reflow profile study of snagcu

Re owpro lestudyoftheSn-Ag-Cusolder

B.Salam

SchoolofEngineering,UniversityofGreenwich,C.Virseda

EuropeanFuelCellH.Da

SchoolofEngineering,UniversityofGreenwich,N.N.Ekere

SchoolofEngineering,UniversityofGreenwich,R.Durairaj

SchoolofEngineering,UniversityofGreenwich,

Kent,UK

Kent,UKKent,UKKent,UK

Keywords

Soldering,Surfacetexture,Surfacetreatment

Introduction

Tinleadsoldersaretheprimarymaterialsusedfor

interconnectingelectroniccomponents.However,thesafeuseanddisposaloflead-containingelectronicproductsisanissuethatiscausingamoveintheconsumerelectronicindustrytoremoveleadfromproducts.Itiswidely

anticipatedthattheuseofleadcontainingsoldersbytheelectronicindustrywillbeseriouslyconstrainedbya

legislativebanonleaduseinsolders.AnexampleofsuchlegislationistheenvironmentallyconsciousengineeringinelectronicscommitteeinJapan,whichhasscheduledthatlead-freesoldersshouldbestandardisedfrom2003.

Furthermore,EuropeanlegislationundertheWastefromElectricalandElectronicEquipment(WEEE)and

RestrictionofHazardousSubstances(RoHS)Directivesisscheduledtoeliminatetheleadfromelectronicproductsbytheyear2006.Besidesthislegislation,someEuropeancountrieshaveconsideredimposingnewregulations

requiringthemanufacturerstotakefullresponsibilityfortherecyclingoftheirproducts.Inaddition,switchingtolead-freesolderingcanbringsomeadvantages,inparticularbyimprovingthereputationofcompanies,whoareenvironmentallyconscious.

Theindustryhasembarkedonanumberofstudiesinsearchofsuitablelead-freealternatives,butyettherearenodrop-insolutionswithrespecttore owtemperature,jointreliabilityandassemblycosts.OursurveyshowsthattheSn-Ag-Cualloyisoneofthemostpromisinglead-freealloyscurrentlybeingevaluatedbytheindustry.However,Sn-Ag-Cuisahighmeltingpointalloy(2178C).Furthermore,themaximumsolderingtemperatureofelectroniccomponents/partsisoften xed.Hence,thesafetywindowisreducedinsize.Therefore,thetemperaturepro lingforthisalloybecomesmoredif cultandchallengingthanfortheeutecticSn-Pballoy.

AstudyoftheSn-Ag-CualloyswasconductedbyHwang(2001).ThecompositionofthestudiedSn-Ag-Cualloyswas96.5-92.3percentSn,3-4.7percentAgand0.5-3percentCu.ThestudyfoundthatthemicrostructureoftheSn-Ag-CualloysconsistedofCu6Sn5andAg3SnintermetallicsinaSn-matrix.Theseintermetallics(theCu6Sn5andAg3Sn)werefoundtoeffectivelystrengthenthealloyandactasabarriertofatiguecrackpropagation.Inaddition,theyalsoactaspartitionswithintheSn-matrixgrains,therebyproducinga nermicrostructure.The nertheAg3SnandCu6Sn5intermetallicsare,themoreeffectivelytheypartitiontheSn-matrixgrains,resultinginanoverall nermicrostructurethatcanfacilitategrainboundaryslidingmechanisms,andwhichinturnwilllengthenthelifeofasolderjoint.

ThepresenceoftheseintermetallicsinthebulksolderforthetraditionalSn-Pballoyishardlyseen;hence,http://wendang.chazidian.com/researchregister

Abstract

AstudyoftheSn-Ag-Culead-freesolderre owpro lehasbeenconducted.ThepurposeoftheworkwastodeterminetheSn-Ag-Cure owpro lethatproducedsolderbumpswithathin

intermetalliccompound(IMC)layerand nemicrostructure.Twotypesofre owpro leswerestudied.Theresultsoftheexperimentindicatedthatthemostsigni cantfactorin

achievingajointwithathinIMClayerand nemicrostructurewasthepeaktemperature.Theresultssuggestthatthepeak

temperaturefortheSn-Ag-Culead-freesoldershouldbe2308C.Therecommendedtimeaboveliquidusis40sfortheRSSre owpro leand50-70sfortheRTSre owpro le.

Received:17March2003Revised:6November2003Accepted:6November2003

microstructureofaSn-Ag-Culead-freesolderjointis

differentfromthatofatraditionalSn-Pbsolderjoint.Sincemostofthecurrentlyavailablestudiesofthere owpro learebasedontheSn-Pbsolderalloy,thein uenceof

differenttypesofthere owpro leonthemicrostructureofaSn-Ag-Cusolderjointisunknown.Thus,thisisoneoftheobjectivesofthispaper.

There owpro lecouldaffectthereliabilityofasolderjoint,becauseitisoneofthefactorsthatin uencetheformationoftheintermetalliclayersinasolderjoint.Theintermetalliclayerisinfactacrucialpartofasolderjoint.Althoughhavinganadvantageinfacilitatingbonding

betweenthesolderandsubstrate,ithasadisadvantagethatitisgenerallythemostbrittlepartofthesolderjoint(Frear,1991).Thus,itmustbeasthinaspossible.Therefore,agoodre owpro lemustproducesolderbumpswithathinintermetalliclayer.

Re owpro lingwasextensivelystudied,forexample,byLee(1999),SkidmoreandWaiters(2000),SuganunaandTamanaha(2001),Suraski(2000)andYangetal.(1995).Suraski(2000)studiedtheramp-to-spike(RTS)re owpro le.SuganunaandTamanaha(2001)discussedtheavailablere owtechnologyforlead-freesoldering.Experimentswerecarriedouttodeterminewhich uxchemistries,lead-freealloysandre owpro leshadthegreatestin uenceonsolderjointqualityintermsofwettingability,solderballs,soldersplashesandvoids(SkidmoreandWaiters,2000).AstudyofeutecticSn-Agre owpro leshasalsobeenreportedbyYangetal.(1995),whostudiedtheeffectofthesolderingtemperature,solderingtimeandcoolingrate.Astudyanalysingthetypesofdefectsaffectedbythere owpro lehasalsobeenreportedbyLee(1999).Someofthesestudies,suchasYangetal.(1995),arecloselyrelatedtotheworkreportedhere.However,therehavebeennoreportsoftheeffectofthere owpro leontheintermetalliclayerthicknessandmicrostructurefortheSn-Ag-Cualloys.Therefore,thispaperpresentsastudyofthere owpro lefortheSn-Ag-Cualloys.ThegoaloftheworkwastodetermineaSn-Ag-Cure owpro lethatresultsinathinintermetalliccompound(IMC)layerand nemicrostructuresolderjoint.

Experimentaldesign

Theexperimentstudiedtwotypesofre owpro les.Theyweretheramp-soak-spike(RSS)andramp-to-spike(RTS)pro les.Examplesofthesepro lesareshowninFigure1.Eachofthesepro leswasinvestigatedusingasetof

factorialdesignexperiments.ThedesignparametersfortheRSSpro lewerethesoaktemperature,http://wendang.chazidian.com/0954-0911.htm

Soldering&SurfaceMountTechnology

16/1[2004]27–34

qEmeraldGroupPublishingLimited

[ISSN0954-0911]

[DOI:10.1108/09540910410517022]

[27]

B.Salam,C.Virseda,H.Da,N.N.EkereandR.Durairaj

Re owpro lestudyoftheSn-Ag-Cusolder

Soldering&SurfaceMountTechnology

16/1[2004]27–34

[28]

Figure1

Typesofre ow

pro les

andcoolingrate.TheparametersfortheRTSpro lewereThecombinationsoftheseparameterswereassignedtothepeaktemperature,timeaboveliquidusandcoolingrate.theL4(23)orthogonalarrayasshowninTableII.Therefore,ThepurposeofthesoaktemperatureintheRSSpro leiseachtypeofre owpro lehadfourcombinationsoffactorstoprovideheattoanassemblygraduallyanduniformly.Intobetested,asshowninTableIII.

generalapplicationswithSn-Pbsolder,thesoaktemperatureisbetween150and1708C(SkidmoreandWaiters,2000).Experimentalprocedure

However,thesoaktemperatureforthehigh-temperaturesolderalloysisbetween170and1908C(Lee,1999).ThisThetestvehiclesusedinthisstudywereFR4substrateswithsoaktemperaturesettingiscriticalbecausetheassemblythreesolderbumpson2.5mmdiametercircularCupads,astemperatureshouldberaisedsmoothlyanduniformlytotheshowninFigure2.Theproceduretomakethesolderbumpspeaktemperaturesothatthetemperaturedifferentialstartedfrommanuallyprintingthesolderpasteusingabetweenthecomponentsisminimised.ThesoakstencilafterthebareCupadsofthetestvehiclehadbeentemperaturewasthereforeinvestigated.

scrubbedandcleanedusingiso-propanol.ThesolderpasteSolderpastemanufacturersrecommendthatthepeakusedforprintingwasSn-3.8Ag-0.7Cu,withatype3particletemperatureisashighaspossibletoaidwetting,butsizedistributionand89wtpercentmetalcontent.Thecomponentmanufacturersadvisetokeepthelowestpeakstencilwas0.7mmthickandhadthreeroundaperturesoftemperaturepossibletopreventcomponentdamage

2.5mmdiameter.Toensureconsistentprinting,printing(WickhamandHunt,2001).Alow-peaktemperatureisalsotestswereconducted.TheheightofthesolderpastedepositbelievedtocreateathinnerIMClayer(Yangetal.,1995).producedbytheprintingtestshowedgoodyieldwith20mmInthisstudy,twopeaktemperatureswereinvestigated:deviationsbetweenthedeposits.Thisrepeatableandgoodhighð250^58CÞandlowð230^58CÞ:

yieldmanualprintingwaspossiblebecausethesizeoftheExtendingthetimeatpeaktemperaturepermitsanystencilaperturewaslargeandtheparticlesizedistributionofcomponentwithalargeheatcapacitytoreachtherequiredthesolderpastewas25-45mm(type3).Thevolumeofthere owtemperature.Thebestresults,intermsofgood

solderbumpsafterre owwasestimatedfromthevolumeofwettability,nosolderballsandnovoids,havebeenfoundtothesolderpasteprintedandthesolderalloyvolumefractionoccurwithaRTSpro lewithtimeaboveliquidusof

ofthepastewasfoundtobe1:66^0:1mm3:

90-120s(SkidmoreandWaiters,2000).However,toomuchThenextprocedurewasformingthesolderbumpsbyheatinputabovethesoldermeltingpointleadstoexcessivere owingthesolderpastedepositaccordingtothelistofintermetallicformation.Therefore,twodifferenttimesre owpro lesinTableIII.Aforcedconvectionre owovenaboveliquidusfortheRTSpro lewereinvestigated:short(batchtype)wasusedtore owthem.The rststepinthis(40-60s)andlong(90-120s).FortheRSScase,thetimewassetting-upthere owovensothatthetestvehiclecouldabovethemeltingpointmustbekeptaslowaspossible.Inbere owedfollowingtherequiredtemperaturepro les.There owpro lingprocedureisdividedintotwosteps:

thispro le,thetimeaboveliquidusdependsonthepeakthermocoupleattachmentprocessandovenset-up.Threetemperature,e.g.forthepeaktemperature2308C,thelowestthermocoupleswereusedtomeasurethePCBtemperatures.timeaboveliquiduswas40s,whereasforthepeakAthermocouplewasattachedoneachpadofthetestvehicletemperature2508C,itwas60s.

asshowninFigure2.Ahighmeltingpointsolder(88Pb/A nemicrostructuresolderjointcouldbeproducedby10Sn/2Ag)wasusedtoattachthethermocouplesafter

afastcoolingrate(Prasad,1989).However,electronicscrubbingandcleaningthepadswithiso-propanol.Oncethecomponentsaregenerallyvulnerabletothermalshockthermocoupleswereattachedtothespecimen,there owcausedbyafastcoolingrate;thus,itisnecessarytoovenwasset-uptocreatetherequiredthermalpro leinvestigatewhichcoolingrateshouldbeappliedandits(TableIII).Theovenparameters,adjustedtocreatetheeffecttowardsthemicrostructureoftheSn-Ag-Culead-freethermalpro le,were:zonesetpointtemperaturesandsolderalloys.Therefore,twocoolingrateswerestudied:conveyorspeed.Thepro lingisaniterativeprocessslow,28C/sandfast,48C/s.

comparingthetestboardtemperatureresultswiththe

Insummary,theexperimenthasthreefactorswithtwodesiredpro le.Iftheresultsdiffer,theovenparametersarelevels,thedetailsofwhichcanbeseeninTableI.

adjustedandthetestisrepeated.

TableI

ExperimentalparametersFactorno.Highlevel(H)Lowlevel(L)RSSre owpro le1Soaktemperature

175-1908C

150-1758C

2Peaktemperatureandtimeaboveliquidus

250^58Cand60^15s

230^58Cand40^15s

3Coolingrate

Fast

Slow

RTSre owpro le1Peaktemperature250^58C230^58C2Timeaboveliquidus

90-120s50-70s3

Coolingrate

Fast

Slow

B.Salam,C.Virseda,H.Da,TableII

N.N.EkereandR.Durairaj

TheLRe owpro lestudyoftheSn-Ag-4(23)orthogonalarrayCusolder

Factornumber

Soldering&SurfaceMountExperimentno.12

3Technology

16/1[2004]27–34

1HHH2HLL3LHL4

L

L

H

TableIII

Parametercombinationsofthere owpro leSoaktemperature

Peaktemperatureandtimeaboveliquidus

RSSpro le(8C)

(8C,s)

Coolingrate

1175-190250^5and60^15Fast2175-190230^5and40^15Slow3150-175250^5and60^15Slow4

150-175

230^5

and

40^15

FastRTSpro lePeaktemperature(8C)

Timeaboveliquidus(s)

5250^590-120Fast6250^550-70Slow7230^590-120Slow8

230^5

50-70

Fast

Figure2

Testvehicle

design

Afterre ow,thespecimenswerecross-sectionedtoobservethemicrostructureandmeasuretheintermetallicthickness.Someofthespecimenswereagedinaclimaticchamberat1508Cfor300htoaccelerategrowthoftheIMClayerandevolutionofthemicrostructure.Theagedspecimenswerealsocross-sectioned.

TheIMClayerthicknesswasmeasuredusingasoftwareattachedtothemicroscope.OntheIMCphototakenalongthebumpsata1,000magni cation,apoly-linewasdrawnalongthetwobordersoftheIMClayer,i.e.atboththesolderandthesubstrateside.Theborderatthesoldersideismostlynon-planarandtheborderatthesubstratesideisoftenplanar.Afterthelinesweredrawn,thesoftwarecalculatedthemaximumdistance,minimumdistanceandaveragedistancebetweenthetwolines.Thesestepswererepeatedthreetimesoneachbump(middlepart,leftsideandrightside)toensureconsistency.TheIMClayerthicknessesreportedinthispaperaretheaveragethickness.

Sixsolderbumpswereformedwitheachre owpro leand,oneachbump,therewerethreemeasurementsoftheIMCthicknesshencetherewere18measurements(N)foreachre owpro le.ThedeviationofthemeasuredIMClayerthicknessvaried,especiallybetweentheas-solderedandtheagedbumps.ThestandarddeviationoftheIMCthicknessfortheas-solderedbumpswasbetween0.086and0.5mmandthatoftheagedbumpswasbetween0.128and0.8mm.Thecon dencelevelofthedatacouldbe

determinedbysubstitutingthehigheststandarddeviations(s)obtainedforeachconditionofthebumpsinthefollowingequation(Ott,1988):

r zNE2

a=2¼

s2ð1ÞTheintervalwidth(E)ischosenas0.25becauseitisareasonablevalueforthefactsthatthethinnestintermetallic

thicknessisaround1mmandthehigheststandarddeviationofthedatais0.8.Furthermore,iftheintervalwidthvalueistoolow,thenumberofsamples(N)havetobeincreasedandiftheintervalwidthvalueistoohigh,itwillnotbe

reasonable,e.g.1^0:8mmðy^EÞ:Thus,thevalueofzcalculatedandconvertedtothecon dencelevelsa=2couldbeofthemeasureddata(Ott,1988)which,fortheas-solderedandagedbumps,were97and82percent,respectively.Inotherwords,weare97percent(fortheas-solderedbumps’data)and82percent(fortheagedbumps’data)surethatthemeasuredintermetallicthicknesseswillbewithinthe

intervalof^0:25mmofthevalueslistedinTablesIVandV.

Resultsanddiscussion

Theresultsofthisstudyaredividedintotwoparts:themicrostructureevaluationandtheIMClayerthicknessmeasurements.

Microstructures

MicrostructureswithintheSn-Ag-Cualloyshavebeenanalysedandidenti edearlierbyHwang(2001)andMoonetal.(2000).Hence,inthispaper,themicrostructuresoftheSn-Ag-CusolderbumpswererecognisedbasedontheirshapesandidentitiesasdescribedbyMoonetal.and

Hwang.TheresultsofthebulkmicrostructureobservationsaresummarisedinTablesIVandV.Themicrostructuresoftheas-solderedbumpsgenerallyshowedlargeareasoftin(Sn)populatedbysmallAg3SnandCu6Sn5intermetallic akes.The nestintermetallic akeswereobservedintheas-solderedbumpsformedwithre owpro le4.AnexampleofthemicrostructurescanbeseeninFigure3.The

microstructuresoftheagedbumpsweregenerallysimilartotheas-solderedbumps,exceptthattheintermetallicsweremoreuniformlydistributedcomparedwiththoseintheas-solderedbumps(Figure4).Hence,theislandsofSnintheagedbumpswerereducedinsize.Theagedbumpsformedwithre owpro le4stillhadthe nestintermetallic akes.Thepresenceofthevery neintermetallicsinthebulksolderofthebumpsformedwithre owpro le4mightbecausedbythelowpeaktemperature,shorttimeabove

liquidusandfastcoolingrate.These neintermetallicswerenotpresentinthebumpsformedwiththeRTSre owpro le,eventhough(RTS)re owpro le8hadparametersverysimilartothe(RSS)re owpro le4.ThismightbebecausetheRTSpro ledidnothaveasshortatimeaboveliquidusastheRSSpro le.Inaddition,theRTSpro lewillneverhavethatshorttimeaboveliquidus,becauseitrequiresalongertimetoachieveaneventemperaturedistributionacrosstheassembly.

TheinterfaceintermetallicsbetweenthebumpandCusubstrateareshowninFigures5-12.The guresshowthatpro les1,2,5and7producedaveryirregularIMClayermorphologyandpro les3,4,6and8formedamore

uniformIMClayer.Frear(1991)reportsthattheslowerthedissolutionratethemoreplanaristheinterfacial

intermetallics.Thisindicatesthatthosere owpro lesmadetheCudissolvefasterintothesolder.FromTableIII,thesimilaritiesbetweenthosepro lescouldbesummarized.Pro les1and2havethesamesoaktemperature(175-1908C)asdopro les3and4(150-1708C),andpro les5and7havethesametimeaboveliquidus(90-120s)asdopro les6and8(50-70s).ItseemsthatthesoaktemperatureintheRSSpro leandthetimeaboveliquidusintheRTS

pro learethefactorsmostaffectingthedissolutionrate.Furtherextensiveinvestigationisneededtocon rmthis nding.

IMClayerthickness

TheIMClayerthicknessesofthesamplesareshowninFigures5-12andTableIV.ThemeasuredIMCthicknessesinTableIVareapproximatelyinagreementwiththat

reportedbyHwang(2001).ThetotalIMCgrowthkineticfortheSn-Ag-Cualloywasparabolicandcanbeexpressedbythefollowingequation(Hwang,2001):

[29]

B.Salam,C.Virseda,H.Da,N.N.EkereandR.Durairaj

Re owpro lestudyoftheSn-Ag-Cusolder

Soldering&SurfaceMountTechnology

16/1[2004]27–34

[30]

TableIV

IMClayerthicknessandmicrostructurefortheRSSre owpro leIMCthicknessandRSSpro leno.standarddeviation(mm)

IMCshapeMicrostructures

Non-aged12.23^0.2VeryirregularLargeareaofSn+alotoflongAg3Sn+Cu6Sn5

21.65^0.1VeryirregularSmallerareaofSnthanno.1+alotoflongAg3Sn+Cu6Sn532.26^0.2IrregularLargeareaofSn+alotofsmallAg3Sn41.19^0.1Irregular

LargeareaofSnandvery neAg3SnAged14.56^0.4Smoothlayer,withLargeareaofSn+smallCu6Sn5andAg3Snseveralspikesmicrostructure23.84^0.5Smoothlayer,withLargeareaofSn+ neCu6Sn5andAg3Snseveralspikesmicrostructure34.20^0.3SmoothlayerLargeareaofSn+smallCu6Sn5andAg3Snmicrostructure4

3.09^0.8

Smoothlayer

LargeareaofSn+very neCu6Sn5andAg3Sn

microstructure

TableV

IMCthicknessandmicrostructurefortheRTSre owpro leIMCthicknessandRTSpro leno.standarddeviation(mm)

IMCshapeMicrostructures

Non-aged52.25^0.4VeryirregularLargeCu6Sn5andAg3Sn62.49^0.5Irregular

LargeCu6Sn5andAg3Sn

71.66^0.2VeryirregularLargeareaofSnandsmallsizeofAg3Sn81.37^0.1Irregular

Large

areaofSnandsmallsizeofAg3Sn

Aged54.65^0.4Smoothlayer,withLargeareaofSn+smallCu6Sn5andAg3Snseveralspikes64.34^0.8Smoothlayer,withLargeareaofSn+smallCu6Sn5andAg3Snseveralspikes73.26^0.1Smoothlayer,withLargeareaofSn+ neCu6Sn5andAg3Snseveralspikes8

3.55^0.5

Smoothlayer

LargeareaofSn+very neCu6Sn5andAg3Sn

Figure3

Microstructuresofnon-agedsolderjointsfordifferentre ow

pro les

effect,varianceanalysiswasconducted,asshownin

X¼X0þ1:78£1022t

0:52

exp

257;700 ð2Þ

TableVI.Theeffect guresinTableVIarethedifferencebetweentheaverageIMCthicknessesforthelowandhighwhereXisthetotalIMClayerthickness(m)afterageingforvalueforeachvariableandthepercentage guresarethetimet(s);XratiooftheseindividualdifferencestothesumoftheeffectsRisthe0istheinitiallayerthicknessattimeequaltozero;universalgasconstant(8.314J/molK);andTisforallthreevariables.

thetemperature(K).ThisparabolicgrowthkineticimpliesAccordingtothevarianceanalysisinTableVI,thethattheIMCgrowthintheSn-Ag-Cualloyiscontrolledbymostsigni cantfactorintheRSSpro lewasthepeakbulkdiffusionofelementstothereactioninterface.

temperature.Beforeageing,thecoolingrategaveaFromthedatashowninTableV,itcanbeseenthatthebiggercontributionthanthepreheattemperature,butre owpro lethatgavethethinnestIMClayerwasre owafterageing,itwasopposite.Theresultssuggestthatthepro le4fortheRSSpro leandre owpro le8fortheRTSpeaktemperatureforathinIMClayerand ne

pro le.Inordertodeterminewhichfactorshadthebiggest

microstructureis2308Cfor40s.Thesuggestedsoak

B.Salam,C.Virseda,H.Da,N.N.EkereandR.Durairaj

Re owpro lestudyoftheSn-Ag-Cusolder

Soldering&SurfaceMountTechnology

16/1[2004]27–34

Figure4

Microstructuresofas-solderedandagedsolderjointsforre owpro le

1

Figure5

TheIMClayerforre owpro le

1

Figure6

TheIMClayerforre owpro le

2

Figure7

TheIMClayerforre owpro le

3

[31]