ContentslistsavailableatSciVerseScienceDirect
Materials
Science
and
Engineering
B
j o ur n a l ho me p a g e :w w w . e l s e v i e r . c o m / l o c a t e / m s e b
Diffusion
barrier
characteristics
and
shear
fracture
behaviors
of
eutectic
PbSn
solder/electroless
Co(W,P)
samples
Hung-Chun
Pan,
Tsung-Eong
Hsieh
∗DepartmentofMaterialsScienceandEngineering,NationalChiaoTungUniversity,1001Ta-HsuehRoad,Hsinchu30010,Taiwan,ROC
a
r
t
i
c
l
e
i
n
f
o
Articlehistory: Received12May2011 Receivedinrevisedform 13September2011 Accepted26September2011 Available online 6 October 2011 Keywords: ElectrolessCo(W,P) IMC Diffusionbarrier Sheartest
a
b
s
t
r
a
c
t
Diffusionbarriercharacteristics,activationenergy(Ea)ofIMCgrowthandbondingpropertiesof
amor-phousandpolycrystallineelectrolessCo(W,P)(termedas˛-Co(W,P)andpoly-Co(W,P))toeutecticPbSn solderarepresented.Intermetalliccompound(IMC)spallationandannano-crystallineP-richlayerwere observedinPbSn/˛-Co(W,P)samplessubjectedtoliquid-stateagingat250◦C.Incontrast,IMCsresided ontheP-richlayerinPbSn/˛-Co(W,P)samplessubjectedtosolid-stateagingat150◦C.ThickIMCs
neigh-boringtoanamorphousW-richlayerwasseeninPbSn/poly-Co(W,P)samplesregardlessoftheaging type.˛-Co(W,P)wasfoundtobeasacrificial-plusstuffed-typebarrierwhilepoly-Co(W,P)ismainlya sacrificial-typebarrier.ThevaluesofEa’sforPbSn/˛-Co(W,P)andPbSn/poly-Co(W,P)systemswere338.6
and167.5kJ/mol,respectively.Sheartestrevealedtheductilemodedominatesthefailureinboth˛-and poly-Co(W,P)samples.AnalyticalresultsindicatedthehighPcontentinelectrolesslayermightenhance thebarriercapabilitybutdegradethebondingstrength.
© 2011 Elsevier B.V. All rights reserved.
1. Introduction
As an essential component of flip-chip (FC) bonding for advancedelectronicpackaging,theunderbumpmetallurgy(UBM) isthemultilayerthin-filmstructurecomprisedofadhesion, dif-fusionbarrierandwetting/protectionlayers.Thediffusionbarrier layerinhibitstheinterdiffusionbetweenbumpbodyandbondpad materialssoastoensurethereliableoperationofintegrated cir-cuits (ICs).Diffusion barriercan beclassifiedassacrificial type, stuffedtype,passive-compoundtypeandamorphoustype accord-ingtothemechanismstoinhibittheinterdiffusion[1].Refractory metalssuchastungsten(W),molybdenum(Mo),platinum(Pt), palladium(Pd)andrhodium(Rh)preparedbyphysicalvapor depo-sition(PVD)haveadoptedasthebarrierlayerinconventionalFC structures.Itwasalsofoundthatthethinlayerspreparedby“wet” process,e.g.,electrolessnickel(EN),mayalsoserveasthe diffu-sionbarrierlayersfortheadvantagesincludinghighthroughput, highstepcoverage,lowerstressstatus,andfreeofgrain bound-aryforshort-circuitdiffusion[2–6].Electrolesscobalt–phosphorus (Co(P))possessesasuperiorbarriercapabilitytoinhibitthe inter-diffusionbetweenCuandtheinterlayerdielectric(ILD)inCu-ICs [7]anditmightalsoinhibitthediffusionofPbSnsolder[8].Kohn et al.reportedthe enhancement ofthermal stability and diffu-sionbarrier capability toCu metallizationby incorporatingthe
∗ Correspondingauthor.Tel.:+88635712121x55306,fax:+88635724727. E-mailaddress:[email protected](T.-E.Hsieh).
W element in electroless Co(P)[9,10] and ourstudy on amor-phouselectrolessCo(W,P)revealeditmayserveasamixedtype barrier,i.e.,acombinationofsacrificialandstuffedtype,toPbSn solderandCu[11].Inthisstudy,wepreparetheamorphousand polycrystallineelectrolessCo(W,P)(termedas˛-Co(W,P)and poly-Co(W,P)hereafter)layersandtheirdiffusionbarriercapabilitiesto eutecticPbSnsolderareevaluatedviatheliquid-andsolid-state agingtests.Scanningelectronmicroscopy(SEM)andtransmission electronmicroscopy(TEM)inconjunctionwithenergydispersive spectroscopy(EDS)wereemployedtoanalyzethemicrostructure evolutionatPbSn/Co(W,P) interfaces.Theexperimentalfindings alsoprovidetheinformationregardingoftheactivationenergies (Ea’s)ofIMCgrowthinthediffusioncouplescontaining˛-Co(W,P)
or poly-Co(W,P) subjected to the solid-state aging. The solder ball shear test was also carried out to elucidate the bonding characteristics of various PbSn/Co(W,P) samples so as to clar-ifythepracticalapplicabilityofelectrolessCo(W,P)layerstoFC bonding.Thefracturesurfacessubjectedtosheartestwerealso examinedbySEMinordertoidentifythecorrespondingfracture mode.
2. Experimentalmethods
Silicon(Si)waferscoatedwithTi(50nm)/Cu(100nm)layerto simulatetheadhesionlayerandtheCuinterconnectswereadopted asthesubstratesofthisstudy.Priortotheelectrolessplating,a pre-treatmentincludingroughening,sensitizationandactivationwas performedinordertoyieldtheCo(W,P)layerswithbetterstructure 0921-5107/$–seefrontmatter © 2011 Elsevier B.V. All rights reserved.
62 H.-C.Pan,T.-E.Hsieh/MaterialsScienceandEngineeringB177 (2012) 61–68
Table1
Chemicalsandprocessingconditionsofpretreatment[11].
Step Component Concentration Immersiontime Roughening H2SO4 5wt.% 30s
Sensitization SnCl2·2H2O 10g/L 5min
HCl 40ml/L
Activation PdClHCl2·2H2O 0.18ml/Lg/L 1min
integrity.Afterward,about6-to8-mthickelectrolessCo(W,P)
layer was deposited on the Cu/Ti/Si substrate. Tables 1 and 2
separatelylisttheformulationsandprocessingconditionsfor pre-treatmentandelectrolessplating process [11].ThepH valueof platingbath wasmonitoredbyusing a pHmeter and adjusted witha3MKOHsolutionsoastoachievetheCo(W,P)layerswith desiredcrystallinities.The˛-Co(W,P)andthepoly-Co(W,P)were separatelyobtainedatpH=8.6and7.6.Subsequentcompositional analysisindicatedthePcontentisabout9.1–10.2at.%andW con-tentisabout0.4–0.9at.%in˛-Co(W,P)whereasthePcontentis about4–6at.%andtheWcontentisabout5–8at.%inpoly-Co(W,P). ImmediatelyaftertheCo(W,P)deposition,anappropriateamount of eutectic PbSn solder pastewas appliedon theCo(W,P) sur-faceandabriefreflowat250◦Cfor30swasperformedinorder tosolidifythesolder.ThePbSn/electrolessCo(W,P)/Cu/Ti/Si sam-ples were annealed in N2 atmosphere for liquid-state aging at
250◦C for up to 5hor vacuum-sealed and sent to furnace for solid-stateagingtestat150◦Cfor1000h.Forthedetermination oftheEa’sofIMCgrowthinvarioussamples,solid-stateagingat
130and170◦C for atleast 500hwere alsoperformedand, for thepurposeofcomparison,thePbSn/pureCosampleswere pre-paredbyapplyingthePbSnsolderpasteonpureCofoilandtested atthesameconditions.SEM(JeolJSM-6500ForHitachiS-4700) and TEM(Philips Tecnai F-20) in conjunction with EDS (Gene-sis)wereadoptedtoexaminethemicrostructureandcomposition changesinthesamples.Thecross-sectionalTEM(XTEM)samples werepreparedbyusingthefocused-ion-beam(FIB,FEI-201) tech-niquesupportedbyMaterialsAnalysisTechnology,Inc.atChupei, Taiwan,ROC. Atleastfourdifferentlocationswereanalyzedby EDSandtheaverageresultswerepresentedasthecomposition dataofsamples.Astosamplesforballsheartest,anarrayof cir-cularbondpadpatternwith200mindiameterwasformedon thesubstratesdepositedwith˛-Co(W,P)orpoly-Co(W,P)bythe photolithography processusing anSU-8permanent photoresist (supplier:MicroChemCorp.;5%-weight-lossthermal decomposi-tiontemperature=279◦Cinair)asthemasklayer.Afterattaching the300-m-diameterPbSnsolderballsonthepads,areflow treat-ment in N2 atmosphere at 250◦C for 1, 10, 20, 30 and 60min
wasthencarriedouttoformthesolderballjoints.Thesheartest wasperformedinaccordwiththeJEDECStandard,JESD22-B117A [12],byusinga Dage4000multipurposebond-testersupported bySchmidtScientific TaiwanLtd.at Hsinchu,Taiwan,ROC.The sheartoolstandoff=30mabovethesamplesurfaceand shear speed=100m/s.Theaverageshearforcewasdeducedfromthe
Table2
ChemicalsandprocessingconditionsofelectrolessCo(W,P)plating[11].
Component Concentration(g/L)
Cosource CoSO4·7H2O 23
Reducingagent NaH2PO2·H2O 18 Complexingagent Na3citrate 144
Bufferagent H3BO3 31
Wsource Na2WO4·2H2O 10
pHvalue 7.6or8.6
Temperature 90◦C
resultsofatleast25bumpsforeachsamplepreparation
condi-tion.
3. Resultsanddiscussion
3.1. PbSn/˛-Co(W,P)samples
Fig.1(a)–(f)presentsthecross-sectional SEMmicrographsof PbSn/˛-Co(W,P) couples subjected to liquid-state aging for as-reflow, 20min, 30min, 1h, 3h and 5h, respectively. We note the prolonged liquid-state aging is for examining a complete microstructure evolution in such a sample type. As shown in Fig.1(a),granularCoSn2intermetalliccompound(IMC)coatedwith
CoSn3 bytheperitecticreaction[11]orCoSn3 IMCsemergedat
thesolder/Co(W,P)interfaceattheearlystageofaging.Voidswere occasionallyobservedandtheymightbecausedbytheevaporation oforganicadditivesinthesolderpaste.Inthesamplesubjected to20-minaging,theIMCscoarsenanddetachmentofIMCsaway fromthereactinginterfaceoccursasshowninFig.1(b).Inaddition, anabout1-mthickcontinuouslayerneighboringtounreacted Co(W,P)canbeobservedandtheEDSanalysisrevealsitcontains about20at.%ofPandabout2.3at.%ofW.Asaresultofthe accu-mulationofPelementsatthereactinginterface,suchaP-richlayer isinfactamixtureofnano-scaleIMCsasreportedbyourprevious study[11].ThespallationofIMCsbecameratherobviousafterthe agingfor30minasshowninFig.1(c)andnearlyallIMCsspalled intosolderafter1-hagingasdepictedbyFig.1(d).TheSEMimages showninFig.1(e)and(f)indicatethatthereisnodramaticchange ininterfacialmorphologyinthesamplessubjectedtoprolonged aginginwhichthethicknessofP-richlayerremainsthesameat about1m.
Aboveresultsseemtoindicatethattheformationof continu-ousP-richlayermaydetersubsequentCo–Snreactions.Inorder toverifythisphenomenon,a250◦C/30-minagedsamplewas fur-theragedat150◦C for200h.AsshowninFig.2,thegrowthof secondaryIMCsfromtheP-richlayerintosolderwasobserved.A supplyofCoandSnelementswouldberequiredforformingsuch bush-likeIMCs,implyingtheP-richlayercannotblocktheCo–Sn interdiffusionandtheIMCspallationisinfactanunceasingprocess atthereactinginterfaceduringliquid-stateaging.Thisalso indi-catesthePaccumulationatthereactinginterfaceplaysakeyrole intheevolutionofinterfacialmorphology.WhenthePcontentat reactinginterfaceremainslowattheearlystageofaging,theIMCs formandresideatthesolder/Co(W,P)interface.Withtheincrease ofagingtime,thePelementsaccumulatetoformthecontinuous P-richlayer.However,thesufficientlyhighamountofPdeteriorates theadherenceofIMCstoP-richlayerandconsequentlyleadstothe IMCspallation.Duringtheprolongedaging,theSnelements con-tinuouslydiffuseacrosstheP-richlayertoreactwithCotoform theIMCsattheinterfaceneighboringtoCo(W,P)whiletheIMCs neighboringtothemoltensolderunceasinglyspallaway.Inthe meantime,thehighlyaccumulatedPelementsinterruptsthe coars-eningofIMCsandresultsinanultrafineIMCmixturecomprising oftheP-richlayer[11].Comprisingofreactinginterfacecontaining theP-richlayerwithfixedthicknessinthesamplessubjectedto prolongedaging.
AnalyticalresultsregardingtothePbSn/˛-Co(W,P)subjectto solid-stateaginghavebeenreportedindetailpreviously[11].In conjunctionwiththeresultsofliquid-stateagingpresentedabove, theformationofIMCsapparentlyimpliesthesacrificial-typebarrier featureof˛-Co(W,P)toeutecticPbSnsolder.Further,ourprevious TEMcharacterizationrevealedafinelydispersedCo2Pprecipitates
inthe˛-Co(W,P)layerandtheelectrolesslayertendsto recrystal-lizeduringtheagingtreatment[11].Theformationofphosphide compoundsandsupersaturatedPelementsin˛-Co(W,P)mayblock
Fig.1.Cross-sectionalSEMmicrographsofPbSn/˛-Co(W,P)samplessubjectedtoliquid-stateagingat250◦Cfor(a)1min,(b)20min,(c)30min,(d)1h,(e)3hand(f)5h.
thediffusionofCuintoCoand,hence, the˛-Co(W,P)mayalso possessthestuffed-typebarriercapabilitysinceEDShasrevealed anegligibleinterdiffusionbetweenCoandCuunderlayer.The ˛-Co(W,P)ishenceacombined-type,i.e.,sacrificial-plusstuffed-type, diffusionbarrier.
Fig.2.Cross-sectionalSEMmicrographofPbSn/˛-Co(W,P)samplesubjectedto 250◦C/30-minliquid-stateagingfollowedby150◦C/200-hsolid-stateaging.
3.2. PbSn/poly-Co(W,P)samples
Fig.3(a)depictsthecross-sectionalSEMimageof PbSn/poly-Co(W,P)sample subjected to liquid-state aging for 1h and the correspondingEDS linescanning profilesis shown inFig. 3(b). Unlike the sample containing ˛-Co(W,P), about 5-m thick, scallop-type CoSn3 IMCs formatthesolder/poly-Co(W,P)
inter-face without spallation. In addition, the XTEM/EDS analysis of PbSn/poly-Co(W,P)samplerevealedanabout500-nmthick, amor-phouslayerwithrelativelyhighWcontent(∼15at.%)emerging inbetweentheCoSn3 IMCs andunreactedCo(W,P)asshownin
Fig.3(c).SincetheWcontentishighincomparisonwiththatin previous˛-Co(W,P)system,wehencetermitastheW-richlayer. SimilartotheaccumulationofP,theW-rich layershouldresult fromtheaccumulationofWattheinterfacewhenCoreactswith SntoformtheIMCsduetothelowsolubilityofWelementsinthe samples.
Sincetheamorphismhasbeencategorizedasaplausiblebarrier mechanism [1], a liquid-state aging up to 5h was hence per-formedtoverifywhethertheW-richlayermayretardsubsequent interdiffusion.Nevertheless,theIMClayerwasfoundtothicken to about 7m after the 5-h aging in the presence of suchan amorphous layer.The decouplingof structure amorphism with
64 H.-C.Pan,T.-E.Hsieh/MaterialsScienceandEngineeringB177 (2012) 61–68
Fig.3.(a)Cross-sectionalSEMmicrographofPbSn/poly-Co(W,P)samplesubjected to250◦C/1-hliquid-stateagingand(b)correspondingEDSlinescanningprofiles.(c) XTEMmicrographofthePbSn/poly-Co(W,P)samplesubjectedto250◦C/1-h liquid-stateaging.Thedottedcircledenotestheareawheretheselectedareaelectron diffraction(SAED)patternwastaken.
thebarriercapabilitywassimilarlyobservedinourstudy regard-ingoftheSnBi/Co(W,P)system[13].Thebarriercapabilityshould thusascribetothenatureofchemicalbonds[14],ratherthanto thestructureamorphism asproposedbypreviousclassification scheme[1].
Fig.4(a)and(b) separatelypresentsthecross-sectional SEM images of PbSn/poly-Co(W,P) samples subjected to solid-state aging for 1000h and the corresponding EDS line scanning
Fig.4.(a)Cross-sectionalSEMmicrographofPbSn/poly-Co(W,P)samplesubjected to150◦C/1000-hsolid-stateagingand(b)correspondingEDSlinescanningprofiles.
profiles.Anabout3minthickness,layer-likeIMCformedatthe solder/poly-Co(W,P)interfaceand,accordingtotheEDSanalysis, it ismainlytheCoSn3 type. Theformation ofIMCs iteratesthe
sacrificialtypebarrierfeatureofpoly-Co(W,P)sample.
AnXTEManalysisofthe150◦C/1000-hagedsampleare pre-sentedinFig.5.Anabout250-nmthickamorphousW-richlayer (Wcontent∼10at.%)inbetweenIMCsandunreactedCo(W,P)was similarlyobserved.ThethinnerW-richlayerinsuchasampleis attributedtothelowertemperatureofsolid-stateaging.
WenotethatTEMandEDSanalysesdetectnegligibleamountof Co2Pprecipitatesand/ortheCo–Walloyphaseinunreacted
poly-Co(W,P).AlthoughsupersaturatedPandWcontentsarepresentin poly-Co(W,P),theireffectsonstuffed-typebarrierseem compara-tivelylessthanthosein˛-Co(W,P).Besides,thepresenceofgrain boundariesmightserveasthefastdiffusionpathsandsubsequent kineticanalysisindicatedthepoly-Co(W,P)exhibitsalower activa-tionenergyofIMCgrowth.Thisweakensthestuffed-typebarrier capabilityinpoly-Co(W,P)and,hence,thepoly-Co(W,P)ismainly asasacrificial-typebarrier.
3.3. DeterminationofEaofIMCgrowth
Fig. 6 presents the thickness consumption of Co(W,P) layer againstthesquare rootof agingtimeinthesamplescontaining variouselectrolessCo(W,P)layersdeducedfromtheSEM charac-terizations.Itcanbereadilyseenthatatthesameagingtimespan, lessamountof˛-Co(W,P)isconsumedregardlessoftheagingtype, indicatingthe˛-Co(W,P)possessesa betterbarriercapabilityin termsofitssacrificialbehavior.Itisbelievedthatthebarrier capa-bilityof˛-Co(W,P)layerisenhancedbyitshighPcontentwhich may block the interdiffusion in a more efficient manner in
Fig. 5.(a) XTEM micrograph of PbSn/poly-Co(W,P) sample subjected to 150◦C/1000-hsolid-stateaging.(b)EnlargedpictureofW-richlayerand corre-spondingSAEDpattern.
comparison with that in poly-Co(W,P) sample. Fig. 6
also indicates that the consumptions of ˛-Co(W,P) and
poly-Co(W,P) layers subjected to 1-h liquid-state aging are about1.98and2.45m,respectively,whereastheconsumptions
2400 1800 1200 600 120 80 40 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Poly-Co(W,P)/liquid-state aging α-Co(W,P)/liquid-state aging Co(W,P) consumption ( μ m)
√
aging time (s) Poly-Co(W,P)/solid-state aging α-Co(W,P)/solid-state agingFig.6.ConsumptionofCo(W,P)layerasafunctionofsquarerootofagingtimein thesamplescontainingvariouselectrolessCo(W,P)layers.
(a)
(b)
1000 1200 1400
800
600
400
0.1
1
5
10
25
50
75
90
PbSn/p-Co(W,P)@170 °C PbSn/p-Co(W,P)@150 °C PbSn/p-Co(W,P)@130 °C PbSn/α-Co(W,P)@170°C PbSn/α-Co(W,P)@150°C PbSn/α-Co(W,P)@130°C PbSn/pure Co @170 °C PbSn/pure Co @150 °C PbSn/pure Co @130 °CIMC Thickness (
μ
m)
√
aging
ti
me (s
)
2.5
2.4
2.3
2.2
-34
-32
-30
-28
-26
-24
-22
PbSn/α-Co(W,P) PbSn/p-Co(W,P) PbSn/pure Co Eαa-Co(W,P)= 338.6 kJ/mol Ep-Co(W,P)a = 167.5 kJ/molln K (cm
2/s)
1000/Τ (Κ
−1)
Epure Coa = 89 kJ/molFig.7.(a)IMCthicknessagainstthesquarerootofagingtimeforvariousPbSn/Co samplessubjectedtosolid-stateagingat130–170◦Cupto500h.(b)PlotsoflnK
versus1/TforthedeterminationofthevaluesofEaforIMCgrowth.
of˛-Co(W,P)andpoly-Co(W,P)layersubjectedto1000-h solid-stateaging are about1.7 and 1.84m,respectively. Regardless itscrystallinity,anabout2m-thickelectrolessCo(W,P)layeris hencesuggestedforUBMappliedtoFCbondingutilizingeutectic PbSnsolderjoints.
PbSn/˛-Co(W,P)andPbSn/poly-Co(W,P)samplessubjectedto solid-state agingat130–170◦C upto500hwerecarriedout to determinetheactivationenergy(Ea)ofIMCgrowth.Forthe
pur-poseofcomparison,PbSn/pureCosampleswerealsoinvestigated. AsindicatedbytheplotofIMCthicknessasafunctionofsquareroot ofagingtimeinFig.7(a),thelinearprofilesrevealthe diffusion-controlledIMCgrowthinthetimespanofinvestigation.Itiswell knownthatthethickness-timerelationcanbeexpressedbythe formula[15–17]
x=√Kt (1)
wherex=totalthicknessofIMClayer,t=agingtimedurationand K=constantcorrelatingtothediffusionalgrowthofIMC.Thevalues ofKcanbedeterminedfromtheslopesofplotspresentedinFig.7(a) byemployingtheArrheniusform:
K=Aexp
−Ea kT (2) where A=constant, Ea=activation energy for IMC growth,k=Boltzmann’sconstantandT=absolutetemperature.According tothelnKversus1/TplotshowninFig.7(b),thevaluesofEaare
foundtobe338.6,167.5,and89kJ/molforelectroless˛-Co(W,P), electroless poly-Co(W,P)and pureCo, respectively. Thehighest value ofEa for ˛-Co(W,P)illustrates thebest barrier efficiency,
66 H.-C.Pan,T.-E.Hsieh/MaterialsScienceandEngineeringB177 (2012) 61–68
60
50
40
30
20
10
0
40
60
80
100
120
140
160
180
200
Shear Stress (MPa)
Aging Ti
me
(min
)
PbSn/α-Co(W,P)-10at.%P PbSn/poly-Co(W,P)-6at.%P PbSn/Ni(P)-15at.%P [17] PbSn/Ni(P)-23at.%P [18] PbSn/Ni(P)-17at.%P [18] PbSn/Ni(P)-14at.%P [18]Fig.8.ShearstressesofPbSn/˛-Co(W,P)andPbSn/poly-Co(W,P)samplessubjected toliquid-stageagingforvarioustimes.
inagreementwiththeslowestCothicknessconsumptionandthe enhancementofbarriercapabilityduetothehighPandWcontents insuchanelectrolesslayer.
3.4. Sheartest
Ballsheartestswerecarriedouttoevaluatethemechanical per-formanceofPbSn/Co(W,P) jointssubjected toliquid-state aging atvarioustimesandtheaverageshearstrengthispresented in Fig.8.Thetestresultsreportedbypreviousstudiesregardingof thePbSn/Ni(P)systems[18,19]werealsoaddedinFig.8forthe purposeof comparison.Fig. 8indicates theelectrolessCo(W,P) layersprovideahigherbondingstrengthincomparisonwiththe PbSn/Ni(P)systems.TheshearstrengthofPbSn/˛-Co(W,P)sample increasesfrom95MPaatas-reflowtoapeakvalueof119MPawhen agedtime is increasedto 10min.The shear strengthdecreases slightly to 111MPa in the20-min aged sample and remains a constant valueafter 60-min aging. Thesuppression of bonding strengthhasbeenascribedtotheIMCspallation[20,21].However, accordingtothefailuremodecharacterizationpresentedbelow, thedecreaseofshearstrengthinPbSn/˛-Co(W,P)samplemight becorrelatedtothehighPcontentinelectrolesslayerwhich sup-pressestheadhesionstrengthonCuunderlayer.Ontheotherhand, thePbSn/poly-Co(W,P)sampleexhibitsabettermechanical perfor-mancethattheshearstrengthincreasesfrom95.7MPaatas-reflow to132.8MPaafter1-haging.Yoon[22]andSharif[23]alsoreported thepersistenceofbondingstrengthintheSn–Znsolder/Ni(P) sys-temsubjectedto1-hreflow.Suchaphenomenonhasbeenascribed tothesagging ofsolderduetotheweight,resultinginalarger contactareaandthusahighershearforceafterprolongedreflow [23].Islametal.proposedtheincrementofbondingstrengthis resultedfromthestrengtheningofsolderduetothe homogeniza-tionduringthereflow[24].Notably,poly-Co(W,P)layerpossesses alowerPcontentincomparisonwith˛-Co(W,P)andthereisno IMCspallationinsucha systemasshowninFig.3.Theadverse effectduetothePaccumulationisexpectedtobelessin PbSn/poly-Co(W,P)systemsothat a higherinterfacial bondingstrength is observed.
FailuremodesforPbSn/˛-Co(W,P)andPbSn/poly-Co(W,P) sam-ples identified in terms of the SEM characterizations and the JESD22-B117AStandardareseparatelypresentedinFig.9(a)and (b).Inas-reflowsample,ductilemode(mode#1),i.e.,failurein sol-derbulkregardlessoftheexistenceofdimplefailures[12,25],was observedinmorethan40%ofthe25-testedPbSn/˛-Co(W,P)and about60%ofthe25-testedPbSn/poly-Co(W,P)samples.A repre-sentativeSEMimageofductilefailureisshowninFig.10(a).Failure atsolder/IMCinterfaces,i.e.,theinterfacialbreak(mode#4),was
60 30 20 10 0 0 20 40 60 80 100
Percentage (
%
)
Percentage (
%
)
Aging Time (min)
Pad Lift Ductile Interfacial Break Ball Lift 60 30 20 10 0 0 20 40 60 80 100 Interfacial Break Ball Lift
Aging Time (min)
Pad Lift Ductile
(a)
(b)
Fig.9. Summaryoffailuremodesfor(a)PbSn/˛-Co(W,P)samplesand(b) PbSn/poly-Co(W,P)samples.
observedintherestofas-reflowPbSn/˛-Co(W,P)and PbSn/poly-Co(W,P)samplesinwhichtheexposureofIMCdebriswasobserved inthefracturesurfaceasshowninFig.10(b).Thepercentageof ductilefailuremodeincreasedwiththeincreaseofagingtimein bothsamples,indicatingagradualcompletionofalloyinteractions andthusanimprovementofbondingstrengthinbetweenthe sol-derandCo(W,P)layer.Dominanceofductilefailurealsoindicated theinteractionsofPbSnandCoarefairlystrongthatitensuresthe occurrenceoffracturemainlyinthebulkofsolder.Wenotethe duc-tilefailureinPbSn/˛-Co(W,P)samplessubjectedtoprolongedaging issomewhatdifferentfromthat observedinPbSn/poly-Co(W,P) samples.Asshown by a SEM imageof PbSn/˛-Co(W,P)sample subjectedto20-minaginginFig.10(c),relativelyflatareaswith irregularperipherieswasoccasionallyobservedinthefracture sur-faces.Solderdewettingmightoccuronthehigh-P-contentsurface andconsequentlyweakenthebondingforceofsolderbump.The fracturewouldinitiateatthesolder/P-richlayerinterfaceandthus causetheexposureofflatP-richlayersurfaceswhensolderball isremoved.Degradationofsolderabilityduetotheaccumulation ofPelementsinENlayerhasbeenreportedpreviously[26]and it might bea cause of bonding strength decrementin PbSn/ ˛-Co(W,P)samplesubjectedtoprolongedagingasshowninFig.8. Furthermore,padlift(mode#2)failureemergedinPbSn/˛-Co(W,P) sampleagedformorethan20min.Arepresentativefracture sur-facemorphologyispresentedinFig.10(d).Thisimpliesthehigh Pcontentin˛-Co(W,P)mightdegradeitsadherenceonCu under-layer.Inaddition,thepresenceofCo2Pprecipitatemightharden
the˛-Co(W,P)layerandthusitcouldnotaccommodatethe ther-malstressinducedbytheglassytransitionofCo(W,P)layerwhich underwentarecrystallizationduringaging.Thiscauseda break-ageofCo(W,P)layerand,hence,thepadliftfailureasillustratedin Fig.10(d).
Fig.10.(a)Ductilefailureinas-reflowPbSn/poly-Co(W,P)sample,(b)interfacialbreakinanas-reflowPbSn/poly-Co(W,P)sample,(c)ductilefailureinPbSn/˛-Co(W,P) samplesubjectedto20-minagingand(d)padliftinaPbSn/˛-Co(W,P)samplesubjectedto30-minaging.Thearrowineachmicrographindicatesthesheardirection.
4. Conclusions
In this study, thediffusion barrier characteristics,activation energies of IMC growth and bonding properties of electroless Co(W,P) layers with various crystallinities to PbSn solder are presented. In PbSn/˛-Co(W,P) sample subjected to liquid-state aging,spallationofIMCsintosolder,formationofpolycrystalline P-richlayerandrecrystallizationof˛-Co(W,P)wereobserved.For thePbSn/˛-Co(W,P)subjectedtosolid-stateaging,P-richlayer sim-ilarlyformedattheinterface,however,analyticalresultsindicated suchaP-richlayercannotretardsubsequentCo–Sninteractions. The˛-Co(W,P)layerwasfoundtoserveasacombinedtype,i.e.,the sacrificial-plusstuffed-type,barrierlayer.AstoPbSn/poly-Co(W,P) samples,theCoSn3IMCsneighboringtoanamorphousW-richlayer
wasobservedatthesolder/Co(W,P)interfaceregardlessofaging types.Thepoly-Co(W,P)layerservedmainlyasasacrificial-type barrier.SimilartotheP-richlayer,theamorphismofW-richlayer didnotprovidethebarriercapabilityinthesamples.
The activation energies of IMC growth in PbSn/˛-Co(W,P), PbSn/poly-Co(W,P)andPbSn/pureCosamplesdeducedfromthe solid-state aging were found tobe 338.6,167.5 and 89kJ/mol, respectively.ThisillustratedthepresenceofPandWelementsin electrolesslayersmayenhancethediffusionbarriercapability.
Shear test revealed a dominance of ductile failure in both PbSn/˛-Co(W,P) and PbSn/poly-Co(W,P) samples, indicating a promising applicationofsuchelectrolesslayersasthediffusion barriersinUBMstructure.However,relativelyhighPcontentin ˛-Co(W,P)mightbeaconcernontheapplicationssinceitcaused adverseeffectsonthesolderabilityandwoulddegradethe mechan-icalperformanceofFCbonding.
Acknowledgments
ThisworkwassupportedbytheNationalScienceCouncilof Republic of China under ContractNo. NSC 95-2221-E-009-130. SEM/TEM/EDSanalysesandtheFIBtechniqueforpreparationof XTEM specimens supported by Materials Analysis Technology, Inc.,Chupei,Taiwan,ROC,andsheartestapparatussupportedby SchmidtScientificTaiwanLtd.atHsinchu,Taiwan,ROC,arealso acknowledged.
References
[1] M.A.Nicolet,ThinSolidFilms52(1978)415–443.
[2] R.H.Uang,K.C.Chen,S.W.Lu,H.T.Hu,S.H.Huang,IEEEElectron.Pack.Technol. Conf.(2000)292.
68 H.-C.Pan,T.-E.Hsieh/MaterialsScienceandEngineeringB177 (2012) 61–68
[3]M.W.Liang,T.E.Hsieh, C.C.Chen,Y.T.Hung,Jpn.J.Appl.Phys.43(2004) 8258–8266.
[4]T.Oppert,E.Zakel,T.Teutsch,Proc.IEMT/IMCSymp.(1998)106.
[5]T.Teutsch,T.Oppert,E.Zakel,E.Klusmann,Electron.Comp.Technol.Conf. (2000)107.
[6]G.O.Mallory,J.B.Hajdu,ElectrolessPlatingFundamentalsandApplications, AESF,Orlando,Florida,1990(Chap.1–7).
[7] E.J.O’Sullivan,A.G.Schott,M.Paunovic,C.J.Sambucetti,J.R.Marino,P.J.Bailey, S.Kaja,K.W.Semkow,IBMJ.Res.Dev.42(1998)607–620.
[8] M.W.Liang,H.T.Yen,T.E.Hsieh,J.Electron.Mater.35(2006)1593–1599. [9]A.Kohn,M.Eizenberg,Y.Shacham-Diamand,Y.Sverdlov,Mater.Sci.Eng.A302
(2001)18–25.
[10]A.Kohn,M.Eizenberg,Y.Shacham-Diamand,B.Israel,Y.Sverdlov, Microelec-tron.Eng.55(2001)297–303.
[11]W.C.Wu,T.E.Hsieh,H.C.Pan,J.Electrochem.Soc.155(2008)D369–D376. [12] JEDECStandard,JESD22-B117A.
[13]H.C.Pan,T.E.Hsieh,J.Electron.Mater.40(2011)330–339.
[14]D.A.Porter,K.E.Easterling,PhaseTransformationsinMetalsandAlloys, Chap-man&Hall,London,1981(Chap.1).
[15]V.I.Dybkov,O.V.Duchenko,J.AlloysCompd.234(1996)295–300. [16]A.Sharif,Y.C.Chen,Mater.Sci.Eng.B106(2004)126–131.
[17]D.Q.Yu,C.M.L.Wu,C.M.T.Law,L.Wang,J.K.L.Lai,J.AlloysCompd.392(2005) 192–199.
[18]D.G.Kim,J.W.Kim,S.S.Ha,B.I.Noh,J.M.Koo,D.W.Park,M.W.Ko,S.B.Jung,J. AlloysCompd.458(2008)253–260.
[19]M.O.Alam,Y.C.Chan,K.C.Hung,Microelectron.Reliab.42(2002)1065–1073. [20] C.B.Lee,S.B.Jung,Y.E.Shin,C.C.Shur,Mater.Trans.43(2002)1858–1863. [21]P.C.Shih,K.L.Lin,J.AlloysCompd.422(2006)153–163.
[22] J.W.Yoon,H.S.Chun,S.B.Jung,Mater.Trans.46(2005)2386–2393. [23]A.Sharif,Y.C.Chan,J.AlloysCompd.440(2007)117–121.
[24] M.N.Islam,Y.C.Chan,A.Sharif,M.O.Alam,Microelectron.Reliab.43(2003) 2031–2037.
[25] S.M.L.Nai,J.Wei,M.Gupta,J.AlloysCompd.473(2009)100–106. [26]J.J.Guo,A.P.Xian,J.K.Shang,Surf.Coat.Technol.202(2007)268–274.
