JOURNAL OF MATERIALS SCIENCE 27 (1992) 458-463 Production of zirconia powders from the basic disintegration of zircon, and their characterization J. M. A Y A L A , L. F. VERDEJA, M. P. GARCIA, M. A. LLAVONA, J. P. SANCHO Departamento de Metalurgia y Materiales, Escuela de Minas, Universidad de Oviedo, 33004 Oviedo, Spain Monoclinic Zr02 has been prepared through the decomposition of ZrSiO4 with soda ash and lime, followed by leaching in hydrochloric acid and sodium hydroxide. The resulting zirconia powders are characterized in terms of their physical and chemical properties. 1. I n t r o d u c t i o n Zircon is the most common and widely distributed raw material for the production of zirconia. The oxide is commercially manufactured from the silicate through a variety of processes, most of which involve a treatment to form an aqueous solution of zirconium from which the fine zirconia particles are precipitated. In each case, the process for the fabrication of the oxide is chosen according to cost, raw material availability, and the purity and properties required in the final product. This paper reports on the development of two novel methods for the fabrication of zirconia powders from zircon which do not involve an aqueous solution of the element. Instead, the silicate raw material is reacted with soda ash or lime to form monoclinic zirconia. Silica, and Na- or Ca-silicozirconate, are also formed as by-products, which are leached out with hydrochloric acid and alkali, respectively. In the first route, the zircon raw material is decomposed with soda ash; in the second route, lime is used. Both approaches are described below. 2. Experimental procedure 2.1. Raw materials and equipment The zircon ores used in this study were commercial grade products; their physical and chemical characterization is summarized in Tables I and II. Chemical composition was determined with X-ray fluorescence for the major elements (Zr, Hf and Si), and atomic absorption spectrometry for Fe, Mg, K, Na, Ti, Ca and A1. True density was measured with a helium autopycnometer (Micromeritics, Inc. Norcross, GA, USA). Specific surface area was measured by the Brunauer-Emmett-Teller adsorption equation (BET method) with nitrogen at the temperature of liquid nitrogen as the adsorbate (Micromeritics, Inc.). As shown in the tables, true density and surface area of the different zircon raw materials were very uniform (around 4.6 gcm- 3 and 1 m 2 g- 1 respectively). They exhibited, however, widely different particle size dis- 458 tributions: average (dso) particle sizes for powders B and D were 15 and 130 gm, respectively. The same analytical techniques were applied to the characterization of the zirconia powders. Sodium carbonate, hydrochloric acid and caustic soda were commercial grade products. Thermal decomposition of the zircon was made in a high-temperature muffle furnace capable of reading 1650 ~ (Lindberg, Inc. Chicago, IL, USA). Preliminary work in the alkali and acid leaching of the decomposition products was carried out in teflon-lined stainless steel acid digestion bombs (Parr Instruments); once the optimum digestion parameters had been established, alkali digestion was made in an autoclave (Parr Instruments), and acid leaching was performed in a glass reactor provided with mechanical stirring. 2.2. Synthesis of zirconia powders by the soda-ash process 2.2.1. Zircon disintegration It is well known that zircon is decomposed at high temperature in reducing media [13. The decomposition temperature, however, is lowered when the silicate is mixed with a fluxing agent such as NaOH [23, Na2CO3 [-3, 43, CaCO3 o r M g C O 3 [-5, 63. Samples of grade G zircon were mixed with 5, 8, 10 and 20 wt % soda ash by dry ball milling with zirconia media; the mixed powder was compacted into small pellets with a hand press. Thermal treatmmit was carried out at temperatures between 1100 and 1600 ~ soak times varied between 5 min and 3 h. As evinced by X-ray diffraction (XRD), reaction products depended on reaction temperature, reaction time, and the concentration of soda ash in the solid mixture. For samples with 20% soda ash, the reaction was not complete at temperatures below 1300 ~ even when long reaction times (180 min) were used. Reaction products included monoclinic zirconia, Na4Zr2Si3012 [73 (Fig. 1), amorphous silica and unreached zircon. At 1350~ decomposition was complete in only 5 min. 0022-2461/92 $03.00 + .12 9 1992 Chapman & Hall T A B L E I Chemical composition of zircon Composition (wt %) Zircon ZrO2 HfO/ SiO2 Fe203 MgO H/O Na20 TiO2 CaO A1203, A B C D E F G 64.30 62.64 59.89 59.89 57.76 56.64 62,60 1.52 1.50 1.52 1.52 1.47 1.42 1.52 32.84 33.00 32.88 32.58 33.42 33.02 32.70 0.14 0.23 0.36 0.36 0.36 0.12 0.50 0.02 0.03 0.04 0.04 0.04 0.05 0,11 0 0.01 0 0 0.03 0.02 0.15 0.03 0.02 0 0 0.04 0.02 0.47 0.04 0.15 0.09 0.09 0.06 0.09 0.18 0.07 0.08 0.07 0.07 0.04 0.07 0.50 0 0 0 0 0 0 0 TAB L E I I Physical characteristics of zirconium silicates Zircon Open porosity (%) Total porosity (%) Macro porosity (%) Micro porosity (%) Apparent density (gcm -3) Bulk density (gcm -3) True density (gcm 3) ds0 porosity (~m) ds0 particle size (llm) BET m 2g - 1 A B C D E F G 35.6 37.1 17.3 33.6 39.2 35.7 43.7 46.5 54.9 46.7 54.6 52.4 50.1 56.4 16.5 7.4 11.2 33.4 8.6 7.4 9.8 20.1 29.7 6.1 0.2 30.4 28.3 33.9 3.91 3.35 3.01 4.11 3.61 3.68 3.57 2.52 2.10 2,49 2,72 2,19 2,37 2.01 4.71 4.66 4.67 4.67 4.60 4.83 4.65 2.19 2.38 2.34 35.32 2.72 2.12 2.11 125 15 80 t30 100 77 30 0.8 0.4 0.9 0.7 0.3 0:6 ! 0.6 M (o) M M S MM SS S ( d ) ~ ~ ~ 55 4'0 3'5 3'0 20 (CuKe) A ' io- Figure 1 X-ray patterns of samples with 20% Na2CO 3 heated at different temperatures. (a) 1350; (b) 1400; (c) 1450; (d) 1500~ M = m-ZrO2; S = Na4ZrzSi3012. Higher temperatures were required to complete the decomposition of samples containing 10% soda ash. Decomposition of this material was complete only after a thermal treatment at 1500 ~ while samples reacted for l h at 1400~ still exhibited 15% of unreacted zircon. Higher reaction temperatures (1600 ~ were necessary tocomplete the decomposition of samples containing even lower amounts (5-8%) of soda ash. 2.2.2. Alkafi leaching The product from the complete decomposition of the silicate with soda ash consists of a mixture of monoclinic zirconia, Na-silicozirconate, and amorphous silica. The purpose of the alkali leaching step is the removal of this amorphous silica through a wet chemistry process. A parametric study was carried out first, in which approximately 0.5 g of the product from the previous step was treated with 10 ml of a NaOH solution in an acid digestion bomb, which was heated in a stove at 200 ~ for 24 h. The leached sample was then filtered and washed with distilled water. It was found that a 20% solution of NaOH was most effective in the removal of silica from the samples that had been decomposed with 10% NazCO 3. In these conditions, silica removal increased with increasing leaching time to an optimum value at 24h, after which it did not improve appreciably any further. Once the parametric study was completed, further research was carried out with products which had been decomposed at 1350, 1400, 1450, 1500 and 1600 ~ Samples were leached in the autoclave. For a given alkali leaching procedure (5 h at 200 ~ with 20% NaOH), the best results were obtained with the material that had been decomposed with 10% Na2CO3, the average ZrO 2 content in the filtered, washed and dried powder was 74%, while the corresponding figure was less than 64% for the material that had been decomposed with 20% NazCO 3. To determine the optimum concentration of caustic soda, leaching experiments were carried out at three different alkali concentrations (10; 15 and 20% NaOH) with the materials which had been decomposed at 1500~ with 10% NazCO 3. As shown in Table III, results improved with increasing concentration of caustic soda: a product of 76% ZrOz was obtained with a 20% NaOH solution at 200 ~ while the zirconia content of the product was only 74% when a 15% NaOH solution was used. The effects of pulp density on the leaching of silica were also studied; No large differences were exhibited 459 T A B L E III Zirconia and silica content averages of products elaborated with mixtures of zircon with 10% NazCO ~ heated at 1500 ~ after caustic leaching NaOH (wt %) Temperature (~ Time (h) ZrO2 (wt %) SiOz (wt %) 10 10 15 15 15 20 20 20 20 20 20 20 20 20 20 20 200 220 200 200 220 180 180 180 200 200 200 200 200 220 220 220 6 6 10 24 10 8 10 24 5 6 8 10 24 8 10 24 72.8 70.1 73.7 70.6 74.5 74.3 74.6 75.1 70.5 72.8 76.0 75,8 75.7 74,1 75.0 75.8 14.36 16.06 11.13 11.71 10.07 12.18 10.01 11.90 16.87 14.32 12.17 10.57 10. t6 11.99 t 1.13 11.42 by treatments involving pulp densities in the 10-30 wt % range. There was in any case enough excess of alkaline reactive. 2.2.3. Acid leaching After the silica had been eliminated from the samples via an alkaline leaching, Na-silicozirconate still remained unattacked in the product. This compound can be removed through solution in hydrochloric acid [8]. As in the previous step, a parametric study was performed first in acid digestion bombs, while subsequent work used a heated glass reactor provided with a mechanical stirrer. The effects of HC1 concentration, temperature and reaction time were studied. Na4Zr2Si3012 is attacked by HC1 preferentially to ZrO2, but dissolution and loss of the zirconia will also take place once the Na-silicozirconate has reacted. For this reason, optimum conditions were found to 100 96 o t,,. N 94 92 BB 90 0 I I [] I 10 20 30 40 HCI. ( wt Old Figure 2 Effect of temperature, time and hydrochloric concentration on acid leaching. Treatments: diamonds, room temperature; squares, boiling temperature; closed symbols, 1 h; open symbols, 2 h. 460 2.3. Synthesis of zirconia powders by the lime process As an alternative to the use of soda ash, zircon can be decomposed by lime [5, 6, 9, 10]. The synthesis of zirconia by the lime process also includes three steps: decomposition of the silicate, acid leaching to remove the resulting wollastonite phase, and removal of the silica with alkali. 2.3. 1. Zircon disintegration 98 ~ include the lower acid concentrations (preferably 10%), room temperature, and shorter reaction times (Fig. 2). As an example, the use of 10% HC1 yielded a product consisting of 98% ZrO2, while this figure decreased to 91.6% ZrO2 when 30% HCI was used in the acid leaching step. Results are summarized in Table IV. Differences in pulp density within the 5-30% range had negligible effect on the results of the acid leaching; there was in any case enough excess of acid reactive. Mixes of zircon powder with 5, 10, 15, 20 and 30 wt % lime were prepared by dry ball milling with zirconia media. Samples of the compacted powder mix were heated in a Lindberg furnace for 2h at 1450, 1500 and 1550~ Results indicated that decomposition of the zircon was enhanced by higher lime contents and higher temperatures. Samples with lime contents below 20% exhibited only partial decomposition in the temperature range studied. For 20% lime content, total decomposition took place only at temperatures of 1500~ or higher. Samples with 30% lime, however, were totally decomposed even at the lowest temperature. Monoclinic zirconia and wollastonite [7] (Ca3Si2ZrO9) were identified by XRD analysis as the two crystalline phases in the reaction products (Fig. 3). 2.3.2. Acid leaching Wollastonite is soluble in hydrochloric acid [8]. When the products of the decomposition of zircon with CaO T A B L E IV Chemical composition of zirconias manufactured by soda ash and lime processes Zircon ZrO 2 + HfO2 (wt %) SiO 2 (wt %) Fe20~ (wt %) MgO (wt %) K20 (wt %) Na20 (wt %) TiO 2 (wt %) CaO (wt %) AI203 (wt %) Ca: A B C D E F G Na:A B C D F G 94.68 91.21 94.27 90.57 92.71 92.39 96.82 90.53 93.66 95.96 90,07 95,48 98,10 2.34 4.83 1.36 4.23 5.81 3.06 1.28 5.05 6.10 1.40 5.58 3.40 0.57 0.14 0.21 0.14 0.14 0.26 0.21 0.14 0.25 0.28 0.14 0.14 0.14 0.14 0.48 0.70 0.29 0.47 0.51 0.42 0.16 0.1l 0.29 0.17 0.16 0.06 0.09 0 0.04 0 0 0.06 0 0.05 0.06 0.06 0.01 0.07 0.15 0.01 0.25 0.64 0.15 0.25 0.46 0.19 0.37 0.12 0.19 0.07 0.23 0.30 0.05 0.04 0.04 0.09 0.04 0.05 0.08 0.05 0 0.05 0.05 0.07 0.04 0.03 1.26 2.59 0.91 1.19 1.34 1.34 0.98 0.21 0.21 0.07 0.14 0.07 0.07 0 0 0 0 0 0 0 0 0 0 0 0 0 within the error of the analytical procedure [11]; shorter or longer residence times resulted in poorer zirconia yields. The preferred reaction time was therefore fixed as 30 min after the mixture had reached the boiling temperature. The effect of pulp density was also studied. There was in any case enough excess of acid reactive. Zirconia yields decreased with increasing pulp density. A solids loading of 20% was preferred in order to optimize the process economics. M H H H 2.3.3. Alkafi leaching Since the products from the acid leaching step still contained large amounts of silica, they were treated with caustic soda in order to produce zirconia powders of high purity. Samples were treated with 20% NaOH solution at 200 ~ for 5 h, according to the optimum conditions that had been derived in the work with soda-decomposed zircon (section 2.2.2.). A study of the effect of pulp density in the 30-50% range revealed that lower densities were preferred. As an example, alkali leaching with a pulp density of 30% yielded a product with zirconia content of 98%, while only 91% ZrO2 was obtained with a pulp density of 50%. A S 25 ZO S 15 zo (CuX~) 3. Results and discussion Figure 3 X-ray patterns of samples with 20% CaO heated at 3.1. Zirconia purity different temperatures. S = ZrSiO4; M = m-ZrOz; W = CaSiO 3. Purity of the zirconia products obtained in this work ranged between 90 and 98%. Reports available in the literature indicate that the purity of zirconias produced by hydrothermal processes range between 84 and 99.9%; plasma methods have produced zirconias in the 94.5-99.6% range. The processes proposed in this work produce zirconia powders in the upper region of those ranges. were leached with HC1 at high temperatures and pressures, a ZrO2-rich gelatinous precipitate was obtained which was not produced when leaching took place at atmospheric pressure. Leaching with 10 and 20% HC1 produced a precipitate containing 74 and 76% ZrO2, respectively; the corresponding values for the silica content were 17 and 14%. An HC1 concentration of 20% was therefore considered optimum. Reaction rate increased with increasing temperature up to the boiling point of the mixture (106 ~ Optimum reaction time at 106 ~ was determined as 30-60 min. Results were similar in this time range 3.2. Zirconia density and porosimetry Apparent density for the zirconias obtained in this study ranged from 3.1 to 4.5 gcm- 3. These values are lower than those reported for the commercial monoclinic zirconias (typically 4.7 gem-3). The same applies to the bulk density: values obtained in this study 461 T A B L E V Physical characteristics of zirconias manufactured by soda ash and lime processes Zircon Open porosity (%) Total porosity (%) Macro porosity (%) Micro porosity (%) Apparent density (g cm- 3) Bulk density (g cm- 3) True density (g cm- 3) dso porosity (gm) Na:A B C D F G 52.1 50.9 59.2 42.9 54.3 52.6 65.5 60.0 67.9 63.0 64.9 68.0 8.4 7.1 6.0 7.2 10.7 8.7 43.7 43.8 53.2 35.7 43.6 45.9 3.66 4.47 4.38 3.49 4.37 4.09 1.75 2.20 1.78 1.99 1.99 1.94 5.09 5.50 5.55 5.39 5.70 5.69 1.13 0.33 0,37 0.49 0.51 0.34 Ca:A B C D E F G Mon. Com.* 53.2 60.9 53.4 52.9 62.4 37.7 53.6 40.0 63.0 69.3 64,0 61.5 70.7 65.7 65.3 52.1 8.2 15.2 6.7 8.1 12.3 7.2 11.4 8.7 45,0 45,7 46.7 44.5 50.1 30.5 42.4 31.3 4,44 4.06 4.35 4.44 4.09 3.06 4.25 4.60 2.07 1.59 2.04 2.09 1.54 1.90 1.97 2.81 5.61 5.18 5.67 5.44 5.25 5.55 5.67 5.87 0.54 0.24 0.27 0.20 0.14 0.13 0.85 1.68 dso particle size (gm) 4.5 5.5 8.0 6.0 8.0 8.0 70 53 14 9 10 7 58 60 BET (m 2 g- ') 5.3 2.8 10.2 3.1 0.5 5.0 3.6 5.5 4.4 6.9 13.4 6.8 2.0 0.6 * Monoclinic commercial product. ranged between 5.1 and 5.6 g cm-3, while the corresponding value" for commercial zirconias is typically 5.9 g cm-3. As shown in Table V, total porosity is approximately the same (about 65%) for all zirconias produced in this study. It should be noted, however, that microporosity (approximately 10%) is small compared with the macroporosity (around 45%). In general, pore size distribution for the zirconias obtained in this work is coarser than for commercial monoclinic zirconias. 3.3. Zirconia surface area: particle size and shape Specific surface aria for the zirconias obtained in this study is generally higher than the typical values for the commercial powders (0.6 mZg-1), and similar to those of the zirconias produced by plasma methods (5-8 m2g2l): Powders derived from hydrothermal processes 1-15-18-] exhibit a much wider range (4~}23 m 2 g-1). Different grain morphologies of our zirc0nias are shown in Fig. 4. In general, powders with high purity such as those derived from ores C, F and G exhibit a spherical grain shape; lower-purity zirconias (such as those derived from ore A) show small impurity grains adhering to the zirconia particles. This applies to zirconias fabricated by both lime and soda methods. Particle size distribution for zirconias fabricated by the lime method varies widely: dso is 70 gm for zirconia A, and only 7 gm for zirconia F. It is much more consistent for the products derived from the soda ash method, with dso ranging between 4.5 and 8 gm (Figs 5 and 6). For comparison purposes, commercial monoclinic zirconias exhibit a dso of approximately 60 gm, and the products obtained from hydrothermal processes have a dso value in the 3-35 gm range. Figure 4 Scanning electron micrographs of zirconias: (a) via soda ash, Na-G; (b) via lime, Ca-G. results of the process. As a general rule, products of higher purity were obtained from ores containing higher amounts of zirconium when leached with alkali solutions of higher concentrations (Table IV). When two zkcon ores with similar particle Sizes were used (ores C and F, dso = 80 gm), a slightly better product was obtained in both methods from the ore C with a higher ZrO2 content. 3.4. Effects of raw material purity 3.5. Influence of zircon particle size The chemical composition of the raw material used, and the amount of lime or soda ash used in the decomposition step, had a large influence on the The particle size distribution of the zircon ore had an important role in the purity of the zirconia manufactured from both the soda and the lime processes. A 462 100. ~ 80. / ff / [] / ~ 40. [] / / 20. tn rl i i I 10 100 Particle diameter (I-tin) 1000 Figure 5 Particle size distribution of zirconia D, N a - D . 100 _ I/i,i.I III'- 80 : j 7 ./ z0. 0 ~'~* I i 10 100 Particle diemeter (I.Lm) i 1000 Figure 6 Particlesize distributionof zirconias via lime, ~1,,Ca-G; Ca-C; I , Ca-F. coarser particle size distribution in the zircon ore usually resulted in a decreased purity of the zirconia powder. This effect is illustrated by a comparison between the behaviour of ores C and D, which exhibited similar chemical composition but different particle size distributions. The zirconia from ore C (the finer of the two) was richer in ZrO2 than the powder resulting from the much coarser ore D. optimum parameters in this step are a pulp density of 30%, temperature of 200 ~ pressure of 14 atm, and residence time of 5 h. This step is followed by an acid leaching of the filtered, washed powder mixture in order to remove the Na4SisZr2012. Optimum parameters for the acid leaching step include the use of a 10% HC1 solution, room temperature, atmospheric pressure, 1 h residence time, and 20% pulp density. For the lime process, the optimum parameters in the disintegration of the zircon ore are: 20 wt % CaO, and a thermal treatment at 1500 ~ for 2h. The resulting product consists of monoclinic zirconia, [3-wollastonite, and amorphous silica. The l~-wollastonite phase is then leached out with hydrochloric acid under the following preferred conditions: 20% HC1 solution, boiling temperature (106~ atmospheric pressure, 30 min residence time, and 20% pulp density. The last step is the removal of the amorphous silica via alkaline leaching with 20 wt % NaOH at 200 ~ and 5 h; preferred pulp density is 15%. The ZrOz content of the zirconias fabricated by these methods lies in the 90-98% range, depending on the chemical purity of the zircon ore. Higher purity can be achieved with lower pulp densities in the leaching steps. True, apparent and bulk densities of the zirconias were measured in the ranges 5.1-5.7, 3.1-4.5, and 1.5-2.2 g cm- 3, respectively. Total porosity was approximately 65%, most of which (45%) was microporosity. Morphological analysis showed a grain shape factor close to 1. The zirconias derived from the soda ash process exhibited smaller and more uniform grain size distribution than those derived from the lime method. Chemical composition of the starting zircon, and the amount of soda ash or lime used for disintegration, had an important effect on the quality of the resulting zirconia. Acknowledgements The authors express their gratitude to FICYT for their financial support of this work. References 3.6. Effects of pulp d e n s i t y Zirconia purity could be improved with lower pulp densities in the leaching steps. 4. C o n c l u s i o n s Two methods are proposed for the fabrication of zirconia powders from zircon. The two methods differ in the fluxing agent used for the decomposition of the silicate: soda ash in the first case, and lime in the second. In both approaches, zircon decomposition is achieved with smaller amounts of fluxing agent than those previously reported. The optimum parameters for the thermal decomposition of zircon with NazCOs were defined as follows: 10 wt % NazCOs atad 2 h at 1500 ~ After this treatment, zirconium is crystallized as monoclinic zirconia, while silica remains in the mixture as both amorphous silica and as Na4Si3Zr2012. This product is reacted in an autoclave with 20 wt % NaOH to remove the silica; 1. A. GARCIA and L. DEL a L M a , Bol. Sac. Esp. Ceram. 9 (1970) 555. 2. M. T R O J A N and Z. SOLC, Silikaty 29 (1985) 309. 3. M. A. Z A G H E T E , J. M. CASTRO e CASTRO and J. A. VARELA, Ceramica 32 (1986) 165. 4. P.R. M E N O N , J. M. J U N E J A and T. S. K R I S H N A N , Amer. Ceram. Sac. Bull. 59 (1980) 635. 5. P . G . USOV and E. P. S O L O M A T I N A , lzv. Tomsk. Politekh. Inst. 185 (1970) 135. 6. I. F. P O L E T A E V , L. V. KRASNENKOV'A and T. V. S M U R O V A , Tsvetn. Met. 11 (1981) 24. 7. "Index (Inorganic) to the Powder Diffraction File", Joint Committee on Powder Diffraction Standards, Pennsylvania (1971). 8. WEAST, R. C., H a n d b o o k of Chemistry and Physics (CRC Press, Florida, 1978/9). 9. I. F. P O L E T A E V , L. V. K R A S N E N K O V A and A. S. B E R E S T Y U K , Tsvetn. Met. 12 (1983) 47. 10. S.R. REDDY and G. M A N D A L , Trans. Ind. Ceram. Sac. 34 (1975) 1. 11. J . P . S A N C H O , L. F. VERDEJA and J. AYALA, in Proceedings of the 8th Congreso Internacional de Mineria y Metalurgia, Nuevos Materiales, Oviedo, 1988 (Asociacibn Nacional de Ingenierus de Minas de Espafia, Ovieda, 1988) p. 73. 463 12. P. REYNEN, H. BASTIUS, B. PAVLOVSKI and D. von MALLINCKRODT, in "Advances in Ceramics, Vol. 3", edited by Heuer and Hobbs (American Ceramic Society, Columbus, Ohio, 1981) p. 464. 13. M: YOSHIMURA and S. SOMIYA, in ibid. p. 455. 14. G. CLARKE, Metal Bull. (1987). 15. M.A. van de GRAAF and A. J. BURGGRAAF, in Advances in Ceramics, Vol 12" edited by Claussen, Ruhle and Heuer (American Ceramic Society, Columbus, Ohio, 1985) p. 744. 464 16. 17. 18. L.S. MILLBERG, J. Metals (1987) 9. M. CIFTCIOGLU, M. AKINC and L. BURKHART, Amer. Ceram. Soc. Bull. 65 (1986) 1591. D.J. CLOUGH, Ceram. Enqng Sci. Proc. 6 (1985) 1244. Received 11 September 1990 and accepted 28 February 1991