Sol-gel-derived binary silica glasses with high refractive index

J O U R N A L OF

Journal of Non-Crystalline Solids 146 (1992) 121-128 North-Holland

N-CltLLESOIDS

Sol-gel-derived binary silica glasses with high refractive index S. Satoh, K. Susa 1 and I. Matsuyama Central Research Laboratory, Hitachi Ltd, Higashikoigakubo 1-280, Kokubunji, Tokyo 185, Japan Received 2 October 1990 Revised manuscript received 6 April 1992

Doped silica glasses, Si02-MOn/m(M = Ga, Gd, Nb, Sb, Sn, Ta, Ti and Zr), were prepared by the sol-gel method and evaluated. Refractive index and X-ray diffraction measurements showed that SiO2-Sb203, SiO2-Ta2Os, SiO2-TiO 2 and SiO2-ZrO 2 glasses are potentially suitable for optical components. Experimentally observed refractive indices are in good agreement with the calculated values using Appen's empirical law.

1. Introduction

High refractive index glasses are suitable for optical components such as micro-lenses and optical fibers [1]. Optical fibers typically have a double-layer structure which consists of core and cladding materials. High refractive index core materials are suitable for high numerical apertures (NA), which form effective light couplings with conventional light emitting diode (LED) light sources and offer several important potential advantages over conventional coaxial links for short-haul data-bussing applications. Silica glass has high thermal stability, high chemical durability and good optical transparency over a wide range of wavelengths [2]. We suggest that these advantages indicate that high refractive index doped silica glasses should have more potential than multi-component glasses for enhancing the characteristics of optical components. Since doped silica glasses have much higher melting temperatures, it is rather difficult to produce pure materials using conventional melting techi Present address: Hitachi Chemical Co. Ltd., Tsukuba Research Lab., 48 Wadai Tsukuba, Ibaraki 300-42, Japan. Correspondence to: Dr S. Satoh, Central Research Laboratory, Hitachi Ltd, Higashikoigakubo 1-280, Kokubunji, Tokyo 185, Japan. Tel: + 81-423 23 111. Telefax: + 81-423 27 7670.

niques. The sol-gel process is one of the most promising techniques for obtaining such pure materials [3]. This process, which permits production of a glass body at low temperatures, consists of three steps: (1) hydrolysis of metal alkoxides to obtain a 'wet gel'; (2) drying the wet gel to obtain a porous 'dry gel' (a low-density glass); (3) sintering the dry gel to produce a glass body. An important feature of this process is that a variety of dopant materials are readily available in the form of alkoxides or soluble salts. Therefore, the sol-gel process makes it easy to produce homogeneous doped silica glasses with higher refractive index. Titania-silica glasses have been investigated [4-7] because of the technological significance of the very low thermal expansion (in some case negative) of refractory glasses. Zirconia-silica glasses, known for their high resistance to attack by alkaline solution, have been also investigated [7,8]. This paper proposes some doped silica glasses with high refractive index suitable for optical component applications. First, the dopant metals from the periodic table were screened to obtain high refractive indices on the basis of Appen's empirical law [9]. Then, glasses of the selected system were experimentally synthesized by the sol-gel process to determine if their refractive

0022-3093/92/$05.00 © 1992 - Elsevier Science Publishers B.V. All rights reserved

122

S. S a t o h et al. / S o l - g e l - d e r i v e d b i n a r y silica glasses

indices and other properties would have practical applications.

Transition metals cause absorption losses due to the d - d electronic transition [11], except Ti, Y, Zr, Nb, La, Ta and Gd which have a closed shell electronic structure. Y [12], Nb [13], La [14] and Gd [15] form two liquid phases with silica, which could lead to crystallization during the processing and result in optical scattering loss. However, by using the sol-gel method, it is expected that homogeneous glasses can be prepared even for these phase-separating systems. The dopant metals were finally chosen using the third criterion, thereby achieving high refractive index and excellent characteristics for the silica glass. Those selected were: Bi, Ga, Gd, In, La, Nb, Sb, Sn, Ta Ti, Y, Zr. The refractive index of doped silica glasses can be estimated by several methods including the Lorentz-Lorenz formula [16,17]. However, these methods are not suitable for the glasses doped with the above-mentioned metal oxides. An empirical method proposed by Appen [9] is generally applied to systems which consist of many types of metal oxides and therefore is suitable for the present study.

2. Preliminary screening of dopant elements Dopant elements for high refractive index silica glass (the SiO2-MOn/m binary system) were chosen from the periodic table using the following criteria. (1) Element should be stable in SiO 2 network and the doped glasses should be durable in the ambient atmosphere. (2) Low optical losses are expected: (a) low absorption loss; (b) low scattering loss. (3) Doped glasses should be non-toxic. Volatile elements such as N, S, halogens, alkaline metals, and alkaline earth metals were eliminated by the first criterion. Both alkaline and alkaline earth metals are well known constituents of multi-component glasses which break the Si-O bond in the silica glass network, and reduce atmosperic durability [10].

Table 1 Solubility of doping materials in various solvents where C n (n = 1-4) represents alcohol, CnH2n+ 1OH Dopant raw

Aspect at

Solubility

material

room temperature

Si(OCH3) 4

Bi(OC4H9) 3 Ga(NO3) 3 . H z O Gd(NO3) 3-6H20 HBiO 3 In(OCH3) s La(OCH3) 3

pasty (black) powder (white) powder (white) solid (dark brown) powder (white) powder (white) m u c o u s (brown) liquid (light yellow)

Nb(OC4H9)

5

Sb(OC2H5) 3

CnH2 n + 10 H

H20

X

0 c4

X

•

0 Cl - C 4

0

•

0 Cl-C 4

0

• •

• c 1- c4 0 £1, C2

• X

•

• q-c

• o

O c 1- c 4

4 X

o c2

X

X Cl~ C3~ C4

Sb(OC4H9) 3 Sn(OCH3) 4 SnHPO 4 Ta(OC2Hs) 5 Ta(OC 4 H 9)5 Ti(OC4H9) 4 Y(OCH3) 3 Zr(OCaH9) 5

liquid (light yellow) powder (white) powder (white) liquid (light yellow) liquid (light yellow) mucous (brown) powder (white) mucous (brown)

o

X £1-c4

X

•

• cl-c 4

•

• o 0 0

• x 0 0

• x X x

Cl- 6.4 c1 6'1-- ¢4 6.I - 6.4

•

• 6.1-c 4

•

0

0 ¢1-6.4

X

O, soluble; o , very slightly soluble; o, insoluble; and, x , precipitated.

S. Satoh et al. / Sol-gel-derived binary silica glasses

3. Experimental 3.1. Synthesis of doped silica gels The samples were prepared by the sol-gel method which has been described previously [10] for the synthesis of pure silica glass. The outline of the process is as follows. A mixture of tetramethoxysilane, Si(OCH3) 4 and dopant metal alkoxide (S), water (H) and alcohol (C) of a given molar ratio S / H / C was cast in glass containers (8 mm diameter × 200 mm long). When kept at 70°C, the solution became viscous and gelled. The gels were then dried for 1 week at 70°C. Finally, porous gel bodies of 4-6 mm in diameter and 100 mm in length were obtained, with apparent densities between 0.6 and 1.8 × 103 k g / m 3. The molar ratio ( S / H / C ) was about 1/2/4.5 and the pH of the solution was usually 7 in this experiment. The rate of the hydrolysis reaction of the dopant alkoxide was generally higher than that of the silicon alkoxide, and this often led to precipitation of dopant metal oxide. These difficulties were usually overcome by a slow hydrolysis procedure in which water diluted with alcohol was slowly added to a mixed solution of silicon alkoxide and dopant alkoxide with alcohol. Raw materials used for dopants are shown in table 1. However, the preparation of doped silica gels was often quite different from the case of pure silica gel. Therefore, preparation processes had to be modified in accordance with dopant alkoxides. The main modifications were as follows.

123

with silicon tetramethoxide under ambient conditions. However, it turned dark but still transparent after adding more methanol. Thus, niobiumdoped gels were successfully formed with an S / H / C ratio of 1 / 2 / 9 .

3.1.3. Sb Antimony triethyloxide also exhibited a fast reaction with water. Therefore, water diluted with ethanol was used for the hydrolysis. After a longterm mixing, precipitate which appeared at the beginning of the hydrolysis reaction was dissolved. Thus, antimony-doped silica gels were successfully prepared with an S / H / C ratio of 1/2/9. 3.1.4. Sn Tin tetramethoxide and tin hydrophosphate were insoluble in solvent or water. Therefore, they were dissolved in sodium hydroxide solution. Silica gels with limited Sn content were then formed using 0.1 N NaOH solvent, having an S / H / C ratio of 1/30/4. 3.1.5. Ta Tantalum pentamethoxide was very sensitive to hydrolysis. It was important to keep the rate of the hydrolysis reaction low using methanol-diluted water for several hours in order to avoid precipitation. 3.1.6. Ti Using titanium tetrabutoxide, titanium-doped silica gels were formed without any difficulty as in the case of pure silica gels.

3.1.1. Gd, Ga Nitrate of gadolinium (Gd(NO3) 3 • 6H20) , and nitrate of gallium (Ga(NO3) 3) were used as dopant materials. They were soluble in both water and alcohol. Doped gels with desired compositions were successfully formed without any difficulty with S / H / C ratio of 1/4/4.5 for gadolinium-doped gel and 1/2/4.5 for gallium-doped gel.

3.1.7. Zr Zirconium tetrabutoxide showed a slight tendency to precipitate even when methanol-diluted water was used. It was important to mix Zr(OC4Hg) 4 with Si(OCH3) 4 and n-C3H7OH and then to hydrolyze with n-propanol-diluted water under controlled hydrolysis conditions.

3.1.2. Nb The dopant for niobium metal was niobium pentabutoxide, which was so reactive that the liquid surface turned white when it was mixed

3.1.8. Bi, In, La, Y Silica gels doped with these metal oxides were not prepared successfully because of poor solubility of the metal alkoxides in alcohol or water.

S. Satoh et al. / Sol-gel-derived binary silica glasses

124

,° m -'_

ing are shown in fig. 1. Doped silica glass samples about 4 mm in diameter were obtained.

/ >/ -[ 1

0.8

•

0.6

.0"

3.3. Characterization of gels and sintered glasses

-

3.3.1. Apparent density measurements of gels 0.4

-

"" _=

0.2

0

Pure

•

10mo1% Ti Doped

SilicaGel

S i l i c a Gel

0 O

I 2OO

I 400

[ 600

I 800

Temperature

I IOOO 1200

(°C)

Fig. i. Shrinking behavior of titanium-oxide-doped silica gels during heating. Relative density of gels is shown as a function of temperature, obtained at a heating rate of 0.06°C/s up to 800°C under 02 atmosphere and 0.02°C/s in the range from 800 to ll00°C under He atmosphere.

3.2. Sintering of gels The gels were sintered by the process described in a previous paper [18]. The gels, about 5 mm in diameter and 100 mm in length, were sintered under constant heating rate of 0.020.06°C/s up to 700°C in O 2 or 0 2 + He atmosphere and then up to 1100°C in He or He + O 2 atmosphere to obtain glasses. Typical shrinking curves of titanium-doped silica gels during heat-

Apparent density of a gel is an important factor in the sintering process, it is known empirically that a gel of lower apparent density generally has larger pores, and is easily sintered into a clear glass [18]. Gels prepared in the present experiments assumed various apparent densities depending on the species of metal oxides and the amount of dopants. The apparent density was calculated from the gel weight and its dimension. The gels were not always regular in shape because of cracking or deformation occurring during the drying process. Therefore, the accuracy of the measurements was about + 5%.

3.3.2. Compositional analysis Dopant concentrations of both dry gels and sintered glasses were analyzed by the induction coupled plasma-activated spectroscopy (ICPS) method with an accuracy within __+5%. A powdered samples (gel or glass) was dissolved in concentrated hydrofluoric acid at a temperature of 80-90°C. The solution was heated with a small

Table 2 Appearance of doped silica gels and glasses Dopant

Ga Gd Nb Sb Sn Ta Ti Zr

Gel and glass

Concentration (as weighed, mol%)

gel glass gel glass gel glass gel glass gel glass gel glass gel glass gel glass

2

5

10

o o © (~

(J o

o © © •

o

O

o

•

•

•

•

•

o o

0 ol

o • o

•

•

©, transparent; (D, translucent; o, opaque.

• •

© o

o ~ o

20

• © o

•

15

o o o

125

S. Satoh et al. / Sol-gel-derived binary silica glasses

amount of H C 1 0 4 t o vaporize the SiF4, a reaction product, and to leave the metal salt behind. The residual metal salt was again dissolved in acid (HCI or HNO 3) and then subjected to ICPS.

2.0 o~

Gd 03/2,- " ~

'E

%

1.5

/ -''0/ ~

\

1.0

.0

Sa03/2J ~ 0 / Nb05/2 ...,...

~ f ~ . - - ~ - : . o..... ..

....

3.3.3. Refractive index measurement

.'~,,.,..~......................• .... Sb03/2

Refractive indices of the glasses were measured by Becke's method (immersion method). The series of index matching oils used in this experiment covered the range from 1.450 to 1.620 with intervals of 0.002 for light of 590 nm wavelength. This method achieves errors of less than + 1% in refractive index.

"o E

.5

0

(a)

0

2.0

I

I

I

,

i

,

/

3.3.4. X-ray diffraction analysis

1.5 L .ll--,. O

20

5 10 15 Bole percent of metal oxide

0

TaOS/Z@

Inhomogeneities such as crystalline phases in the glass matrix were examined by means of X-ray diffraction.

-

I,,",o,-2"*..-.....- ~ .... -:~,~.~_

.~.x

\

"O-.::'::'"":~ ...............i r O2

,~"~ 1'i t °

o / (b)

0

~

~o--.-----o-Ti02 Sn02

,

4. Results i

4.1. Gel properties

,

5 10 15 20 Bole percent of metal oxide

The appearance of synthesized gels is shown in table 2. All the doped gels except Sn-doped are

Fig. 2. Apparent density of the doped silica gels.

~ 20 t

' ' J//'l Ga03/

~ElN:>,15 p -~ ~¢10

0

=.¢0

"

0

:/J I

~20 ~15 u--El >" o.(o IO

t

a 5

~t5

- N b ~

I

¢

I

.~15oo SnO2 / El=

-~ ~ 5 0

0 I0 15 20 Mole percent

,:~......, .....

(as weighed)

,.o 20 [

I

[

I

O~

I

~20

-C~

'

Ti02

~

I

J

0

~//

~20

El E

¢z~lO

-6~5 0

(a)

10 15 20 Mole percent (as w e i g h e d )

0

5

i 10

I 15

Mole percent (as w e i g h e d )

0 L/ 20

(b)

i

I

i

=

(5"

5 10 15 20 Mole percent (as weighed }

-: .......

o.o=

5

0

_ Zr02 O /

'0

~El~5

i ~,

ggl0

/

"'1

,

2~

5 10 15 20 Mole percent ( as weighed )

5

i

C:N15

0

~.~10

~20 /

|

"°

~E

I I 10 15 20 Mole percent

,,¾

~20 ~O

i

IE

(as weighed )

~20

i

.......,...... "'"" I I I 5 10 15 20 Mole percent ( as weighed )

0 5 10 15 20 Mole percent . ( as weighed)

Fig. 3. Compositional change of dopant metals (a) Ga, Gd, Nb and Sb and (b) Sn, Ta, Ti and Zr, in the gels (©) and glasses (e).

S. Satoh et al. / Sol-gel-derived binary silica glasses

126

translucent similar to pure silica gels. Sn-doped gels were opaque because of the large pores in the gels. The apparent density of the doped silica gels is shown in fig. 2. A gel with low apparent density was composed of highly aggregated particles and large pores. Therefore, the gel is opaque due to strong light scattering. Chemical compositions of metal-doped gels were, within errors of measurement, urichanged during gelation as shown in figs. 3(a) and (b), where gels are indicated by open circles. It is apparent from this result that the dopants did not segregate and were uniformly taken into the gel network. However, chemical compositions of glasses changed, from the composition of the metal-doped silica gels, after vitrification in the cases of Gd, Sn and Zr as shown in figs. 3(a) and (b) by solid circles for glasses and open circles for gels. The decrease in dopant concentration after vitrification was particularly large for Gd. This phenomenon is due to the vaporization of the dopant during the sintering process. Tin partly formed under reducing atmosphere [19] in the imperfect combustion of hydrocarbon in the gel is a volatile material [20], which leads to reduced concentration during sintering. However, gadolin-

1.6,

7

,

,

f

1.4

O

~

.........

I

I

I

lO

15

>~

== 1,4

0

percent

I

I

I

5

I0

15

Mole

>~ 1.5

20

(a)

i

I

i

5

10

15

Mole p e r c e n t

1.4 20

0

Mole

10

1.6

,

~f =: 1.4

Mole

1,4~ 0

20

I

I

I

5

10

15

Mole

15

percent

20

0

,

{b)

................

"~'t 1 i 5

I IO

i 15

Mole percent

20

percent

......

I I i'"l 10

15

percent

......

5

.......

.........................

5

percent

........ 0

i

=f14f 1

"~. 1,5 ~ee/'r

1.5

20

]

Srl02

Gd03/2

x 1.6

1.4

The existence of inhomogeneities such as crystalline phases in the glass matrix causes light scattering. Therefore, glasses containing crystalline materials are not suitable for optical components. Of the systems studied, crystalline materials were found in the Sn, Ta and Nb systems. These were identified as oxides of dopant metals by powder X-ray diffraction. In the Nb system, an unknown phase was observed in addition to the cristobalite. From the width of the diffraction line, the crystal size of the hexagonal tantalium oxide in the case of Ta-doped silica glass was estimated to be 7 nm. These crystalline particles may have

|.6,

-~-

5 Mole

4.2. Glass properties

1.6

Ga 03/2

>~ 1.5 L

ium and zirconium oxides are stable at high temperatures up to around 1000°C. Gadolinium nitrate and zirconium tetrabutoxide usually change to metal oxides through oxidation reactions at high temperatures. One possibility to explain the decrease in dopant concentration is that a part of those dopant metals remained in the gels in the form of the original materials and the remnant dopants vaporized during the sintering process.

l 20

zo, 1.4 O

i 5 Mole

i IO

t 15

20

percent

Fig. 4. Refractive index of silica glasses doped with (a) Ga, Gd, Nb and Sb, and (b) Sn, Ta, Ti and Zr. Dotted lines show refractive index calculated by Appen's method; see eq. (1) and (2) in the text.

S. Satoh et al. / Sol-gel-derived binary silica glasses

been caused by segregated tantalum oxide formed in inhomogeneous gelation. If so, improved gelling techniques may make it possible to produce crystal-free glass in the tantalum system. On the other hand, Sn- and Nb-doped silica glasses both gave sharp and high X-ray diffraction peaks due to tin oxides or unknown crystalline phases and the cristobalite phase even when the dopant content was low, 2 tool%. Precipitation leads to opalization. The phase diagram [13] shows that in the N b 2 O s - S i O 2 system, Nb205 and tridymite phases are stable at temperatures up to 1450°C. Therefore, an unknown phase and cristobalite phase in the N b 2 O s - S i O 2 system must be metastable and assume very similar structures to the starting material (Nb2Os-doped silica gel). Gd-, Ti- and Zr-doped silica glasses which were transformed from gels of high apparent density (fig. 2) were opaque or translucent because of foaming produced during sintering. Therefore, for making transparent glasses, it is important to prepare gels with low apparent density. It is expected that some improvements in the preparation process might make it possible to produce gels of low apparent density and transparent glasses. Crystallization of the Z r O 2 - S i O 2 system has been reported by Kamiya et al. [8]. The crystallization temperature was decreased with increase in Z r O 2 content, and the gel containing 7 mol% ZrO 2 partially crystallized at 800°C, which disagrees with the results of our experiment. The reason for this disagreement is still unknown. One reason may be the differences in glass homogeneity when prepared under different conditions.

127

above. Therefore, non-linear increase in the refractive index may be caused by the crystallized particles of the metal oxides. Existence of the cristobalite phase of high refractive index may be responsible for the observed higher refractive index. However, the phenomenon in the Sb system is not understood. One possibility is inhomogeneity of doped silica glass due to heterogeneous hydrolysis, where doping metal alcohoxide hydrolizes much faster than silicon tetramethoxide.

5. Discussion

The dotted line in figs. 4(a) and (b) is the calculated refractive index from Appen's method. Appen's method is easily handled and widely applied to many elements. Supposing the SiO 2MOn/m system, refractive index, g, is represented by the equation (1)

g = g l a l + gza2,

where a I and a 2 are the molar fractions of silica and the doping metal oxide, respectively, and ga and g2 are the partial coefficients for the refractive index of these conponents (silica and doping metal oxide). Appen found gl statistically for SiO 2 to be represented by ga = 1.475 - 0.0005(a I - 0.67)

(for a t > 0.67),

(2) whereas g2 is constant and nearly equal to the refractive index of the doping metal oxide, as listed in table 3. Refractive index changes linearly with concentrations of metal oxides as shown in eq. (1). The

4.3. Refractive index of doped silica glasses The refractive index of doped silica glasses is shown in figs. 4(a) and (b). Each dopant concentration is shown as mol% of metal oxide in the Si02-MOn/m systems (M - Ga, Gd, Nb, Sb, Sn, Ta, Ti, Zr). Refractive indices of Ga-, Ta-, Zrand Ti-doped glasses increase linearly with dopant concentrations, whereas those of Nb-, Sb- and Sn-doped glasses do not. Of these systems, the Nb and Sn systems crystallized as mentioned

Table 3 Partial coefficient for dopant; g2 used for the calculation of refractive index [7] Dopant metal oxide

g2

Dopant metal oxide

g2

Ga203

1.77 2.82 2.57 1.94

Ta205 TiO 2 ZrO 2

2.74 2.13 2.20

Nb205 Sb203 SnO 2

128

S. Satoh et al. / Sol-gel-derived binary silica glasses

calculated and experimental values are in good agreement. The doped silica binary system, SiO2-MOn/m (M = Sb, Ta, Ti and Zr), is suitable for application to optical components.

6. Conclusions

High refractive index glasses suitable for optical components such as micro-lenses and optical fibers were prepared by the sol-gel method. The doped silica glasses, S i O 2 - M O n / m , m a i n t a i n e d silica glass characteristics such as thermal stability, chemical durability and excellent optical transparency. Refractive index and X-ray diffraction analysis of these glass systems showed that the glasses in the S i O 2 - S b 2 0 3 , S i O 2 - T a z O s , SiO2-TiO 2 and S i O 2 - Z r O 2 systems had potential for use in optical components. The experimental results for the refractive index closely followed Appen's empirical law. The authors are indebted to Mr Hisao Kojima for compositional analysis of gels and glasses.

References [1] S. Konishi, K. Shingyochi and A. Makishima, Electron. Lett. 22 (1986) 1108.

[2] R. Briickner, J. Non-Cryst. Solids, 5 (1970) 123. [3] K. Susa, I. Matsuyama, S. Satoh and T. Suganuma, Electron. Lett. 18 (1982) 499. [4] K. Kamiya, S. Sakka and I. Yamane, in: Proc. 10th Int. Congr. on Glass, Kyoto, Vol. 13 (Japan Ceramic Society, Tokyo, 1974) p. 44. [5] P.C. Shultz, J. Am. Ceram. Soc. 59 (1976) 214. [6] B.E. Yoldas, J. Non-Cryst. Solids 38&39 (1980) 81. [7] M. Nogami and Y. Moriya, Yogyo-Kyokai-shi (J. Ceram. Soc. Jpn.) 85 (1977) 448. [8] K. Kamiya, S. Sakka and Y. Tatemichi, J. Mater. Sci. 15 (1980) 1765. [9] A.A. Appen, The Chemistry of Glass (Japanese Edition) (Nisso-Tsushinsha, Wakayama, 1974). [10] V. Dimbleby and W.E.S. Turner, J. Soc. Glass. Tech. 10 (1926) 304. [11] P.C. Schultz, J. Am. Ceram. Soc. 57 (1974) 309. [12] N.A, Toropov and I.A. Bondar, Izv, Akad. Nauk SSSR, Otd. Khim. Nauk 4 (1961) 547. [13] M. Ibrahim and N.F.H. Bright, J. Am. Ceram. Soc. 45 (1962) 222. [14] N.A. Toropov and I.A. Bondar, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk 5 (1961) 740. [15] N.A. Toropov, F.Y. Galakhovand and S.F. Konovalova, Izv. Akad. Nauk SSSR, Otd. Khim. Nauk 4 (1960) 541. [16] M.L. Huggings and H. Sun, J. Am. Ceram. Soc. 6 (1943) 4. [17] J.M. Stevels, Progress in the Theory Of the Physical Properties of Glass (Elsevier, Amsterdam, 1948). [18] I. Matsuyama, K. Susa, S. Satoh and T. Suganuma, Am. Ceram. Soc. Bull. 63 (1984) 1408. [19] T. Maeda, Bull. Jpn. Inst. Phys. Chem. 2 (1923) 350. [20] J.H. Hildebrand, J. Am. Chem. Soc. 40 (1918) 84.