A high thermal gradient directional solidification method for growing superalloy single crystals

ContentslistsavailableatScienceDirect

JournalofMaterialsProcessingTechnology

journalhomepage:www.elsevier.com/locate/jmatprotec

Ahighthermalgradientdirectionalsolidificationmethodforgrowingsuperalloysinglecrystals

F.Wanga,b,∗,D.X.Maa,J.Zhangb,S.Bognera,A.Bührig-Polaczeka

ab

FoundryInstitute,RWTHUniversity,52072Aachen,Germany

StateKeyLaboratoryofSolidificationProcessing,NorthwesternPolytechnicalUniversity,Shaanxi,Xi’an710072,PRChina

article

info

abstract

Articlehistory:

Received17May2014

Receivedinrevisedform18July2014Accepted19July2014

Availableonline27July2014

Keywords:

DownwarddirectionsolidificationSinglecrystal

Ni-basesuperalloysBridgemanprocess

Theexperimentsherewereconductedatwithdrawalratesof3mm/minand1mm/minusingaCMSX-6andaCMSX-4superalloy,respectively.Theprocesswasassessedintermsofthethermalgradient(GL),structuralrefinement,microsegregationandporositydistribution,andcomparedtothoseusingaBridgmanprocess.TheGLoftheprocesswas200–236K/cm,whichwas10–12timeshigherthanthatintheBridgmanprocess.Amorerefinedmicrostructurewasproducedhavingaverageprimaryandsecondarydendritearmspacingvaluesaslowas243␮mand72␮m,aswellas272␮mand76␮mintheCMSX-6andtheCMSX-4castings,respectively.Thediameterof␥󰀄phaseinthedendritecoreofCMSX-6andCMSX-4castingswasreducedfrom0.8␮mto0.3␮mandfrom1.2␮mto0.6␮m,respectively.Theaverageareasof(␥󰀄+␥)eutecticpoolsbecamesmallerandmorehomogeneouslydistributed.Themeanporesizesinthecastingswerereducedby57%and43%fortheCMSX-6andCMSX-4superalloys,respectively,andtheareafractionsoftheporesintheCMSX-6andCMSX-4sampleswere16%and12%ofthoseproducedintheBridgmansamples.ThesegregationcoefficientsofthemajoralloyingelementswereclosertounitythanthoseintheBridgmanprocess,whichindicatesthatthecompositiondistributionismoreuniform.

©2014ElsevierB.V.Allrightsreserved.

1.Introduction

Thesinglecrystalsuperalloybladeisoneofthemostimpor-tantcomponentsinthehigh-efficiencyturbine.Thepropertiesofthesebladesdeterminethedevelopmentoftheturbine.Theperformancesofthesebladescanbeimprovedviadirectionalsolid-ificationprocessing(DS)andalloydevelopment.Theinnovationofthisprocesscentersonincreasingthethermalgradient(GL)atthesolidificationfront.Ahighthermalgradientduringsolidi-ficationnotonlyassuressequentialsolidificationalongtheaxialdirectionandpreventsequiaxedgrainsfrominitiatinginconsti-tutionalundercoolingzoneswithinthemeltbut,asreportedbyBrundidge(2011),thisgradientalsoreducessegregationandallowstheoperatingtemperatureofthematerialstobeincreased.

SinceBridgmanandStockbarger(1926)proposedtheBridgmandirectionalsolidificationprocessinthe1920s,todate,aseriesofdirectionalsolidificationprocesses(DS)havebeendeveloped.LauxandTingquist(1974)presentedthehighratesolidification(HRS)

∗Correspondingauthorat:FoundryInstitute,RWTHUniversity,52072Aachen,Germany.Tel.:+492418095903;fax:+492418092276.

E-mailaddresses:F.Wang@gi.rwth-aachen.de,darrel0112038@hotmail.com(F.Wang).

processintheirpatent.InthestudyofGiameiandTschinkel(1976),theliquid-metalcooling(LMC)processwasdescribed,andKonteretal.(2000)developedthegascoolingcastingprocess(GCC).How-ever,duetothelowthermalconductivityofmostsuperalloys,heatextractionintheHRSprocessbyconductionthroughthecastingtoacooledchillplatequicklybecomesinefficientwithincreas-ingdistancebetweenthesolidificationfrontandthechill.This,asdemonstratedbyLundandHockin(1972)aswellasKermanpuretal.(2000),thereforeresultsinlowthermalgradientsattheliquid/solid(L/S)interface.TomaintainastableL/Sinterface,with-drawalratesmustbereducedandthusincreasingthefrequencyofdefectformation,suchasfrecklesandstraygrains.AsreportedbyD’Souzaetal.(2000)andHugoetal.(1999),thesedefectsresultinahighrejectionrateforthecasting.IntheLMCprocess,astudybyMaetal.(2012)showedthatthethickandnon-uniformceramicmoldinfluencesthethermaluniformity,andcausesdefectssuchasstraygrains.Besidesthis,thecastingscouldbecontaminatedbythecoolant(SnorGa–Inalloy)duetofractureofthemoldduringthecoolingprocess.DuetotheopenstructureofthefurnaceintheGCCprocess,thecoolinggasmaychillthefurnaceandleadtotheformationofdefects.TheseshortcomingslimitthewideindustrialapplicationoftheconventionalDSprocesses.

TomitigatethesedisadvantagesintheconventionalDSpro-cessesandtomeetthedemandsofhighlyefficientturbines,the

http://dx.doi.org/10.1016/j.jmatprotec.2014.07.0200924-0136/©2014ElsevierB.V.Allrightsreserved.

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

3113

downwarddirectionalsolidificationprocess(DWDS)waspre-sentedbyMaetal.(2012).Inthisprocess,theceramicmoldisthinner(1mm)thanthatinconventionalprocesses(8mm)thusimprovingtheheatextractionincomparisontotheconventionalDSprocessesthusgivingrisetoanenhancedGL.Thecastingscannotbecontaminatedbythecoolantbecauseoftheapplicationofthegascoolingmethod.However,thepressureofthecoolinggasshouldbecontrolledtoanappropriatevaluebecauseexcessivepressurescancauseatransversaldiffusionofheatandanoccurrenceofstraygrains.Indeed,thecontaminationofthemoldcannotbeavoidedasinthecasewiththeconventionalDSprocesses.

Theoccurrenceoffrecklescanbeeffectivelyreducedbecause,inthedownwardgrowth,thedensityinversedemonstratedbyCopleyetal.(1970)isconstrained.GiameiandKear(1970)suggeststhatfrecklesaremacroscopicdefectswhichhavebeenobservedinunidirectionallysolidifiedmonocrystallinerods,bars,ingotsandshapedpartsofseveralnickelbasealloys,manyofthesuperalloytype.Tinetal.(2001)illustratethatsuperalloyshavingdifferentcompositionshaveadistincttendencytoformthefreckles.Forexample,Schadtetal.(2000)indicatesthatCMSX-4isatypicalfrecklepronesuperalloy.ThisdefectwasalsofoundinourearlierinvestigatedCMSX-6castings.

Itisimperativetoperformfundamentalstudiesoftheeffectofthesolidificationprocessparametersonthemicrostructuraldevel-opmentduringthecastingofsuperalloyssincetheas-solidifiedmaterialdictatesthesubsequentmicrostructuraldevelopmentandtheultimateengineeringperformance.Howeverhitherto,investi-gationshavestillnotbeencarriedouttoassesstheDWDSprocesswithrespecttothermalgradientsandmicrostructures.Theobjec-tiveofthecurrentstudyistocomparetheseaspectsintheBridgmanandtheDWDSprocesses.ThesuperioradvantagesoftheDWDSprocessovertheBridgmanprocessarediscussedbasedontheexperimentalresults.

2.Experimentalequipmentandprocedure

Cylindrical,single-crystalbarsofCMSX-6(Ni-10.0Cr-5.0Co-3.0Mo-2.0Ta-4.8Al-4.7Ti-0.01Hf-0.02C,wt%)andCMSX-4(Ni-6.5Cr-9.0Co-0.6Mo-6.5Ta-5.6Al-6.0W-3.0Re,wt%)wereusedforthisinvestigation.AsreportedbyMaandSahm(1996)andHeckletal.(2010),theliquidustemperatures(TL)ofCMSX-6andCMSX-4are1609Kand1653K,respectively.

DWDSbarsweresolidifiedusingin-housedesignedequipment,shownschematicallyinFig.1.Thisequipmentiscomposedofathermalsystemandaverticaltransmissionsystem.Thethermalsystemconsistsofacontrollableelectricresistancefurnace,apro-tectivepartandacoolingpartwhichcontainsawater-cooledchillrodandgas-coolednozzles.Theverticaltransmissionsystemcanwithdrawbarsataconstantspeedthatrangesfrom0.2mm/minto19.8mm/min.

IntheDWDSprocess,theCMSX-6ortheCMSX-4superalloywasoverheatedto1773KinacrucibleandcoveredwithhollowAl2O3particles(1–3mmindiameter)asadynamicbaffle.Aceramictube(1mmthick,200mmhigh,and9mminsidediameter),whichwasconnectedatoneendtothechillrod,havingasinglecrystalseedofCMSX-6orCMSX-4wasinsertedintothemelt.Theotherendofthetubewaswrappedandsealedusinganickelfoilinordertopreventthepenetrationofthedynamicbaffle.Whenthefoilmelted,thealloymeltflowedintothemoldandmadecontactwiththeseed.Eachseedwasinstrumentedwithathermocouplelocatedalongitscenterline.Thethermocouplewasinsertedintoanaluminatubewhichwassealedatoneend.Thissealedendwaslocated65mmfromthechillend.Thermocouplereadingswererecordedeachsec-ondduringtheentireprocess.Toobtainthethermalgradientahead

oftheL/Sinterface(GL),thecoolingrate(T

˙)wasfirstcalculatedfromFig.1.SchematicoftheequipmentusedintheDWDSprocess.

Fig.2.Positionofsectionedsamplesinthebars.

thecoolingcurve’sslopeattheliquidustemperature.Thethermal

gradientwasthencalculatedbydividingthecoolingratebythe

withdrawalrate(V):GL=T/V˙.Thethermocoupleusedwasatype-B(Pt-30%Rh/Pt-6%Rh).Whenaportionoftheseedhadmelted,the

tubewaselevatedatawithdrawalrateof3mm/minforCMSX-6and1mm/minforCMSX-4superalloysandcooledbygas(argon),andthesinglecrystalbarswerethensolidified.

ConventionalBridgmanbarsweresolidifiedinanALDVacuumTechnologies,Inc.furnace.Theparameters,suchasmeltingtem-perature,thepositionofthemetallurgicalsamples,thepositionofthethermocoupleandthewithdrawalrates,werethesameasthoseemployedintheDWDSprocess.Theceramicmoldusedinthispro-cessisaconventionalcylindricalcluster.Theinsidediameterofthecylinderis9mm.Thewallthicknessofthemoldis8mm.

Aftersolidification,thebarsweresectionedlongitudinally(par-alleltothegrowthdirection)andtransversely(perpendiculartothegrowthdirection),andsamplesweremountedandpolishedformicrostructuralanalyses.Fig.2showsthepositionofthesectionedsamples.

Themicrostructuralanalysesincludedmeasurementsoftheprimarydendritearmspacing(PDAS)(󰀃1),thesecondaryden-dritearmspacing(SDAS)(󰀃2),thesizeofthe␥󰀄phase,themeanareasof(␥󰀄+␥)eutecticpoolandthemicrosegregation.A60mLC2H5OH+40mLHCl+2g(Cu2Cl·2H2O)etchantwasusedtorevealthemicrostructure.Thetrianglemethod,asdescribedbyGündüzandCadirli(2002)aswellasRochaetal.(2003),wasusedformeasuringthePDAS(󰀃1)onthetransversesectionA(inFig.2).Thetriangleisformedbyjoiningthethreeneighboringdendrite

3114

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

Fig.3.Thecoolingcurvesandcoolingratesof(aandc)CMSX-6superalloyand(bandd)CMSX-4superalloy.(aandb)TheBridgmanprocess;(candd)TheDWDSprocess.

centers,thesidesofthetrianglecorrespondingto󰀃1.TheSDAS(󰀃2)wasmeasuredbasedontheline-interceptmethodreportedbySpearandGardner(1963),where󰀃2=L/(n−1);thelengthoflineLintersectsnsecondaryarms.Usingthesemethods,30val-uesof󰀃1and󰀃2weremeasured.Themeanareasofeutecticpools

Imageweredeterminedbythemetallographicanalyticalsoftware󰀇ST/n,ProPlus.Thesizeofthe␥󰀄phasewasdeterminedbyA=

󰀄whereSTisthecumulativetotal␥areaofthewholefieldofview

ofthemicrograph(inthedendritecore)andnisthenumberof␥󰀄variations.Thetransversesectionofthenon-etchedsampleswasphotographedtorevealpores.PoresizemeasurementswereacquiredusingtheImageProPlussoftware.Atleast80poresweremeasuredineachofthebarssolidifiedduringtheDWDSandtheBridgmanprocesses.Thesegregationbehaviorofdifferentalloy-ingelementswasdeterminedwiththeaidofaJAX-8100electronmicroprobeanalysis(EPMA)device.Threepointsinthedendritecoreandinterdendriticregionswereexamined,andtheaveragevaluesofthecontentsofthealloyingelementswereusedtocalcu-latethesegregationcoefficient.

gradientsaheadoftheL/SinterfacefortheCMSX-6superalloyinthetwoprocesseswerecalculatedasGL=236K/cm(DWDSprocess)andGL=20K/cm(Bridgmanprocess),andfortheCMSX-4super-alloyGL=200K/cm(DWDSprocess)andGL=20K/cm(Bridgmanprocess).ThethermalgradientsintheDWDSprocesswere10–12timeshigherthanthoseintheBridgmanprocess.Theresultssug-gestthattheDWDSprocessdemonstratedabetterheatextractionabilitythantheBridgmanprocess.Owingtothehighthermalgradi-entintheDWDSprocess,astableL/Sinterfacecanbemaintainedatahigherwithdrawalrate,whichcanreducetheoccurrenceofdefects.

3.2.ComparisonofBridgmanandDWDSmicrostructures

3.Results

3.1.Thermalgradientatthesolidificationfront(GL)

Fig.3(a)–(d)showsthemeasuredcoolingcurvesatthewith-drawalrateof3mm/minforCMSX-6superalloyand1mm/minforCMSX-4superalloyduringtheBridgmanandtheDWDSpro-cesses.Usingthesecurves,coolingratesof1.18K/sand0.1K/swereobtainedforCMSX-6superalloyintheDWDSandtheBridg-manprocesses,respectively.ForCMSX-4superalloy,thecoolingrateswere0.3K/sand0.03K/sintheDWDSandBridgmanpro-cesses,respectively.Accordingtothesecoolingrates,thethermal

ThedifferencesintheprimarydendritemorphologybetweentheBridgmanandDWDSsamplesattheinvestigatedwithdrawalratesforthesetwosuperalloysareclearlyvisibleintheopticalmicrographs(Fig.4).SignificantrefinementofthedendritewasobservedwithinthesamplesoftheDWDSprocess.IncomparisontotheBridgmanprocess,a53%reductionin󰀃1valuewasmeasuredfortheDWDSsolidifiedCMSX-6andCMSX-4samples(Fig.5).

AsubstantialrefinementinthesecondarydendritewasalsoobservedintheDWDSsamplescomparedtothoseintheBridgmansamples(Figs.6and7).SamplesproducedbyDWDSpossessedamuchsmalleraverage󰀃2thantheBridgmansamples.Themea-suredaverage󰀃2ofCMSX-6samplestakenfromtheDWDScastingwas72␮m,and97␮mintheBridgmansamples.Thisreductionintheaverage󰀃2valuewasalsofoundinCMSX-4samples.A󰀃2valueof76␮mwasobservedintheDWDSsamples,and99␮mintheBridgmansamples.

ThecrystalssolidifyingduringtheDWDSandtheBridgmanpro-cesseshavea␥matrixwithadisorderedface-centeredcubic(fcc)

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

3115

Fig.4.Theprimarydendritemorphologiesof(aandc)CMSX-6superalloyand(bandd)CMSX-4superalloy.(aandb)theBridgmanprocess;(candd)theDWDSprocess.

showsthedifferencesinaverageareaof(␥+␥󰀄)eutecticpoolforDWDSsamplesandBridgmansamples.Theaveragearea’svaluesofeutecticpoolsfortheDWDScastCMSX-6superalloysam-plesandCMSX-4superalloysampleswere481␮m2and495␮m2,respectively,and1377␮m2and1619␮m2,respectively,forthesematerialscastusingtheBridgmanprocess.

3.3.ComparisonofBridgmanandDWDSmicroporosities

Fig.5.Variationsin󰀃1ofDWDSandBridgmansolidifiedsamples.

structureandadispersionoforderedintermetallicprecipitatepar-ticlesofthetypeNi3(Al,Ti,Ta)(␥󰀄phase).Themorphologiesof␥󰀄phaseinthedendritecorefortheBridgmanandDWDSsamplesareshowninthescanningelectronmicroscopy(SEM)micrographs(Fig.8).Thedifferenceinthesizeof␥󰀄phaseisclearlyvisibleinFig.9.Themeasureddiameterof␥󰀄phaseintheCMSX-6superal-loysamplessolidifiedusingtheDWDSprocesswasonly42%ofthatusingtheBridgmanprocess.TheanalogousvalueintheCMSX-4superalloysampleswas48%.

Fig.10showsthemorphologiesof(␥+␥󰀄)eutecticforthesam-plesproducedbytheDWDSandtheBridgmanprocesses.Bridgmansampleshadfewer,butlarger,eutecticpools,whereasDWDSsam-pleshadmorenumerous,butsmaller,eutecticregions.Fig.11

Asaresultoftheliquidtosolidcontractionduringthefinalstagesofsolidification,alargeshrinkageporesizeoccursaccord-ingtotheprocessingconditions.Fig.12depictsshrinkageporeswithintwo-dimensionalsectionsoftheDWDSandBridgmansolid-ifiedsamples.Fig.13(a)and(b)showstheaveragediametersandtheareafractionsoftheporesmeasuredinthemetallographicsec-tions.Differencesintheporesizesandtheareafractionscanbeobservedinthesefigures.IncontrasttotheBridgmansamples,a57%anda43%reductionintheporesizeswasmeasuredintheDWDSsolidifiedCMSX-6andCMSX-4superalloysamples,respec-tively.Fig.13alsoshowsthattheareafractionsofporesinDWDSsolidifiedCMSX-6andCMSX-4sampleswereonly16%and12%ofthoseintheBridgmansamples,respectively.

3.4.Solutesegregationbehavior

Thesegregationofelementsischaracterizedbythesoluteseg-regationcoefficient,k󰀄,asgivenbyk󰀄=CDC/CIDwhereCDCistheaverageconcentration(wt%)oftheelementinthedendriticcoreandCIDisthatintheinterdendriticregion.Whenk󰀄equalsunity,theelementsarehomogeneouslydistributed.Ifthesoluteseg-regationcoefficientdeviatesfromunity,thisindicatesthatthe

3116

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

Fig.6.Thesecondarydendritemorphologiesof(aandc)CMSX-6superalloyand(bandd)CMSX-4superalloy.(aandb)theBridgmanprocess;(candd)theDWDSprocess.

segregateintotheinterdendriticregion.IntheBridgmansolidi-fiedsamples,thesegregationcoefficientsofCr,MoandCoweremuchgreaterthanone,whichindicatedthattheseelementsexhib-itedsignificantsegregationintothedendritecore.Incontrasttothis,thesegregationcoefficientsforthesethreeelementsfluctu-atedaroundunityinDWDSsolidifiedsamples.Moreover,theydidnotexhibitastrongtendencytosegregateintothedendritecoreorintotheinterdendriticregion.Inadditiontothis,thesegrega-tioncoefficientsofWandReinDWDSsolidifiedCMSX-4samplesshowasmallerdeviationfromunitythanthoseintheBridgmanprocess.

4.Discussion

Fig.7.Variationsin󰀃2ofDWDSandBridgmansolidifiedsamples.

4.1.ComparisonofheatfluxintheDWDSandconventionalDSsystems

correspondingelementispartitioningpreferentiallyintotheden-dritecoreorintotheinterdendriticregionsduringsolidification.Thelargerthedifferencebetweenk󰀄andunitysuggestsamoresubstantialsegregation.Fig.14showsthesegregationbehaviorsofthesamples’alloyingelementsproducedbytheBridgmanandtheDWDSprocesses.Amongtheelementsthatpartitiontotheinterdendriticregion,TiandTaexhibitedastrongdegreeofseg-regationinthesetwoprocesses.However,incomparisontotheBridgmanprocess,thesegregationcoefficientsofthesetwoele-mentsinDWDSsolidifiedCMSX-6andCMSX-4superalloysamplesweremuchclosertoone.Thisindicatedthatthedegreeofseg-regationofthesetwoelementswasreducedbyusingtheDWDSprocess.Inadditiontothis,theAlelementalsowaslessproneto

InordertocomprehendthehighthermalgradientintheDWDSprocess,weanalyzedtheheatfluxinthissystem,andcomparedittothoseintheconventionaldirectionalsolidificationprocesses.AccordingtoEq.(1)givenbyFuetal.(2008),

1GL=

kL

󰀅2h(T−T)a󰀆

s0

Vr−󰀄sLV

(1)

where󰀄sisthedensity,kListheliquidthermalconductivity,Vis

thecrystalgrowthvelocity,ristheradiusofthecylindricalrodcasting,TsandT0arethetemperaturesofthesolidcastingandthecoolingmedium,respectively,histhecombinedcoefficientofheattransfer,andListhelatentheat.

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

3117

Fig.8.Morphologiesof␥󰀄phaseinthedendritecoreof(aandc)CMSX-6superalloyand(bandd)CMSX-4superalloy.(aandb)TheBridgmanprocess;(candd)theDWDSprocess.

(4)Radiationfromthemold’soutersurfacetothesurroundings:

havingtheheattransfercoefficienthr.

(5)Heattransferfromthemold’soutersurfacetothesurroundings

byargongas:havingtheheattransfercoefficienthgas.

ThecombinedcoefficientofheattransferderivedbyReed(2006)isdefinedbyEq.(3)

󰀃

hDWDS=

1111+++hchgaphmold(hr+hgas)

󰀄−1

(3)

Theconductiveheattransferofthesolidifiedmetalhcisgivenby

ElliottandPollock(2007)as1500W/(m2K).Theconductiveheattransfercoefficienthmoldcanbecalculatedby

hmold=

Fig.9.Theaveragediametersof␥󰀄phaseofDWDSandBridgmansolidifiedsamples.

󰀃l

(4)

ItisdeducedthatahigherheattransfercoefficienthcangiverisetoahigherthermalgradientGL.TheheatfluxinthecoolingprocesscanbedescribedbyEq.(2)

q=h(T−T0)

(2)

InthecaseoftheDWDSprocess,thepredominantheatlossisacombinationof:

where󰀃isthethermalconductivitylyingintherangeof1.5–2.5W/(mK).IntheDWDSsystem,itisassumedthat󰀃=2W/(mK).Thewallthicknessofthemoldisl=1mm(0.001m),givinghmold=2000W/(m2K).

TheheattransfercoefficientinthegapbetweenthemoldandthemetalisevaluatedbyKonteretal.(2000)ashgap=300W/(m2K).Theradiationheattransfercoefficientfromthemold’soutersurfacetothesurroundingsisevaluatedbyKonterashr=90W/(m2K).Theaverageheattransfercoefficientfromthemold’soutersurfacetothesurroundingscausedbytheimpingementofargonisalsogivenbyKonterashgas=510W/(m2K).

(1)Conductionthroughthesolidifiedmetaltothewater-cooled

rod:havingtheheatexchangecoefficienthc.

(2)Radiationfromthemetalthroughthegapbetweenthemetal

andthemold’sinnersurfaces:havingtheheatexchangecoef-ficienthgap.

(3)Conductionthroughtheceramicmoldtothemold’soutersur-face:havingtheheatexchangecoefficienthmold.

1hDWDS

=

1111

+++15003002000600

Then,hDWDSis162W/(m2K).

Konteretal.(2000)andLiuetal.(2010)presentthecom-binedcoefficientsofheattransferintheBridgman,LMCandGCCprocessesashHRS=71W/(m2K),hLMC=110W/(m2K)(thecoolant

3118

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

Fig.10.Morphologiesof(␥+␥󰀄)eutecticformedin(aandc)CMSX-6superalloysamplesand(bandd)CMSX-4superalloysamples.(aandb)TheBridgmanprocess;(candd)theDWDSprocess

foundthat,inthesystemofsuperalloys,thesoluteislighterthanthesolvent.AsaresultofthedownwardgrowthdirectionofDWDSsamples,thesoluteformedatthetiptendstoflowupwardsandfillstheinterdendriticregions(Fig.16).Spinellietal.(2004)foundthatthisflowcontributestotheradialtransportofsolute,andleadstotheformationoflargerprimarydendritearmspacing.

Thefavorablefactoristhethermalgradient.Thetheoreticalmodels(formulatedinEq.(5))ofprimarydendritespacingpro-posedbyHunt(1979),KurzandFisher(1981)and(1984),andTrivedi(1984)suggestthatataconstantgrowthrate,theprimarydendritearmspacingisreducedbyincreasingthethermalgradi-ent.ThethermalgradientintheDWDSprocessismuchlargerthanthatintheBridgmanprocess(Fig.3).Thishigherthermalgradi-entproducedthesmallerprimarydendritearmspacinginDWDSsamples.

󰀃1=kGL

−1/2

V−1/4

(5)

Fig.11.Theaverageareasof(␥+␥󰀄)eutecticpoolsformedintheBridgmanandtheDWDSprocesses.

beingliquidgallium)andhGCC=81W/(m2K),respectively.Fig.15showsthedifferenceintheheattransfercoefficientsfordifferentprocesses.TheDWDSprocesspossessesthelargestheattransfercoefficient,whichillustratesthehigherthermalgradientinthissystem.

where󰀃1istheprimarydendritearmspacing,andkisaconstantwhichdependsonthealloycomposition.IntheDWDSprocess,theeffectofthehigherthermalgradientdominatestheunfavorableeffectofconvection.Therefore,DWDSsampleshavefinerprimarydendritesandthevalueofPDASissmallerthanthatforBridgmansamples(Figs.4and5).

Wagneretal.(2004),FeurerandWunderlin(1986),andKirkwood(1985)havederivedthesecondarydendritearmspacingformulagivenbyEq.(6).

4.2.DWDScastmicrostructurerefinement

󰀃2=5.5

TwofactorsarebelievedtoinfluencethemicrostructuresofDWDSsamples:Theunfavorablefactoristheconvection.IntheverticalDSprocess,themodeofconvectiondependsonwhetherthesoluteislighterorheavierthanthesolvent.Trivedietal.(2001)

󰀁M󰀅T󰀂ˇ

GLV

(6)

where󰀃2isthesecondarydendritearmspacing,M=(11±2)/s,ˇ=0.27±0.05,and󰀅Tisthesolidificationinterval.Thisrelation-shipbetween󰀃2andGLataconstantgrowthratesuggeststhata

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

3119

Fig.12.Opticalmicroscopeimagesofsamplessectionedperpendiculartothegrowthdirection.Poresareshownasdarkvoidsfor(aandc)CMSX-6superalloyand(bandd)CMSX-4superalloy.(aandb)TheBridgmanprocess;(candd)theDWDSprocess.

higherthermalgradientgivesrisetoasmallerSDAS.TheDWDSprocesspossessesahighertemperaturethanthatoftheBridg-man,thus,incomparisontoBridgmansamples,afinersecondarydendritecanbeobservedinDWDSsamples(Figs.6and7).

Theprecipitationofthe␥󰀄phasefromthe␥matrixisadiffu-siontransformationprocess.Theshapeandsizeofthe␥󰀄phaseareaffectedbynucleationandgrowthconditions.Fuetal.(2008)hasgiventherelationshipbetweentheaveragesizeofprecipitatedphaseandGLandVasEq.(7).

ˇıDRT¯=2r(GLV)−1

󰀃Q

(7)

¯istheaveragesizeoftheprecipitatedphase,Disthediffu-wherer

sioncoefficient,ıisthethicknessofeffectiveboundarylayer,Qistheactivationenergy,andˇisthedrivingforceofphasetransition.Thisrelationshipindicatesthattheaveragesizeof␥󰀄phasereduceswithanincreaseinthermalgradient.Xiao(2004)suggeststhatahigherthermalgradientgivesrisetoahighercoolingrate,whichlowerstheprecipitationtemperatureofthe␥󰀄phaseleadingtoanincreaseinundercoolingofthe␥solidsolution,andthenucleationrateof␥󰀄isincreased.Therefore,incomparisontotheBridgmansamples,asubstantialreductioninthesizeof␥󰀄phaseisobservedintheDWDSsamples(Figs.8and9).

Fig.13.Theaveragediameters(a)andareafractionsofpores(b)fromDWDSandBridgmansolidifiedsamples.

3120

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

Fig.14.SegregationcoefficientsofalloyingelementsintheCMSX-6superalloy(a)andCMSX-4superalloy(b)samplescastedbytheBridgmanandDWDSprocesses.

4.3.Microporosityanalysis

Solidificationshrinkageisthoughttobeacriticalcontributingcauseofmicroporosity.Themicroporositymayactasnucleationsitesforcracksthatpropagateunderfatigueloadingconditions.Poirieretal.(1987)suggestthatthenumberofporesissensitivetothethermalgradient,andanincreaseinthermalgradientdecreasesthequantityofpores.Asaconsequence,alowernumberofporeswereformedinDWDSsamplesduringsolidificationbecauseofthehigherthermalgradient(Fig.12).Moreover,hisresearchalsoindi-catesthattheporesdevelopfromtheshrinkageofthefinaleutecticliquidasafractionoffl,andtheporediameter,dpore,willincreasewith󰀃1:

dpore=

fl󰀃12

(8)

Fig.15.TheheattransfercoefficientsfordifferentDSprocesses.

SamplesproducedbytheDWDSprocesspossessasmallerpri-marydendritearmspacingincomparisontothatofBridgman;thisexplainswhythemeansizeandtheareafractionoftheporesintheDWDSsamplesissmallerthanthosefoundintheBridgmanprocess.

4.4.Solutesegregationbehavior

Fig.16.Schematicdiagramshowingsoluteflowinthepresentexperimentalsys-tem.

Thesizeofthe(␥󰀄+␥)eutecticpoolisassociatedwiththeper-centagecontentoftheconstituentelements.Inthesuperalloys,theconstituentelementsof(␥󰀄+␥)eutecticareAl,TiandTawhichseg-regateintotheinterdendriticregions.Whenthethermalgradientishigher,thesolidificationratebecomeshigher,andasmallquantityofAl,TiandTacandiffuseintotheinterdendriticregionsformingasmalleranddispersive(␥󰀄+␥)eutecticpool.FromFig.14wecanseethatthesegregationcoefficientsofAl,TiandTaintheDWDSsolidi-fiedsamplesareclosertounitythanthoseoftheBridgmanprocess.Itindicatesthatthesealloyingelementsdistributemoreuniformly.ComparedtotheBridgmanprocess,therearefeweramountsoftheseelementssegregatingintotheinterdendriticregions.Forthisreason,samplesproducedbytheDWDSprocesshavesmallerandmoredispersive(␥󰀄+␥)eutecticpoolsthanthoseproducedbytheBridgmanprocess.ThisscenarioisclearlyshowninFig.10andFig.11.

Seoetal.(2011)andWangetal.(2014)foundthattheinflu-enceofelevatedthermalgradientonthesegregationbehaviorfortheconstituentelementscanbelargelyattributedtothehomoge-neousback-diffusioninthesolidifiedsolidwhichwasputforwardbyHaroldandMerton(1966).Acriticalfactorforsolidback-diffusionisthediffusiondistance(approximatelyequalto󰀃1/2).Underhigherthermalgradients,thediffusiondistancereducessig-nificantlyduetothegreatlyrefineddendriticstructures.Fig.3showsthatthethermalgradientaheadoftheliquid/solidinter-faceintheDWDSprocesswasalmosttwelvetimesgreaterthanthatintheBridgmanprocess.Subjecttothishighthermalgradient,themicrostructuresweresignificantlyrefined.Therefore,thesam-plesproducedbyusingthisprocesshadasmallerprimarydendritearmspacingwhichresultedinashorterbackdiffusiondistance.Forthisreason,thesegregationcoefficientsofthealloyingelementsareclosertounitythanthoseintheBridgmanprocess(showninFig.14),andthedegreeofsegregationoftheseelementswasreduced.

5.Conclusions

1.Athermalgradientwhichis10–12timeslargerthanthatoftheBridgmanprocesscanbeachievedbyusingtheDWDSprocess.

F.Wangetal./JournalofMaterialsProcessingTechnology214(2014)3112–3121

3121

2.Thedendriticstructuresaresignificantlyrefined.58%and68%reductioninthediameterof␥󰀄phasewasobservedintheCMSX-6andCMSX-4superalloysamples,respectively.Smallerandmorehomogeneouslydistributed(␥󰀄+␥)eutecticpoolsareobtainedintheDWDSsamples.

3.ComparedtotheBridgmanprocess,theaverageporesizesintheCMSX-6andCMSX-4superalloysampleswerereducedby57%and43%,respectively.Theareafractionoftheporeswas15.6%and11.9%ofthoseintheBridgmanprocessfortheCMSX-6andCMSX-4superalloys,respectively.Thesegregationofthealloyingelementswasdecreasedbyusingtheprocess.

Acknowledgements

ThisresearchwassupportedbytheGermanResearchFounda-tion(DFG)throughGrantNo.MA2505/3-1.Oneoftheauthors(FW)wouldliketoacknowledgetheChinaScholarshipCouncilforsupportinghisstaysinGermany.TheauthorswouldliketoacknowledgetheaidofElkeSchaberger-ZimmermannandElkeBreuer.

References

Bridgman,P.W.,1926.Crystalsandtheirmanufacture.USPatent1,793,672(Filedon

February16,1926.GrantedonFebruary24,1931).

Brundidge,C.L.,2011.Developmentofdendriticstructureintheliquid-metal-cooled,directional-solidificationprocess.Metall.Mater.Trans.A42,2723–2732.Copley,S.M.,Giamei,A.F.,Johnson,S.M.,Hornbecker,M.F.,1970.Theoriginoffreck-lesinunidirectionallysolidifiedcastings.Metall.Trans.1,2197–2204.

D’Souza,N.,Ardakani,M.G.,McLean,M.,Shollock,B.A.,2000.Directionalandsingle-crystalsolidificationofNi-basesuperalloys:PartI.Theroleofcurvedisothermsongrainselection.Metall.Mater.Trans.A31,2877–2886.

Elliott,A.J.,Pollock,T.M.,2007.ThermalanalysisoftheBridgmanandliquid-metal-cooleddirectionalsolidificationinvestmentcastingprocesses.Metall.Mater.Trans.A38,871–882.

Feurer,U.,Wunderlin,R.,1986.FundamentalsofSolidification.TransTech.Publi-cationsLtd.,Aedermannsdorf,Switzerland,pp.214–216(Appendix8)(citedbyKurz,W.,Fisher,D.).

Fu,H.Z.,Guo,J.J.,Liu,L.,Li,J.S.,2008.DirectionalSolidificationandProcessingof

AdvancedMaterials,firsted.TheSciencePress,Beijing,pp.498.

Giamei,A.F.,Kear,B.H.,1970.Onthenatureoffrecklesinnickelbasesuperalloys.

Metall.Trans.1,2185–2192.

Giamei,A.F.,Tschinkel,J.G.,1976.Liquidmetalcooling:anewsolidificationtech-nique.Metall.Trans.A7,1427–1434.

Gündüz,M.,Cadirli,E.,2002.Directionalsolidificationofaluminium–copperalloys.

Mater.Sci.Eng.A327,167–185.

Harold,D.B.,Merton,C.F.,1966.Soluteredistributionindendriticsolidification.

Trans.Metall.Soc.AIME236,615–623.

Heckl,A.,Rettig,R.,Singer,R.F.,2010.Solidificationcharacteristicsandsegregation

behaviorofnickel-basesuperalloysindependenceondifferentrheniumandrutheniumcontents.Metall.Mater.Trans.A41,202–211.

Hugo,F.,Betz,U.,Ren,J.,Huang,S.-C.,Bondarenko,J.A.,Gerasimov,V.,1999.Cast-ingofdirectionallysolidifiedandsinglecrystalcomponentsusingliquidmetalcooling(LMC).In:Mitchell,A.,Ridgway,L.,Baldwin,M.(Eds.),International

SymposiumonLiquidMetalProcessingandCasting.AmericanVacuumSociety,NewYork,USA,pp.16–30.

Hunt,J.D.,1979.InternationalConferenceonSolidificationandCastingofMetals,

firsted.TheMetalsSociety,London,pp.3–9.

Kermanpur,A.,Rappaz,M.,Varahram,N.,Davami,P.,2000.Thermalandgrain-structuresimulationinaland-basedturbinebladedirectionallysolidifiedwiththeliquidmetalcoolingprocess.Metall.Mater.Trans.B31,1293–1304.

Kirkwood,D.H.,1985.Asimplemodelfordendritearmcoarseningduringsolidifi-cation.Mater.Sci.Eng.A73,L1–L4.

Konter,M.,Kats,E.,Hofmann,N.,2000.Anovelcastingprocessforsinglecrystalgas

turbinecomponents.In:Pollock,T.M.,Kissinger,R.D.,Bowman,R.R.,Green,K.A.,McLean,M.,Olson,S.,Schirra,J.J.(Eds.),Superalloys2000.TheMinerals,MetalsandMaterialsSociety(TMS),Warrendale,PA,pp.189–200.

Kurz,W.,Fisher,J.D.,1981.Dendritegrowthatthelimitofstability:tipradiusand

spacing.ActaMetall.29,11–20.

Laux,E.G.,Tingquist,S.C.,1974.Methodandapparatusformeltingandcastingmetal.

U.S.PatentNo.3,841,384(FiledonFebruary21,1973.GrantedonOctober15,1974).

Liu,L.,Huang,T.W.,Qu,M.,Liu,G.,Zhang,J.,Fu,H.Z.,2010.Highthermalgradient

directionalsolidificationanditsapplicationintheprocessingofnickel-basedsuperalloys.J.Mater.Process.Technol.210,159–165.

Lund,C.H.,Hockin,J.,1972.Investmentcasting.In:Sims,C.T.,Hagel,W.C.(Eds.),The

Superalloys.JohnWiley&Sons,NewYork,USA,pp.403–425.

Ma,D.,Sahm,P.R.,1996.Microstructuralevolutionduringdirectionalsolidification

ofsuperalloyCMSX-6.Mater.Sci.Forum215–216,223–228.

Ma,D.,Lu,H.,Bürig-polaczek,A.,2012.Experimentaltrialsofthethinshellcasting

(TSC)technologyfordirectionalsolidification.IOPConf.Ser.Mater.Sci.Eng.27,012036.

Poirier,D.R.,Yeum,K.,Maples,A.L.,1987.Metall.Trans.A18,1979–1987.

Reed,R.C.,2006.TheSuperalloys:FundamentalandApplication,firsted.Cambridge

UniversityPress,Cambridge,pp.137.

Rocha,O.L.,Siqueira,C.A.,Garcia,A.,2003.Heatflowparametersaffectingdendrite

spacingsduringunsteady-statesolidificationofSn–PbandAl–Cualloys.Metall.Mater.Trans.A34,995–1006.

Schadt,R.,Wagner,I.,Preuhs,J.,Sahm,P.R.,2000.Newaspectsoffreckleformation

duringsinglecrystalsolidificationofCMSX-4.In:Pollock,T.M.,Kissinger,R.D.,Bowman,R.R.,Green,K.A.,McLean,M.,Olson,S.,Schirra,J.J.(Eds.),Superalloys2000.TheMinerals,MetalsandMaterialsSociety,Warrendale,PA,pp.211–218.Seo,S.M.,Lee,J.H.,Yoo,Y.S.,Jo,C.Y.,Miyahara,H.,Ogi,K.,2011.Acomparativestudy

ofthe␥/␥󰀄eutecticevolutionduringthesolidificationofNi-basesuperalloy.Metall.Mater.Trans.A42,3150–3159.

Spear,R.E.,Gardner,G.R.,1963.Dendritecellsize.AFSTrans.71,209–215.

Spinelli,J.E.,Rosa,D.M.,Ferreira,I.L.,Garcia,A.,2004.Influenceofmeltconvection

ondendriticspacingsofdownwardunsteady-statedirectionallysolidifiedAl–Cualloys.Mater.Sci.Eng.A383,271–282.

Tin,S.,Pollock,T.M.,Murphy,W.,2001.Stabilizationofthermosolutalconvective

instabilitiesinNi-basedsingle-crystalsuperalloys:carbonadditionsandfreckleformation.Metall.Mater.Trans.A32,1743–1753.

Trivedi,R.,1984.Interdendriticspacing:PartII.Acomparisonoftheoryandexperi-ment.Metall.Mater.Trans.A15,977–982.

Trivedi,R.,Liu,S.,Mazumder,P.,Simsek,E.,2001.Microstructuredevelopmentin

thedirectionallysolidifiedAl–4.0wt%Cualloysystem.Sci.Technol.Adv.Mater.2,309–320.

Wagner,A.,Shollock,B.A.,McLean,M.,2004.Grainstructuredevelopmentindirec-tionalsolidificationofnickel-basesuperalloys.Mater.Sci.Eng.A374,270–279.Wang,F.,Ma,D.,Zhang,J.,Liu,L.,Hong,J.,Bogner,S.,Bührig-Polaczek,A.,2014.

EffectofsolidificationparametersonthemicrostructureofsuperalloyCMSX-4formedduringthedownwarddirectionalsolidificationprocess.J.Cryst.Growth389,47–54.

Xiao,J.M.,2004.AlloysandPhaseTransition,seconded.MetallurgicalIndustryPress,

Beijing,pp.233–235.