Precipitation of large Ag3Sn intermetallic compounds in SnAg2.5 microbumps after multiple reflows in 3D-IC packaging

ContentslistsavailableatSciVerseScienceDirect

Materials

Chemistry

and

Physics

j our na l h o 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 a t c h e m p h y s

Precipitation

of

large

Ag

3

Sn

intermetallic

compounds

in

SnAg2.5

microbumps

after

multiple

reflows

in

3D-IC

packaging

Ruo-Wei

Yang,

Yuan-Wei

Chang,

Wei-Chi

Sung,

Chih

Chen

a

r

t

i

c

l

e

i

n

f

o

Articlehistory: Received6May2011

Receivedinrevisedform30January2012 Accepted27February2012

Keywords: Intermetallics Pb-freesolderalloys Joining

a

b

s

t

r

a

c

t

MicrobumpshavebeenadoptedasinterconnectsbetweenSichipsin3Dintegrated-circuitpackaging.

Thesoldervolumeofamicrobumpdecreasesdramaticallyduetofine-pitchrequirementanditis

approx-imatelytwoorderssmallerinmagnitudethanthatofatraditionalflip-chipsolderjoint.Themetallurgical

reactionsinthemicrobumpsmaybehavequitedifferentlytothoseinflip-shipbumps.Liquid-state

metal-lurgicalreactionswereexaminedinSnAg2.5microbumpswithNimetallization.Theresultsindicatethat

largeparticlesofAg3Snintermetalliccompounds(IMCs)precipitateaftera10-minreflowonmicrobumps

with4.0-␮m-thicksolder,whichdoesnotoccurwithflip-chipsolderbumps.ItisproposedthattheAg

concentrationintheremainingsoldermayincreaseasSnreactswithNi.TheincreaseintheAg

concen-trationismainlyresponsiblefortheoccurrenceofthelargeAg3Snprecipitates.Theformationofthese

Ag3SnIMCswouldbedetrimentaltothemechanicalpropertiesofthemicrobumps.

© 2012 Elsevier B.V. All rights reserved.

1. Introduction

MetallurgicalreactionsbetweenPb-freesoldersandthe met-allizedpadsonintegrated-circuit(IC)chipshavebeenthefocus ofmuchattentioninrecentyears[1–7].Pb-freeSnAgalloyshave beenadoptedassoldermaterialsinflip-chipsolderjointsdueto theirexcellentmechanicalpropertiesandtheirabilitytobe electro-plated.DispersedAg3Snintermetalliccompounds(IMCs)areused

toenhancethemechanicalpropertiesoftheseSnAgalloys[8–10]. However,large,plate-likeAg3SnIMCsforminsidethesolderjoints

whentheconcentrationofAgishigherthan3.5wt%[11],andcracks mayinitiateattheinterfaceoftheseplate-likeAg3SnIMCsandthe

solderwhenthesolderjointsaresubjectedtostress[1]. There-fore,theICpackagingindustryhasadoptedSnAgalloyswithalow Agconcentration(∼2.5wt%)toavoidtheformationoftheseAg3Sn

IMCs.

Traditionalflip-chipsolderjointshavegreatersoldervolumes thandothosewithunder-bump-metallurgy(UBM).AtypicalSnAg solderjointconsistsofanapproximately100-␮m-thicklayerof solder alloys and several microns to 20␮m of UBM materials [12].Therefore,astheUBMmaterials(CuorNi)reactwithSnto formmicron-sizedCu–SnorNi–SnIMCsduringreflowingor solid-stateaging,thecompositionofthesolderalloysdoesnotchange much.However,asthepackagingindustryhasmovedto three-dimensional(3D)packaging,microbumpshavebeenadoptedasthe

∗ Correspondingauthor.Tel.:+88635731814;fax:+88635724727. E-mailaddresses:[email protected],[email protected](C.Chen).

interconnectsbetweenchips[13–15].Thesoldervolumeshereare dramaticallydecreased,andthethicknessofthesolderisreducedto arangebetweenafewmicronsand10␮m.Themicrobumpvolume isapproximatelytwoordersofmagnitudelessthaninatraditional flip-chipsolderjoint.Incontrast,thethicknessoftheUBMlayers remainsalmostthesameasthatinflip-chipsolderjoints.Therefore, thesoldercompositionmaychangesignificantlyduringmultiple reflows,andthemicrostructuresoftheSnAgsolderalloysmaybe differentinmicrobumps.However,therehavebeennoreported studiesonthisissue.

In this study, we investigated themetallurgical reactionsin SnAg2.5microbumpswithNiUBMsduringreflows.Wefoundthat theAg3SnIMCsprecipitatetoformlargeAg3Snparticlesafter

10-minreflowswith4-␮m-thick SnAg2.5microbumps. Theoretical calculationswereperformedtoshowthattheAgconcentration increasesastheSninSnAgalloysreactswithNi,anditexceeds 3.5wt%aftera10-minreflow.TheincreaseintheAgconcentration mayberesponsiblefortheprecipitationofthelargeAg3SnIMCs.

2. Experimental

Themicrobumpsusedinthisstudyconsistedof4.0-␮mor 6.2-␮mSnAg2.5alloyssandwichedbetween5-␮mCu/3-␮mNiUBMs ontwoSichips,asillustratedinFig.1.The6.2-␮mmicrobump hadmoreamountofSnAg2.5solderthanthe4.0-␮mone.Here, theupperchipisdenotedasthetopchip,whereasthebottomone isdenotedastheinterposerchip.Thesoldervolumeislessthan thetotalUBMvolume.Thediameterofthemicrobumpsis18␮m. TheSnAgalloysandtheUBMmaterialswereelectroplatedandthe 0254-0584/$–seefrontmatter © 2012 Elsevier B.V. All rights reserved.

Fig.1.SchematicdiagramoftheSnAg2.5microbumpswithCu/NiUBMsonboth thetopchipandtheinterposerchip.

microbumpswerejoinedbythermo-compressionat285◦Cfor25s [16].Tostudythemetallurgicalreactionsintheliquidstate,the sampleswerereflowedonahotplatemaintainedat260◦Cfor var-ioustimeperiods.Thereflowtimeswere5,10,and30minforthe 4.0-␮m-thickmicrobumpsand0,5,10,40,and90minforthe 6.2-␮mmicrobumps.Aftertheallottedreflowtime,thesampleswere removedfromthehotplateandcooledinair.Themicrostructures ofthemicrobumpswere examinedwithaJSM-6500F scanning electronmicroscope(SEM).Compositionanalysiswasperformed usingenergydispersivespectroscopy(EDS)andanelectronprobe microanalyzer(EPMA,JXA-800M,JEOL).

3. Resultsanddiscussion

Inthe4.0-␮msample,finelydispersedAg3SnIMCs

agglomer-atedintoafewlargeprecipitatesafterreflowingfor10min.Forthe as-fabricatedsampleshowninFig.2(a),theAg3SnIMCswere

ran-domlydistributedwithintheSnmatrixintheformoftinyparticles. Thesampleshowedaneckingnearthecenterofthesolderjoint, whichiscausedbyalowcompressiveforceduringthebonding process.Thesoldervolumewasestimatedfromthecross-sectional SEMimage,andtheequivalentbumpheightwascalculatedtobe approximately4.0␮m.Fig.2(b)showsback-scatteredSEMimages foranothersamplereflowedfor5min.TheSnintheSnAg2.5 sol-derreactedwiththeNilayerstoformNi3Sn4 IMCsontheboth

sides,resultinginthethickeningoftheNi3Sn4 layers.TheAg3Sn

IMCswerestillfinelydispersedintheremainingsolder.However, largeAg3Snparticlesweresometimesobservedafterreflowingfor

10min,asindicatedbythearrowsinFig.2(c).Therewasalarge Ag3SnIMCprecipitatesemergingfromtheremainingsolderlayer.

Whenthereflowtimewasincreasedto30min,theprecipitation oflargeAg3SnIMCsbecamemoreobvious.Fig.2(d)illustratesthe

microstructureafterreflowfor30min.Here,thesolderlayerwas almostcompletelytransformedintoNi3Sn4IMCs.Itisinteresting

thatlargeAg3Sn IMCswerefrequentlyfoundinside theNi3Sn4

IMCs.ThelargeAg3SnIMCsmaybedetrimentaltothemechanical

propertiesofthemicrobumps.

ItisintriguingthatprecipitationofthelargeAg3SnIMCs did

nothappenwithin40minofreflowwiththemicrobumpswitha 6.2-␮m-thicksolderlayer.Fig.3(a)isanSEMimageshowingthe microstructureof theas-fabricated6.2-␮m-soldermicrobumps. SimilarlytotheresultsinFig.2(a),theAg3SnIMCswerescattered

throughouttheSnmatrix. After5-minand40-min reflows,the interfacialNi3Sn4IMCsgrewthicker,asdepictedinFig.3(b)and(c),

butnolargeAg3SnIMCswereobserved.However,whenthereflow

timewasincreasedto90min,largeAg3SnIMCsstartedtoemerge.

Fig.3(d)showsanSEMimageofa6.2-_␮mmicrobumpaftera 90-minreflow.AlargeAg3Sninclusionappearedintheright-handside

ofthejoint,asindicatedbythearrowinthefigure.Allthesolder hadreactedwithNitoformNi3Sn4IMCsatthisstage.

Atleast50microbumpswereinspectedforeachconditionto ensurethattheobservedcross-sectionalmicrostructureswere rep-resentativeofeachcondition.Herein,alargeAg3SnIMCisdefined

Fig.3.Cross-sectionalback-scatteredSEMimagesof6.2-␮m-thickSnAg2.5microbumpsafterreflowingfor(a)0min,(b)5min,(c)40min,and(d)90minat260◦_C. Table1

ProbabilityofobservingAg3SnIMCslargerthan2␮mforeachspecimeninthestudy.

Reflowtime(min) 4.0-␮mmicrobumps 6.2-␮mmicrobumps

Sampleamount Probability(%) Sampleamount Probability(%)

0 74 6.8 50 0 5 77 7.8 – – 10 95 35.8 53 0 30 90 40.0 – – 40 – – 50 0 90 – – 58 31

asonelongerthan2␮m.TheprobabilityofobservingalargeAg3Sn IMCislistedinTable1.Theprobabilitywasalsoplottedagainst reflowingtime,asdepictedinFig.4.Therewasanapproximately 7%chanceoffindingalargeAg3SnIMCintheas-fabricatedand

5-min-reflowed4.0-_␮msamples.However,theprobabilityincreased significantlyforthe10-min-reflowedsamples,toapproximately 36%,andfurtherincreasedto40%aftera40-minreflow.In con-trast,noneofthe6.2-␮msampleshadlargeAg3SnIMCswhenthe

Fig.4.Thecurvefortheprobabilityagainstreflowingtimeforthe4.0-␮msamples. Thereisanabruptincreasearound10-minreflowingtime.

reflowtimewaslessthan40min;however,theprobabilityroseto 31%afterreflowingfor90min.

TheincreaseinAgconcentrationmayhavecausedthe precip-itationofthelargeAg3SnIMCsatextendedreflowtimes.Asthe

SnreactedwiththeNimetallizationlayer,theamountsofSnin theSnAgsolder decreased.Conversely,Ag doesnot form inter-metalliccompoundswithNi,andtherearenoternarycompounds fortheSn–Ni–Agsystem[8].Therefore,theAgconcentrationin thesolderwouldbeexpectedtoincreasewithincreasingreflow time.IthasbeenreportedthatlargeAg3SnplatesforminSnAg

sol-deratAgconcentrationsover3.5%[1,11].Thus,largeAg3SnIMCs

wereobservedinSnAg2.5microbumpsatextendedreflowtimes. Forflip-chipsolderjoints,thesoldervolumeisapproximately100 timesgreaterthanthatofthemicrobumps.Therefore,theAg con-centrationremainsunchangedaftermetallurgicalreaction;there havebeennoreportsoftheincreaseinAgconcentrationafter met-allurgicalreactionsinconventionalflip-chippackaging.However, thisissuewillbecriticalformicrobumpsin3DICpackaging.

ToverifywhethertheAgconcentrationisgreaterthan3.5wt% after a specific reflow time, thefollowing calculation was pre-formed to examine the evolution of Ag concentration in the remainingsolder.Itisnoteworthythattheremaynotbeacritical Agconcentration,abovewhichtheprecipitationofAg3Snwould

occur.Thisisbecausetheprecipitationalsodependsonthecooling rate.Wewilldiscussthispointlater.Yet,inthispaper,wechoose theeutecticconcentrationasthecriticalconcentration,sincethe Ag3SnprecipitationwasfoundinSn3.5AgandSn3.8Ag0.7Cu

sol-ders[1,11].BecauseAgdoesnotreactwithNi,alltheAgatoms shouldremainintheSnAgsolder,asrevealedinFigs.2and3.The AgtherereactswithSntoformAg3SnIMCswhenthespecimensare

Fig.5. ThemeasuredthicknessofNi3Sn4IMCsagainstreflowingtimeforthechip side,theinterposersideandthesumofthebothsides.

cooled.However,accordingtoSn–Agphasediagram[8],alltheAg atomsshoulddissolveinthemoltenSnAgsolderat260◦C.Froma massbalanceofAgatoms,theAgconcentrationCintheremaining soldercanbeexpressedas:

C= VssI

Vss−Vssf

(1) where Iis theoriginal Agconcentration inthe SnAg2.5 solder, Vs is theoriginalvolume ofthesolder, s isthedensity ofthe

SnAg2.5solder(7.34gcm−3),iisthedensityoftheNi3Sn4IMCs

(8.64gcm−3[9]),ViisthevolumeincreaseoftheNi3Sn4IMCs,and

fistheweightfractionofSnintheNi3Sn4IMCs(here72.93%).The

shapesofbothsolderandtheNi3Sn4layersarenearlycylindrical,

asshowninFig.3.BecausethediametersoftheNi3Sn4IMCand

solderlayersareroughlythesame,Eq.(1)maybereducedto: C= hssI

hss−hiif (2)

wherehsistheoriginalthicknessoftheSnAgsolder,andhiisthe

increaseinthicknessoftheNi3Sn4layeronbothtopandbottom

sides.Inaddition,thetime-dependentthicknessoftheNi3Sn4layer

canbeexpressedas:

ht−h0=(kt)1/2 (3)

whereh0istheoriginalthickness,htisthethicknessafterreflowfor

tmin,tisreflowtimeinminutes,andkisthegrowth-rateconstant. Fig.5(a)showsthemeasuredNi3Sn4thicknessasfunctionof

reflowtimeonthetopandinterposersidesforthe6.0-␮msamples. ThethicknessoftheNi3Sn4IMCsontheinterposersideisthicker

thanthatonthechipside.Theexactreasonforcausingthis differ-enceisnotclear.Itmayberelatedtothebondingprocess.Thetotal thicknessrepresentsthesumofthethicknessonthebothsides.

Fig.6. TheevolutionofAgconcentrationintheremainingsolderaftervariousreflow timesforboththe4.0-␮mand6.2-␮mmicrobumps.TheAgconcentrationexceeded 3.5wt%after13.0minwiththe4.0-␮m-thickmicrobumps,whereasittook25.0min todosowiththe6.2-␮m-thickmicrobumps.

Therefore,wecanplottheincreaseintheNi3Sn4thicknessagainst

thesquarerootofreflowtimeandtheresultsareshowninFig.5(b). Withthefittingcurve,therateconstantwascalculatedtobe0.44. CombiningEqs.(2)and(3),thetime-dependentAgconcentration canbeexpressedas:

C= I

1−((kt)0.5if/hss)

(4) Fig.6showsthecalculatedAgconcentrationintheremaining solderas afunction ofreflowtime forthe4.0-␮mand 6.2-␮m microbumpsusingtherateconstantfromFig.4.Withthe4.0- ␮m-thickmicrobumps,theAgconcentrationincreasestoabove3.5wt% afterareflowofapproximately9.0min,whereasittakes approxi-mately22.0minfortheAgconcentrationtoexceed3.5wt%withthe 6.2-␮m-thicksolder.Thisisbecausethe4.0-␮m-thickmicrobump simplyhaslesssolder.Therefore,theconsumptionofSnhasamore obviouseffectontheAgconcentration.

Thecalculateddatafittheexperimentalresultsquitewell.As shownin Fig.4,theprobabilityforobserving largeAg3Sn IMCs

increasedsignificantlyaftera10-minreflowforthe4.0-␮m-thick microbumps.ThecalculateddataindicatethattheAg concentra-tionwas over 3.5wt% afteran 13.0-min reflow. Approximately 28.5wt%oftheSninthesolderwasconsumedtoformNi3Sn4IMCs,

resultingintheAgconcentrationexceeding3.5wt%inthe remain-ingsolder.Additionally,theexperimentalresultsshowedthatno largeAg3SnIMCswerefoundaftera40-minreflowwiththe

6.2-␮m-thickmicrobumps,buttheprobabilityroseabruptlyaftera 90-minreflow.Notethatherethecalculatedvaluesshowthatthe Agconcentrationshouldexceed3.5wt%aftera25.0-minreflow. Thisdifferencemaybeattributedtotheassumptionofaconstant reactionrateinthecalculation.Inreality,thereactionrateslowsas thereflowtimeisincreased.Thismayberesponsibleforthe dis-crepancybetweentheexperimentalandtheoreticalresultsforthe 6.2-␮m-thickmicrobumps.

Fig.7showsthecalculatedconcentrationofAginthe remain-ingsolderasafunctionofsolderthicknessafter10-minreflowtime whenSn2.5Ag,SnAg2.0,andSnAg1.5Agsolderswereadopted.The reasonwechoose10ministhatsolderbumpsneedstopass relia-bilitytestsafterapproximately10-minreflow.Theresultsindicate thatthethinnerthesolder,themorerapidintheconcentration increaseofAgintheremainingsolder.Forexample,whenthesolder thicknessis14.0-␮mthick,theconcentrationofAgonlyincreases from2.5wt%toapproximately2.75wt%after10-minreflow. How-ever,whenthesolderthicknessdecreaseto4.0-␮mthick,theAg concentrationwouldincreaseover3.5wt%after10-minreflow.The calculatedcriticalsolderthicknessis4.2_␮mfortheSnAg2.5solder, belowwhichtheconcentrationoftheAgintheremainingsolder wouldexceed3.5wt%after10minreflow.

Fig.7.ThecalculatedAgconcentrationintheremainingsolderagainstthicknessof solderafter10-minreflowforSnAg2.5,SnAg2.0andSnAg1.5solders.

TheaboveresultsalsosuggestthatreducingtheAg concentra-tioninSnAgsolderwilleasetheprecipitationofthelargeAg3Sn

IMCs.Fig.7alsoshowsthecalculatedconcentrationofAginthe remainingsolder asafunction ofsolder thicknessafter10-min reflowtimeforSnAg2.0,andSnAg1.5Agsolders.Thecriticalolder thicknessis2.8and2.1_␮mfortheSnAg2.0andSnAg1.5solders, respectively.Thatis,whenSnAg2.0solderisused,theprecipitation ofAg3Snmayoccurafter10minreflowwhenthesolderthickness

islessthan2.8␮m.

Shenetal.studiedthegrowthmechanismofAg3Snprecipitation

inSnAgsolder[17].TheyproposedthattheprimaryAg3Sn

parti-clesformbeforetheonset oftheeutecticreaction,moreAg3Sn phasecouldnucleateadjacenttotheprimaryAg3Snprecipitation,

resultingintheformationoflargeAg3SnIMCs.It isnoteworthy

thatcoolingratesmayhaveaprofoundeffectontheprecipitation ofAg3SnIMCs[9–12,18].BecauseAg3SnIMCsareformedduring

thecoolingprocess,aslowcoolingratefacilitatestheformationof largeAg3SnIMCs.Inthisstudy,thesampleswerecooledinairafter

reflowatafastercoolingratethanistypicallyusedinthe packag-ingindustry.Therefore,theprecipitationoflargeAg3SnIMCscould

beacriticalreliabilityissueformicrobumpsifthinsolderlayers areadopted.Inmicrobumpswithfewamountofsolder,thesolder reactswithNitoformNi3Sn4IMCsafterreflowandtheentiresolder

layermaytransformintoIMCs,asshowninFigs.2and3.TheAg3Sn

particlesadheretotheNi3Sn4IMCs.Therefore,theAg3Sn/Ni3Sn4

interfaceswillplayacrucialroleinthemechanicalpropertiesof microbumps.Thisdeservesmorefuturestudies.

4. Conclusion

Insummary,weinvestigatedtheliquid-statemetallurgical reac-tionsin microbumpswith two solder thicknesses. LargeAg3Sn

precipitatesemergedaftera10-minreflowwiththe4.0-␮msolder microbumps.Whenthesolderthicknesswasincreasedto6.2␮m, nolargeAg3SnIMCswerefoundinthesamplesforreflowtimes

below40min.TheoreticalcalculationsindicatedthattheAg con-centrationincreaseswithreflow,becauseSnreactswiththeNiUBM toformNi3Sn4IMCs,thusincreasingtheAgconcentrationinthe

remainingsolderwithreflowtime,whichfacilitatestheformation oflargeAg3Snprecipitatesduringthecoolingstage.Theselarge

Ag3Sninclusionsmaybedetrimentaltothemechanicalproperties

ofthemicrobumps;additionalstudyisneededtoinvestigatethese effects.

Acknowledgements

The authors would like to thank the Ministry of Economic Affairs,R.O.C.fortheirfinancialsupportundergrantno. 98-EC-17-A-05-S2-0051.Inaddition,theauthorsthankDr.Tao-ChihChang, Mr. Chau-Jie Zhan, and Mr. Jin-Ye Juang in Industrial Technol-ogyResearch Instituteof Taiwan for providing themicrobump samples.

References

[1]K.Zeng,K.N.Tu,J.Mater.Sci.Eng.Rep.38(2002)55.

[2] K.S.Kim,S.H.Huh,K.Suganuma,J.AlloysCompd.352(2003)226. [3]S.J.Wang,C.Y.Liu,Scr.Mater.49(2003)813.

[4]N.S.Liu,K.L.Lin,J.AlloysCompd.456(1)(2008)466.

[5]C.F.Yang,F.L.Chen,W.Gierlotka,S.W.Chen,K.C.Hsieh,L.L.Huang,Mater.Chem. Phys.112(1)(2008)94.

[6]S.W.Chen,W.Y.Lee,C.M.Hsu,C.F.Yang,H.Y.Hsu,H.J.Wu,Mater.Chem.Phys. 128(2011)357.

[7]C.E.Ho,Y.W.Lin,S.C.Yang,C.R.Kao,D.S.Jiang,J.Electron.Mater.35(5)(2006) 1017.

[8] S.K.Kang,A.K.Sarkhel,J.Electron.Mater.23(8)(1994)701. [9]K.S.Kim,S.H.Huh,K.Suganuma,Microelectron.Reliab.43(2003)259. [10]C.P.Huang,C.Chen,C.Y.Liu,S.S.Lin,K.H.Chen,J.Mater.Res.20(2005)2772. [11] J.W.Jang,D.R.Frear,T.Y.Lee,K.N.Tu,J.Appl.Phys.88(2000)6359. [12] C.Chen,H.M.Tong,K.N.Tu,Annu.Rev.Mater.Res.40(2010)531. [13]K.N.Tu,Microelectron.Reliab.51(2011)517.

[14]R.Agarwal,W.Zhang,P.Limaye,R.Labie,B.Dimcic,A.Phommahaxay,P. Soussan,Proceedingsofthe60thElectronicComponentsandTechnology Con-ference,LasVegas,NV,USA,June,2010,pp.858–863.

[15]J.Hwang,J.Kim,W.Kwon,U.Kang,T.Cho,S.Kang,Proceedingsofthe60th ElectronicComponentsandTechnologyConference,LasVegas,NV,USA,June, 2010,pp.1399–1403.

[16]C.J.Zhan,C.C.Chuang,J.Y.Juang,S.T.T.C.Lu,T.C.Chang,Proceedingsofthe60th ElectronicComponentsandTechnologyConference,LasVegas,NY,USA,June, 2010,pp.1043–1049.

[17] J.Shen,Y.C.Chen,S.Y.Liu,Intermetallics16(2008)1142.

[18]R.S.Sidhu,S.V.Madge,X.Deng,N.Chawla,J.Electron.Mater.36(12)(2007) 1615.