Alumina journal - Sabancı Üniversitesipeople.sabanciuniv.edu/~m-gulgun/Ca-doped-Al2O3.pdf-alumina (0.0212 nm3/cation), and G is the grain size. The grain boundary excess concentrations

Microstructural Evolution of Calcium-Doped �-Alumina

Arzu Altay* and Mehmet Ali Gulgun*Sabanci University, FENS, 34956 Orhanli, Tuzla, Istanbul, Turkey

Effect of different calcium doping levels on the microstructureof high-purity �-alumina was studied as a function of sinteringtime and temperature using scanning electron microscopy(SEM). Microstructural evolution was related to hypotheticalcalcium excess at the grain boundaries (�Ca) that was calcu-lated assuming zero solubility of calcium in bulk �-alumina.Under all sintering conditions, grains were uniform in size andequiaxed for low calcium concentrations (<3 Ca atoms/nm2).The grain morphology became elongated when the calciumconcentration at the grain boundaries reached calcium excessof �Ca � 3–3.5 Ca atoms/nm2 in all samples. The average grainsizes of undoped samples were �10% larger than the averagegrain sizes of low-calcium-doped samples. This decrease isbelieved to be due to solute drag effect of segregated Caimpurities on the grain boundary mobility. For the samplesthat were sintered at 1500° and 1600°C, slablike abnormallygrown grains appeared for critical calcium excess concentra-tions of �Ca � 4.5–8 Ca atoms/nm2. With abnormally growngrains a dramatic increase in average grain size was observed.However, when the calcium concentration was increased fur-ther, above certain calcium excess concentration depending onsintering temperature, a significant decrease in grain size wasobserved. In contrast to samples sintered at 1500° and 1600°C,when the samples sintered at 1400°C, although the hypothet-ical calcium coverage exceeded �Ca � 11 Ca atoms/nm2, onlyfew grains grew abnormally without significantly affecting theaverage grain size. Observations clearly indicated that calciumimpurities caused elongated (slablike) grain morphology whentheir excess concentrations reached a critical level at the grainboundaries.

I. Introduction

FOR many years the effects of various impurities such as Ca, Si,Mg, and Y on the microstructure and related properties of

Al2O3 have been studied extensively.1–22 Most of the studies onthe Ca-doped �-Al2O3 were focused on the anisotropic segregationof calcium to the surfaces and grain boundaries of alumina. Thereare still some disagreements among the scientists on this subject.Baik and White1 could not observe Ca segregation to the (0001)basal plane in the temperature range 800° to 1500°C althoughstrong enrichment of Ca on the (101�0) plane occurred between1300° and 1500°C. A small but noticeable amount of Ca wasdetected even below 1300°C.1 Similar experiments were done onthe segregation of magnesium and calcium to the (101�0) prismaticplane of magnesium-doped sapphire.2 In contradiction with theresults of these studies, the presence of Ca at the embedded basal

surfaces of �-Al2O3 was shown by high-resolution transmissionelectron microscopy (HRTEM) combined with analytical electronmicroscopy (AEM).3 It was suggested that Ca existed not only atthe interface plane, but rather was spread over four cation layers,which resulted in a surface phase having the nominal compositionof CaO�6Al2O3. It was also shown that alumina grains becomeelongated with Ca segregation.3

The combined effect of some impurities on the microstructureof alumina is dramatic. The presence of impurities such as SiO2

and CaO or other glass-modifying ions tends to promote abnormalgrain growth (AGG) in Al2O3.4–6 However, addition of a smallamount of MgO is a key step to control AGG.6–20 Occurrence ofAGG was related to the formation of glassy films at the grainboundaries when the amount of calcium and silica content togetherexceeded a critical concentration.4

Bae and Baik determined the minimum bulk concentration ofSiO2 and/or CaO required to trigger AGG in alumina.5 It wasreported that either Si or Ca caused AGG; however, when theycoexisted, the threshold concentration for AGG was lower.�-Al2O3 has very limited solubility for most of the impurities.This results in strong segregation of impurities (and/or dopants)to grain boundaries, which affects the sintering and microstruc-tural development of the material.21 Thus, the amount of impuri-ties at the grain boundaries plays a decisive role in determiningthe critical concentrations rather than the impurity concentra-tion in the bulk. Therefore, it is more relevant to give criticalconcentrations in terms of excess impurity atom concentrationat grain boundaries21,22

Grain boundary microstructures in a commercial 99.8% aluminaceramic were analyzed.23 Transmission electron microscopy re-vealed that all grain boundaries were wetted by an amorphous film.In the same samples several triple point pockets containing acalcium aluminosilicate glass were revealed by energy dispersiveX-ray microanalysis in the STEM. In another study, for an aluminasample that was sintered with the additions of calcium silicate, anamorphous grain boundary film with a nominal composition ofCaO�6Al2O3 was reported. In the same system both silicon andcalcium segregated to triple point pockets.24 In pure �-aluminapowder that was mixed with Si and Ca, large elongated grains withfaceted grain boundaries were observed without any frozen liquidat the triple point junctions and grain boundaries.6 In this study,addition of MgO suppressed the AGG and the grain boundariesbecame curved. The occurrence of AGG in alumina was correlatedwith the formation of faceted and straight grain boundaries.6 It wasproposed that these grain boundaries had singular ordered struc-tures with low boundary energies and their growth by lateral stepmovement could cause AGG. The addition of MgO caused grainboundary roughening and, thus, normal grain growth.6

Besides the dramatic effects of MgO on the sintering behaviorand microstructure of alumina, it was discovered that doping ofrare-earth elements such as yttrium and lanthanum has a tremen-dous influence on the microstructure and creep properties of theceramic.21,25–31 The beneficial effect of yttrium in alumina isclosely related to the segregation and/or precipitation behavior ofthe dopant.

In this study, the influence of Ca doping in the microstructuraldevelopment of an otherwise pure �-Al2O3 was investigated. The

S. M. Wiederhorn—contributing editor

Manuscript No. 186790. Received August 6, 2002; approved December 10, 2002.Presented at the International Symposium on Science and Technology of Alumina

(Schloss-Ringberg, Germany, March 17–22, 2002.Based in part on the thesis submitted by A. Altay for the M.S. degree in Material

Science and Engineering, Sabanci University, Istanbul, Turkey, 2002.Supported by the Tubitak MISAG under Grant No. MISAG-181.*Member, American Ceramic Society.

Alumina

J. Am. Ceram. Soc., 86 [4] 623–29 (2003)

623

journal

observed changes in the morphology and grain growth behaviorwere related to the calculated excess concentration at the grainboundaries.

II. Experimental Procedure

Samples were prepared from high-purity AKP-500 (Sumitomo,Osaka, Japan) �-alumina powder which contained maximum 17ppm total cation impurity initially (8 ppm Fe, 8 ppm Si, �1 ppmCu). ACS-grade calcium nitrate tetrahydrate (Merck, Germany)was used as the calcium doping source. Alumina powders withcalcium concentrations varying from 0 to 1000 ppm (molar ratio ofCa/Al2O3) were dispersed in ACS-grade 2-propanol (J. T. Baker,Phillipsburg, NJ) and ball-milled for 12 h with 99.7% pure aluminaballs (Friatec, Mannheim, Germany). After the evaporation of2-propanol, powders were ground lightly in an agate mortar, thenpressed first unidirectionally into disks under 28 MPa and coldisostatically pressed at 250 MPa for 1 min. Undoped samples wereprocessed in the same way.

Bulk chemical analyses of doped powders were done byinductively coupled plasma optical emission spectroscopy (ICP-OES). To protect the samples from impurities during sintering,they were embedded in their native powders in high-purityalumina crucibles. Sintering of samples was conducted in air at1400°, 1500°, and 1600°C for 1 and 12 h.

After sintering, densities of the samples were measured by theArchimedes method using distilled water. For microstructuralanalysis, samples were cut in half perpendicular to the axialdirection and inner surfaces of the samples were polished first withSiC emery papers and then diamond pastes. After polishing,samples were thermally etched to reveal the grain boundaries.Samples that were sintered at 1500° and 1600°C were etched at1400°C and samples that were sintered at 1400°C were etched at1300°C for 8 h inside covered alumina crucibles.

Microstructural analysis was conducted by SEM (JSM 840A,JEOL, Tokyo, Japan) and X-ray spectral measurements were doneby an attached EDS system (Oxford Link, Oxfordshire, U.K.). Allof the micrographs were taken at 10 kV and 8 mm workingdistance. From each individual sample several micrographs weretaken at various magnifications. Particles/grains with a possibleprecipitate morphology were analyzed for their chemistry usingthe EDS. Backscattered electron imaging was used to detectpossible second-phase precipitates or pockets.

Grain sizes of the samples were determined by the mean linearintercept method. The average grain size was calculated bymultiplying the mean linear intercept by 1.5. More than 400 grainsin three to four micrographs were counted for each sample. For thesamples that contained abnormally grown grains the average grainsizes of small and large grains were also determined. The averagegrain size of small grains was calculated by the mean linearintercept method. Grain sizes of large grains were measured one byone, and then the arithmetic average was taken. Always the largestdimension, i.e., the diameter of the enclosing “sphere,” of theanisotropic grains was taken as the grain size of large grains. Agrain was decided to be large if the longest observed dimensionwas approximately 3 times the average grain size of the smallgrains.

Microstructural evolution under these conditions was related tocalcium excess at the grain boundaries (�Ca). �Ca was calculatedusing a simple adsorption model as used by Gulgun et al.21 forcalculating the yttrium excess at the grain boundaries in the diluteregime:

� � Xt/Sv� � XtG/3� (1)

where � is the planar density of the dopant ion at the boundary, Xt

is the total concentration of dopant ion, Sv is the total grainboundary area per unit volume, � is the volume per cation in�-alumina (0.0212 nm3/cation), and G is the grain size. The grainboundary excess concentrations reported in this study are calcu-lated using the overall average grain size, including the samplesthat showed strong AGG. Thus, the reported values of hypothetical

Ca-excess concentrations for samples with AGG shall be regardedwith this caution in mind.

III. Results

(1) ChemistryA polycrystalline alumina ceramic was doped with controlled

amounts of calcium without any significant contamination. Bulkchemical analyses of the powders were done by ICP-OES. Theresults of the analysis are shown in Table I.

In this table the left column contains the intended amounts ofcalcium doping. In the central column the actual experimentallydetermined amounts of calcium concentrations present in thepowders are shown. Although the reason is not clear, all of themeasured calcium levels are less than intended. The trends in theobserved microstructures confirmed the results of chemical anal-ysis. Thus, all the calculations and comparisons are based on theexperimentally determined concentrations. The right column ex-hibits the bulk concentration of silicon contamination, which isestimated to be less than 5 ppm for all doping levels. A molybde-num blue indicator method was used with samples that wereintentionally doped with 5 ppm silicon. Those specimens dopedintentionally with 5 ppm silicon showed blue color although thepowders prepared in this study did not show any coloration.

(2) DensificationIt was not possible to reach 100% density in these samples

during sintering. The densities varied between 96.5% and 98.5% ofthe theoretical density of �-alumina (3.986 g/cm3). Several factsabout the ultraclean processing of the samples can be listed aspossible reasons for relatively low final densities:

(i) The powders were ultrapure and no sintering aids wereadded. Thus, liquid-phase sintering to enhance densification wasnot used in this study.

(ii) To avoid the further contamination of the powders, noorganic lubricants were used to facilitate particle packing. Most ofthe commercially available organic surfactants contain siliconimpurities. Only pure liquid paraffin was used to lubricate the diewalls during uniaxial pressing.

(iii) The pressure of the cold isostatic pressing was not highenough to pack the “dry” compacts to higher green densities beforesintering.

Although it is very well known that porosity can affect themicrostructural development dramatically, the main results re-ported here are a comparison between the microstructures of thesamples with various calcium doping levels for a specific sinteringtemperature and time. Other parameters in all of the experimentssuch as the amount of porosity for all samples after sintering wereapproximately the same. Therefore, the presence of �2% porositydid not receive a special emphasis while discussing the results.However, the absolute grain growth rates may have been reducedby the presence of this much porosity.

(3) Microstructural EvolutionThe average grain sizes of the samples are given in Table II. For

samples that showed AGG, besides the overall average, the

Table I. ICP-OES Results of theSamples before Sintering

Doping level (ppm) Ca (�g/g) Si (�g/g)†

Undoped 0 �510 7.1 �520 10.8 �550 23 �5

100 60 �5200 133 �5500 344 �5

1000 650 �5†Detection limit of the instrument.

624 Journal of the American Ceramic Society—Altay and Gulgun Vol. 86, No. 4

average size of the small grains and the average size of the large(abnormally grown, elongated) grains are also reported.

Segregation behavior of calcium in polycrystalline �-alumina atleast in the dilute regime can be modeled by a simple adsorptionmodel like most of the surface active impurities with very low bulksolubility.21 For the calculations, it was assumed that the calciumcoverage increases with increasing total atomic concentration ofcalcium in the alumina (Xt) and there was no second-phaseprecipitate formation. Thus, most of the calcium could be adsorbedat the grain boundaries without saturation and it was possible toobtain hypothetical �Ca values that would correspond to multiplelayer coverage at the boundaries.

Gulgun et al.21 used these approximations only for the calcu-lations of �Y at very low yttrium doping levels. However, thisrelation was also used for high calcium doping levels in this study.The main reason for this extrapolation is that the exact amounts ofcalcium at the grain boundaries could not be experimentallydetermined. However, there is the possibility of precipitation of asecond phase or formation of amorphous triple point pocketphases. Moreover, the bulk solubility (which was assumed to bezero in this consideration) may be changing with temperaturebetween 1400° and 1600°C. These will be discussed further in thefollowing section. Thus, initially it was assumed that all calciumcould be accommodated at the grain boundaries in a multilayergrain boundary film without reaching the saturation level.

(A) Microstructural Evolution of the Samples Sintered at1500°C: In undoped and low-calcium-doped samples, grainswere equiaxed with 1.7 and 2.5 �m average grain sizes forsintering times 1 and 12 h, respectively. The average grain sizes ofundoped samples were �10%–20% larger than the average grainsizes of low-calcium-doped samples (Figs. 1 and 2).

Fig. 1. SEM micrograph of undoped �-Al2O3 sintered at 1500°C for 1 hshowing equiaxed grains.

Fig. 2. SEM micrograph of 10.8-ppm-Ca-doped �-Al2O3 sintered at1500°C for 1 h showing equiaxed grains.

Tab

leII

.A

vera

geG

rain

Size

s†of

the

Sam

ples

Ca

(ppm

)

1400

°C/1

h15

00°C

/1h

1500

°C/1

2h

1600

°C/1

h

Ave

Gof

smal

lgr

ains

Ave

GA

veG

ofla

rge

grai

nsA

veG

ofsm

all

grai

nsA

veG

Ave

Gof

larg

egr

ains

Ave

Gof

smal

lgr

ains

Ave

GA

veG

ofla

rge

grai

nsA

veG

ofsm

all

grai

nsA

veG

Ave

Gof

larg

egr

ains

0–

0.90

�0.

02–

–2.

14�

0.07

––

2.76

�0.

1–

–2.

90�

0.07

–7.

1–

0.93

�0.

02–

–1.

74�

0.05

––

2.52

�0.

08–

–2.

95�

0.07

–10

.8–

0.89

�0.

02–

–1.

73�

0.05

––

2.46

�0.

07–

–2.

89�

0.07

–23

–0.

85�

0.01

––

1.74

�0.

06–

–2.

53�

0.09

––

2.60

�0.

06–

60–

0.73

�0.

01–

–1.

65�

0.05

––

2.53

�0.

09–

–2.

70�

0.06

–13

3–

0.71

�0.

01–

–1.

68�

0.02

–1.

53�

0.09

3.03

�0.

116.

82�

0.38

1.56

�0.

093.

04�

0.09

8.42

�0.

4634

4–

0.81

�0.

01–

1.26

�0.

052.

74�

0.09

7.61

�0.

321.

58�

0.08

3.27

�0.

147.

89�

0.43

2.03

�0.

114.

79�

0.22

9.84

�0.

7465

0–

0,87

�0.

01–

1.26

�0.

092.

36�

0.05

6.91

�0.

341.

54�

0.05

3.02

�0.

117.

31�

0.58

1.65

�0.

093.

63�

0.17

6.43

�0.

82†T

hegr

ain

size

sar

egi

ven

inm

icro

met

ers.

April 2003 Microstructural Evolution of Calcium-Doped �-Alumina 625

It can be seen in Table II that by increasing sintering time grainsbecame coarser as expected. However, when one tried to relate thegrain sizes and morphologies to bulk calcium concentrations, noclear trends could be seen as a function of bulk calcium concen-trations for different sintering conditions. For example, 133-ppm-calcium-doped samples had small grains when sintered at 1500°Cfor 1 h (Fig. 3). However, another sample with same amount ofcalcium had large elongated grains when sintered at 1500°C for12 h (Fig. 4). As described previously, the morphology and grainsizes were not controlled by the bulk calcium concentrations alone.

When the calcium concentration exceeded a certain limit, firstthe grain morphology started to change from equiaxial to elon-gated without any significant change in the average grain size.With a further increase in calcium concentration, AGG wasobserved with again elongated morphology (Figs. 4 and 5).However, grain size is not monotonically correlated with calciumconcentration. If the calcium concentration was increased evenfurther, the overall average grain size decreased again when thecalcium excess concentration at grain boundaries reached around�Ca � 20 Ca atoms/nm2 (650 ppm Ca) (Fig. 6 and Table II). Nosecond-phase precipitates or triple point pockets could be detectedwith SEM analysis. It was also noted that in abnormally grownsamples there were regions of small grains closed in by the largegrains. The average grain size in these small grain regions wassmaller than the average grain size in the samples that did not showAGG, i.e., low-calcium-doped samples (Fig. 5). Average grainsizes of large grains and small grains are given in Table II.

(B) Microstructural Evolution of the Samples Sintered at1600°C: In the samples that were sintered at 1600°C for 1 h,grains were equiaxed and the average grain size was around 2.9

�m for undoped and low-calcium-doped samples. A 10% decreasein the average grain size which was suggested to be due to solutedrag was first observed with 23-ppm-calcium-doped sample andthe average grain size dropped to 2.6 �m.

After a critical calcium concentration of �3 Ca atoms/nm2,grains again became elongated and abnormally grew up to anaverage grain size of 4.8 �m. With a further increase in thecalcium dopant concentration a similar decrease in the averagegrain size, which was observed with samples sintered at 1500°C,was observed when the calcium excess concentration at grainboundaries reached a value of �Ca � 33 Ca atoms/nm2.

Bimodal grain size distribution can be seen clearly in theabnormally grown samples. The average grain size of the smallgrains was �2.03 �m, while the average grain size of large grainswas around 9.84 �m for the samples doped with 344 ppm Ca(Table II). Again the average grain size in small grain regions inAGG samples was smaller than the average grain size in equiaxedlow-calcium-doped morphologies.

(C) Microstructural Evolution of the Samples Sintered at1400°C: Microstructural evolution was rather different for thesamples that were sintered at 1400°C for 1 h. All the samples hadalmost the same overall average grain sizes around 0.8 �m. Nosignificant drop in the average grain size was observed due tosolute drag at low calcium doping levels. Grains were mostlyequiaxed up to a certain calcium doping level. When the calciumconcentration at the grain boundaries reached a critical value ofabout 3–5 Ca atoms/nm2 grains started to become elongated (Fig.7). It was surprising to observe that although the calcium level atthe grain boundaries exceeded a rather high concentration of 11 Caatoms/nm2, only a few grains grew abnormally without anysignificant change on the average grain size (Fig. 7).

Fig. 3. SEM micrograph of 133-ppm-Ca-doped Al2O3 sintered at 1500°Cfor 1 h showing the onset of elongated grain morphology without affectingthe overall average grain size significantly.

Fig. 4. SEM micrograph of 133-ppm-Ca-doped Al2O3 sintered at 1500°Cfor 12 h showing abnormally grown grains.

Fig. 5. SEM micrograph of 344-ppm-Ca-doped �-Al2O3 sintered at1500°C for 12 h showing abnormally grown grains.

Fig. 6. SEM micrograph of 650-ppm-Ca-doped �-Al2O3 sintered at1500°C for 12 h with abnormally grown, elongated grains.

626 Journal of the American Ceramic Society—Altay and Gulgun Vol. 86, No. 4

IV. Discussion

To compare the samples sintered at different temperatures andpredict the critical grain boundary excess calcium concentration,�Ca values where the grain size and morphology changed, grainsize versus �Ca curves were plotted (Fig. 8). These curves togetherwith the micrographs were used to determine the critical calciumexcess concentrations for elongated grain morphology and anaccelerated grain growth rate to set in.

In Fig. 8, the dashed curve belongs to the set of samples sinteredat 1600°C for 1 h. Solid and dotted lines are for the samplessintered at 1500°C for 12 and 1 h, respectively. The solid curve atthe bottom of the figure, i.e., smallest grain sizes, representssamples sintered at 1400°C for 1 h. This order of the curves alsoconfirmed that as the sintering temperature and time increased,grains became coarser. The grain size played a key role todetermine the level of calcium segregation at the grain boundaries.The calculated calcium concentration at the grain boundariesincreased sharply as the grain size increased according to Eq. (1).

With the help of these curves and the corresponding micro-graphs, several facts can be discerned:

(1) Below certain calcium excess concentrations dependingon temperature, equiaxed, small grain morphology was obtained.

(2) Although the average grain size remained the same whenthe calcium excess at the grain boundaries was around �Ca � 3 Caatoms/nm2, the grain morphology started to change from equiaxedto elongated for all sintering temperatures (Fig. 3).

In all of these samples with elongated morphology sintered at1400°C for 1 h, at 1500°C for 1 and 12 h, and at 1600°C for 1 h,the grain sizes were rather different, 0.8, 1.6, 2.5, and 2.7 �m,respectively. Thus, the concentration of any other trace impurity(e.g., perhaps silicon) at grain boundaries would have been verydifferent for these four different grain sizes. Therefore, the onlycommon denominator for the elongated morphology in thesesamples is the 3–4 Ca atoms/nm2 �Ca. For example it is safe toassume that all samples would have had similar amounts of silicontrace impurity. Then the �Si for 1400°C samples will be 1/3 of the�Si in 1600°C samples where elongated morphology was ob-served. Thus, a critical �Si as a possible trigger for elongatedmorphology can be ruled out by this observation.

In samples that were sintered at 1500° and 1600°C when thecalcium at the grain boundaries reached a critical concentrationbetween �Ca � 4.5–8 Ca atoms/nm2, AGG occurred with elon-gated morphology (Figs. 4 and 5). Assuming that calcium cationswould substitute for the aluminum cations, this critical concentra-tion would correspond approximately to 0.5 monolayer of calciumcoverage at the grain boundaries. An additional assumption madehere is that all calcium atoms were confined to one plane in thegrain boundary.

The bottom curve in Fig. 8 showed that all samples sintered at1400°C for 1 h had almost same average grain sizes. Although the

average calcium excess at the grain boundaries exceeded �Ca � 11Ca atoms/nm2, only few grains grew abnormally without anysignificant change of the average grain size (Fig. 7). It appears thatfor the AGG to occur the condition �Ca � 4–8 Ca atoms/nm2 isnot enough at this temperature. This condition can be a necessarybut not a sufficient condition for AGG in these samples. Mostlikely, the synergy of an additional trace impurity whose excessbecomes critical as the grain size increases is necessary for theAGG to start. The obvious candidate for this trace impurity issilicon, which may have been there in the powders below 5 ppmconcentration. Furthermore, silicon may have contaminated sam-ples during sintering in the furnace with MoSi2 heating elementsdespite the fact that substantial precautions were taken against thiscontamination possibility. It should be mentioned here again thatall of the observed microstructures are from the center of samplesthat were 10–11 mm in diameter and 8 mm in height. In thesamples sintered at 1400°C, the existence of elongated but yetsmall grains with facetted grain boundaries without AGG showedthat occurrence of AGG in alumina cannot be directly correlatedwith the formation of facetted and straight grain boundaries asPark and Yoon claimed.6 Although facetted and straight bound-aries formed with a 3 Ca atoms/nm2 calcium excess concentrationat grain boundaries, almost all of these elongated grains remainedrather small (0.8 �m). Maclaren et al.32 and Gulgun et al.22

previously showed that AGG could happen with equiaxed mor-phology and curved boundaries in Y- and Si-codoped alumina.

The existence of small grains with an average grain size smallerthan the average grain size of the low-calcium-doped samples wasalso observed in abnormally grown samples sintered at 1500° and1600°C. The mechanisms that were responsible for the AGG ofthese samples are still elusive.

When the calcium doping level was increased to above �Ca �20 Ca atoms/nm2 or above �Ca � 30 Ca atoms/nm2 for samplessintered at 1500° and 1600°C, respectively, the average grain sizestarted to decrease again (Fig. 6). This drop in the average grainsize after a certain calcium excess at the grain boundaries could beindicative of the formation of some second- phase precipitatesand/or some calcium-rich film and phases at the grain boundariesand triple point pockets. However, with SEM/EDS point analysisand EDS calcium mapping no second-phase precipitates orcalcium-rich phases at triple point pockets were observed. Toclarify the reason of this behavior, chemical analysis with highspatial resolution and sensitivity needs to be performed.

Given the fact that the system is at the Al2O3-rich end of thebinary phase diagram, the expected composition of the precipitatesor phases was CaO�6Al2O3 (CA6). The studies that were done byKaplan3 and Brydson et al.24 on other alumina systems alsosuggested indirectly the formation of CA6-like phases although itwas never observed as a second-phase precipitate. Brydson et al.claimed that in their �-alumina/calcium silicate system there wasan amorphous grain boundary film with a nominal composition ofCA6 in the samples sintered at 1400°C. However, the phasediagram indicates that the melting point of CA6 is 1850°C. Thus,it may seem unlikely to have an amorphous CA6 grain boundaryfilm in calcium-doped samples sintered below 1850°C. However,Brydson et al. started with anorthite glass (CaO�Al2O3�2SiO2)alumina mixture. An anorthite, CA6, and alumina phase mixturehas a liquidus temperature at 1380°C.24 In the system that wasstudied here, if there was any CA6 precipitate or phase, it wasexpected to be crystalline. The highly Ca only doped samples with�Ca reaching 11 Ca atoms/nm2 are the best candidates to verify thestructure of calcium-rich grain boundary films.

All of the calcium excess values given here were the calculateddata as described before. However, to interpret the results moreeffectively the actual calcium excess values at the grain boundariesshould be determined. According to the calculations, calciumcoverage would have increased up to �Ca � 33 Ca atoms/nm2 atthe point where the average grain size reached a maximum on thedashed curve (i.e., 1600°C) in Fig. 8. Most likely, this value washigher than the saturation level of alumina grain boundaries.However, limited SEM analysis done on the samples was not ableto clarify this important issue. The question is whether there were

Fig. 7. SEM micrograph of 344-ppm-Ca-doped �-Al2O3 sintered at1400°C for 1 h showing elongated grain morphology. Few abnormallygrown grains did not affect the average grain size significantly.

April 2003 Microstructural Evolution of Calcium-Doped �-Alumina 627

a submonolayer segregant and second-phase precipitates/pocketsor a multilayer grain boundary film. According to Brydson et al.24

calcium segregation of �ca � 6.1 Ca atoms/nm2 spread over six toseven cation planes. Kaplan et al.3 observed that �ca � 2.5 Caatoms/nm2 distributed over 4 � 1 cation planes. In light of thesearguments in the literature, a �Ca � 33 Ca atoms/nm2 shouldspread over 30–40 cation planes which shall be easily detectablewith analytical TEM.

When the grain size versus �Ca plot (Fig. 8) was examined moreclosely, it was recognized that there were some interesting variationsbetween the behavior of 1500° and 1600°C sintered samples althoughtheir curves look similar. One of the clearest differences was themaximum calcium excess values at the point where the average grainsize reached its maximum. For 1500°C sintered samples this pointcorresponded to calcium excess of �Ca � 18–22 Ca atoms/nm2;however, this value increased up to 33 calcium atoms/nm2 for thesamples sintered at 1600°C. Furthermore, as discussed before forthese temperatures, there was also a difference at the calcium excesslevel where a 10% decrease in grain size was observed in the averagegrain size. The 10% decrease in grain size came at a higher �ca valueat 1600°C. Based on these two facts it appears plausible that thesolubility of calcium in the bulk �-alumina may be different at 1400°,1500°, and 1600°C. However, the solubility experiments for calciumin bulk alumina are best performed with a single-crystal sapphire.Such experiments will establish the solubility of calcium in�-alumina at different temperatures. The other possibility fordifferent �Ca values at the maximum of the grain size versus�Ca curve is that at these temperatures the equilibrium thicknessof the calcium-rich grain boundary film—if any— could bedifferent. Transmission electron microscopy (TEM) investiga-tions of these samples will clarify these issues along with thepossibility of second-phase precipitation.

IV. Conclusions

Microstructural development of a material is one of the keyconcerns in developing the desired properties. To control the grainstructure in a material, it is crucial to control the effects of impurities.This study investigated the effects of calcium impurities on themicrostructure of alumina that is believed to cause AGG.

For all sintering conditions, the grains were small and equiaxedfor low calcium concentrations. Only when the calcium excess atthe grain boundaries exceeded the critical concentration of �Ca �

3–4 Ca atoms/nm2 did the grains became elongated. Calcium ionsat the grain boundaries at these levels, i.e., between 3 Caatoms/nm2 or higher, cause solute drag on the grain boundarymotion. The elongated morphology is believed to be due topreferential segregation of calcium to basal planes (0001) inalumina grain boundaries.

AGG with elongated morphology was observed in thesamples that were sintered at 1500° and 1600°C above a criticalcalcium coverage of �Ca � 4.5 Ca atoms/nm2. In these samplesthe calcium concentrations exceeding a value of �Ca � 20 Caatoms/nm2 and �Ca � 30 Ca atoms/nm2 for the sinteringtemperatures 1500° and 1600°C, respectively, caused an appre-ciable decrease in grain size. Precipitation of a second phase orformation of triple point pockets were suspected for thisbehavior. Limited SEM/EDS studies of the microstructureshowed no evidence of a second-phase precipitate or calcium-rich triple point pockets phases.

For the samples that were sintered at 1400°C, despite theexistence of elongated grains, there was no significant AGG at thecalcium excess concentration of �Ca � 11 Ca atoms/nm2. Onlyfew grains grew abnormally without affecting the overall averagegrain size. It is concluded that above a certain grain boundaryconcentration calcium is responsible for elongated grain morphol-ogy in �-alumina. However, for the occurrence of AGG, thepresence of calcium as the only impurity in alumina is not asufficient condition at this temperature. It appears that the influ-ence of an additional impurity that reaches a critical concentrationat grain boundaries for grain growth may be necessary. Thisimpurity is suspected to be silicon that could have been introducedin small amounts during sintering.

Acknowledgment

We would like thank to BRISA for letting us use their electron microscopefacilities and Mr. A. Alkan and Dr. R. M. Cannon for very insightful discussions.

References

1S. Baik and C. L. White, “Anisotropic Calcium Segregation to the Surface ofAl2O3,” J. Am. Ceram. Soc., 70 [9] 682–88 (1987).

2S. M. Mukhopadhyay, A. P. Jardine, J. M. Blakely, and S. Baik, “Segregation ofMagnesium and Calcium to the (101�0) Prismatic Surface of Magnesium-ImplantedSapphire,” J. Am. Ceram. Soc., 71 [5] 358–62 (1988).

Fig. 8. Grain size versus �Ca semilog plot sharing the changes in the overall average grain sizes as a function of calcium excess concentrations at the grainboundaries three different temperatures and two different sintering times at 1500°C.

628 Journal of the American Ceramic Society—Altay and Gulgun Vol. 86, No. 4

3W. D. Kaplan, H. Mullejans, and M. Ruhle, “Ca Segregation to Basal Surfaces in�-Alumina,” J. Am. Ceram. Soc., 78 [10] 2841–44 (1995).

4J. Bae and S. Baik, “Abnormal Grain Growth of Alumina,” J. Am. Ceram. Soc.,80 [5] 1149–56 (1997).

5S. I. Bae and S. Baik, “Determination of Critical Concentrations of Silica and/or Calcia forAbnormal Grain Growth in Alumina,” J. Am. Ceram. Soc., 76 [4] 1065–67 (1993).

6C. W. Park and D. Y. Yoon, “Effects of SiO2, CaO and MgO Additions on theGrain Growth of Alumina,” J. Am. Ceram. Soc., 83 [10] 2605–609 (2000).

7R. L. Coble, “The LucaloxTM process,” U.S. Pat. No. 3 026 210.8S. J. Bennison and M. P. Harmer, “A History of the Role of MgO in the Sintering

of �-Al2O3,” Ceram. Trans., 7, 13–49 (1990).9R. L. Coble, “Sintering of Crystalline Solids—II, Experimental Test of Diffusion

Models in Porous Compacts, ”J. Appl. Phys., 32 [5] 793–99 (1961).10A. H. Heuer, “The Role of Magnesia in the Sintering of Alumina,” J. Am. Ceram.

Soc., 62 [5–6] 317–18 (1979).11A. H. Heuer, “Reply,” J. Am. Ceram. Soc., 63 [3–4] 230–31 (1980).12W. C. Johnson and R. L. Coble, “A Test of the Second Phase and Impurity

Segregation Models for MgO Enhanced Densificaiton of Sintered Alumina,” J. Am.Ceram. Soc., 61 [3–4] 110 (1978).

13S. J. Bennison and M. P. Harmer, “Grain Growth Kinetics for Alumina in theAbsence of a Liquid Phase,” J. Am. Ceram. Soc., 68 [1] C-22–C-24 (1985).

14K. A. Berry and M. P. Harmer, “Effect of MgO Solute on MicrostructureDevelopment in Al2O3,” J. Am. Ceram. Soc., 69 [2] 143–49 (1986).

15C. A. Handwerker, J. M. Dynys, R. M. Cannon, and R. L. Coble, “DihedralAngles in MgO and Al2O3,” J. Am. Ceram. Soc., 73, 1371 (1990).

16K. L. Gavrilov, S. J. Bennison, K. R. Mikeska, J. M. Chabala, and R. Levi-Setti,“Silica and Magnesia Dopant Distributions in Alumina by High-Resolution ScanningSecondary Ion Mass Spectrometry,” J. Am. Ceram. Soc., 82 [4] 1001–1008 (1999).

17C. A. Handwerker, P. A. Morris, and R. L. Coble, “Effects of ChemicalInhomogeneities on Grain Growth and Microstructure in Al2O3,” J. Am. Ceram. Soc.,72 [1] 130–36 (1989).

18Sakaguchi, H. Haneda, V. Srikanth, and T. Ikegami, “An Effect of Second Phaseon Oxygen Grain Boundary Diffusion in MgO-Doped Polycrystalline Alumina,”Mater. Res. Bull., 31 [7] 837–43 (1996).

19C. A. Bateman, S. J. Bennison, and M. P. Harmer, “Mechanisms for the Role ofMagnesia in the Sintering of Alumina Containing Small Amounts of a Liquid Phase,”J. Am. Ceram. Soc., 72 [7] 1241–44 (1989).

20J. Zhao and M. P. Harmer, “Sintering of Ultra-High-Purity Alumina DopedSimultaneously with MgO and FeO,” J. Am. Ceram. Soc., 70 [12] 860–66 (1987).

21M. A. Gulgun, R. Voytovych, I. Maclaren, and M. Ruhle, “Cation Segregation inan Oxide Ceramic with Low Solubility: Yttrium Doped �-alumina,” Interface Sci.,40, 99–110 (2002).

22M. A. Gulgun, R. Voytovych, E. Bischoff, R. M. Cannon, and M. Ruhle,“Microstructural Development in �-Al2O3,” Ceram. Trans., 118, 115–25 (2000).

23S. C. Hansen and D. S. Phillips, “Grain Boundary Microstructure in a Liquid-Phase Sintered Alumina (�-Al2O3),” Philos. Mag. A, 47 [2] 209–34 (1983).

24R. Brydson, S. C. Chen, F. L. Riley, and S. J. Milne, “Microstructure andChemistry of Intergranular Glassy Films in Liquid-Phase-Sintered Alumina,” J. Am.Ceram. Soc., 81 [2] 369–79 (1998).

25C. M. Wang, G. S. Cargill III, H. M. Chan, and M. P. Harmer, “StructuralFeatures of Y-Saturated and Supersaturated Grain Boundaries in Alumina,” ActaMater., 48, 2579–91 (2000).

26J. Cho, C. M. Wang, H. M. Chan, J. M. Rickman, and M. P. Harmer, “Role ofSegregating Dopants on the Improved Creep Resistance of Aluminum Oxide,” ActaMater., 47 [15] 4197–207 (1999).

27J. Bruley, J. Cho, H. M. Chan, M. P. Harmer, and J. M. Rickman, “ScanningTransmission Electron Microscopy Analysis of Grain Boundaries in Creep-ResistantYttrium- and Lanthanum-Doped Alumina Microstructures,” J. Am. Ceram. Soc., 82[10] 2865–70 (1999).

28M. A. Gulgun, V. Putlayev, and M. Ruhle, “Effects of Yttrium Doping�-Alumina: I, Microstructure and Microchemistry,” J. Am. Ceram. Soc., 82 [7]1849–56 (1999).

29R. Voytovych, I. MacLaren, M. A. Gulgun, R. M. Cannon, and M. Ruhle, “TheEffect of Yttrium on Densification and Grain Growth in �-Alumina,” Acta Mater., 50[13] 3453–63 (2002).

30H. Yoshida, Y. Ikuhara, and T. Sakuma, “High Temperature Creep Resistance inRare-Earth Doped, Fine Grained Al2O3,” J.Mater. Res., 13, 2597 (1998).

31H. Yoshida, Y. Ikuhara, and T. Sakuma, “Transient Creep in Fine GrainedPolycrystalline Al2O3 with Lu3 Ion Segregation at the Grain Boundaries,” J. Mater.Res., 16, 716 (2001).

32I. MacLaren, R. Voytovych, M. A. Gulgun, N. Popescu-Pogrion, R. M. Cannon,C. Scheu, and M. Ruhle, “Abnormal Grain Growth in Y-Doped Alumina and theEffect of Si,” submitted to J. Am. Ceram. Soc. �

April 2003 Microstructural Evolution of Calcium-Doped �-Alumina 629