Effects of the nature of the doping salt and of the thermal pre-treatment and sintering temperature on spark plasma sintering of transparent alumina

To cite this version

_{: Roussel, Nicolas and Lallemant, Lucile and Durand,}

Bernard and Guillemet, Sophie and Ching, Jean-Yves Chane andFantozzi,

Gilbert and Garnier, Vincent and Bonnefont, Guillaume Effects of the nature

of the doping salt and of the thermal pre-treatment and sintering temperature

on spark plasma sintering of transparent alumina. (2011) Ceramics

International, vol. 37 (n° 8). pp. 3565-3573. ISSN 0272-8842

O

pen

A

rchive

T

OULOUSE

A

rchive

O

uverte (

OATAO

)

OATAO is an open access repository that collects the work of Toulouse researchers and

makes it freely available over the web where possible.

This is an author-deposited version published in :

http://oatao.univ-toulouse.fr/

Eprints ID : 8689

To link to this article

_{: DOI:10.1016/j.ceramint.2011.05.152}

URL :

http://dx.doi.org/10.1016/j.ceramint.2011.05.152

Any correspondance concerning this service should be sent to the repository

administrator:

staff-oatao@listes-diff.inp-toulouse.fr

Effects

of

the

nature

of

the

doping

salt

and

of

the

thermal

pre-treatment

and

sintering

temperature

on

Spark

Plasma

Sintering

of

transparent

alumina

Nicolas

Roussel

,

Lucile

Lallemant

,

Bernard

Durand

*

,

Sophie

Guillemet

,

Jean-Yves

Chane

Ching

,

Gilbert

Fantozzi

,

Vincent

Garnier

,

Guillaume

Bonnefont

a

CIRIMAT,UMRCNRS5085,Universite´ PaulSabatier–118routedeNarbonne–F-31062TOULOUSECedex9,France b

Universite´ deLyon,CNRS,France c

INSA-Lyon,MATEISUMR5510,F-69621Villeurbanne,France

Abstract

Aslurryofa-Al2O3
wasdopedwithMg,ZrandLanitratesorchlorides,invariousamountsintherange150–500wtppmandthenfreeze-dried

toproducenanosizeddopedpowder(150nm).ThepowderwassinteredbySPStoyieldtransparentpolycrystallinealphaalumina.Theinfluence ofthenatureofthedopingelementandthestartingsalt,thethermaltreatmentbeforesinteringandthesinteringtemperatureonthetransparencyof theceramicswereinvestigated.ThetransparencyoftheceramicsofnanosizedAl2O3
wasshowntodependmainlyonthewaythepowderwas

prepared,thenatureofthedopingsaltalsohadaneffect.Finally,ahighrealinlinetransmittance,reaching48.1%wasachievedafteroptimization.

Keywords:B.Grainsize;B.Porosity;C.Opticalproperties;D.Al2O3;Doping;SparkPlasmaSintering

1. Introduction

For both mechanical and economical reasons, numerous studies have been performed over recent years to replace sapphire by transparent polycrystalline alumina (PCA) in variousopticalapplicationssuchasdischargelampenvelopes andopticalwindowsorarmour [1–5].The lighttransmission propertiesoffine-grainedPCAcanbedescribedbytheApetz and van Bruggen model [6], based on the Rayleigh-Gans-Debye approximation. The realinline transmittance (RIT) is stronglydependentongrainsize(i.e.lightscatteringbygrain boundariesgG)andporosity(i.e.lightscatteringbyporesgp)as

canbeseenfrom thefollowingequations: RIT¼I2 I1 ¼ð1ÿRsÞ
expðÿgtot
DÞ (1) gtot¼gGþgp¼ 3p2_rDn2 l20 þ₄ p 3p
r3p
Csca;p (2)

with I1 and I2 the light beam intensities before and after travellingthroughasampleofthicknessD;Rsthetotalnormal

surface reflectance (=0.14 for PCA); gtot the total scattering coefficient;rtheaveragegrainradius;Dntheaveragerefractive indexchangebetweentwoadjacentgrains(=0.005forPCA),l0 the wavelength of incident light under vacuum; p the total porosity, rpthe average pore radiusand Csca,p the scattering crosssectionofonesphericalpore[6,7].AsPCAisa birefrin-gentmaterial,themodelpredictsthattoobtainhighrealin-line transmittanceoverthespectrum,bothgrainsize(<0.5mm)and porosity (<0.05% with a narrow distribution of nanometric pores)havetobecarefullycontrolled.Todoso,twostrategies werecombined:dopingaluminawithmetaloxides(ppmrange) andsinteringthe powdersbySparkPlasmaSintering(SPS).

The effectsof the dopingelements onalumina have been studied for many years but are still the subject of diverse research.Thefirstruletocontrolthedensificationofaluminais tostartfromahighpuritypowder.Theprocessisverysensitive to impurities, especially silicon, calcium and sodium [8,9], whichcanleadtoinhomogeneousdensificationwithabnormal grain growth [9]. Small amounts of various doping oxides (MgO,La2O3,Y2O3)canbeintroducedinthealuminatohavea bettercontrolofdensification.Mostofthesedopingagents,due

* Correspondingauthor.Tel.:+33661567751;fax:+33661556163.

to their very low solubility in alumina, segregate at grain boundaries during the sintering. Some Secondary-Ion Mass Spectrometry (SIMS)images illustratethisphenomenonvery clearly;afterdensification,thedopingagent(MgO,La2O3and Y2O3)islocatedatgrainboundariesandporesurfaces[10,11]. For La2O3 and Y2O3 doping, the segregation leads to an inhibitionofgraingrowthanddecreasesthedensificationrate

[11].Therate-controller,duringsinteringoffinealumina,was reportedbysomeauthorstobegrainboundarydiffusion[12– 15].Yoshida’sgrouptriedtoexplainhowthedopingelements actongrainboundarydiffusivity.Itisclearthatthevalenceof the cation cannot explain this mechanism because Pt4+ increases the grain boundary diffusion while Zr4+ decreases it. According tothesame authors,grainboundarydiffusivity cannotbepredictedbythesizeofthecation,butbyionicityin thevicinityofthegrainboundaries.Thisionicityimprovesthe ionic bondstrengthandcanlimit atomicdiffusion[16].

SPSisaprocessingtechniquethathasbeenwidelyusedover the past decades [1,5,17–20]. It was shown that dense, fine-grained materials canbe developed at low temperaturesand shortsinteringtimescomparedtoconventionaltechniquessuch as hot isostatic pressing (HIP). Grain growth is then significantly reduced. Several studies have been performed over recent yearstoobtain transparentalumina bySPS. The heating rate has been found to be a critical parameter in obtainingsuchamaterial.AccordingtoAmanetal.[20], grain-boundary diffusion probably dominates at low heating rates whereas grain coarsening, in respect of thermal equilibrium considerations,isunavoidableathighheatingratesduringthe initialstagesoflowtemperaturessintering.AccordingtoKim etal.[1],whensinteringatunder12508C,rapidheatingcreates highdefectconcentrationsinducingdynamicgraingrowthand thus decreasing the transparencyof the material. The in-line transmissionofpurePCAwasincreasedbyabout15%(upto 46%atawavelengthof640nm)ondecreasingtheheatingrate from10to28C/min.Recently,Stueretal.[5]increasedtheRIT ofSpark PlasmaSintered PCAup to57%at 640nm,by tri-dopingthepowderwithMg,LaandYwhichprovesthebenefit ofcombiningthestrategyofdopingwithSPS.However,their RITvaluesarestillfarfromthosefoundbyKrelletal.[2]using HIP(72%at640nm).Greenbodyshapingwasshowntobeof crucial importance by Krell et al. [21] for natural sintering. Recently,Amanetal.[22]haveshownthatprocessingthegreen state also has astrong influence onthe optical propertiesof Spark Plasma Sintered PCA. However, some aspects of the powder preparation such as the thermal treatment before sintering and the nature of the starting salt, have not been studiedyet.Theaimofthisworkistogiveabasicoverviewof thepowderpreparationparameterswhichcanplayaroleduring the SparkPlasmaSinteringof transparentPCA.

2. Experimental

The starting material was a commercial (BA15PSH, Baı¨kowski) high purity a-Al2O3 slurry (solid content 73.5wt%) with a median particle size D50 of 150nm. The totalamountofimpuritywaslessthan0.01wt%(14ppmNa,

60ppmK,7.1ppmFe,13ppmSi,4ppmCa)asreportedbythe manufacturer.

The doping agents were introduced by weighing out the requiredamountofwater-solublesalts(nitratesorchlorides)of lanthanum,zirconiumandmagnesium,pouringtheminto the aluminaslurryandstirringfor24hbyrotationofthecontainer. Then,theslurrywasfrozeninliquidnitrogenandfreeze-dried forapproximately48h (ÿ408C,0.1mbar,Alpha2-4,Christ). Theresultingpowdersweresinteredeitheras-obtainedorafter thermalpre-treatmenttotransformthesaltinto oxide,at450, 500and6508CforMg,ZrandLarespectively,withnoholding time.

Thechosendopingagentamounts(wtppm)were:100,200, 500forZrO2,100,200,500forLa2O3and150,300,500 for MgO.Allppmgiveninthefollowingareexpressedinweight. Specific surface areas were measured by BET (Micro-meritics FlowSorb II 2300), pore sizes were determined by mercuryinfiltration(MicromeriticsInstrumentCorp.)andTGA wasperformedusing aSetaramTAG24.

Densification was carried out by either conventional sinteringorSPS.Inthefirstcase,6mmdiameterpelletswere uniaxially pressed at 200MPa for 2min and sintered in a conventionalfurnaceat13508Cinstaticairfor2hataheating rateof 2008C/handthenfinaldensitieswereestimatedfrom size and weight measurements. For SPS the powders were sieved (<500mm) and sintered (HP D 25/1, FCT System, Rauenstein, Germany) using the following cycle: applied uniaxialpressureof80MPathroughoutthecycle,rapidheating up to 8008C, a heating rate of 108C/min from 8008C to 11008Cfollowedbyaslowerheating(18C/min)uptothefinal sintering temperature (Tf) in order to remove the residual

porosity[1].The finalsintering temperaturewas inthe range 1180–12808C.Rapidcoolingendedthecycle,interruptedbya 10-min dwell at 10008Cto release the residual stresses [1]. Then,thesampleswerecarefullymirror-polishedonbothsides usingdiamondslurriesandthetransparencywasevaluatedbya real in-line transmittance (RIT) measurement (Jasco V-670) which only takes into account the unscattered light passing straightthroughthesample (i.e. therealtransmittedlight)as explainedbyApetzandvan Bruggen [6].Allthe RITvalues giveninthispaperwere measuredatl=640nm andEq. (3)

wasusedtoobtaintheRITatthesamethicknessof0.88mmto comparetheresults:

RITðt2Þ¼ð1ÿRSÞ

RITðt1Þ

1ÿRS

 t2=t1

(3) whereRSisthetotalnormalsurfacereflectance(=0.14forPCA)

andRIT (ti)istheRITfor asampleofthicknessti.

SETARAMSetsysevolutionTMA-16/18wasusedtocarryout

dilatometricmeasurementsinairwithaheatingrateof2.58C/ minandafinaltemperatureof16008C.

SEMJEOLJSM-6510LVandSEMPHILIPSESEM-FEGFEI

XL30 were used to investigate the microstructure of the samples.The averagelineinterceptmethodhasbeen usedto estimategrainsizesonfracturesurfaces,applyingacorrection factorof1.56[23].

3. Resultsanddiscussion

3.1. Effectofsintering temperatureand doping

As too higha temperaturecan cause graingrowth during sintering,thisparameterneedstobecarefullycontrolled.Five samplesoffreezedriedpurealuminaweresinteredbySPSat differentTf (1180–1200–1230–1250–12808C).RITandgrain

sizeweredetermined(Table1).AmaximumRITof33.6%was reachedat11808C. Butatthistemperatureandat12008C,a gradient of transparency was noted, indicating that the temperaturewas not the same at the centre (very dense and transparent)andattheedge(notdenseandtranslucent)ofthe pellet [24–26].The gradient was nolonger observedfor the sample sintered at 12308C which presented a high and homogeneousRITof 30.5%. Moreover,when comparing the RITallover thespectrum (300–2500nm)(Fig.1),itappears thattheRITofthesamplesinteredat11808Cwaslowerthan thatofthesamplesinteredat12308CintheIRwavelengths(at

l=2000nm, RIT (11808C)=64.9% and RIT

(12308C)=75.5%). The grain size of the ceramicincreased regularlyfrom0.52to1.93mmasthetemperaturewasraised from1180to12808C.Thisincreasecontributestothedecrease inRITas thesinteringtemperaturerises.

ThemeasuredRITofsamplessinteredbelow12308Cwere significantly lower than the theoretical values calculated for non-porous ceramics with the same grain size which corroborates the presence of residual porosity (Fig. 2). The RITmeasuredforceramicssinteredat1230,1250and12808C were very close to the theoretical ones, so 12308C can be consideredas the optimized sinteringtemperature.TheSEM micrographs(Fig.3) showthatthe grain size inall undoped ceramicsampleswasnothomogeneous.Thesamplesinteredat

12308C, wherethe mediangrain sizewas 1.17mm, exhibits grainswithdiameters rangingfrom0.8mmto1.5mm.

The grain size and the transparency of the pure alumina sintered at 12308C are in good agreement with the results foundbyKimetal.[18].Whenthetemperaturewasdecreased to11508C,thetransparencywasimprovedinthevisiblerange butdecreased inthe IRcorroboratingour resultsat11808C. Theseauthors were alsoable toincrease the transparencyof 30mm diameter pellets sintered at 11508C up to 47% and decrease grain size to 300nm [1,17]. However, these measurementsweretakenatthecentreofthepellets.According tothesameauthors[18],rapidgraingrowthoccursinthecentre of30mmdiameterpelletscomparedtotheouterpart,probably resulting in inhomogeneous transparency of the samples. Decreasingthetemperatureisbeneficialforadecreaseingrain sizeandthusanincreaseintheRIT.Buttoolowatemperature will lead to inhomogeneous samples. At low temperatures,

Table1

RITat640nmfor0.88mmthickpelletsandgrainsizeofundopedceramicssinteredbySPSatdifferenttemperatures.

Sinteringtemperature(8C) 1280 1250 1230 1200 1180 RIT(%)(l=640nm) 18.9 23 30.5 27.4 33.6 Grainsize(mm) 1.930.35 1.500.23 1.170.20 0.800.19 0.520.11 0 500 1000 1500 2000 2500 0 10 20 30 40 50 60 70 80 90 RIT (%) λ (nm) 1180°C 1230°C

Fig.1.Wholespectrum(300–2500nm)RIToftwoundopedaluminasamples, sinteredat11808C(grey)and12308C(black).

0,0 0,5 1,0 1,5 2,0 2,5 3,0 0 10 20 30 40 50 60 70 80 90 RIT (%) grain size (µm) pure alumina

theoretical without porosity theoretical 0.01% porosity theoretical 0.05% porosity 1180°C 1200°C 1230°C 1250°C 1280°C

Fig.2. Comparison of theRITofpure aluminasinteredby SPSwith the theoreticalcurves(l=640nmth=0.88mm).

temperature gradient led tomore porosity atthe edge of the pelletexplainingthelossofRIT.Whenincreasingtemperature, lastporositieswillshrinkandanoptimumbetweenlowergrain size withahigheramountof porosityat theedge andhigher grainsizewithaloweramountofporosityatthecentrewillbe found,explainingthehomogenizationoftheRIT.Finally,when increasingagaintemperature,thesameamountof porosityis foundalloverthesample.ItmeansthattheRITdifferenceis onlyduetograinsizedifference.Thisdifferenceisdecreasing whenincreasinggrainsizeexplainingwhythesamplesarestill homogeneous (Fig. 2). Optimizing temperature is a critical pointtoobtainanoptimumbetweenanhomogeneoussample havingahighRIT.

It has been proved over the past decadesthat the use of appropriate doping agents can decrease the grain size of sintered alumina and thus improve the RIT. Conventional sinteringat13508Cfor2hunderstaticairwasperformedon4 non-thermallypre-treatedfreeze-driedaluminapowders:pure aluminaandaluminadopedwithMg2+,Zr4+andLa3+nitrates. Therelativedensityoftheundopedceramicwas93.80.5%. For MgO doped ceramics, a regular increase of density was observed(93.70.3%,94.30.4%and95.70.5%for150, 300 and 500ppm respectively). With ZrO2 and La2O3, the inverse effect was noted: for 500ppm ZrO2 and La2O3 the densitywas92.30.5%and93.10.8%respectively.These results indicate that MgO doping enhances the densification whereasZrO2andLa2O3reduceitatthissinteringtemperature. Theseresultswereconfirmedbythedilatometricmeasurements thatshowthatthedensificationoftheZrO2andLa2O3-doped samplesbeginsathighertemperaturesthanthatofMgO-doped andundopedaluminasamples(Fig.4).Thedensificationofthe MgO-dopedsamplesbeganatnearlythesametemperatureas that of undoped samples but the shrinkage of MgO-doped ceramicwasgreaterabove12808C.Thelaterofdensification of ZrO2 and La2O3-containing samples is explained by the segregationof thedopingagentsatgrainboundaries(insolid solution and/or as precipitates). Indeed some works have already reported this segregation for all the doping agents considered[10,12,27–31]duringconventionalsintering.Thus, the diffusion of the Al ions is reduced and grain boundary

slidingis limited. The result iscontrolled grain growth with densificationathighertemperatures.Thiseffectdependsonthe amount of doping agents included: as the dosage of doping agentdecreased,theshrinkagecurvesbecameclosertothoseof undopedsamples.

Accordingtotheseresults,optimizingthefinaltemperature duringthe‘‘SPScycle’’foreachdopingagentisnecessaryto achievefully densepellets.

3.2. Effectofthermalpre-treatment ofthe powder

Powders obtained by freeze-drying were thermally pre-treated attemperaturescorresponding tothe thermal decom-positionof the dopingelementsalts(nitratesor chlorides)to yieldoxides,i.e.450,500,650and7008CforMg(nitrate),Zr (nitrate and chloride), La (nitrate and chloride) and Mg (chloride) respectively, with no holding time. Preliminary thermal pre-treatment at 2008C showed no influence on the specificsurfaceareas(SSA)ofpowderswhichremainedinthe same range as those of non-thermally treated powders, 19– 20m2gÿ1. Lower SSA, between 14 and 16m2gÿ1 were obtainedafter thermalpre-treatmentat6508Cwithoutdwell. Thermalpre-treatment at7008Cwith1h dwell timeweakly

800 1000 1200 1400 1600 60 65 70 75 80 85 90 95 100 Relative density (% TD) Temperature (°C) Undoped MgO ZrO₂ La₂O₃ 800 1000 1200 1400 1600 0,000 0,005 0,010 0,015 0,020 densification rate (g.cm .° C ) Temperature (°C)

Fig. 4. Dilatometric measurements under air of 500ppm doped alumina powders(MgO,ZrO2,La2O3)preparedwithnitratesalts.

Fig.5. SEMpicturesofthefreeze-dried500ppmMgO-dopedaluminapowder (a)non-thermallypre-treated(b)thermallypre-treatedat7008Cfor1h.

decreasedtheSSAto12–13m2gÿ1.SEManalysesshowedthat agglomerationappearedatthiselevated temperature(Fig.5). Porositymeasurementsbymercuryinfiltrationwereperformed on200ppmZrO2-dopedaluminapowder(nitratebased)either thermallypre-treatedat5008Cornot(Fig.6a).Bothpowders presentedthesamerangeofintra-granularporosity(20–60nm) indicating the same primary particle arrangement. However, bigger inter-granular porosity was found on thermally pre-treatedpowder(10–300mm)indicatingthepresenceoflarger agglomerates.Thetwokindsofpowders(thermallypre-treated ornot)werethenuniaxiallypressedat50MPa.Fig.6bshows the subsequent results of mercury infiltration measurement. The larger agglomerates were no longer present in the thermally pre-treated sample after pressing but porosity (100nm–1mm) was larger than that measured for the non-thermally treated samples (<200nm). It is well known that largepores are harder to removeduring sintering than small ones.

Three samples (undoped, 500ppm ZrO2-doped and 500ppmMgO-doped)werepreparedwithandwithoutthermal pre-treatmentat6508C.Nitratesaltswereusedforthedoping. After sintering in a conventional furnace at 13508C, the

relative densities of all thermally pre-treated samples were lowerthanthoseof the non-thermallytreatedsamples. Thus, the densityof theundoped samples fellfrom93.80.5% to 87.20.2%,thatoftheMgO-dopedsamplesfrom95.70.1 to 90.30.4 and that of the ZrO2-doped samples from 92.30.5 to 84.60.3. When thermally pre-treating the powder,densificationoccurredatthesametemperatureas non-thermally pre-treated powder (Fig. 7) but the density of the greenbodywas higherfor non-thermallypre-treatedsamples leading to an improvement of densification (higher relative densitythroughoutdensification).Furthermorethehighergreen bodydensityfornon-thermallypre-treatedpowderwasingood agreementwiththemercuryinfiltrationporositymeasurements. Agglomeration during thermal pre-treatment led to poor particle packing so the pores were harder to remove during sinteringas explainedbyAzaretal.[32].

3.3. Nitrateorchloridedopingsalteffect

Conventional sintering (13508C/2h) was performed on 300ppm MgO-doped alumina and 200ppm ZrO2-doped alumina,nonthermallypre-treated.Inbothcases,twosamples wereprepared:onedopedusingthechloridesaltandtheother using thenitrate salt.The samplesdopedusingchloride salts

Fig.6. Porosimetrymeasurementbymercuryinfiltrationofa200ppmZrO2

-dopedaluminapowder(nitrate)(a)incrementalporevolumeofthepowder(b) cumulatedporevolumeafterpressing(50MPa).

800 1000 1200 1400 1600 60 65 70 75 80 85 90 95 100 R e la ti v e d e n s it y (% T D ) Temperature (°C) NO₃- salt Cl- salt 0,000 0,005 0,010 0,015 0,020 d e n s if ic a ti o n r a te ( g .cm -3 .° C -1 )

Fig.8.Dilatometricmeasurementofnon-thermallypre-treatedalumina pow-dersdopedwith300ppmofeitherMgnitrateorchloridesalt.

800 1000 1200 1400 1600 55 60 65 70 75 80 85 90 95 100 R e la ti v e d e n s it y ( % T D ) Temperature (°C) BA15PSH non calcined BA15PSH 650°C 0,000 0,005 0,010 0,015 0,020 d e n s if ic a ti o n r a te ( g .c m -3 .° C -1 )

Fig. 7. Dilatometric measurements of pure freeze-dried alumina powders thermallypre-treatedornot.

exhibitslightlyhigherdensitiesthanthosedopedusingnitrate salts(96.1%TDagainst94.3%TDfortheMgO-dopedalumina and94.9%TDagainst92.9%TDfortheZrO2-dopedalumina). Theresultsareconfirmedbythedilatometriccurves(Fig.8)as shown for the MgO-doped sample: sintering occurred atthe same temperature for both salts(nitrate and chloride), while dopingwithchloridesaltslightlyimprovedtherelativedensity attheendofsintering.Moreover,200ppmLa2O3-doped non-thermallytreatedaluminapowderspreparedeitherwithnitrate

orchloridesaltsweresinteredbySPSatafinaltemperatureof 12808C.Forl=640nmandathicknessof0.88mm,theRIT measurementwasslightlyimprovedforthesampledopedwith chloride salt (48.1% against 45.9% for sample doped with nitratesalt).Sincethesampleswerenotthermallypre-treated, chloride or nitrate ions may have played a role during the sintering.

To understand the slightly better sintering behaviour of samplespreparedwithchloridesalt,porositywasmeasuredby mercury infiltration on 200ppm ZrO2-doped green bodies (non-thermallypre-treatedpowderswereuniaxiallypressedat

1E-3 0,01 0,1 1 10 100 1000 0,00 0,05 0,10 0,15 0,20 0,25 C u m u la te d p o re v o lu m e ( m L /g )

diameter to access porosity (µm) nitrate salt chloride salt

Fig.9. Porosimetrymeasurementbymercuryinfiltrationofnon-thermally pre-treated200ppmZrO2-dopedaluminapowder(nitrateorchloride)–cumulated

porevolumeafterpressing(50MPa).

Yes NO3 -Cl -1280°C 1280°C No Cl -1280°C RIT = 36.3 % RIT = 27.3 % RIT = 48.1 % NO3- 1280°C RIT = 45.9 %

200ppm La2O3-doping – RIT Measurement (λ = 640nm – th. = 0.88mm)

Yes NO3 -Cl -1280°C 1270°C No Cl -1270°C RIT = 33.9 % RIT = 28.8 % RIT = 40.2 % NO3- 1260°C RIT = 35.5 %

200ppm ZrO2-doping – RIT Measurement (λ = 640nm – th. = 0.88mm)

Yes NO3 -Cl -1240°C 1260°C No Cl- 1240°C RIT = 16.8 % RIT = 27.5 % RIT = 44.1 % NO3- 1240°C RIT = 36.6 %

300ppm MgO-doping – RIT Measurement ( λ = 640nm – th. = 0.88mm)

Thermal pre-treatment Thermal pre-treatment Thermal pre-treatment

(a)

(b)

(c)

Fig.11.RITmeasurements(l=640nmth=0.88mm)of(a)300ppmMgO-doped(b)200ppmZrO2-doped(c)200ppmLa2O3-doped.

0 200 400 600 800 1000 1200 1400 -3 -2 -1 0 d W t/ d t W e ig h t lo s s ( % ) Temperature (°C) Mg(NO ) precursor MgCl precursor -0,0008 -0,0006 -0,0004 -0,0002 0,0000 0,0002

Fig.10. Thermogravimetricmeasurements(TGAandDTG)of2350ppmMgO dopedalumina(nitrateorchloride).

50MPa) prepared either with nitrate or chloride salt. No significant difference was observed (Fig. 9) and sintering behaviour of powders doped with chloride salts cannot be explained by a better pressing behaviour or final packing homogeneity.

Thermogravimetricanalysis under air at aheating rate of 2.58C/min was also performed on 2350ppm MgO-doped, 7250ppm ZrO2-doped and 9650ppm La2O3-doped alumina powders. Comparable behaviour was observed for the three doping elements. As shown in Fig. 10 for the MgO-doped samples, the chlorides and the nitrates are transformed into oxides inthe same temperature range. The main part of the weightlossoccursduringthethreestepsbelow3508Candthen the weight slowly and regularly decreasesuntil it reaches a constantvalueatabout10008C.Thewholeweightloss,close to3wt%wasfivetotenfoldhigherthanthatcalculatedforthe transformationintooxideoftheamountofdopingchlorideor nitrate, i.e. respectively 0.30 and 0.59wt% for MgCl2 and Mg(NO3)2.Below2508C,thetwofirstlossesareundoubtedly attributedtothereleaseofadsorbedwaterandabove2508Cit isdifficulttodistinguishtheendofdehydrationandthestartof decompositionofthedopingsalts.Althoughthermogravimetric investigationdidnotclarifythedifferencebetweennitratesand chlorides, doping with chloride salts enables slightly higher densitiestobereachedinnaturalsinteringandaslightlybetter RIT after SPS sintering. Thus, it was decided to introduce dopingelementsintotheslurryintheformoftheirchloridesalt forthe followingstepsof thisstudy.

3.4. Sinteringoptimizationofdopedpowders

Dopedsamplespreparedwitheitherchlorideornitratesalts (300ppm MgO, 200ppm ZrO2, 200ppm La2O3) either thermally pre-treated or not were sintered by SPS. Final sinteringtemperaturewasoptimisedfor eachdopingagent as densification was delayed by their presence (fig. 4). RIT measurements were performed on these samples and the optimisedresultsaregiveninFig.11.Theresultsareingood agreementwiththoseof theconventional sintering presented above. The SPS efficiency is also sensitive to the powder preparation, as previously shown in Section 3.2 for natural sintering.Thepresenceoflargeagglomeratesinthethermally pre-treated powderis detrimental to full densificationof the ceramic.Theresidualporosityisassociatedtotheformerlarge inter-agglomerate pores, reducing the RIT value. Moreover, doping with chloride salt has slightly increased the RIT, as previouslyshowninSection3.3,eventhoughthiseffectwas lesspronouncedthanthatofthermalpre-treatment.Finally,we wereabletoincrease theRITofdoped-PCAupto48.1%for La2O3-dopedalumina.

TheSEMmicrographsofceramicssinteredattheoptimised temperaturefornon-thermallypre-treatedpowdersdopedwith chloridesaltsarereportedinFig.12.Allthedopingagentsled toadecreaseingrainsize (fG)comparedtothe freeze-dried

purealuminasinteredat12308C.(fGwasequalto1.170.20,

0.630.24, 0.790.13, 0.810.06mm respectively for pure,MgO,ZrO2,La2O3-dopedalumina).However,thegrain

Fig.12. SEMpicturesofnon-thermallypre-treatedchloridebased(a)300ppm MgO-(b)200ppmZrO2-(c)200ppmLa2O3-dopedalumina(sintering

size was still heterogeneous even though an improvement occurredespeciallyfortheLa2O3-dopedsample.Moreover,the porosity remaininginthesedopedsamples,exceptfor La2O3 samples, was in the same range as that of optimized pure aluminaatcloseto0.01%(Fig.13).FinallyLa2O3seemstobe more efficient thanthe other dopingagents atincreasingthe RIT of PCA but the underlying reasons are still under investigation.

TheRITvaluefoundfortheLa2O3-dopedsampleiscloseto thatfoundbyStueretal.[5](50%at640nmforaLa2O3-doped samplewithathicknessof0.88mm)witha20-minSPScycle (heatingrate=1008C/min)andafinaltemperatureof13508C. Thecycleusedinourstudylasted4htoobtainhighdensityat low temperature. We attempted to reproduce Stuer’s results usinghisspecificthermalcycle,butnotransparencywasfound withcyclesattemperatureshigherthantheoneweoptimised. This can be explained by the amount of dopingagent used. Stuerdopedaluminawith3.6timesmoreLa2O3thanwedid, probably delayingthe sinteringover 1250–13008C. Another explanation can befound inthe different size of the starting powderparticles: around5 timeslargerinStuer’sstudy. The higherparticlesizemaydelaysinteringtohighertemperatures. Our RIT values can still be improved by optimizing the amount of each dopingagent. Thiseffect is currently under study.

4. Conclusion

Obtaining transparent polycrystalline alumina requires a verylittleamountofporosityandgrainsasfineaspossible.This particularpointcanbeachievedbyboththeuseofdopingagent and the SPS sintering technique for the densification of the sampleasalreadypublishedbyStueretal.[5].However,other parametershavetobeoptimisedallover theprocesstoavoid defectsasagglomeratesorporosity.Inthisstudy,thethermal pre-treatment,thenatureofthedopingsaltandelementandthe sinteringtemperaturehavebeeninvestigatedandtheireffects on the transparency of polycrystalline alumina (PCA) have been characterized. It has been shown that the sintering temperature has to be carefully optimised for each powder (dopedornot)inordertoobtainahomogeneoussamplewitha

RITashighaspossible.Then,fordopedsamples,thethermal pre-treatment will lead to the formation of agglomerates, decreasing the specific surface areas and increasing sample inhomogeneity,makingdensificationharderandloweringthe resulting RIT. Moreover, doping with chlorides instead of nitratescan helptoslightlyimprovedensificationin conven-tionalsinteringandleadtoaslightimprovementoftheRITby SPSsintering.

Finally, each dopingagent enablesto increase the RITof PCAat640nm(40.1%forZrO2,44.1%forMgOand48.1%for La2O3against30.5%forpurealumina).IfLa2O3appearstobe themostefficientone,thereasonshavenotbeendeterminedyet andarestillunderinvestigation.Nevertheless,theresultsinthis study should highlight the fact thateach step of the process (from the powderpreparation to the sintering of transparent polycrystallinealuminasamples)hastobecarefullyoptimised toavoidanydefects astemperature/microstructuregradients, agglomerationorporositywhichwillinducealossof RIT. Acknowledgements

TheauthorsaregratefulforthefinancialsupportoftheANR CERATRANS and to all the participants, Lionel Bonneau

(Baikowski),JohanPetit(ONERA),Ste´phaneChaillot(BOOSTEC)

andDenisLanglade(BTS Industrie). References

[1] B.N.Kim,K.Hiraga,K.Morita,H.Yoshida,Effectsofheatingrateon microstructureandtransparencyofspark-plasma-sinteredalumina,J.Eur. Ceram.Soc.29(2009)323–327.

[2] A.Krell,J.Klimke,T.Hutzler,Advancedspinelandsub-mmAl2O3for

transparentarmourapplications,J.Eur.Ceram.Soc.29(2009)275–281. [3] A.Krell,T.Hutzler,J.Klimke,Transmissionphysicsandconsequences formaterialsselection,manufacturing,andapplications,J.Eur.Ceram. Soc.29(2009)207–221.

[4] A.Krell,J.Klimke,T.Hutzler,Transparentcompactceramics:inherent physicalissues,Opt.Mater.31(8)(2009)1144–1150.

[5] M.Stuer,Z.Zhao,U.Aschauer,P.Bowen,Transparentpolycrystalline aluminausingsparkplasmasintering:effectofMg,YandLadoping,J. Eur.Ceram.Soc.30(6)(2010)1335–1343.

[6] R.Apetz,M.P.B.vanBruggen,Transparentalumina:alight-scattering model,J.Am.Ceram.Soc.86(3)(2003)480–486.

[7] www.lightscattering.de/MieCalc/.

[8] S.J.Bennison,M.P.Harmer,Grain-growthkineticsforaluminain the absenceofaliquidphase,J.Am.Ceram.Soc.68(1)(1985)C22–C24. [9] S.J.Bennison,M.P.Harmer,EffectofMgOsoluteonthekineticsofgrain

growthinAl2O3,J.Am.Ceram.Soc.66(5)(1983)C90.

[10] A.M.Thompson,K.K.Soni,H.M.Chan,M.P.Harmer,D.B.Williams, J.M. Chabala, R. Levi-Setti,Dopant distributionsin rare-earth-doped alumina,J.Am.Ceram.Soc.80(2)(1997)373–376.

[11] K.K.Soni,A.M.Thompson,M.P.Harmer,D.B.Williams,J.M.Chabala, R. Levi-Setti, Solutesegregation to grain boundaries in MgO-doped alumina,Appl.Phys.Lett.66(21)(1995)2795–2797.

[12] J.Fang,A.M.Thompson,M.P.Harmer,H.M.Chan,Effectofyttriumand lanthanumonthefinal-stagesinteringbehaviorofultrahigh-purity alumi-na,J.Am.Ceram.Soc.80(8)(1997)2005–2012.

[13] K.A. Berry, M.P. Harmer, Effect of MgO solute on microstructure developmentinAl2O3,J.Am.Ceram.Soc.69(2)(1986)143–149.

[14] A.H.Chokshi,Anevaluationofthegrain-boundaryslidingcontributionto creepdeformationinpolycrystallinealumina,J.Mater.Sci.25(7)(1990) 3221–3228. 0,0 0,5 1,0 1,5 2,0 0 10 20 30 40 50 60 70 80 90 RIT (%) grain size (µm) pure alumina MgO doping ZrO2 doping La2O3 doping

thereotical without porosity thereotical 0.01% porosity thereotical 0.05% porosity

Fig.13. ComparisonoftheRITvaluesofnon-thermallypre-treatedchloride baseddopedsampleswiththetheoreticalcurves(l=640nmth=0.88mm).

[15]J.D.Wang,R.Raj,Estimateoftheactivationenergiesforboundarydiffusion fromrate-controlledsinteringofpurealumina,andaluminadopedwith zirconiaortitania,J.Am.Ceram.Soc.73(5)(1990)1172–1175. [16] H.Yoshida,S.Hashimoto,T.Yamamoto,Dopanteffectongrainboundary

diffusivityinpolycrystallinealumina,ActaMater.53(2005)433–440. [17] B.N.Kim,K.Hiraga,K.Morita,H.Yoshida,Sparkplasmasinteringof

transparentalumina,ScriptaMater.57(7)(2007)607–610.

[18] B.N.Kim,K.Hiraga,K.Morita,H.Yoshida,T.Miyazaki,Y.Kagawa, Microstructureandopticalpropertiesoftransparentalumina,ActaMater. 57(2009)1319–1326.

[19] Z.Shen,etal.,Sparkplasmasinteringofalumina,J.Am.Ceram.Soc.85 (8)(2002)1921–1927.

[20] Y.Aman,V.Garnier,E.Djurado,Sparkplasmasinteringkineticsofpure a-alumina,J.Am.Ceram.Soc.94(4)(2011).

[21] A.Krell,P.Blank,H.Ma,T.Hutzler,M.Nebelung,Processingof high-densitysubmicrometerAl2O3fornewapplications,J.Am.Ceram.Soc. 86(4)(2003)546–553.

[22] Y.Aman,V.Garnier,E.Djurado,Influenceofgreenstateprocessesonthe sinteringbehaviourandthesubsequentopticalpropertiesofsparkplasma sinteredalumina,J.Eur.Ceram.Soc.29(2009)3363–3370.

[23] M.I.Mendelson,Averagegrainsizeinpolycrystallineceramics,JACS52 (1969)443–446.

[24] K.Vanmeensel,A.Laptev,J.Hennicke,J.Vleugels,O.VanderBiest, Modellingofthetemperaturedistributionduringfieldassistedsintering, ActaMater.53(2005)4379–4388.

[25] U.Anselmi-Tamburini,S.Gennari,J.E.garay,Z.A.Munir,Fundamental investigationsonthesparkplasmasintering/synthesisprocessII. Model-lingofcurrentand temperaturedistributions,Mater. Sci.Eng. A394 (2005)139–148.

[26] S.Mun˜oz,U.Anselmi-Tamburini,Temperatureandstressfieldevolutions insparkplasmasintering,J.Mater.Sci.45(2010)6528–6539. [27] J.T. Lin, H.Y. Lu, Grain growth inhibition and mechanical property

enhancementbyaddingZrO2toAl2O3 matrix,Ceram.Int. 14(1988)

251–258.

[28] D.Monceau, C.Petot,G.Petot-Ervas, J.W.Fraser, M.J.Graham,G.I. Sproule,SurfacesegregationandmorphologyofMg-dopeda-alumina powders,J.Eur.Ceram.Soc.15(1995)851–858.

[29] H.Yoshida,Y.Ikuhara,T.Sakuma,Grainboundaryelectronicstructure relatedtothehightemperaturecreepresistanceinpolycrystallineAl2O3,

ActaMater.50(2002)2955–2966.

[30] H.Yoshida,Y.Ikuhara,T.Sakuma,Vacancyeffectofdopantcationonthe high-temperaturecreepresistanceinpolycrystallineAl2O3,Mater.Sci.

Eng.A319–321(2001)843–848.

[31] J.Cho,M.P.Harmer,H.M.Chan,J.M.Rickman,A.M.Thompson,Effect ofyttriumandlanthanumonthetensilecreepbehaviorofaluminumoxide, J.Am.Ceram.Soc.80(4)(1997)1013–1017.

[32] M. Azar, P. Palmero, M. Lombardi, V. Garnier, L. Montanaro, G. Fantozzi,J.Chevalier,Effectofinitialparticlepackingonthesintering of nanostructuredtransitionalumina,J.Eur. Ceram.Soc.28 (2008) 1121–1128.