of WC Grains in Liquid Co
-andanevaluationofexistinggrowththeories
EMANUEL EKSTRÖM
Master of Science Thesis, Stockholm 2007
School of Industrial Engineering and Management
Department of Materials Science and Engineering
Royal Institute of Technology
SE-100 44 Stockholm
Cemented carbides (WC-Co) are powder metallurgical products produced by
liquid phase sintering. WC-Co is widely used for making a large variety of
cutting tools, such as drills and insertsturning applications, due to its great
mechanicalproperties,where thehardnessof theWCgrains iscombinedwith
thetoughnessoftheoftheCobinder. WCgrainsizeandgrainsizedistribution
arethetwomostimportantfactorsto controlthemechanicalpropertiesofthe
products.
This study examinedthe grain growth dependence of dierent millingand
sintering times. Theresultinggrain size andgrain size distribution were mea-
suredusingimageanalysisonscanningelectronmicroscopyimages(SEM)and
by using electronbackscatter diraction (EBSD). Inaddition, the correlation
betweenhardnessand coercivity,themost commonindirectmeasures ofgrain
size,anddierentmethodsofcalculatingaveragegrainradiuswereinvestigated.
Anattemptwasalsomadetostudythecontributionofdefectstograingrowth.
This work also includes an overview of various grain growthequations and a
numericalimplementationofthese.
Experimental resultsshow that forshorter sintering times, powders milled
for short times (15 min and 1 h) have larger average grain radii. There is a
crossoverafter 6to 8h of sintering, wherethe powdersmilled foralongtime
(40hand200h), havelargeraverageradii. Themeasuredhardnessvaluescor-
relatewellwiththeaveragegrainradiuscalculatedfrom thegrain surfacearea
and the coercivity correlates with the establishedequations. EBSD measure-
mentsdetected boundariesthat could notbedetected by image analysis,and
that were not
Σ
2boundaries. It is likelythat these boundariesareeither low energyboundariesorboundariesbetweengrainsthat areverycloselyoriented.Comparing heat-treated powder with the untreated resulted in a lower aver-
agegrain size after sintering for theheat-treatedpowder. Noneof thegrowth
equationsinvestigatedin thisworkcouldfullydescribetheexperimentalgrain
growth.
Through increased understanding of the grain growth, the growth can be
controlled and the end product can havethe desiredtool properties. The oc-
currence of abnormal grains in cutting tool applications can cause breakage,
which is especially important to avoid in applications such as PCB drills. A
correlation between hardnessand grain size provides further means forcheap
andfastindirectmeasures ofthegrain sizeinproduction.
Keywords: Cementedcarbide;WC-Co;Abnormalgrain growth;Nucleation;
Defects;Coarsening;Milling;Sintering;EBSD; Mechanicalproperties;
Hardness;Simulation;Modeling; Cuttingtools
Hårdmetallärettpulvermetallurgisktmaterialsomtillverkasgenomsmältfass-
intring och som kännetecknas av hårdhet, styvhet och god slitstyrka. Vol-
framkarbidens (WC)kornstorlekochkornstorleksfördelning ärtvåviktiga fak-
torerförattkontrollerademekaniskaegenskapernai hårdmetall.
I denhärstudien harkorntillväxtensberoendepåmalningochsintringun-
dersökts. WC-Comaldesochsintradesfyraolikatiderochkornstorleksfördelnin-
genmättesmedbildanalyspåsvepelektronmikroskopbildersamtmedelectron
backscatterdiraction (EBSD).I arbetetharävenkorrelationenmellan hård-
het,koercivitetocholikasättatt beräknamedelkornstorlekenundersökts. Ett
försökharocksågenomförtsförattstuderahurdefekterna idetmaldapulvret
påverkar korntillväxten. I arbetet har även ett ertal olika tillväxtekvationer
modelleratsnumeriskt ochför ochnackdelarmed deolikatillväxtekvationerna
harvägtsmotvarandra.
En långmaltid (40h och200h) visadesigge litenkornstorlekförsintring
kortare än 6h, men för sintringarlängre än 8h gavistället kort malning (15
min och 1h) den mindre kornstorleken. Det visade sig att uppmätt hårdhet
korrelerarbästmeddenmedelkornstorleksradiesomräknatsframfrånkornytan.
I EBSD mätningarna kunde man observeraett ertal korngränser, utöver
Σ
2korngränser,somintehadedetekteratsmedbildanalys. Värmebehandlingenav
detmaldapulvretminskadekorntillväxtenunderefterföljandesintring. Ingenav
deundersöktatillväxtekvationernakundebeskrivadeexperimentellaresultaten
fulltut.
Genom ökad förståelse för korntillväxt kan man kontrollera tillväxten och
slutproduktenkanfåönskadeegenskaper. Förekomstenavabnormkorntillväxt
i skärverktyg i hårdmetall är en av de vanligaste kritiska defekterna och det
är speciellt viktigt är undvika korntillväxt i tillverkning av små verktyg, som
tillexempelkretskortsborrar. Hårdhetochkoercivitetärdevanligasteindirekta
mätmetodernaförattmätakornstorlekiproduktion. Enbrakorrelationmellan
kornstorlekochindirektamätmetodergerutökadeverktygförsnabbaochbilliga
mätningar.
Nyckelord: Hårdmetall;WC-Co;Abnormkorntillväxt;Kärnbildning;
Defekter;Förgrovning;Malning;Sintring;Mekaniskaegenskaper;Hårdhet;
EBSD;Simulering;Modellering;Skärverktyg
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Controllingtheproperties . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Thecurrentwork . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Literature review 5 2.1 Normalgraingrowth . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Newanalysismethods . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Abnormalgraingrowth . . . . . . . . . . . . . . . . . . . . . . . 6
2.4 Growthmodels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.5 Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.6 Summaryanddiscussion . . . . . . . . . . . . . . . . . . . . . . . 10
3 Experimentalprocedure 11 3.1 Specimenpreparation . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2 Measuringphysicalproperties . . . . . . . . . . . . . . . . . . . . 12
3.3 Analysispreparation . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4 Analysistechniques . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.5 Removinggraindefectsbeforesintering . . . . . . . . . . . . . . 16
4 Data Analysis 17 4.1 Makingadistributionplot . . . . . . . . . . . . . . . . . . . . . . 17
4.2 Determiningthesmallestgrainsize . . . . . . . . . . . . . . . . . 19
4.3 Methodstocalculate theaveragegrainsize . . . . . . . . . . . . 20
5 Results 23 5.1 Averagegrainsize . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5.2 Distributionplots. . . . . . . . . . . . . . . . . . . . . . . . . . . 24
5.3 Materialproperties . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.4 EBSDmeasurements . . . . . . . . . . . . . . . . . . . . . . . . . 28
5.5 Impactofdefects . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6 Discussion 31 6.1 Eectofmilling. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6.2 Hardness,coercivityandaveragegrainsize . . . . . . . . . . . . 32
6.3 EBSD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.4 Impactofdefects . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
7 Numericalsimulations 39
7.1 Basicmodelforgraingrowth . . . . . . . . . . . . . . . . . . . . 39
7.2 ClassicLSW-theory . . . . . . . . . . . . . . . . . . . . . . . . . 40
7.3 Implementationofvariousgrowthmodels . . . . . . . . . . . . . 41
7.4 Generaldiscussion . . . . . . . . . . . . . . . . . . . . . . . . . . 49
8 Conclusions 51
9 Suggestionsfor future work 52
Acknowledgements 53
References 54
A Theoretical calculations 59
B Tables 63
C SEM images 69
D Additionalresults 77
E Additionalnumericalsimulations 83
Introduction
W
C-Cois interestingformakingcutting tools,dueit itsgreat mechanical
properties. This introduction aims to provide abackground about the
properties and productionof hardmetals. Theobjectivesand structure of the
thesisarealsostated.
1.1 Background
Metal machining is widely used in industry, becauseof the great possibilities
regarding shapeand itshigh precision. Cutting tools areusually produced as
asmallinsertof dierentshapesandcan bexedto atoolholder(seeFigure
1.1). Common cutting tool materials are hardmetals, ceramics and diamond.
Processessuchas cutting anddrillingrequirehigh performancematerialswith
themostimportant propertiesbeinghardness, toughnessand wear resistance.
Eachkindofcutting putsdierentdemandsonthetool,such assharpcutting
edgesor thermalfatigue. Unfortunately, high hardnessusually correspondsto
poor toughness. A successful combination of the two is found in cemented
carbides,whichwillbestudiedinthepresentwork.
Commercial hardmetals mainly consist of WC, TiC or TiCN. WC-Co ce-
mentedcarbidesconsistofWCgrainsembeddedinaCobinder. Thisstructure
oersgoodmechanicalpropertiescombiningthehardnessoftheWCgrainsand
the toughness of the Co binder. A ne homogenous microstructure provides
high potential for sharp cutting tools while a coarser structure is better for
applications working in a high temperature environment. The coarser grains
conduct heat betterthan the nergrains. Therefore, thevarious applications
requiredierentgrainsizes.
The process of making WC-Co cemented carbides begins with milling of
theraw materials into ane powder. Themilled powderis then pressedinto
thedesiredgeometricalshape. Finally,theproductis producedbyliquidstate
sintering. Thesinteringtemperaturerangesfrom 1300to1500
◦
C,andtypical
isothermal dwells are around 60 minutes long. During the sintering process,
somegrainsdissolve,whichmakesitpossibleforotherstogrow.
Changingthecuttingtoolinsertsinamachineinterruptsthemanufacturing
process,andisalargeeconomicfactor. Beingabletousehighercuttingspeeds
and thereby speedup production is anotherfactor for increased productivity.
Figure 1.1: Cuttingtoolinsert(yellow) foraturningapplication. (CourtesyofAB
SandvikCoromant)
Figure1.2: Aprintedcircuitboarddrill.
There is a high industry demand for longer-lasting tools that can withstand
higher temperatures and strongercutting forces. Today coatings are used on
almostallinserts,andthecarbideisnolongerincontactwiththecuttingpiece,
as longas thecoatingis intact. Thelifetime of theinsertis thereforedecided
eitherbythewearofthecoatingorbytheplastic deformationinthecemented
carbide.
1.2 Controlling the properties
The WC grain size and grain size distribution are two of themost important
parametersin thecontrolthemechanicalpropertiesof thetools. Factors often
studied to determine grain size are sintering temperature, pressure and time,
and also the choice of raw materials. Research during the past few decades
has been dedicated to improve the production of cemented carbides. Other
important factors in controlling the properties are the binder phasesand the
alloyingelementcontents.
In the grain growth process, grains tend to coarsendierently; individual
grainscanbecomemuchlargerthantheaverage,calledabnormalgraingrowth
(AGG). Theabnormal grains will reduce theperformance of theproductand
should therefore be avoided. There is also an industry demand for making
smallertools,such as printed circuitboard drillswith drilldiametersas small
as 75
µ
m (see Figure 1.2). In applications like this, it is important to avoid failures such as drill breakage. WC grains that grow abnormallylarge act asinitiationpointsforbreakage.
ThephysicalmechanismofAGGisnotwellunderstoodandtherehasbeen
alotofresearchonhowtopreventAGG.Onewayistoadd smallamountsof
alloyelements, such asV or Cr,that work asinhibitors,preventingthegrains
from growing. Even thoughthis method is currently used, the physics of the
methodis notfullyunderstood. Anothermethodofreducingthegraingrowth
1.3 The current work
Theaim ofthis work was to studyhowthegrain size distribution depends on
and changes withvariation in themillingprocess. An attempt was also made
toseparatetheeectofinitialgrainsizefromtheeectofincreasednumberof
defects. Thegrain size and grain sizedistribution were measuredusingimage
analysis from a scanning electron microscope as well as electron backscatter
diraction.
Another objectiveof this thesis was to investigate thecorrelation between
indirect measures of grain radius, the hardness and coercivity, with dierent
waysofcalculatingtheaveragegrainradius.
This thesishasalsoevaluatedcurrentgrowthmodels,andinvestigatedhow
wellthemodelsaccountforthedefectsanddierenceingrainsizedistribution.
ThemodelswereimplementednumericallyinMATLABandcomparedwiththe
experimentaldata.
Chapter 2containsaliteraturereviewofthemostinterestingpapersinthe
eld andadiscussionofthendings. Thepaperswillinclude studiesofAGG,
variousgrowthequationsandprevioussimulationsofAGG.
Chapter 3willexplaintheexperimentalprocedure,themillingandsinter-
ingprocesses,andthetechniques usedforanalyzingthesamples.
Chapter 4will describethehow therawdata from theexperimentswere
analyzed. It will explain how the data were organized and dierent ways of
calculatingtheaveragegrain size. Theresultswill bepresentedin Chapter 5
andbediscussedin Chapter 6.
Chapter 7will explainhowthegrowththeorieswereimplementednumer-
icallyandit will alsoshowthe resultofthevarious growthequations. Finally
allndingsinthecurrentworkwillbeconcludedin Chapter 8.
Literature review
T
hischapterisareviewofthepastandpresentresearchintheeldofvarious
growththeoriesofWCgrains. Thereviewwillgiveabroaderbackground
totheareaofgrain growth. Thedierentgrowthequationsandtheirphysical
backgrounds will be covered and previous numerical simulations will also be
discussed. Inthiswork,whenevertheaverageradiusis mentioned,itrefersto
averagebasedonthenumberofparticles,unlessotherwisestated.
2.1 Normal grain growth
Duringsintering of cemented carbides,theaveragegrain size increasesdue to
coarsening of the grains. This phenomenon is called Ostwald ripening, and
the theory statesthat largegrains grow at the expense of thesmaller grains.
Diusionandtheinterfacialreactionsarelimitingfactorsofthecoarseningrate.
Ostwaldripeningwas studiedbyLifshitz, SlyozovandWagner [1,2]resulting
intheLSW-theory. Thetheoryproposesastationaryparticledistributionwith
the averagegrain radius increasing with time. The basic assumptionof LSW
theoryisthatthegrowthrateisproportionaltoitsdrivingforce. Theclassical
growthequationis
r
n− r
0n= 2Kt,
(2.1)where
r
0 is the average initial grain radius and K a temperature-dependant coecient. Avalueofn = 2
appliesforsystemslimitedbytheratethatatomsleavethesolid/liquidinterface,andavalueof
n = 3
appliesforwhenthegrowthrateislimitedbydiusion ofatomsofthesolidorliquidphase[2]. Mullins[ 3]
predicted that the distribution of crystal shapes does not aect the growth
exponent,
n
.WC hasa simplehexagonal crystal structure and astoichiometric compo-
sition. Christensen et al. [4] derived a methodology for using rst principle
calculations to estimate the energy of an arbitrary unknown structure. This
methodwasapplied todierentinterfacialenergiesin aWC-Cocementedcar-
bide and to quantify theshape of theWC grains. Two newparameters were
denedtodescribetheequilibriumsize:
r
astheratiobetweenthetotallengthofthe longandshort prismfacets and
k
as theratiobetween thethicknessoftheprismandtheheightofthetruncatedtriangle. Thearticleshowedalarger
accuracy in determining the
r
parameter than thek
parameter, and that itisthereforereasonableto usethesamevalueoftheinterfacialenergyforevery
grainboundary. Theinterfaceenergy
γ
Co/W C wasdeterminedtorangefrom0.5J/m
2
to 1-2J/m
2
. ThecalculationsinthearticleshowedtheshapeofWC-Co
tobetruncatedtriangularprismswithatfacets.
2.2 New analysis methods
Kumar et al. [ 5] used electronbackscattering diraction (EBSD), further ex-
plained in Section 3.4.3, to characterize grain boundaries of WC-Co during
sintering. Theyobservedboundariesthat exhibitlowinterfacialenergy,called
Σ
2boundaries. Theboundariesareusually notwetted byCo in thesintering process. They occurfrequentlyin theearly stage of thesintering processbutareannihilatedathighertemperaturesandlongersinteringtimes. Theauthors
believe that
Σ
2 boundaries originate in the raw powder, and that was laterconrmedbyKimandKang[6] .
Mannesson et al. [ 7] have analyzed the WC grain size distribution after
milling and sintering at three dierent times, 0.25 h, 1h and 8 h. The sam-
pleswereanalyzedbyimageanalysisusingascanningelectronmicroscopeand
EBSD. Whenanalyzingthepowderbeforesintering, itbecomesclearthat the
EBSD techniqueistheonlytechniqueavailabletodetect
Σ
2boundaries.The resultof thestudy byMannesson etal. [ 7]was that EBSD, excluding
the
Σ
2boundaries,givesaresultthat isinagreementwiththeimageanalysis.Therewasalsoalimitationin thedetectionofsmallgrainsfortheimageanal-
ysis. When including the
Σ
2 boundaries, the distributions varied because of thedetectionofalargernumberofsmallgrains.TheamountofΣ
2boundariesdecreasedquicklyduring theinitial sintering. Theauthorssuggestedtheeect
of
Σ
2 boundariesdoes not have to be considered in future modeling of graingrowth.
When comparing the averagegrain size based on the number of grains in
the distribution to thehardness, alack of correlation hasbeen demonstrated.
Engqvistand Uhrenius[8]suggestan alternatewayof calculatingtheaverage
grain size, based on the volume or center of gravity. When calculating the
averageoftheobservedgrainsizes,moreweightisgiventothenergrains. By
usingtheequation
d
avg= P d
4iP d
3i,
(2.2)where
d
is the equivalent diameter of a grain and the sum is of all grains.The authors concluded that this method gives the larger grains more weight
in calculatingtheaveragegrainradius. This volume-basedaverageresultedin
abettercorrelation with hardnessvalues. Thesamples used byEngqvist and
Uhrenius[8]showedbimodaldistributions.
2.3 Abnormal grain growth
Abnormalgrain growth(AGG)iswhen fewgrainsgrowveryfastcomparedto
thegrowthofthesurroundinggrainsduringthesinteringprocess. Thisgrowth