

) (This

Æ has, of ourse, nothing to do with the parameterÆ

of NEKIM). In systems with innitely many

absorb-ing states, where Æ and depend on the density of

partilesintheinitialonguration,ageneralised

sal-ingrelationwasfoundto replaetheoriginal relation,

2Æ+ =z=2,namely[41℄:

2( 0

+) 1

? Z

+2 0

=z 0

(48)

whih is also appliable to the present situation. In

eq(48) 0

is dened through lim

t!1

P(t;) /

0

and

thesalingrelation 0

=Æ 0

?

Z holds. A prime onan

exponent indiates that it may depend on the initial

state. ForthePCtransitiontheexponentzhasproven

(withinerror)tobeindependentoftheinitial

(12)

Inaseofasinglekinkinitialstateallsamples

sur-vivethusÆ 0

=0,andviatheabovesalingrelationalso

0

= 0. Eq(48) then beomes: 2

?Z +2

0

= z. For

t ! 1, however, lim

t!1

N(t;) /

has to hold (

the steady statereahed annot depend onthe initial

stateprovidedsamplessurvive). Thususingtheabove

saling form forN(t;), =

? Z

follows,and the

hy-persalinglawanbeastintotheform:

4

? Z

=z: (49)

Startingwithatwo-kinkinitialstate,however, 0

=

and 0

= 0 holds [12℄. With these valuesEq. (48)

leadsagaintoEq. (49). Asz=2=Z,Eq. (49)involves

onlysuhquantities,whiharedenedalsowhen

start-ing with a random initial state. Using Eq. (17), Eq.

(49)beomes

2y=1=Z =x (50)

Inthiswaythefatoroftwobetweenxandyappearing

atthePCpointbetweentheexponentsofaspin-bound

andakink-boundquantity(andwhihhasbeenfound

also for avarietyof ritial exponents, see [22℄) ould

beexplainedasfollowingfrom(hyper)saling.

V Eet of thePC transition on

theproperties of the

underly-ing spin system

Time-dependentsimulations,nite-sizesalingandthe

dynamiearly-timeMCmethod[43℄havebeenapplied

to investigate thebehaviourof the1dspinsystem

un-dertheeetofthePCtransition. Wefound[22℄that

withinerrorofthenumerialstudies,thedynami

rit-ialexponent of thekinks Z and that of the spinsZ

agree, and have the value: Z = Z

= 1:75. Thus in

omparison with the Glauber-Ising value Z = 2:0, Z

dereases, as do and : = = :444 (instead of

1/2). The PC point is the endpointof aline of

rst-orderphasetransitions(bykeepingp

ex

and xedand

hangingÆthroughnegativevaluestoÆ

),where=0

stillholds,asdoesFisher'ssalinglaw: =d 2. To

obtainthesevaluesthePCpointhasbeenapproahed

fromthediretionofnitetemperaturesofthespin-ip

proess,byvaryingp

T =e

4J=k T

.

Thedynamialpersistenyexponentandthe

ex-ponent [44℄ haraterisingthe two-time

autoorrela-tionfuntionofthetotalmagnetizationunder

nonequi-librium onditions havealso been investigated

numer-ially at the PC point. It has been found that the

PCtransitionhasastrongeet: theproessbeomes

non-Markovianand theaboveexponentsexhibit

dras-tihangesasomparedtotheGlauber-Isingase[23℄.

Theseresultstogetherwithritialexponentsobtained

earlierin[22℄ aresummarizedinTableIII.

s

Z

Glauber-Ising 0 1=2 1=2 2 1=4 1

PC :00(1) .444(2) :444(2) 1.75(1) :67(1) 1.50(2)

TableIII.SimulationdataforstatianddynamiritialexponentsforNEKIM

Wehavealsoinvestigatedtheeet ofritial

u-tuations at the PC point on the spreading of spins,

and found the analogue of ompat direted

perola-tion (CDP) [46℄ to exist. Compat direted

perola-tion is known to appear at the endpoint of the

di-reted perolation ritial line of the Domany-Kinzel

ellular automaton in 1+1 dimensions [46℄.

Equiva-lently, suhatransition ours at zerotemperature in

amagnetield h,uponhangingthesignofh,inthe

one-dimensionalGlauber-Isingmodel,withwellknown

exponentsharaterisingspin-lustergrowth[42℄.

in the sea of downward spins (s(j) = 1 for j 6= i),

the exponentsÆ

s ,

s and z

s

aredened at the

transi-tion point bythe power-laws governingthe densityof

1'sn

s /t

s

, thesurvivalprobabilityP

s (t)/t

Æ

s

,and

themean-squaredistaneofspreading<R 2

s

(t)>t zs

.

Weobtained=0,Æ=1=2andz=1,in aordwith

the hypersalingrelation [42℄ in the form appropriate

forompatlusters:

s +Æ

s =z

s

=2 (51)

(13)

ouldbeexpeted. Butbasisimilaritiesremain. Thus

thetransitionwhihtakesplaeuponhangingthesign

of themagnetield isofrst-orderanditsexponents

satisfy Eq. (51). Aordingly it an be termed as

'ompat',andweallittheompatparity-onserving

transition (CPC).Forthephasediagram,seeFig. 10.

On the basis of Eq. (3), the t-dependene of the

magnetization insalingformmaybewrittenas

m(t;h)t

s

Z

~ g(ht

Z

) (52)

Wehaveinvestigated theevolutionofthe

nonequi-libriumsystemfromanalmostperfetlymagnetized

ini-tialstate(or ratheran ensemble ofsuh states). This

state is prepared in suh a way that asingle up-spin

is plaed in the sea of down-spins at L=2. Using the

languageofkinks this orrespondsto the usualinitial

stateof twonearestneighbourkinksplaedatthe

ori-gin. At theritial point thedensity of spinsis given

as

n

s

(t;h)t

s

g

1 (ht

Z

) (53)

for the deviation of the spin density from its initial

value, n

s

= m(t;h) m(0). The results are

summa-rizedin TableIV.

s

s 0

s

Æ

s

z

s

NGI-CDP 0 :99(2) 1=2 :0006(4) :500(5) 1(=2=Z)

CPC :00(2) :45(1) :49(1) .288(4) :287(3) 1:14(=2=Z)

TableIV.Spin-lusterritialexponentsforNEKIMin amagnetieldTheNGI-CDPdataareresultsof

simulationsatthepointÆ=0,p

ex =0:3

Figure10. Phase diagramofNEKIM inthe(h; ~

Æ) plane.

ThehosenPC-pointisat ~

Æ= 0:395. Forsmallervaluesof

~

Æ,intheGlauber-Isingregime,thevertiallineonneting

thePCandthenonequilibriumGlauberIsing(NGI)points

(ath=0)onsistsofallCDPpoints,withitsharateristi

ritial exponents. The simulations around the PC

tran-sition weredone for the interval0 h 0:1. The other

NEKIMparameters,inthewhole plane,are: =0:35and

p

ex =0:3.

Conerningstatiexponents,onlythemagneti

ex-ponent remains unhanged under the eet of the

ritialutuations. Thisisnotanobviousresult(see,

e.g., theoherene-length exponentabove).

s 0

,

har-aterizingthelevel-ovaluesofthesurvivalprobability

of thespinlustersisdened through

lim

t!1 P

s

(t;h)/h s

0

; (54)

thisisanewstatiexponentonnetedwithothersvia

thesalinglaw

s 0

= Æ

s Z

(55)

whih is satised by the exponents obtained

numeri-ally,withinerror.

VI Damage-spreading in

vesti-gations

Whiledamagespreading(DS) wasintrodued in

biol-ogy [47℄ it has beome an interesting topi in physis

as well [48-50℄. Themain questionis ifdamage

intro-dued in a dynamial system survives or disappears.

Toinvestigatethistheusualtehniqueistoonstruta

pairofinitialongurationsthatdieratasinglesite,

and let them evolvewith the samedynamis and

ex-ternal noise. This method has been foundveryuseful

foraurate measurementsof dynamialexponents of

equilibriumsystems[51℄.

In[25℄weinvestigatedtheDSpropertiesof several

1dmodelsexhibitingaPC-lassphasetransition. The

orderparameterthatwemeasuredinsimulationsisthe

Hammingdistanebetweenthereplias:

D(t)= *

1

L L

X

i=1 js

i s

;

i j

+

(56)

wheres

i

andthereplias ;

i

maydenotenowspinorkink

variables.

If there is aphase transition point, the Hamming

distanebehavesin apowerlawmanneratthatpoint.

Inaseofinitialreplias dieringat asinglesite(seed

simulations)thislookslike:

D(t)/t

(14)

Similarly the survival probability of damage variables

behavesas:

P

s (t)/t

Æ

(58)

and the average mean square distane of damage

spreadingfrom theentersalesas:

R 2

(t)/t z

: (59)

Inaseof theNEKIM model wehave investigated

theDSonboththespinandkinklevels. Wefoundthat

the damage-spreading point oinides with the

riti-al point, and that the kink-DS transition belongs to

thePC lass. The spindamage exhibits a

disontinu-ousphasetransition,withompatlustersandPC-like

spreadingexponents. Thestatiexponentsdetermined

bynite-sizesaling areonsistentwith thoseof spins

of the NEKIM model at the PC transition point and

thegeneralisedhypersalinglawissatised.

Byinspetingourlargedataset forvariousmodels

exhibitingmultipleabsorbingstates,weareledtoa

-nal onjeture: BAWe dynamis and Z

2

symmetry of

the absorbingstates together formthe neessary

ondi-tiontohaveaPC-lasstransition.

VII Summary

Thepresentpaperhasbeendevotedtoreviewingmost

of the authors' investigations of a nonequilibrium

ki-netiIsingmodel(NEKIM).

NEKIM has proven to be a good testing ground

forideas,espeially onnetedwithdynamialsaling,

in the eld of nonequilibrium phase transitions. The

eldoveredherebelongstothelassofabsorbing-state

phasetransitionswhihareusuallytreatedonthelevel

ofpartiles, thebest known examplesbeingbranhing

annihilatingrandomwalkmodels. Thephasetransition

on the level of kinks in the Ising system belongs the

the so-alledparity-onserving universalitylass. The

NEKIM model oers the possibility of revealing and

larifying features and properties whih takeplae on

the level of the underlying spin system as well. The

introdution of a magneti eld in the Ising problem

has also led to understand dierent features, e.g.,the

hange of theuniversalitylassfrom PC toDPin the

ritial behaviorat (and in the viinity of) the phase

transitionofkinks.

Investigations of various ritial properties have

beenarriedoutmainlywiththehelpofnumerial

sim-ulations(time-dependentsimulations,nite-sizesaling

and dynami early-time MC methods have been

ap-plied). In addition, a generalized mean-eld

approxi-a limiting situation of the phase diagram of NEKIM,

where the usually seond-order phase transition line

ends at aMF-typerst-order point. Someof the

rit-ialexponents,espeially that oftheorder parameter,

ouldbepreditedtohighaurayin thisway.

Whilefromthesideofnumerisalotofinformation

hasaumulatedoverthelastveyearsonerningthe

PCuniversalitylass,furtherinvestigationsareneeded

at themirosopilevel.

Aknowledgements

Theauthors would liketothanktheHungarian

re-searhfund OTKA (Nos. T-23791,T-025286 and

T-023552)forsupportduringthisstudy. Oneofus(N.M.)

would like to aknowledge support by SFB341 of the

Deutshe Forshungsgemeinshaft, during her stay in

Koln, where this work wasompleted. G.

O

aknowl-edges support from Hungarian researh fund Bolyai

(No. BO/00142/99)aswell. Thesimulationswere

per-formed partially on the FUJITSU AP-1000 and

AP-3000 superomputers and Aspex's System-V parallel

proessingsystem(www.aspex.o.uk).

Referenes

[1℄ R.J.Glauber, J.Math.Phys.4,191(1963).

[2℄ See,e.g.,K.KawasakiinPhaseTransitionsand

Crit-ialPhenomena,editedbyC.DombandM.S.Green

(AademiPress,NewYork,1972),vol.2.

[3℄ R.A.Ferrell,N.Menyhard,H.Shmidt,F.Shwabl,

andP.Szepfalusy,Phys.Rev.Lett.18,891(1967); B.

I.HalperinandP.C.HohenbergPhys.Rev.177,952

(1969).

[4℄ Z.RazandR.K.PZia,Phys.Rev.E49,139(1994),

andreferenestherein.

[5℄ A. DeMasi, P.A. Ferrari,and J.L. Lebowitz, Phys.

Rev. Lett. 55 1947 (1985); J. Stat. Phys. 44, 589

(1986).

[6℄ J.M.Gonzalez-Miranda,P.L.Garrido,J.Marro,and

J.L.Lebowitz,Phys.Rev.Lett.59,1934(1987).

[7℄ J.S.WangandJ.L.Lebowitz,J.Stat.Phys.51,893

(1988).

[8℄ M. Droz,Z. Razand J.Shmidt,Phys.Rev.A 39,

2141(1989).

[9℄ Z. Raz, \Kineti Ising models with ompeting

dy-namis: mappings, steady states and phase

transi-tions,"inNonequilibriumStatistialMehanisinone

(15)

Univer-[10℄ S. J. Cornell, \One-dimensional kineti Ising

mod-els at low temperatures - ritial dynamis, domain

growth and freezing," in Nonequilibrium Statistial

Physis in one Dimension, edited by V. Privman

(CambridgeUniversityPress, Cambridge,1996).

[11℄ Variouspapersondiretedperolationaretobefound

inPerolationStruturesandProesses,editedbyG.

Deutsher,R.ZallenandJ.Adler,AnnalsoftheIsrael

PhysialSoiety,Vol.5(Hilger,Bristol,1983).

[12℄ I.Jensen,Phys.Rev.E50,3623(1994).

[13℄ H.TakayasuandA. Yu.Tretyakov,Phys.Rev.Lett.

68,3060(1992).

[14℄ M. H. Kimand H. Park, Phys.Rev.Lett. 73, 2579

(1994);H.ParkandS.Kwon,thisvolume.

[15℄ K.E.BasslerandD.A.Browne,Phys.Rev.Lett.77,

4094 (1996); K.E.Basslerand D.A.Browne,Phys.

Rev.E55,2552(1997).

[16℄ J. Cardy and U.Tauber, Phys.Rev.Lett. 77, 4780

(1996).

[17℄ P. Grassberger, F.Krause, and T. vonder Twer, J.

Phys.A:Math.Gen. 17,L105(1984).

[18℄ P. Grassberger, J. Phys. A:Math.Gen. 22, L1103

(1989).

[19℄ H.Hinrihsen,Phys.Rev.E55,219(1997).

[20℄ N.Menyhard,J.Phys.A:Math.Gen.27,6139(1994).

[21℄ N. Menyhard and G.

Odor, J. Phys. A 28, 4505

(1995).

[22℄ N. Menyhard and G.

Odor, J.Phys.A: Math. Gen.

29,7739(1996).

[23℄ N. Menyhard and G.

Odor, J.Phys.A: Math. Gen.

30,8515(1997).

[24℄ N. Menyhard and G.

Odor, J.Phys.A: Math. Gen.

31,6771(1998).

[25℄ G.

Odor and N. Menyhard, Phys. Rev. E 57, 5168

(1998).

[26℄ N.Metropolis,A.W.Rosenbluth,M.N.Rosenbluth,

A.H. Teller,andE.Teller, J.Chem.Phys.21,1087

(1953).

[27℄ K.Mussawisade,J. E.Santos, and G.M. Shutz, J.

Phys.A31,4381 (1998).

[28℄ H. Hinrihsen and G.

Odor, Phys. Rev. E 58, 311

(1998).

[29℄ K. K. Mon and K. Binder, Phys. Rev. E 48, 2498

(1993).

[30℄ G.Y.Vihnia,PhysiaD10,96(1984).

[31℄ H.ParkandH.Park,PhysiaA221,97(1995).

[32℄ G.

Odor, A. Krikelis, G. Vesztergombi and F.

Rohrbah, Proeedings of the 7-th Euromiro

work-shop on parallel and distributed proessing,

Fun-hal (Portugal), edited by B. Werner, (IEEE

Com-puter soiety press, Los Alamos, 1999); e-print:

physis/9909054

[33℄ H.A.Gutowitz,J.D.Vitor,andB.W.Knight

Phys-iaD28,18(1987).

[34℄ R.Dikman,Phys.Rev.A38,2588 (1988).

[35℄ G. Szabo, A. Szolnoki, and L. Bodos, Phys. Rev.

A44,6375 (1991);G.

Odor,Phys.Rev.E 51,6261

(1995);G.SzaboandG.

OdorPhys.Rev.E59,2764

(1994),andreferenestherein.

[36℄ M.Suzuki,J.Phys.So.Jpn.55,4205(1986).

[37℄ M.KolesikandM.Suzuki,PhysiaA215,138(1995).

[38℄ A. Sadiq and K. Binder, Phys. Rev. Lett. 51, 674

(1983).

[39℄ P.Grassberger,J.Phys.A22,3673(1989).

[40℄ P.GrassbergerandA.delaTorre,Ann.Phys.(NY)

122,373(1979).

[41℄ J.F.F.Mendes,R.Dikman,M.Henkel,and M.C.

Marques,J.Phys.A27,3019 (1994).

[42℄ R.DikmanandA.Yu.Tretyakov,Phys.Rev.E52,

3218(1995).

[43℄ R.E.Blundell,K.Hamayun,andA.J.Bray,J.Phys.

A25,L733(1992);Z.B.Li,L.Shulke,andB.Zheng,

Phys.Rev.Lett.74,3396 (1995); L. Shulke andB.

Zheng,Phys.Lett.A204,295(1995).

[44℄ S.N.Majumdar,A.J.Bray,S.J.Cornell,andC.Sire,

Phys.Rev.Lett.77,3704(1996).

[45℄ H. K. Janssen, B. Shaub, and B. Shmittman, Z.

Phys.43,539(1989).

[46℄ E.DomanyandW.Kinzel,Phys.Rev.Lett.53,311

(1984).

[47℄ S.A.Kauman,J.Theor.Biol.22,437(1969).

[48℄ M.Creutz,Ann.Phys.167,62(1986).

[49℄ H.Stanley,D.Stauer,J.KerteszandH.Herrmann,

Phys.Rev.Lett.59,2326(1986).

[50℄ B.Derridaand G.Weisbuh,Europhys.Lett.4,657

(1987).

Imagem

Figure 1. Phase diagram of the two-parameter model. The
Figure 2. Eetive exponent
Figure 3. log(
Figure 5. Phase diagram of the CA version of NEKIM . The
+5

Referências

Documentos relacionados

experiment is known whih learly reprodues the ritial exponents of direted

time version of ontat proesses, whih is alled the.. Domany-Kinzel model [2℄ and a related

In the particlewise case the existence of stationary product measures in the homogeneous phase implies that the critical density can be computed exactly, while for sitewise

The original idea of this model was to clarify whether, as in the TTP, PCP and other models without PC conservation, dynamic critical exponents will also depend on the initial

The asymptotic time behavior of the density of particles is derived by mapping the reaction-diusion process into the problem of the reunion of k random walkers bounded to move in

While the models with the two equivalent ab- sorbing states or two equivalent groups of absorbing states belong to the DI universality class, the mod- els with a single absorbing

tions and a oarse-grained eld theory, we show that the late stages of this proess exhibit dynami.. saling, haraterized by a set of saling funtions

It was also been veried [13] from Monte Carlo simulations that the ma- jority vote model dened on a square lattice has the same critical behavior, with respect to the static criti-