3
Volume 76A, number 1 PHYSICS LETTERS 3 March 1980 CORRELATION OF SURFACE SEGREGATION WITH BULK DIFFUSION FOR BINARY METAL ALLOYS Dennis G. SWARTZFAGER Central Research and Development Department, E.I. du Pont de Nemours & Co., Inc., Wilmington, DE 19898, USA and Michael J. KELLEY Engineering Technology Laboratory, Experimental Station, E.I. du Pont de Nemours & Co., Inc., Wilmington, DE 19898, USA Received 17 September 1979 Revised manuscript received 12 October 1979 Previous treatments of surface segregation in metal alloys have all failed in some systems, due either to inadequacies in the theories or in the thermochemical data. A new model is proposed here using only alloy diffusion data. Predicted surface compositions agree quantitatively with ISS measurements and qualitatively AES results. The surface composition of metal alloys and its and the energy to create it can then be related to the relationship to that of the underlying bulk have attract- surface energy of the solid [8]. More recently, bond- ed much experimental and theoretical interest since breaking arguments based on such a concept have been surface compositions were first measured some ten used to estimate this formation energy and other years ago. Now, though surface compositions for many energies [4,29], such as that binding a solute atom binary alloys can be calculated from basic thermoche- to the vacancy or the surface. Thus the free energy mical data, that for many systems of practical interest change for interchanging solvent and solute surface cannot [1]. Even where surface compositions can be atoms can be shown to be proporional to that for inter- calculated, agreement with experiment is not entirely changing solvent and solute atoms adjacent to a vacancy. satisfying [2,31. Moreover, the theories that give better G G 1 agreement with experiment [1,4] require extensive GAS BS ~~GAV BV computer calculations and so are not easily accessible where the value of the proportionality constant k de- to most investigators. pends on the details of the particular model used. Now A simplified theory or empirical correlation which the left-hand side of eq. (1) relates to the surface com- can be applied broadly thus has considerable appeal. position through Attempts [5,30] to at least predict which member of x 5 Xb i G~5 GBS \ a binary alloy will concentrate on the surface have met = exp ) . (2) with only limited success [6,7]. We here offer another “correlation” to estimate binary alloy surface compo- The difference in vacancy binding energies on the right- sition explicitly, not merely which element will segre- hand side of eq. (1) is related to the tracer diffusion gate. coefficients when B is an impurity in A by [9] Twenty-five years ago, it was suggested that a va- D f w G G cancy may be thought of as a small cavity in the solid DA = fAwA exp (— AV BY) (3)

Correlation of surface segregation with bulk diffusion for binary metal alloys

Embed Size (px)

Citation preview

Page 1: Correlation of surface segregation with bulk diffusion for binary metal alloys

Volume76A, number1 PHYSICSLETTERS 3 March 1980

CORRELATION OF SURFACE SEGREGATION WITHBULK DIFFUSION FOR BINARY METAL ALLOYS

DennisG. SWARTZFAGERCentralResearchandDevelopmentDepartment,E.I. duPontdeNemours& Co., Inc.,Wilmington,DE 19898,USA

and

MichaelJ.KELLEYEngineeringTechnologyLaboratory, ExperimentalStation,E.I. du PontdeNemours& Co., Inc.,Wilmington,DE 19898,USA

Received17 September1979Revisedmanuscriptreceived12 October1979

Previoustreatmentsof surfacesegregationin metalalloyshaveall failed in somesystems,dueeither to inadequaciesinthe theoriesor in thethermochemicaldata.A new model is proposedhereusingonly alloy diffusion data.Predictedsurfacecompositionsagreequantitativelywith ISS measurementsandqualitatively AES results.

Thesurfacecompositionof metalalloysandits andthe energyto createit canthenbe relatedto therelationshipto thatof the underlyingbulkhaveattract- surfaceenergyof thesolid [8]. More recently,bond-edmuchexperimentalandtheoreticalinterestsince breakingargumentsbasedon sucha concepthavebeensurfacecompositionswere first measuredsometen usedto estimatethis formationenergyandotheryearsago.Now,thoughsurfacecompositionsfor many energies[4,29], suchas thatbindinga soluteatombinary alloyscanbecalculatedfrom basicthermoche- to thevacancyor thesurface.Thus the freeenergymical data,that for manysystemsof practicalinterest changefor interchangingsolventandsolutesurfacecannot[1].Evenwheresurfacecompositionscanbe atomscanbe shownto beproporionalto that for inter-calculated,agreementwithexperimentis notentirely changingsolventandsoluteatomsadjacentto avacancy.satisfying [2,31.Moreover,the theoriesthat give better G — G 1agreementwith experiment[1,4] requireextensive GAS — BS — ~~GAV BV

computercalculationsand so arenot easilyaccessible wherethevalueof the proportionalityconstantk de-to mostinvestigators. pendson thedetailsof the particularmodelused.Now

A simplified theoryor empiricalcorrelationwhich the left-handsideof eq.(1) relatesto thesurfacecom-canbe appliedbroadlythushasconsiderableappeal. positionthroughAttempts[5,30] to at leastpredictwhichmemberof x5 Xb i G~5— GBS\abinary alloy will concentrateon the surfacehavemet — = exp ) . (2)with only limited success[6,7]. Wehereoffer another“correlation” to estimatebinary alloy surfacecompo- The differencein vacancybindingenergieson the right-sition explicitly, notmerelywhich elementwill segre- handside of eq.(1)is relatedto thetracerdiffusiongate. coefficientswhenB is animpurity in A by [9]

Twenty-fiveyearsago,it wassuggestedthata va- D f w G — Gcancymaybethoughtof asa smallcavity in the solid DA = fAwA exp (— AV BY) (3)

Page 2: Correlation of surface segregation with bulk diffusion for binary metal alloys

Volume 76A, number1 PHYSICSLETTERS 3 March1980

wherethef’s arethe correlationfactors,thew’s the 1.0 I

jumpfrequenciesanddiffusion is by a vacancymecha-nism.Combiningthesethreeequationsgives:

X~ / Xb - k’ DB(Xb) 0.8 0 -

1—XS/l—Xb DA(Xb)’ ~

wheretheD’s arethe tracerdiffusion coefficientsof 0.6 -

A and B in thebulk alloy containingatomfractionXb ~5

of elementB at thetemperaturewhereX~,thesurface Ag

concentrationof elementB, is measuredandall the 0.4 ~ -

previousconstantsare containedin k’.Forsimplicity, wehavetakenk’ = 1 in makingthe °

calculationsshownin figs. 1, 2 to testwhethereq.(4) 0.2 .

canaccuratelypredictsurfacecomposition.The diffu-siondataweremainly takenfrom Askill’s compilation 7[10] with the individual referencesindicated.Wehave /‘ I I I I

chosenion-scatteringspectroscopy(ISS) measurements 0 0. 2 0.4 0.8 0.8 1.0

of surfacecompositionsto comparewith thecalcula-tionsbecauseof its superiority [11] to othertechniques F.2ft Ag—Au alloys..Diffusioncal-for measurmgfirst monolayercompositions.Thiscom-parisonis frequentlycomplicatedby the unavailabilityof diffusion and ISSdatain thesametemperature range.Comparisonof the compositionestimatesforrange. Cu-richCu—Ni at 730 and1025°C(1003 and 1298K)

Theresultsshownin fig. 1 for severalcopper-based suggeststhatsomealloysmayshownearlyathermalalloysshow remarkableagreementwhenthe diffusion behaviorathighconcentrationof the surfaceactivedataare notextrapolatedoveran extremetemperature element.Theresultsfor Ag—Au (fig. 2) showequally

goodagreement.Finally, table 1 showsgoodagreementfor severalotheralloy systemswhenthe soluteis in very

1.0 --~ low concentration.The diffusion correlationpredictsandISS measure-

7 — — ~2,A —— mentsshowno segregationof Ni to the surfaceof Au-

y _~——v // rich Ni—Au alloys, in contrastwith earlierAESresults/

7~,~ Tablel

0.5 - / -‘ Surfacesegregationfor dilute alloys.

xs ~PvCu / Solvent DB/DA T(°C) Ref. Segregationa) Ref.

(solute)

Fe(Cr) 1.3 800 [101 yes [5]Fe(Ni) 1.4 700 [19,21] ?Fe(Sn) 27.7 700 [19,211 yes [5]

0 Ni(Pd) 259Ob) 800 [10,23]. yes [24]I Pt(Fe) 144 1250 [10,22] no [5]

0 xb 0.5 10 Zr(Fe) 3.88 750 [26] yes [5]Cu Ag(Pd) 0.042 750 [10,26] no [25]

Fig. 1. Surfacesegregationin Cu alloys. Diffusion calculations:• Cu—Pt 730°C, ~ Cu—Ni730°C, ~ Cu—Ni1025°C.ISS results: a)As determinedby AugerElectronSpectroscopy.o Cu—Pt 730°C[2], A Cu—Ni500°C[16], A Cu—Ni 500°C b) The value forD (Pd in Fe) reported in ref. [23] seems un-[16], vCu—Ni730°C[17]. reasonably high.

87

Page 3: Correlation of surface segregation with bulk diffusion for binary metal alloys

Volume76A, number1 PHYSICSLETFERS 3 March1980

[7]. Further,the diffusion correlationandothers[30] platinumgroupmetals,andto examinenonmetalsasdisagreewith the AESfinding [28] of the absence well. Nonetheless,thediffusion correlationoffers aof Fe enrichmentin Pt(Fe).However,AEShasproven usefultool for anticipatingboththe directionandex-insufficiently sensitiveto detectAg enrichmentin tent of surfacesegregationevenwith the limited experi-Ag—Au [2] or measurethe extentof Cuenrichment mentaldataavailable.in Cu—Ni [1], comparedto ISS.ThoughthePt(Fe)diffusion data [22] canbe questioned,webelievea Referencescareful ISSstudyis merited.

Now eq.(4) mayberewritten to expressthe temper- [1] M.J. Kelley, J. Catal.57 (1979) 113.

aturedependenceof the diffusion coefficientdirectly: [2] M.J. Kelley, D.G. SwartzfagerandV.S. Sundaram,J. Vac.Sci. Tech.16 (1979)664.

X5 / Xb , Dçj~ ( QA — QB “ [3] M.J. Keiley, D.G. SwartzfagerandP.W. Gilmour, J. Vac.exp

1—X~/1 — Xb = ‘~ Ri’ ) . (5) Sci. Tech.,to bepublished.[4] R.G. DonnellyandT.S.King, Surf. Sci. 74 (1978) 89.

Its formalsimilarity to eq.(1) leadsusto comparethe [5] J.J.Burton andE.S.Machim,Phys.Rev. Lett. 37 (1976)

differencein activationenergies(strictly,enthalpies) 1433.to theenthalpyfor surfacesegregationandthepre- [6] N.S. Tsai,G.M.PoundandF.F.Abraham,J. Catal. 50

(1977) 200.exponentialfactorsto theentropy. [7] P.Wynblatt andR.C.Ku, in: Interfacial segregation(1977)

Table 2 showsthecomparisonfor the few systems (AmericanSocietyfor Metals,MetalsPark,OH).wheredataareavailable.Agreementis bestwhere [8] H. Brooks, in: Impurities andimperfections(1955)

critically evaluateddiffusion data [10,26] areavailable. (AmericanSocietyfor Metals,MetalsPark,OH, 1).

In Pt(Fe)andNi(Pd) wherethey arenot,agreement [9] A.D. LeClaire, J. Nucl.Mat. 69—70 (1978) 70.[10] J. Askill, Tracerdiffusiondatafor metals, alloys and simple

is vague.Forthe segregationentropy,theratio of dif. oxides(IFI/Plenum, New York, 1970).

fusion pre-exponentialfactorspredicts4.9eu (SsegIR), [11] D.P. Smith, Surf.Sci. 25 (1971)171.

while 2.6 euismeasuredby AES for Ni(Au) [15]. In [12] A.D. Kurtz, B.L. AverbachandM. Cohen,Acta Metal!.view of theproblemswith AESnotedabove,we find 3 (1955)442.

this agreementencouraging. [13] J.E. Reynolds,B.L. AverbachandM. Cohen,Acta Metall.

Finally, theprincipaldifficulty limiting moreexten- 5 (1957) 29.[14] J.J. Burton,C.R.HelmsandR.S.Polizzotti,J. Vac.Sd.sivetestinganduseof this correlationis, asalways,the Tech. 13(1976)204.

sparsityof good experimentaldata.Thougha recent [15] P. WynblattandR.C.Ku, Surf. Sci. 65 (1977)511.

critical compilation [26] providesdiffusion datafor [16] H.H. BrongersmaandT.M. Buck,Surf. Sci. 53 (1975)649.298 solvent(solute)systems,surfacecompositions [17] H.H. Brongersma,M.J. SparnaayandT.M. Buck,Surf.

Sci. 71(1978)657.havebeendeterminedonly for thosenotedabove.It [181 G.C. Nelson,Surf. Sci. 59 (1976)310.

thereforeoffers a challengeto physicistsandothers [19] D. Treheux,D. Marchire,J. DelagrangeandP. Giraldenq,

concernedwith diffusion measurementsandsurface C.R. Acad.Sci. ParisC274(1972) 1260.

scientiststo makemeasurementsin a commontempera. [20] F.S.Buffington, K. Hirano andM. Cohen,Acta Metall.

ture rangeon importantalloys,suchasthoseof the 9 (1961)434.[21] K. Hirano, B.L. Averbach and M. Cohen, Acta Metall.

9 (1961) 440.[22] M.I. Dekhtyar,V.N. Kolesnik,V.1. Patoka,V.1.Silantev

Table2 andI.Y. Dekhtyar,Phys.Stat.Sol. (a) 24 (1974)699.Segregationenthalpyfor dilute alloys. [23] G.A. Vorontsova,I.Y. Dekhtyar,A.M. Shalayeand

V.S. Karacye,Metallogizika(Kiev)48 (1973) 102.System QB—QA Hsega) Ref. [24] D.A. Mervyn, R.J.Baird andP. Wynblatt,Surf.Sd.82

(1979) 79.Ni(Au) 15 15 [7] [25] F.J. Kuijers andV. Ponce,I. Catal.60 (1979)100.Au(Ni) 0 1 [7] [26] C.J. Smithells,Metalsreferencebook, 5th edn.(Butter-Fe(Sn) 5 11 [7] worths,London,1976).Zr(Fe) 21 17 [27] [27] R.S.Polizotti andJ.J.Burton, J. Yac.Sci. Technol. 14Pt(Fe) —26 0 [28] (1977) 347.Ni(Pd) 21 7 [7] [281 J.J.Burton andR.S.Polizotti, Surf.Sci. 66 (1977)1.

[29] R.L. Schwoebel,J. Appl. Phys.38 (1968) 3154.a) In kcal/gmatom.A positivevalue favorssegregation. [30] D.F. Ollis, J. Catal.59 (1979)430.

88