florida society for materials simulations reu … carlo simulation of defect diffusion in fcc...

13
Monte Carlo Simulation of Defect Diffusion in FCC Crystals Florida Society for Materials Simulations REU Program Research Report Chera Rogers West Virginia Wesleyan College Buckhannon, West Virginia Host: Dr. Anter El‐Azab Department of Scientific Computing Florida State University Tallahassee, Florida

Upload: phungthuan

Post on 21-Mar-2018

217 views

Category:

Documents


2 download

TRANSCRIPT

MonteCarloSimulationofDefectDiffusion

inFCCCrystals

Florida Society for Materials Simulations REU Program

Research Report

CheraRogers

WestVirginiaWesleyanCollege

Buckhannon,WestVirginia

Host:Dr.AnterEl‐Azab

DepartmentofScientificComputing

FloridaStateUniversity

Tallahassee,Florida

Abstract

Westudiedthediffusionofpointdefectsona2DsquarelatticebyMonteCarlo

method(Randomwalktheory).Inthisstudywehavestudiedtheeffectofensemble

sizeonacalculationofdiffusioncoefficient.Thesimulationsshowthatwhenwe

increasetheensemblesize,weconvergetostatisticallyreliableestimateofdiffusion

coefficient.Thisworkisafirststeptowardsunderstandingtheclusteringofdefects

insolids.

Introduction

Inaperfectcrystal,massandchargedensityhavetheperiodicityofthe

lattice.Theatomsinsolidsarrangethemselvesintoacrystallinestructure;perfect

crystalshavetheiratomsarrangedalongaperiodiclattice.Howeversolidsin

naturearenotperfect.Theyhavedefects.Thecreationofapointdefector

extendeddefectdisturbsthisperiodicity.Twoimportanttypesofpointdefectsare

vacancies,whereanatomismissingfromalatticesite,andinterstitials,wherean

atomisplacedbetweenlatticesites.Insolids,pointdefectsarescattered

throughoutthematerial,andtheirconcentrationisexponentiallydependentonthe

temperatureandtheenergyittakestoformthecrystal.Foraparticularmaterial

theformationenthalpy,Gfj,isconstant. Theconcentrationforpointdefectoftype

jisgivenbythefollowing.

C j expG j

f

kBT

Asyoucanseefromtheconcentrationequation,asweincreasethe

temperature,T,theconcentrationalsoincreases.Soinorderfortheatomsto

overcometheenergybarrierkeepingthemintheirlatticesites,weneedtoincrease

thetemperaturetocausetheatomstobecomemoreenergetic.

Inordertounderstanddefectsincrystalsweneedtounderstandhowthey

moveandspreadthroughthecrystal,soweneedtounderstanddiffusion.

Diffusion

Diffusionisthespreadofparticlesthroughrandommotionfromregionsof

highconcentrationtoregionsoflowconcentration.Diffusionisallaroundus,from

microscopicsystemstosituationsinoureverydaylives.Forillustration,imaginean

elevatoriscrowdedwithpeople(highconcentrationregion).Assoonasthe

elevatordoorsopen,thepeoplepushoutthroughthedoorsanddisperseinthe

emptyhall(lowconcentrationregion).Diffusionisalsopresentwhenmixingtwo

misciblefluids,asininkinwaterorcreaminyourcoffee.Intheseexamples,the

diffusingbodiesmoveinanydirectionandrandomly.Incontrast,inacrystalline

material,themovementofdiffusingdefectsisrestrictedbythesurroundinglattice.

AdolfFickfirststudiedtheprocessofdiffusion,andhedevelopedthe

mathematicalframeworktodescribethephenomenonofdiffusion.Heintroduced

theconceptofdiffusioncoefficient,D,andsuggestedalinearresponsebetweenthe

concentrationgradientandflux,J.ThisisknownasFick’sfirstlaw:

CDJ

Fick’sfirstlawissimilartothelawsgoverningothertypesofflowinnaturesuchas

Fourier’sLawandOhm’slaw.

Inthediffusionprocessthenumberofdiffusingparticlesisconserved,which

meansthatthedifferencebetweenthenumberofparticlesflowingintoaregionand

outofaregionresultsinanaccumulationofparticlesintheregion.Therateofnet

inflowtoaregiongivesthetimerateofchangeofconcentration;thisisthe

continuityequation.

t

CJ

BycombiningFick’sfirstlawwiththecontinuityequation,weobtainFick’ssecond

law,alsoknownasthediffusionequation.

C

t (DC)

DiffusionMechanismsinSolids

Diffusionincrystalshasdifferenttypesofmechanisms.Thesimplest

mechanismsareexchangeandringmechanisms.Exchangemechanismisthe

exchangeoflatticepositionsoftwoatomslocatedinadjacentsites.Ringmechanism

requiresthecoordinatedcirculatingmovementofthreetofiveatoms.However,

sincetheenergyrequiredfortheExchangeandRingmechanismsistoohighthey

arenotlikelytooccur.

OthertypesofdiffusionmechanismsareVacancy,Interstitial,and

Interstitialcymechanisms.Vacancymechanismoccurswhenanatomjumpsfrom

itslatticesitetoavacantsite.Interstitialmechanismiswhenanatombetween

latticesitesjumpstoanotherinterstitialsite.Interstitialcymechanismismuchlike

aninvasion;it’swhenaninterstitialatombumpsaneighboringlatticeatomoutof

itsplaceandthentakeitsplace.

Thediffusionofdefectsinthecrystalisgovernedbytheenergylandscape

seenbythedefects.Toillustratethis,thediagrambelowshowsthatinorderforan

atominsiteAtogettositeB,itmuchcrosstheenergybarrierofheightGM,called

themigrationoractivationenergy.Thehigherthebarrierthemoredifficultitisto

cross.ThelikelihoodofdiffusionisexpressedbythediffusivityD,andisdependent

onthemigrationenergyandthetemperatureT.

Tk

GDD

B

m

exp0

DisproportionaltotheBoltzmannprobabilityandD0isaproportionalityconstant.

RandomWalkTheory

Einsteinreasonedthatmoleculesarealwayssubjecttothermalmovements

ofstatisticalnatureduetotheirBoltzmanndistributionofenergy.Hederivedthat

themeansquaredisplacement,<R2>,ofarandomlymovingatomwasrelatedtothe

diffusioncoefficient,D.isthetimeelapsedfortherandommotion.Einstein’s

relationisgivenby

D R 2

4

Inordertocalculate<R2>,aseriesofrandomwalksofidenticalatomscanbe

analyzedaccordingtoRandomWalktheory.TheRandomWalktheoryreasonsthat

thediffusioninsolidsresultsfromparticlesjumpingfromsitetositerandomlyas

showninthisdiagram.

Inordertofindthetotaldisplacementbytheparticle,addalltheindividual

vectorstepstakenbytherandomlywalkingparticle.Thesquareddisplacement

wouldthentakethisform.

R r1l1

nstep

R2 rl2

l1

nstep1

rl rjJ l1

nstep

l1

nstep

Thisexperimentisrepeatedmanytimes,andanensembleaverageisfound.

R2 rl2 2

l1

nstep1

rl rJJ l1

nstep

l1

nstep

ThediffusioncoefficientDcanbere‐expressedinanotherusefulform.The

traveltimeisrelatedtotheaveragenumberofsteps<n>,thejumprate,andthe

coordinationnumberZ.

n

Z ,

BysubstitutingintoEinstein’srelation,Dcannowbewritinglikethis:

D R2

n

ThejumprateinthediffusionequationdependsontheactivationenergyG,the

attemptfrequency0,andthetemperatureT.

vo exp G

kBT

NumericalApproachtoCalculatingD

AMonteCarlomethodwasusedtosimulatetherandomwalkofdefectsona

2Dsquarelattice.Thejumpdistanceisalwaysthesame,butthedirectionofeach

jumpwasdecidedbythevalueofarandomnumbersampledfromauniform

distribution.Therandomwalkercangoalonganydirection[+x,‐x,+y,‐y]withequal

probability[1/4].ThisisshownintheflowdiagramfortheMatlabcode

implementationwrittenforthisproject.

MonteCarloSimulationResults

SamplesofParticleTrajectories

Thefollowingfiguresshowsnapshotsofrandomwalkon2Dlattice.Plotted

herearetheparticletrajectoriesofdifferentwalks.Thefirstisforonewalk,then

increasingto10,then50.Fromthesnapshots,itcanbeseenthatasthenumberof

walksisincreased,thepathsoftherandomwalkersseemtoconcentrateinamore

orlesscircularregionwiththecenteratthestartingpoint.

Nwalk=1:

Nwalk=10,Nwalk=50:

EffectofEnsembleSize

Whenweincreasetheensemblesize,statisticallytheresultshouldbemore

accurate.WecalculatedDfordifferentsizedensemblesofwalksasshowninthe

followingplot.

Whentheensemblesizeislargeenough,thevalueofDbecomesmoreorless

constant.

Asweincreasedthenumberofwalks,theensembleaveragedresultantX&Y

componentsapproachedzero,asexpected.Wefindthatbeyond15,000walksthe

fluctuationsconvergetozero.

HistogramofresultantX&Ycomponents

ToseethespreadoftheX(orY)componentsaboutthemean,theX(orY)

directionwasdividedintoseveralbins,andthenumberoftimesaresultantX(orY)

component“fell”inthebinwascounted.Fornwalk=15,000,wehaveplotteda

histogramoffrequencywithrespecttovariousbins:

WeseethatthefrequencydistributionisclosetoaGaussiandistribution.

Also,thehistogramforresultantXandYvaluesareplacedsymmetricallyaboutthe

meandisplacementcomponent(0).Ifweweretocontinuetoincreasethenumber

ofwalks,thehistogramwouldapproachaperfectGaussiandistribution.

HistogramofDisplacementdistance

Ifweplotthehistogramofdisplacementdistanceinthesamewayasbefore,

weexpecttogetaskeweddistributioninsteadofGaussiandistributionasshownin

belowfigure.Thisisbecauseallthevaluesofthedisplacementdistanceare

positive.Themostfrequentdistanceis15angstromsinallthreeensembleswith

thesamenumberofsteps/walkwhichisconsistentwiththeconclusionthat

fluctuationsinensembleaveragedvaluesapproacheszerowhenthenumberof

walksisgreaterthanorequalto15,000.

Conclusion

WeusedaMonteCarlomethodtocalculatediffusioncoefficientofFCCCuby

usingrandomwalktheory.

Fromtheresultswesawthatwhenweincreasetheensemblesizethediffusion

coefficientbecomesmoreorlessconstant.Anaturalextensionofthisworkwould

betostudyclusteringofpointdefectsinsolids.

Acknowledgements

SpecialthankstoDr.AnterEl‐Azabandhisgraduatestudentsforguidance,

referencematerials,andinformativediscussions.Thisworkwasdoneaspartofthe

FloridaSocietyforMaterialsSimulationsREUProgram.Iwouldalsoliketothank

thefacultyandstaffoftheDepartmentofScientificComputingatFloridaState

Universityfortheireffortstomakemysummerresearchexperiencemore

enjoyable.