TiO2 powders were obtained by calcination of base hydrolysis products of titanium(IV) isopropoxide, Ti(OPr~)a at 400°C for 3 h. The hydrolysis was carded out ...
POWDER TECHNOLOGY ELSEVIER
PowderTechnology92 (1997) 233-239
Synthesis of high surface area titania powders via basic hydrolysis of titanium(IV) isopropoxide Kamal M.S. Khalil a, Mohamed I. Zaki b., a Chemistry Department, Faculty of Science, South Valley University, Sohag 82524, Egypt b Chemistry Department, Faculty of Science, Kuwait University, PO Box 5969, Safat 13060, Kuwait
Received 17 January 1997;revised 14 April 1997; accepted 15 April 1997
Abstract
TiO2 powders were obtained by calcination of base hydrolysis products of titanium(IV) isopropoxide, Ti(OPr~)a at 400°C for 3 h. The hydrolysis was carded out at room temperature in two different solvents, namely isopropanol and n-heptane, and in the presence of low and high ammonia contents. X-ray powder diffractometry showed the resulting TiO2 powders to consist of anatase crystallites, irrespective of the hydrolysis conditions applied. In contrast, N2 adsorption isotherms (determined at liquid nitrogen temperature) probed notably different surface textures for the test powders, depending on the amount of base added and the solvent used. These results were confirmed by transmission electron microscopy, and attributed to the solvent-permitted ammonia interactions with the titania precursors formed during the course of the alkoxide hydrolysis. Accordingly, preparative conditions can be resolved for producing titania powders of higher specific surface area (67-73 m2/g) than the commercial titanias described by manufacturers as being high surface area powders ( ~ 50 m2/g). Keywords: Titaniapowders; Hydrolysis;Calcination;Surfacearea
1. I n t r o d u c t i o n Titanium(IV) isopropoxide, Ti(OPta)4, is well known as a feasible parent material for the synthesis of pure titania [ 16], TiO2, or titania-containing powders [7-10], for catalytic and various other applications [ 1-5,7-10]. Different preparative routes have been adopted, however hydrolysis has been the most favoured [ 1-4,7-10]. Ti(OPr~)4 hydrolysis occurs readily in neutral media [ 1-4], but it may be modified by acidic [1,7] and basic [4,9] additives. Surface and bulk properties of titania thus produced are affected considerably by such modifications [ 11 ]. The hydrolysis of Ti(OPr')4 in neutral media is quite fast and results in colloidal particles [1-3]. The particle size decreases down to nano-dimensions with increasing value of the water/alkoxy molar ratio [2]. The results of a recent investigation of ours [ 1 ] showed the specific surface area of the titania powders produced to decrease as the water/alkoxy ratio increased and as the solvent type was switched from polar to non-polar. Moreover, acetic acid additives during the hydrolysis have been found to enhance aggregation of colloidal particles [ 1 ]. * Corresponding author.
0032-5910/97 / $17.00 © 1997ElsevierScience S.A. All rights reserved PHS0032-59 10(97) 0325 0-6
The present paper communicates bulk and surface characterization results for TiOe powders obtained by calcination of Ti ( OPr i) 4 hydroly sis products in the presence of different solvents and amounts of ammonia additives. The solvents tested were isopropanol (polar) and n-heptane (non-polar). Following drying, the hydrolysis products were examined using elemental analysis (carbon and nitrogen contents), thermogravimetry (TG), and infrared spectroscopy (IR). On the other hand, X-ray diffractometry (XRD), nitrogen adsorption measurements, and transmission electron microscopy (TEM) were applied to characterize the calcination products, i.e. the TiO2 powders. The principal goal was to test the above procedure as a synthetic route to high surface area titania powders. Currently, high area titania powders are produced and distributed commercially (such as, for instance, TiO2 P25 by Degussa, Germany). The nominal specific surface area as given by the manufacturers is < 5 0 m2/g [ 12].
2. Experimental 2.1. Preparation o f titania powders
Various titania (TiO2) powders were prepared by calcination of the dried hydrolysis products of Ti (OPP) 4 at 400°C
234
K.M.S. Khalil, M.I. Zaki / Powder Technology 92 (1997)233-239
Table 1 Titania precursor materials as a function of hydrolysis conditions, and corresponding carbon (C) and nitrogen (N) analysis results, calcination products, and surface texture characteristics of the calcination products Titania precursor
P1 P2 N1 N2
Mixing molar ratio (alkoxy: ammonia: water)
1.00:0.125:0.387 1.00:1.000:3.112 1.00:0.125:0.387 1.00:1.000:3.112
Solvent
isopropanol isopropanol n-heptane n-heptane
Analysis results (%) C
N
0.57 0.81 1.27 0.00
0.75 2.86 2.18 3.02
for 3 h. Hydrolysis of Ti(OPr')4, 99.9% pure (Aldrich, Milwaukee, WI, USA), was effected by aqueous ammonia in isopropanol (or n-heptane) solvent. First, a 200 ml portion of 0.4 M solution of Ti ( OPP ) 4 in i sopropanol ( or n-heptane) was prepared. Then, the desired amount of ammonia (as a 25% aqueous solution) was added, while the solution was being stirred at 400 rpm at room temperature. The stirring was maintained for one minute for preparations carried out in isopropanol, and for 60 minutes for preparations carried out in n-heptane [ 1 ]. The resulting suspensions were aged for three days and then filtered. The precipitates obtained - without washing - - were allowed to dry overnight at room temperature in the open air, followed by heating at 60°C for 24 h. Table 1 exhibits mixing molar ratios (alkoxy groups:ammonia:water) and the solvents, along with the designations given to the titania precursors (i.e. T i ( O P r i ) 4 hydrolysis products) and their calcination products ( i.e. TiO: powders).
2.2. Thermogravimetry (TG) TG of the titania precursor materials was carried out using a thermal analyser (model DT-40, Shimadzu, Japan). Small portions (8-10 mg) of test samples placed in a shallow Pt cell were analysed by heating them up to 500°C (10°C/min) in a40 ml/min flow of N2. The results (weight change versus temperature) were acquired, processed and presented by means of an online computer system.
2.3. Infrared spectroscopy (IR) IR transmission spectra were taken from test samples, pressed into KBr-supported discs, at 4000-400 cm- 1 with a resolution of 4 cm -~, using a spectrophotometer (model FTIR-8101 Shimadzu, Japan) equipped with a data acquisition and handling system.
2.4. X-ray powder diffractometry (XRD) XRD powder diffractograms were recorded by means of a diffractometer (model JSX-60 PA Jeol, Japan) equipped with Ni-filtered Co Ket radiation (h = 0.179 nm). For identification purposes, the diffraction patterns thus obtained were matched with ASTM standards [ 13].
Calcination product
P1 (400) P2(400) N1 (400) N2(400)
Surface texure
aBET(m2/g)
Porosity
23 67 73 30
non-porous mesoporous mesoporous non-porous
2.5. Nitrogen adsorption volumetry Nz adsorption-desorption isotherms were determined on thoroughly outgassed test materials at liquid nitrogen temperature, employing a conventional volumetric apparatus [14]. Specific surface areas (mZ/g) were derived by BET analysis [ 15] of the adsorption isotherms. Pore size distribution curves [16,17] were also calculated using data determined by the adsorption branch of the isotherm [ 18 ].
2.6. Transmission electron microscopy (TEM) TEM images were obtained, using a microscope (model JEM- 1010 Jeol, Japan). Test samples were prepared by ultrasonic dispersion of the solid in isopropanol. A drop of the resulting suspension was placed on a carbon-coated grid and allowed to dry at 60°C. A number of grids were prepared accordingly for each sample and examined by TEM at 100 kV. Electron diffraction (ED) patterns were taken whenever necessary.
3. Results
3.1. Carbon and nitrogen contents Table 1 reveals that hydrolysis of Ti(OPP)4 in n-heptane results in materials (N1 and N2) containing relatively larger proportions of carbon and nitrogen residues than those (P1 and P2) obtained in isopropanol. In a given solvent, however, it appears that the higher the ammonia content of the hydrolysis medium the larger the proportions of carbon and nitrogen, with the carbon content of N2 being the sole exception. This might imply that the presence of ammonia somehow modifies the course of the alkoxide hydrolysis. When heated in air at 400°C for 3 h, the hydrolysis products resulted in materials that were free of carbon and nitrogen residues. Thus, under the calcination conditions applied, the carbon- and nitrogen-containing species are burned off completely.
3.2. TG curves Fig. 1 displays TG curves obtained for the four hydrolysis products P1 (a), P2 (b), N1 (c) and N2 (d). It is obvious that the four materials exhibit a similar thermal behaviour,
K.M.S. Khalil, M.I. Zaki / Powder Technology 92 (1997) 233-239
235
100
80 10
\
60
20
\
~ 40
b
p-
30
20
A
¢o 0
i
i./
_
r-
"~
,
40
I
,
I
,
I
,
I
,
I
i
4000
3500
3000
2500
2000
l
l
J
I
1500
I
I
I
I
1000
,
,
i
500
0
Wavenumberlem-1 Fig. 2. FT-IR spectra of the Ti(OPr~)4 hydrolysis products indicated.
10
played for the P1 precursor at 3380, 3220, 1630 and 1400 cm-1, together with two weak absorptions at 2950-2850 c m - 1. Similarly, four main absorptions are also displayed for P2, though the former two peaks are shifted to 3375 and 3165 cm-1, respectively. The main absorption peaks are due to vO-H (H-bonded), vNH4 ÷, 60-H, and 6NH4 + vibrations [ 19-21 ], respectively. The weak peaks at 2950-2850 cm-1 are assignable to vC-H vibrations of organic moeities. Inspection of the spectra reveals relatively stronger absorptions due to NH4 ÷ vibrations than to O-H vibrations on going from P1 to P2. This accounts for increasing amounts of surface NH4 + species, at the expense of O-H groups, with increasing ammonia/alkoxy ratio. IR spectra obtained for N1 and N2 (not shown) also showed the amount of surface NH4 ÷ species to increase with the amount of ammonia in the hydrolysis medium. The calcination products of P1, P2, N1 and N2 at 400°C exhibited IR spectra displaying no absorptions due to C-H and NH4 + vibrations. Thus the IR spectra are in line with the elemental analysis results in indicating that under the calcination conditions applied C- and N-containing species are burned off completely.
20
30
40
, 0
I 100
,
I 200
,
I 300
,
A 400
I 500
800
Temperature/~C Fig. 1. TG curves of the Ti(OPri)4 hydrolysis products PI (a), P2 (b), N1 (c) and N2 (d) obtained by heating at 10°C/min in Nz at 40 ml/min.
including a gradual weight loss commencing near 500C and coming to completion near 500°C. The initial weight loss at < 200°C is shown to occur at higher rates than that taking place at higher temperatures. Thus, most of the weight loss completed near 5000C is shown to occur at < 200°C. Upon complete hydrolysis of Ti(OPri)4 the formation of TiO2. nH20 would be expected [ 1 ]. Accordingly, a total weight loss of about 31% would pertain to the decomposition of the hydrolysis products (TiO2.2HzO) into T i Q . Fig. 1 indicates that this magnitude of weight loss is satisfactorily met by the hydrolysis products (30.8% P1 and 27.8% N1) obtained at the low ammonia/alkoxy ratio (Table 1 ). At the higher ammonia/alkoxy ratio, however, the products give rise to significantly lower total weight losses, namely, 24% P2 and 26% N2. Table 1 also implies that about 2-3% of the observed weight losses for P2 and N2 are due to species containing other than water elements, namely, carbon- and nitrogen-containing species, provided that these species are removable in the non-oxidizing atmosphere (nitrogen) of the present measurements. Hence it is obvious that the presence of ammonia either suppresses the hydrolysis of Ti(OPr')4, thus leading to incomplete elimination of the alkoxy groups, or blocks binding sites for volatile components such as those of water and alcohol molecules. It is worth noting that the influence of ammonia is relatively more obvious in n-heptane than in isopropanol.
3.3. IR spectra Fig. 2 shows the FT-IR spectra for the precursor materials P1 and P2 as indicated. Four main absorption peaks are dis-
3.4. XRD patterns The hydrolysis products of Ti(OPr')4 (P1, P2, N1 and N2) were all non-crystalline to XRD. However, their calcination products at 400°C were crystalline (see typical XRD diffractograms given in Fig. 3). The diffraction patterns displayed were similar in characterizing anatase-structured TiOz. Observed d-spacing values are cited in Table 2, together with the standard data [ 13] for anatase TiOz.
3.5. Nitrogen adsorption data Nitrogen adsorption-desorption isotherms determined on the 400°C calcination products are shown in Fig. 4. The isotherms are similarly of type IV [ 17,18]. They are also similar in exhibiting hysteresis loops of type H2 [ 17]. It is worth noting, however, that the loops exhibited for P1 (400) and N2(400) are much smaller than those observed for the corresponding P2(400) and Nl(400) samples. BET surface
K.M.S. Khalil, M.I. Zaki / Powder Technology 92 (1997) 233-239
236
6O
E
m
5O
o
40 =E
=E
~__e
E" P2(400) J t,..
e)
0
~ 8o l
10
i
20
i
30
I
40
t
I
SO
I
I
60
J
'~ 80
70
O
2e/Oegrees
E
Fig. 3. XRD powderdiffractograms( Co Kc(radiation)of the testmaterials. d-values correspondingto diffractionpeaks are due to anatase TiOz (see Table 2).
O
>
4O
N1(400) 20
Table 2 XRD data for the 400°C calcination products (see Table 1 ) and the standard ASTM data for anatase TiO2 Observed
i
i
t
i
t
i
0.0 0.1 0,2 0.3 0.4 0,5 0,6 0.7 0.8 0.9 1.0
ASTM
p/po 20
d-value (nm)
d-value (nm)
hkl
29.5 44.5 56.4 63.7 65.0 74.6
0.352 0.236 0.189 0.170 0.166 0.148
0.352 0.238 0.189 0.170 0.167 0.148
101 004 200 105 211 204
areas calculated for P1 (400) and P2(400) were 23 and 67 m2/g, respectively, whereas the areas calculated for N 1 (400) and N2(400) were 73 and 30 m2/g, respectively. Fig. 5 shows pore size distribution curves for the calcination products, and Table 1 cites porosity characteristics revealed therein. Accordingly, N 1 (4130) and P2 (400) expose mesoporous surfaces, whereas N2(400) and P1 (400) expose largely non-porous surfaces. These results are consistent with the high specific areas determined for N1 (400) and P2(400) versus the low areas assumed by N2(400) and P1 (400).
3.6. TEM images The TEM image (Fig. 6) taken from Pl(400) reveals large ( > 100 nm) overlaying particles in a matrix of varying smaller ( ~ 100-10 nm) particles. The precolloidal nature (not gel) of the small particles can be realized by considering the clarity of the particle edges. Typically, a spotty electron diffraction pattern has been obtained for the large particles (Fig. 7 (a)), and a ring type pattern for the general matrix of
Fig. 4. N2 adsorption-desorption isotherms determined at liquid nitrogen temperature on the 400°C calcination products indicated.
4 --
~,,
• P1(400) r~ P2(400)
\
o N1(400)
2
o
,r~'-~-, 2
I 3
'-
I 4
'-
~ ' 5
r,/nm Fig. 5. Pore size distribution curves as derived from the nitrogen adsorption data (Fig. 4) of the materials indicated.
the material (Fig. 7 ( b ) ) . Both of the ED patterns were indexed as being for anatase TiO2. For P2(400), Fig. 8 demonstrates mesoporous aggregates of fine ( < 20 nm) loose colloidal particles. A ring type ED pattern of anatase TiO:
K.M.S. Khalil, M.L Zaki / Powder Technology 92 (1997)233-239
Fig. 6. TEM imageof the calcinationproduct P1(400). (similar to that given in Fig. 7 ( b ) ) was exhibited by P2(400). TEM images obtained for N l ( 4 0 0 ) and N2(400) (not shown) were rather similar to that obtained for P2(400) (Fig. 8) in showing mesoporous coalesced aggregates formed seemingly via coalescence of small or elementary particles. It is worth reporting, however, that the elementary particles of N I ( 4 0 0 ) were relatively larger than those of N2 (400), and the aggregates of the latter were more compact and of much less obvious mesoporous nature. For both of the materials, however, ring type ED patterns assignable to anatase TiO2 were obtained.
237
Fig. 8. TEM imageof the calcinationproduct P2(400). 4. Discussion
The above results reveal that TiO2 powders of considerably different surface textural characteristics can be produced, depending on the aqueous ammonia/alkoxy ratio and the solvent in the hydrolysis medium of Ti(OPP)4. Despite the progress made toward a better understanding of the reaction mechanism of the alkoxide hydrolysis, a direct relation of the reaction variables to material properties is still far from being established [ 22]. Therefore, the following discussion is an attempt toward an explanation of the role of the present two
Fig. 7. Electron diffractionpatternsobtainedfor (a) large particlesof PI (400) and (b) the general matrixof the material.
238
K.M.S. Khalil, M.L Zaki / Powder Technology 92 (1997)233-239
variables (i.e. the amount of ammonia and the nature of the solvent) in the hydrolysis and condensation of Ti (OPr')4. Recalling the hydrolysis of tetraethyl orthosilicate (TEOS), the most investigated case in the sol-gel processing of metal alkoxides [23], aqueous ammonia additives have been found by Scherer and Luong [ 24 ] to improve the gelling process. This is believed [24] to result from charge created by deprotonation of silanol groups (Si-OH ~ S i - O + H ÷ ), with the positive ammonium ions being responsible for drawing the resulting negatively charged particles together, thus overcoming the generally weak repulsive bartier. According to calculations by Feat and Levine [25], this sort of ammonium bridging function is expected to operate unrestrictively in non-polar solvents. Feat and Levine [25] indicated, moreover, that the role of ammonia does not terminate at the point of gelling improvement, but it extends to influence subsequent drying and calcination of hydrolysis products. The role of the solvent may be understood in terms of its polarity. Polar solvents tend to stabilize gel or colloidal particles more than non-polar solvents. This has been attributed to hydrogen bonding and steric effects [ 26 ]. However, since Ti(IV) is a transition metal ion with available d-orbitals, some differences from the case of silicon can be perceived. Accordingly, the role of different modifiers has been explained in terms of the chemical reactivity of the new ligands towards hydrolysis of Ti(OPP)4 [ 11 ]. According to Ragai [ 27 ], aging of titania hydrogels in aqueous ammonia leads to a significant loss of pore volume and a consequent drop of surface area (from 240 to 43 m2/g). The significant loss of pore volume has been considered [ 27 ] to suggest that the rigidity of the solid network is reduced by aging, possibly as a result of replacement of hydroxyls by ammonia ligands in the inner coordination sphere of Ti(IV) ions. Elemental analysis ( carbon and nitrogen contents) and FFIR results show the hydrolysis products of Ti(OPP)4 to retain appreciable amounts of NH4 + and isopropoxide related species. NH 4 ÷ species may be held by counter-charges generated following a surface deprotonation of Ti-OH groups analogous to the model of Scherer and Luong [24] for TEOS. Deprotonation of surface hydroxyl groups should decrease the amount of hydrogen-bonded water or alcohol molecules. Thus the observed decrease in the total weight loss of Ti(OPr')4 hydrolysis products (Fig. 1) with increase of the ammonia/alkoxy ratio may be understood in terms of an enhanced surface deprotonation, and a consequent decrease of the amount of H-bonded water and alcohol molecules. Another consequence of the increase of surface deprotonation (i.e. decrease of surface hydroxylation) would be a decrease in the structure cross-linking following calcination at high temperatures. TEM results show that the materials formed at the low aqueous ammonia/alkoxy ratio, i.e. P1 and N1 (Fig. 6), irrespective of the solvent, are more condensed than those produced at the high ammonia/alkoxy ratio, i.e. P2 and N2 (Fig. 8).
•~ - 1 1
+
= -~-~'E: NH 3
:NH 3
Scheme 1. H
H
T~---O----'fi
~
Ti
+
"1i
Scheme 2.
NH2
1-12N '13
•
T'r----N--'n
+ NH 3
~. Tr~N--II
+ NH 3
Scheme 3.
NH
H2N
1i
Scheme 4.
The low porosity of TiO2 powders prepared at high aqueous ammonia/alkoxy ratio may be explained at a molecular level in terms of the surface nitridation scheme proposed by Brinker and Haaland [ 28 ]. Accordingly, electrophilic metals capable of increasing their coordination number may interact with ammonia via the Lewis acid adsorption mechanism (Scheme 1). This may be followed by dissociative chemisorption (Scheme 2). Pore collapse is likely to occur following loss of NH3 at high temperatures (Schemes 3 and 4). It has been reported [4] that titanium nitrides are unstable to heating to high temperatures in air. Thus, following the present calcination procedure, the above schemes are succeeded by oxidation and desorption of nitrogen-containing species, eventually giving nitrogen-free calcination products.
5. Conclusions
Whether in a polar (isopropanol) or a non-polar (n-heptane ) solvent, the hydrolysis product of titanium (IV) isopropoxide on calcination at 400°C for 3 h produces anatase TiO2 powders. A controlled addition of NH4OH solution seems to catalyse the hydrolysis, leading eventually to a marked development in the specific surface area (up to 73 m2/g) of the TiO2. It is apparent that the higher the polarity of the hydrolysis medium, the higher the amount of hydroxide required.
References [ 1] K.M.S. Khalil, T. Baird, M.I. Zaki, A.A. E1-Samahy and A.M. Awad, Colloids Surf, (1997) in print. [2] K. Nishiwaki, N. Kakuta, A. Ueno and H. Nakabayashi, J. Catal., 118 (1989) 498. [3l S. Nishimoto, B. Ohtani, H. Kajiwara and T. Kagiya, J. Chem. Soc. Faraday Trans. 1, 81 (1985) 61. [4] K. Kohno, Z Mater. Sci., 27 (1992) 658. [5] K. Morishige, F. Kanno, S. Ogawara and S. Sasaki, J. Phys. Chem., 89 (1985) 4404.
K.M.S. Khalil, M.L Zala"/ Powder Technology 92 (1997) 233-239 [6] H.Y. Ha, S.W. Nam, T.H. Lim, I.H. Oh and S.A. Hong, Membrane Sci., 111 (1996) 81. [7] D.C.M. Dutoit, M. Schneider and A. Balker, J. Catal., 153 (1995) 165. [8] R. Hutter, T. Mallat and A. Balker, J. Catal., 157 (1995) 665. [9] J.H. Kwak, S.J. Cho and R. Ryoo, Catal. Lett., 37 (1996) 217. [ 10 ] B. Lantelme, M. Dumon, C. Mai and J.P. Pascault, J. Non-Cryst. Solids, 194 (1996) 63. [ 11 ] C. Sanchez, J. Livage, M. Henry and F. Babonneau, J. Non-Cryst. Solids, 100 (1988) 65. [ 12] M.1. Zaki, B. Vielhaber and H. KnOzinger, J. Phys. Chem., 90 (1986) 3176. [13] J,V. Smith (ed.), X-ray Powder Data File, American Society for Testing and Materials (ASTM), Philadelphia, PA, 1960. [ 14] International Union of Pure and Applied Chemistry, IUPAC, Pure Appl. Chem., 57 (1985) 603. [ 15 ] B. Brunauer, P.H. Emmett and E. Teller, J. Am. Chem. Soc., 60 (1938) 309. [ 16] C. Orr and J.M. Dalla Valle, Fine Particles Measurement, Macmillan, New York, 1959, p. 271.
239
[ 17] S.J. Gregg and K~S.W. Sing, Adsorption, Surface Area and Porosity, Academic Press, London, 1982. [18]K,S.W. Sing, in J.M. Thomas and R.M. Lambert (eds.), Characterization of Catalysts, Wiley, New York, 1980, p.12. [ 19] S.A.A. Mansour, G.A.M. Hussein and M.I. Zaki, Thermochim. Acta, 150 (1989) 153. [20] K. Nakamoto, Infrared and Raman Spectra of Inorganic and Coordination Compounds, Wiley, New York, 3rd edn., 1978, pp. 132135. [21] J.A. Gadsden, Infrared Spectra of Minerals and Related Inorganic Compounds, Butterworths, London, 1975, p. 15. [22] H. Schmidt, J. Non-Cryst. Solids, 100 (1988) 51. [23] C.J. Brinker and G.W. Scherer, Sol-Gel Science, The Physics and Chemistry of Sol-Gel Processing, Academic Press, London, 1990, p. 20. [24] G.W. Scherer and J.C. Luong, J. Non-Cryst. Solids, 63 (1984) 163. [25] G. Feat and N. Levine, J. CoUoidlnterface Sci., 54 (1976) 34. [26] M.T. Harris, R.R. Brunson and C.H. Byers, J. Non-Cryst. Solids, 121 (1990) 397. [27] J. Ragai, Chem. Tech. Biotechnol. A, 35 (1985) 263. [28] C.J. Brinker and D.M. Haaland, J. Am. Ceram. Soc., 66 (1983) 758.
