Influence of Surface Segregation on Wetting of Sn-Ag-Cu (SAC) Series and Pb-Containing Solder Alloys

Influence of Surface Segregation on Wetting of Sn-Ag-Cu (SAC)Series and Pb-Containing Solder Alloys

M.J. BOZACK,1,3 J.C. SUHLING,2 Y. ZHANG,2 Z. CAI,2 and P. LALL2

1.—Surface Science Laboratory, Department of Physics, Center for Advanced Vehicle andExtreme Environment Electronics (CAVE3), Auburn University, Auburn, AL 36849, USA.2.—Department of Mechanical Engineering, Center for Advanced Vehicle and Extreme Environ-ment Electronics (CAVE3), Auburn University, Auburn, AL 36849, USA. 3.—e-mail: [email protected]

Wetting of Sn-Ag-Cu (SAC) series solder alloys to solid substrates is stronglyinfluenced by surface segregation of low-level bulk impurities in the alloys. Wereport in situ and real-time Auger electron spectroscopy measurements ofSAC alloy surface compositions as a function of temperature as the alloys aretaken through the melting point. A dramatic increase in the amount of surfaceC (and frequently O) is observed with temperature, and in some cases the alloysurface is nearly 80 at.% C at the melting point. The C originates from low-level impurities incorporated during alloy synthesis and inhibits wettingbecause C acts as a blocking layer to reaction between the alloy and substrate.A similar phenomenon has been observed over a wide range of (SAC and non-SAC) alloys synthesized by a variety of techniques. That solder alloy surfacesat melting have a radically different composition from the bulk uncovers a keyvariable that helps to explain the wide variability in contact angles reported inprevious studies of wetting and adhesion.

Key words: Pb-free solder, wetting, tin-silver-copper (Sn-Ag-Cu), SAC,wetting, Auger electron spectroscopy (AES), surface segregation

INTRODUCTION

The wetting of solid surfaces by liquid metals andalloys is important in several areas of technologyand materials science. Liquid-phase sintering,detergency, adhesion, wear, catalysis, impregnationof porous materials by liquid binders, crystal growthfrom the melt, and solderability are largely deter-mined by the ability of a liquid alloy to wet andspread over a solid. Despite the importance ofinterfacial surface chemistry involved in wetting,there have been comparatively few studies of liquidsurfaces using surface analytical techniques, largelydue to the risk involved when heating materials tomelting inside expensive ultrahigh-vacuum sys-tems. Hardy and Fine1 noted in 1982 that they wereaware of only two previous applications of Augerspectroscopy to the study of liquid surfaces. The

pioneering works were by Berglund and Somorjai,2

who studied surface segregation in Pb-In alloys, andGoumiri and Joud,3 who studied the influence ofoxygen coverage on the surface tension and contactangle of Al-Sn alloys. The situation has changedlittle since that time.

In the case of electronic assembly, the wetting ofsolid board finishes by liquid solder alloys is a keyreliability issue. The criteria for wetting andspreading of a liquid on an ideally flat, solid surfaceat equilibrium is given in terms of the Young–Dupreequation, which relates the liquid–vapor clv, solid–vapor csv, and solid–liquid csl surface tensions to thewetting angle h:

cos h ¼ ðcsv � cslÞ=clv: (1)

This relation shows that, for good wetting, thesurface tensions (csl and clv) associated with theliquid phase should be low whereas the solid–vaporsurface tension (csv) should be high. For wettingsystems far from equilibrium, csl is lowered

(Received January 20, 2011; accepted July 20, 2011;published online August 9, 2011)

Journal of ELECTRONIC MATERIALS, Vol. 40, No. 10, 2011

DOI: 10.1007/s11664-011-1725-7� 2011 TMS

2093

primarily by dissolution and chemical reaction withthe substrate; csv is raised by having a clean solidsurface, since adsorbed gases (unclean surface)lower the surface energy; and clv can be lowered bygas adsorption or diffusion from the liquid bulk tothe surface. For regular solutions, for example, ithas been shown4 that the lowering of csl is related tothe mole fraction Xs of solid dissolved in the liquid:

csl ¼ �RTðln XsÞ=4Am; (2)

where Am is the molar surface area of the liquid. Ahigh degree of wettability will be exhibited by solderalloys which react strongly with and/or dissolveindustry board finishes, which explains why Ni-Aufinishes are such prolific wetting surfaces.

For spreading, it is desirable that cos h = 1, or interms of a spreading coefficient S,

S ¼ csv � csl � clv; (3)

where S > 0. Both wetting and spreading arefavored by low values of the liquid surface tensions.One may also cast the Young–Dupre relationshipin terms of the adhesion energy Wad between thesolid–liquid interface, where

Wsd ¼ clv þ csv � csl: (4)

From (1) and (4) one obtains

cos h ¼Wad=clv � 1; (5)

which means that spreading occurs whenWad ‡ 2clv, where 2clv is the cohesion energy of theliquid. Spreading is favored when the adhesionbetween the liquid alloy and substrate is greaterthan the liquid alloy cohesion.

Applied studies of soldering are best performed inreal time using surface analysis techniques andcommercial board finishes rather than laboratory-created substrates. Previous solderability studies5

in our laboratory and others have been conductedusing one system for wetting but then another forAES/x-ray photoelectron spectroscopy (XPS) analy-sis. This is typically done due to limitations involvedwhen heating a laminate-based board finish in anultrahigh-vacuum (UHV) environment, which posesrisks due to excessive substrate/solder outgassingduring heating of the contact system to the meltingpoint. We have recently overcome this difficulty andnow do everything in one UHV surface analysissystem without breaking vacuum or using cooling/reheating cycles.

We focus here on the wetting of SAC electronicsolder alloys. In recent years, there has been amandated shift from Sn-37Pb solder to emergingPb-free solders in electronics applications. The shifthas occurred due to political concerns about thehealth hazards of Pb in the electronics supply chain.The principal Pb-free solders currently under con-sideration are coined SAC alloys (e.g., SAC405 =96.5Sn-4.0Ag-0.5Cu, etc.), which are ternary alloyscontaining the elements Sn, Ag, and Cu in various

mole fractions. While there have been considerableefforts6,7 to elucidate the mechanical, thermal, andreliability properties of SAC alloys under isother-mal, cycling, and field conditions, there has beencomparatively little work on the wetting behav-ior. In the present effort, we report experimentsusing Auger electron spectroscopy (AES) which aredesigned to study the surfaces of SAC solder alloysunder real-time, in situ conditions as the alloys areheated through the melting point. This is a primaryfactor in wetting, since the condition of the moltensurface near and above the melting point affects theliquid–solid and liquid–vapor interfacial tensionsand consequent ability of the alloy to react chemi-cally with the joining solid surface. Our resultsshow that the surface composition of the moltensurface of SAC solder alloys is radically differentfrom the surface composition at room temperature,with a direct and deleterious effect on wetting andspreading.

EXPERIMENTAL PROCEDURES

Studies of the molten solder alloy surfaces werecarried out in a Kratos XSAM 800 surface analysissystem (Fig. 1). The base pressure of this load-locked, ion- and turbo-pumped system is7 9 10�7 Pa as read on an uncalibrated, nude ion

Fig. 1. Schematic diagram of the ultrahigh-vacuum section of thesurface analytical system used for in situ, real-time AES liquid sur-face studies of SAC-based alloys.

Bozack, Suhling, Zhang, Cai, and Lall2094

gauge. The electron energy analyzer for the AES/XPS portion of the system is a 127-mm-radiusdouble focusing concentric hemispherical energyanalyzer (CHA) equipped with an aberration-compensated input lens (ACIL). AES spectra wererecorded in fixed retard ratio (FRR) mode withretard ratio of 10. Such a retard ratio represents acompromise between sensitivity and resolution andis appropriate for acquisition of survey spectra. TheAES energy axis was calibrated by setting theCu (LMM) line to 914.4 eV referenced to the Fermilevel.

To clean the starting alloy surfaces of adsorbedgases, light Ar+-ion sputter cleaning was accom-plished by a differentially pumped Kratos Mini-beam I plasma discharge ion source with rasteringand focusing capabilities. A variable scan voltageprovides a high-frequency scan that yields a sput-tered area of 1 cm2 at the specimen position with asputter rate of �2.5 nm/min measured on a stan-dard SiO2 film. Auger spectra were taken on thesolid and molten alloy surfaces from a focused spot(�0.1 mm) at a position near the center of thesputtered area. The angle of incidence of the ionbeam with respect to the surface normal was 55�.All Auger spectra were recorded at 3.0 keV beamenergy and 0.7 lA primary beam current, measuredwith applied +90 V bias. The detection system of theXSAM 800 consists of a single-channel multiplierand a fast-response head amplifier. Detector outputmodes include direct pulse counting and currentdetection with voltage-to-frequency (V–F) conver-sion. Due to the large exciting currents used, allspectra were recorded in V–F detection mode.

The raw Auger data were collected in EÆN(E)integral mode by acquiring the spectrum of elec-trons emitted from the surface from energy of 50 eVto 600 eV, with starting energy of 50 eV to avoidcollecting data from the intense, broad secondary-electron peak occurring in this energy region. Formost SAC alloys, the atomic fraction of Ag and Cu isso small that the Auger spectrum is dominated bythe triplet Sn (MNN) peaks from 320 eV to 440 eV.The low-weight-percent components Ag, having aprincipal (MNN) peak at 356 eV, and Cu, with aprimary Cu (LMM) peak at 920 eV, are barely vis-ible. Subsequently, the Savitsky–Golay8 algorithmwas employed to smooth the data and calculate thederivative Auger spectrum. Surface elementalcompositions were determined by standard relativeAuger sensitivity factors9 supplied by the instru-ment manufacturer. It is the trend in relative Augerpeak-to-peak heights during heating that is signifi-cant rather than the accuracy (�10% to 20%) of thesurface composition numbers. Accurate stoichiome-tric calculations in AES are problematic due to thecomplex nature of Auger electron production,transmission, and detection.

The SAC alloys were eutectic or near-eutecticcompositions of Sn-Ag-Cu synthesized commerciallyby Cookson Electronics10 and supplied in bar form.

These alloys have narrow or zero pasty ranges withbroad, low-melting (MP � 220�C) compositions overwhich the alloys remain liquid. Eutectic Sn-37Pb(MP = 183�C) was carried along in the experimentalmatrix for comparison with a leaded solder alloy.While we report solely on commercial alloys, similarexperiments in our laboratory on special SAC com-positions and mixed formulation solder alloys(mixtures of Sn-37Pb and SAC305) gave similarresults.11 The wetting substrate was aluminumcovered with a native aluminum oxide, verified byx-ray photoelectron spectroscopy (XPS). This selec-tion was made because aluminum oxide has anextremely large and negative Gibbs free energy offormation,12 which produces a high wetting contactangle for most contacting materials due to the lackof chemical reaction and interdiffusion on such astable substrate. High rates of chemical reactionand low contact angles would tend to confuse theinterpretation of the data due to substrate elementswhich may diffuse into the alloy. Our goal was tocompletely isolate the contact system from anyoutside elemental sources.

In a typical composition versus temperatureexperiment, the SAC alloy was laid atop the alu-minum oxide substrate and introduced into theultrahigh vacuum system. The alloy fragment wasusually rectangular in shape (2 mm 9 3 mm 91 mm in height) and held in contact with the sub-strate by gravity. No solder flux was employed.After the base pressure was achieved (7 9 10�7 Pa),the alloy surface was sputter cleaned and thenheated in 50�C increments (1�C/s rate) until thealloy melted. The heating was accomplished insidethe vacuum system by a high-resistivity heaterelement inserted into the bottom of the stainless-steel sample mount holding the aluminum oxideand alloy. The approximate temperature of the alloywas indicated by a thermocouple built into theheater element. The melting point was easily veri-fied by simultaneous optical and scanning electronmicroscopy (SEM) observations of morphologicalchanges in the alloy at melting (the XSAM 800 is ascanning Auger system) and by the pressure burstin the system at melting due to trapped gases in thealloy. For most of the data reported herein, theAuger beam was positioned near the top of the alloyfragment, although the beam position had to bealtered occasionally as the alloy changed shape fromrectangular to circular during melting. In nearly allcases, the contact angle of SAC alloys on aluminumoxide changed from 0� before heating to �180� atmelting. It took a few minutes at the melting pointfor the alloy to attain a spherical liquid shape due tothe sluggish kinetics and alloy viscosity. In a fewcases, the adhesive forces were so low that thespherical molten alloy droplet rolled off the alumi-num oxide substrate and into the bottom of thevacuum chamber. Acquisition of each Auger spectrarequired about 20 min, during which time the alloywas maintained at the target temperature.

Influence of Surface Segregation on Wetting of Sn-Ag-Cu (SAC) Series and Pb-Containing Solder Alloys 2095

RESULTS

Alloy Surface Composition VersusTemperature

The surface composition as a function of temper-ature for SAC305 and SAC405 on aluminum oxide isshown in Figs. 2 and 3, respectively. The result isrepresentative for all SAC series alloys (SAC105,SAC205, SAC305, and SAC405). An increase in thefraction of surface C and O (with accompanyingdecrease in Sn) is typically observed with increasingtemperature. In most cases, the increase is from�0 at.% C (sputter cleaned) to �50 at.% to 70 at.%C at the alloy melting temperature, and the increasebegins before reaching the melting point. Figures 2and 3a plot the trend with temperature, whileFigs. 2 and 3b display individual survey AES spec-tra at three temperatures. Clearly, the most dis-tinctive feature during the annealing is thedramatic growth of the C (KLL) Auger peak at272 eV. The appearance of low concentrations ofother low-level alloy impurities such as S, Cl, Ca,and P are also observed, particularly as the alloycrosses the melting point. In situ optical and elec-tron microscopy observations of the molten, poorlywetted droplet frequently showed evidence of sec-ond-phase material floating on the surface of theliquid alloy (Figs. 1, 5). Cooling the alloy back toroom temperature did not reverse the surface com-position. Figure 4 shows similar trends and con-clusions for Sn-37Pb during heating to and abovethe melting point. We include these data to showthat impurity segregation is not distinctive to SACseries alloys but occurs for a more general range ofmetal alloys.

Statistics plays a role when measuring the sur-face properties of liquid metal alloys, observedduring repeated measurements on alloy fragmentscut from the same bar of as-cast solder. The ele-mental trends seen in Figs. 2–4 are nearly alwaysobserved, but as the electron micrographs in Figs. 1and 5 show, poorly wetted alloy droplet surfacesafter heating are frequently inhomogeneous due tothe shell of segregated carbon. Real-time observa-tions of the molten droplets by low-power zoomoptical microscopy during heating in the vacuumsystem revealed that parts of the C surface crustcan dynamically migrate during the course of themeasurements, bringing different portions underthe AES beam, resulting in statistical compositionalvariations over repeated experiments (microscaleplate tectonics). A good example is given in Fig. 5b,which shows the segregated C shell appearance (atroom temperature) in low-power optical (whitelight) microscopy after wetting a mixed formulationsolder alloy MIX 10–90 (10% Sn-37Pb + 90%SAC305 by weight) to a Ni-Au board finish (no fluxapplied) showing an area of the alloy surface with�100% surface-segregated C content (poor wettingand high contact angle) and low C content (goodwetting and low contact angle). The AES spectra

were recorded directly after wetting in the surfaceanalysis system (no breaking of vacuum).

Origin of the Surface Carbon

The most ideal experimental setup for studies ofsolder alloy surface composition versus temperaturewould involve levitating the alloy within an ultra-high-vacuum system during Auger analysis, which(barring adsorbed gas atoms) would guarantee alloyisolation from all outside sources of other elements.In lieu of this, wetting to aluminum oxide and theability to sputter clean the alloy surface of adsorbedgases before heating proved to be essential featuresof the experiment. Since the alloy was completelyisolated from outside sources of carbon and otherelemental species, the material observed at the alloysurfaces during heating could only originate fromthe alloy itself in the form of low-level surface-seg-regated impurities. Similar behavior11 has beenfound when SAC alloys are wetted (without flux) tomore reactive, common board finishes [Ni-Au, Pd-Ni, Ag-Cu, hot air solder leveled (HASL), etc.].

Sanity-check calculations of surface enrichmentlend support to the experimental observation ofsurface segregation in molten, spherical specimensof SAC alloys. If we assume a level of bulk alloycontamination just below the limit of detection ofenergy-dispersive x-ray (EDX) analysis (�0.02wt.%) and assume the entire weight percent toaggregate to the surface of a spherical droplet ofSAC alloy whose radius is 0.5 mm, the thickness ofthe carbon shell formed is �100 nm. This shows it isfeasible for the high concentrations of C to originatefrom impurity surface-segregated material. Emis-sion spectroscopic analysis from a variety of com-mercial solder alloys shows that such levels of bulkcarbon contamination are the rule rather than theexception. To verify the order of magnitude calcu-lation, we measured the thickness of the surfacecarbon shell formed on several poorly wettedspheres of SAC/Sn-37Pb alloys by Auger depthprofiling. The carbon shell was �80 nm to 120 nm inthickness on most solder alloys.

DISCUSSION

The primary goal of this work is to determinewhat a solid substrate sees during wetting of moltenSAC solder alloys. Since the soldering processinvolves heating the solder to a temperature greaterthan the alloy melting point, the substrate sees theliquid surface above the melting point, not the solidsurface at room temperature. If the solid and liquidsurfaces are elementally different, presumptionsmade by flux chemists about what surface speciesexist on solder alloy surfaces will result in ineffi-cient flux formulations. A secondary goal of thisresearch was to determine whether the surface-segregation effects observed by one of us (M.J.B.)over 20 years ago in boride and arsenide alloys are


also exhibited in modern electronic solder alloys.The answer is yes.

Two decades ago, the context was to produce high-brightness, focused ion beams using liquid metal ionsources (LMIS), which are employed today as field-

ion sources in secondary-ion mass spectroscopy(SIMS) and focused ion beam (FIB) instrumenta-tion. In a LMIS, a low-melting liquid such as gal-lium is required to wet a solid substrate without theassistance of flux, which would degrade the ion

Fig. 2. (a) Auger surface elemental composition of SAC305 alloy versus temperature; (b) individual AES survey spectra for three temperatures(RT, 150�C, MP).


beam brightness and purity. At the time, we werestudying boride and arsenide alloys in order todevelop a LMIS of B and As for fine-focused dopant(B and As) ion implantation applications.13 Theboride and arsenide alloys were made by the

Materials Chemistry Group at Los Alamos NationalLaboratory by arc melting and combustion synthe-sis techniques, respectively. The greatest impedi-ment to the development process was the inabilityof the alloys to wet compatible field-emission

Fig. 3. (a) Auger surface elemental composition of SAC405 alloy versus temperature; (b) individual AES survey spectra for three temperatures(RT, 150�C, MP).


substrates due to surface-segregated carbon origi-nating from low-level impurities in the alloys, pre-cisely the same phenomenon we have observed herefor SAC and eutectic Sn-37Pb solder alloys.

It is likely that surface segregation of low-levelbulk impurities is a ubiquitous phenomenon span-ning a wide variety of alloys and alloy synthesistechniques. We have now observed it in boride

Fig. 4. (a) Auger surface elemental composition of Sn-37Pb alloy versus temperature; (b) individual AES survey spectra for three temperatures(RT, 150�C, MP).


alloys manufactured by the arc melting technique,arsenide alloys manufactured by the combustionsynthesis technique, and solder alloys made byinduction heating of the elemental metals in avariety of academic and commercial laboratories.

Surface analytical techniques are required toobserve the segregated impurity layer, which istypically too thin (<100 nm) to observe with stan-dard SEM/EDX techniques, where the samplingvolume is on the order of 2000 nm to 3000 nm

Fig. 5. Surface inhomogeneities after wetting. (a) Optical microscope photograph of mixed formulation alloy MIX 10–90 (10% Sn-37Pb + 90%SAC305 by weight) wetted to Ni-Au (no flux applied) after heating to>MP and subsequent cooling to room temperature. The numbers indicateAES analysis positions; (b) Auger spectra at positions–1–3 on the alloy after wetting, showing areas of the alloy surface with �100% surface-segregated C content (poor wetting and high contact angle) and low C content (good wetting and low contact angle). The AES spectra wererecorded immediately after wetting in the surface analysis system.


beneath the surface when using conventionalaccelerating voltages (�20 kV) employed by mostinvestigators who do EDX analysis.

The widespread nature of impurity segregationoccurs because it is difficult to synthesize high-purity alloys in conventional furnaces owing toeffects traceable to poor chamber vacuum and im-pure starting materials. A typical arc meltingapparatus, for example, operates under an inert gasatmosphere of at least hundreds of Pa in pressurewith no provision for evacuation to high or ultrahighvacuum. Common impurities such as nitrogen,oxygen, and carbon can readily get incorporated intothe alloy bulk during synthesis, which can seriouslydegrade the results of subsequent materials studies.It is economically unfeasible to ensure sufficientlyclean conditions during alloy synthesis to com-pletely nullify the effects of surface segregation.

One method to exploit, rather than mitigate, theeffects of impurity surface segregation14 is to reheatthe synthesized alloys, allow segregation to occur,and then skim or react off the impurity shell. It hasbeen shown that, for many alloys, the interiors ofpoorly wetted droplets of alloy are pure to thedetection limit (�0.1 wt.% or �20 ppm to200 ppm)15 of Auger and EDX spectroscopy. Anexample is seen in Fig. 5b, which shows that, in thecase of wetting to a reactive substrate such asNi-Au, there are regions where the wetting isexcellent with little or no evidence of segregatedcarbon, while the C shell remnant is left behind atanother part of the alloy.

The implications for technological applicationswhich depend on the wetting of molten metal alloysare obvious; e.g., in the LMIS case, good wettingwas eventually achieved by introducing boron intothe contact system, which tied up the segregatedcarbon during formation of boron carbide.16 Processengineers who deal with solderability, adhesion,and assembly issues should be aware that segre-gated impurities (rather than solely native oxides)may be the root cause of poor wetting and adhesion.The existence of surface segregation in molten liq-uids further calls into question many previousstudies of wetting that were conducted blind with-out the advantage of surface analytical tools such asAuger spectroscopy.

There are several practical questions whichemerge concerning impurity segregation. First, ifimpurity surface segregation is so prevalent inmolten metal alloys, why do current solder alloysexhibit such high joint reliability? If an inert,impurity blocking layer exists on the surfaces ofmany molten solder materials, why does anythingstick to them? The answer is twofold. In manyhighly reactive wetting systems (e.g., wetting toNi-Au), the high rate of Au dissolution and strongchemical reaction with the substrate is sufficient toovercome the blocking effect of segregated carbon,which is often found to reside on the alloy surface asa cracked, inert mass of material. In our early

experiments (1988) on this issue, we were able topuncture the impurity shell by a sharp stylus withinthe vacuum chamber while the alloy was molten,which emitted pure alloy from the droplet interiorwhich wet and spread beautifully on the substrate,the same substrate that initially formed a sphericalball. Further, modern solder pastes contain fluxeswhich clean the contacting surfaces of carbon aswell as oxygen and other surface impurities. Weknow this is true due to wetting studies in our lab-oratory11 and others involving organic solder pre-servative (OSP). Most solder pastes wet boardfinishes of OSP, which consists of a thin film ofprotective organic (carbon) covering Cu that pro-tects it from oxidation and volatilizes during reflow.Thus, in the case of solder pastes, process engineersescape the more challenging problem of wetting incases where flux usage is prohibited. It is indeed afortuitous circumstance that the fluxes developed toclean the surfaces of native oxides also work to cleanthe segregated, impurity carbon.

Second, which segregated element (C or O) hasthe largest impact on wetting and spreading?Clearly, both elements are detrimental to goodwetting, but, as noted above, flux chemists havepresumed that surface oxygen in the form of nativeoxides is the major impediment to wetting. Thissupposition is due to the well-established fact thatcommon joining surfaces wetted in air are alwayscovered with a native oxide. While this is true, thedata from surface science investigations show thatsurface carbon is at least as important and proba-bly the biggest offender when considering theliquid–vapor and liquid–solid interfaces in Young’sequation. We say this because: (1) while we oftenobserve increases in both surface impurities C andO with temperature, it is often the case that thegrowth of the Auger C(KLL) peak is more pro-nounced than the growth of the O(KLL) peak; (2)eyeball comparisons of the raw peak-to-peakheights of the O (KLL) and C(KLL) peaks in Augerspectra give an overestimation of the absoluteamount of O on the surface. The Auger elementalsensitivity factors of O (0.5) and C (0.18) at the3.0 kV incident electron beam energy used in ourexperiments are nearly 3-to-1, which means thatthe Auger spectrometer is roughly three times moresensitive to O in comparison to C. Calculations ofAuger surface elemental compositions, such asthose in Figs. 2–4, take this factor into account;(3) controlled, in situ studies of wetting, such asreported in Fig. 5, characteristically show a bettercorrelation between poor wetting and the amount ofsurface C (rather than surface O); and (4) experi-ence in our surface laboratory and others overseveral years has shown that the major commonsurface impurity for the vast number of solid andliquid surfaces where adhesion is a problem containa large fraction of surface C.

Third, what elements are likely to surface segre-gate and what is the driving force for surface


segregation? Is the observed temperature depen-dence predictable? Has impurity segregation beenobserved by other investigators? There are a numberof theoretical models that involve surface segrega-tion ranging from simple to complex and whichaddress equilibrium and nonequilibrium situations.A brief discussion of two of the models provide thetheory and intuitive insight to help explain impuritysegregation in complex, nonideal, far-from-equilib-rium materials involved with solder alloys.

Surface segregation is not a new phenomenon;indeed, Gibbs17 developed a comprehensive ther-modynamic formalism for it over a century ago.Experimental verifications of the Gibbs theory,however, awaited the development of surface-sen-sitive spectroscopies during the 1970s and the con-nection of the Gibbs formalism to quantities that aredirectly observable in surface experiments. Thecomplex nature of multicomponent solder alloysposes a challenge for the theories of surface segre-gation, most of which are based on monolayer-scalesegregation, ideal solutions, and ideal surfaces atequilibrium. Accordingly, there have been severalattempts to develop Ising-like atomic statisticalmodels, most of which have used a regular solutionapproximation.18 The regular solution model allowsfor only two contributions to the free energy of thesolution, an entropy contribution assumed to arisefrom the random distribution of atoms (ideal solu-tions) and an enthalpy-of-mixing contribution. Thetreatments address a system of two phases, a ‘‘bulkphase’’ and a ‘‘surface phase,’’ and consider a semi-infinite system consisting of a two-component solid.The common procedure minimizes the total freeenergy of the system with respect to the compositionsof both phases. The general result can be written as

XsA=X

sB ¼ Xb

A=XbB

� �exp �DHs=kT½ �; (6)

where XAs , XB

s (XAb , and XB

b ) are the equilibrium atomfractions of components A and B in the surface (bulk)phases, respectively, for the two constituents, andDHs is the heat of segregation. The heat of segrega-tion is the driving force for the segregation processand represents the enthalpy change which resultswhen an atom of type A lying in the bulk phaseexchanges positions with an atom of type B lying inthe surface phase. Differences between the varioustheories of surface segregation lie in the form of theexpression for DHs. We briefly discuss two forms ofDHs given by the regular solution model.

The simplest monolayer ideal-solution model19 ofsurface segregation shows that low-surface-tensionelements are likely to surface segregate. The surfacecomposition of an ideal binary solution is related tothe bulk composition by

XsA=X

sB ¼ Xb

A=XbB

� �exp cB � cAð Þa=kT½ �; (7)

where a is the average surface area occupied peratom, cA and cB are the surface tensions of the pure

components A and B, k is the Boltzmann constant,and T is the temperature. If cA < cB, the surfacefraction of component A will increase exponentially,resulting in marked surface segregation. It is clearthat the constituent with the lower surface tensionwill have a higher concentration on the surface. Thisis intuitive since the surface tension is a measure ofthe force per unit length or energy per unit arearequired to produce a surface. The driving force forsurface segregation in this model relates to the dif-ference in the binding energies between the twometal atoms, A–B and the binding energies in thepure components, A–A and B–B. The change inbinding energy gives rise to a change in the surfacetension of the binary system compared with thesurface tension for the pure constituents. Metalalloys are not ideal solutions, since they generallyhave some finite heat of mixing, which is ignored inthe ideal-solution monolayer approach; further, themonolayer approximation strictly applies only tosolids where all layers below the top monolayer havethe bulk composition. While the ideal solutionmonolayer model shows that surface tension differ-ences are a key driving force in segregation, it fails towork for impurity segregation by giving the wrongtemperature dependence, showing that the surfaceexcess concentration of the segregating constituentshould diminish with increasing temperature.

Owing in part to the difficulties in measuringaccurate surface tension values, standard compila-tions of surface tension are lacking or inadequate.For metals, however, there is excellent correlationbetween surface tension and the heat of sublimationDHsubl, because both properties are related to crea-tion of a unit area of surface. For semiconductors,semimetals, and insulators, most tabulated valuesin the literature show that the elements of lowestsurface tension are nonmetals (e.g., C, N, O, F, P,Cl, S). Nonmetallic elements are frequently dubbedsurface active since they are the elements commonlyobserved on ‘‘clean’’ (99.999% pure) metal surfacesby surface analytical techniques. Figures 2–5 showseveral of these nonmetals appearing on the sur-faces of heated SAC alloys. Bulk contaminants,many of which are pinned in metal alloys at thegrain boundaries, are released at melting when theboundaries disappear and the low-surface-tensioncomponents surface segregate in order to minimizethe overall free energy of the system.

A second, more realistic regular solution modelcorrects the monolayer, ideal-solution model andshows that more solute should segregate withincreasing temperature. This was demonstrated byLea and Seah,20 who developed a model for surfacesegregation based on a solution of the diffusionequation by McLean21 for the case of grain bound-ary segregation. The general solution can be written

XsðtÞXsð1Þ ¼ 1� exp

D � ta2 � d2

� �erfc

D � ta2 � d2

� �1=2

; (8)


where XsðtÞ and Xsð1Þ are the surface atom frac-tions of the segregating component at times t and1, respectively (i.e., Xsð1Þ is the equilibrium atomfraction of the segregating component), a is the ratioat equilibrium of the surface atom fraction to thebulk atom fraction, D is the bulk diffusion coeffi-cient of the segregant through the alloy, d is the‘‘thickness’’ of the segregated surface region (assumedto be one monolayer), and t is the time. In this model,bulk diffusion is the rate-determining step for sur-face segregation. Furthermore, local equilibrium isassumed between the surface and the nearby bulkregion, which is described by the linear relationXsð1Þ ¼ a � Xbð1Þ; where Xbð1Þ is the solute con-centration of the segregant and a is called theenrichment factor. Equation (8) is strictly valid forlow surface coverage because a is assumed to beindependent of Xs; which is not true for high surfaceconcentrations near saturation. By expanding using

a power series in P ¼ D�ta2�d2

� �1=2for small P values,

higher-order terms can be neglected to give

XsðtÞXsð1Þ ¼

2ffiffiffipp D � t

a2 � d2

� �1=2

; (9)

and by introducing the relation Xsð1Þ ¼ a � Xbð1Þone obtains

XsðtÞ ¼ ½2=dffiffiffipp�ðXbÞ½

ffiffiffiffiffiffiffiffiffiffiðDtÞ

p�; (10)

where XsðtÞ is the surface solute concentration attime t, Xb is the solute concentration in the bulk, dis the thickness of a monolayer of segregant, and Dis the effective diffusion coefficient. This shows thatthe free solute concentration in the bulk and thetemperature are the primary factors controllingsolute accumulation at the surface. Since the gen-eral form of the diffusion coefficient isD � Do exp(�EA/RT) m2/s, where Do is the diffusioncoefficient at infinite temperature and EA is theactivation energy, there is an exponential depen-dence of D on T. For a fixed hold time t at anytemperature T, the element under considerationshould show pronounced enrichment on the surfaceupon heating. The Lea–Seah formalism has beenapplied to a few solid surfaces and appears to givereasonable agreement with diffusion coefficientsmeasured by tracer diffusion methods.22 While it ishazardous to apply the theory to complex, nonideal,liquid solder alloys far from equilibrium, it is nev-ertheless useful to see the roles of temperature andtime on segregation.

Impurity segregation has been observed since theinfancy of experimental surface science. In their1979 paper, Wynblatt and Ku23 report that surfacesegregation experiments had been done on a dozenor so binary alloy systems at that time. Impuritysegregation, where C, O, Ca, S, and other nonmetalsrepeatedly appeared on the surfaces of the alloys,was considered to be a nuisance in the earlyexperiments, whose goal was to examine monolayer-

scale segregation. Removal of these strongly segre-gating contaminants (usually by repeated sputter/heating cycles or reaction with oxygen) was impor-tant before starting the ‘‘true’’ segregation experi-ment involving the binary alloy components. Thiswas because the impurities could interact prefer-entially with one or the other of the binary alloyelements and either enhance or suppress the natu-ral segregating tendency of the uncontaminatedalloy. In fact, Wynblatt and Ku24 remark that ‘‘thesespecies arise from strongly segregating trampimpurities present in the bulk alloy in trace quan-tities, and they can sometimes prove very difficult toremove.’’ Clearly, the significance of impurity seg-regation in metal alloys was recognized early in thedevelopment of modern surface science.

CONCLUSIONS

Wetting of SAC series solder alloys to board fin-ishes is directly influenced by surface segregation oflow-level bulk alloy impurities. The major segre-gating impurities are C and O at the molten solderalloy surface. The impurity segregation is not tem-perature reversible.

The impurities form a high-surface-tension shellaround the alloy and inhibit chemical reaction withthe substrate, resulting in poor wettability. Theinterior of poorly wetted droplets of alloy is com-posed of pure alloy material which allows for vary-ing amounts of good wetting and spreadingdepending on the contacting materials. Wetting issluggish in locations where the carbon shell rem-nant is located.

Impurity surface segregation has been observedover a wide range of alloys synthesized by a varietyof different alloy manufacturing techniques. Solderfluxing remedies the problem with impurity segre-gation by cleaning both carbon and oxygen fromsurfaces. The clear message from this work is thatsurface analytical tools are essential for studies ofwetting and adhesion.

ACKNOWLEDGEMENTS

We gratefully acknowledge the industrial mem-bers of the Center for Advanced Vehicle Electronicsand Extreme Environment Electronics (CAVE3) forcontinued support of this work.

REFERENCES

1. S. Hardy and J. Fine, Materials Processing in the ReducedGravity Environment of Space, ed. G.E. Rindone (Amster-dam: Elsevier, 1982), p. 503.

2. S. Berglund and G.A. Somorjai, J. Chem. Phys. 59, 5537(1973).

3. L. Goumiri and J.C. Joud, Acta Metall. 30, 1397 (1982).4. Ju.V. Naidich, Prog. Surf. Membr. Sci. 14, 353 (1981).5. S.V. Sattiraju, B. Dang, R.W. Johnson, Y. Li, J.S. Smith, and M.J.

Bozack, IEEE Trans. Electron. Packag. Manuf. 25, 168 (2002).6. H. Ma, J.C. Suhling, P. Lall, and M.J. Bozack, Proc. 56th

Electronic Components and Technology Conference, 2006,p. 16ff.


7. Y. Zhang, Z. Cai, J.C. Suhling, P. Lall, and M.J. Bozack,Proc. 58th IEEE Electronic Components and TechnologyConference, Orlando, FL, May 28–30, 2008, pp. 99–112.

8. A. Savitzky and M. Golay, Anal. Chem. 36, 1627 (1964).9. P.W. Palmberg, G.E. Riach, R.E. Weber, and N.C. MacDonald,

Handbook of Auger Electron Spectroscopy (Edina, MN: Physi-cal Electronics Industries, 1972).

10. Cookson Electronics Assembly Materials, South Plainfield,NJ.

11. M.J. Bozack, J.C. Suhling, Y. Zhang, Z. Cai, and P. Lall,J. Vac. Sci. Technol. A 29 (4), 041401 (2011).

12. C.E. Wicks and F.E. Block, Thermodynamic Properties of 65Elements—Their Oxides, Halides, Carbides, and Nitrides(Washington, DC: Bureau of Mines Bulletin 605, U.S. Gov-ernment Printing Office, 1963).

13. M.J. Bozack, L.W. Swanson, and A.E. Bell, J. Mater. Sci. 22,2421 (1987).

14. M.J. Bozack and P.R. Davis, Mater. Sci. Eng. A 150, 255(1992).

15. P. Kuisma-Kursula, X-ray Spectrom. 29, 111 (2000).

16. M.J. Bozack, A.E. Bell, and L.W. Swanson, J. Phys. Chem.92, 3925 (1988).

17. J.W. Gibbs, The Scientific Papers of J. Willard Gibbs, Vol. 1(New York: Dover, 1961), p. 219.

18. W.C. Johnson and J.M. Blakely, eds., Interfacial Segrega-tion (Metals Park, OH: American Society for Metals,1979).

19. G.A. Somorjai, Introduction to Surface Chemistry andCatalysis (New York: Wiley, 1994), p. 285ff.

20. C. Lea and M.P. Seah, Philos. Mag. 35, 213 (1977).21. D. McLean in, Grain Boundaries in Metals (Oxford: Clar-

endon, 1957).22. K. Hennesen, H. Keller, and H. Viefhaus, Scripta Metall. 18,

1319 (1984).23. P. Wynblatt and R.C. Ku, Interfacial Segregation, ed. W.C.

Johnson and J.M. Blakely (Metals Park, OH: AmericanSociety for Metals, 1979), p. 127.

24. P. Wynblatt and R.C. Ku, Interfacial Segregation, ed. W.C.Johnson and J.M. Blakely (Metals Park, OH: AmericanSociety for Metals, 1979), p. 117.


Documents

Influence of Surface Segregation on Wetting of Sn-Ag-Cu (SAC) Series and Pb-Containing Solder Alloys