J. MATER. CHEM., 1994, 4(8), 1343-1348 Novel Preparation of Highly Dispersed Tungsten Oxide on Silica Sophia Colque," Edmond Payen*b and Paul Grangea a Universite Catholique de Louvain, Groupe de Physico-Chimie Minerale et de Catalyse, 1348 Louvain-la-Neuve, Belgium Laboratoire de Catalyse Heterogene et Homogene & LASIR Universite des Sciences et Techniques de Lille, 59655 Villeneuve d 'Ascq Cedex, France W03/Si0, samples prepared by three different methods and calcined at two temperatures (450 and 900 "C) are studied by different physicochemical techniques [X-ray diffraction (XRD), specific surface area, X-ray photoelectron spec- troscopy (XPS), laser Raman spectroscopy (LRS), electron microscopy and temperature-programmed reduction (TPR)]. When the solids were prepared by mixing two gels (tungsten and silica gel) a good WO, dispersion was achieved and calcination at high temperature gave an amorphous phase.The impregnation of silica gel by ammonium paratungstate allows the interaction of tungsten with silica as dispersed oxoanions and, upon calcination, smaller WO, particles and amorphous glassy particles were formed. In contrast, when the solid was prepared by the wet impregnation method using a silica carrier the interaction of the oxotungsten species with the support was weak. When the sample was calcined the silica and the oxides were segregated. Supported oxides and sulfides of molybdenum and tungsten are well known for catalysing a large variety of reactions. Consequently many studies have been devoted to the prep- aration, characterization and activity of these solids.Most attention has been paid to the alumina-supported catalysts whereas silica-supported catalysts were less studied. Most studies of the latter catalysts deal with MOO,-based solids. These showed that a high dispersion of the molybdenum oxide is never achieved. The formation of crystalline MOO, is observed in Mo03/Si02 catalysts at low Mo loadings,'-3 Fewer studies have been devoted to W03/Si02 catalyst^,^-^ and it has been shown that tungstate is even less well dispersed than molybdate on the same silica ~arrier.~ For a WO, concentration as low as 5 wt.% a crystalline WO, phase is formed.5,7 These solids were generally prepared by impreg- nation of the oxide silica carrier followed by drying and calcination.This classical method is limited for providing good dispersions of the tungsten oxide species owing to the low level of interaction between the oxotungstate species and the silica carrier. The tungsten species tend to agglomerate even at low loadings. The final aim of our work is to prepare dispersed carbide phases using tungsten oxide as a precursor. One of the most important aims for catalytic purposes is to obtain finely divided carbides. In addition to the experimental conditions of the carburization process, the precursor plays an important role in the final properties of the carbide. In order to achieve our purpose, a novel preparation of tungsten oxide supported on silica has therefore been carried out by a sol-gel method which allows simultaneously the preparation of the carrier and the introduction of the oxotungstate species.We recently applied this method of preparation to the syn- thesis of Mo/A1203 oxides and showed that it improves the dispersion limit of oxomolybdate entities." This paper deals with the synthesis of samples obtained by using two new preparation methods, including their solid-state characteriz- ation by different physicochemical techniques. These solids will be compared to those obtained by the traditional impreg- nation method. Tungsten is introduced either as a gel or as an ammonium tungstate salt. The silica is either a high-surface-area Si02 or gel of silica. Experimental Preparation Three different methods of preparation have been used.Mechanical Mixture of Tungsten Gel with Silica Gel Gels of tungsten and silica were prepared separately and then mixed. The silica gel was prepared by adding slowly 95 ml of Si(OEt), to 450 ml of water. After 14 h at 60 "C under stirring, this solution became jelly-like. Ethanol was then removed under reduced pressure in a rotary evaporator. The gel obtained was washed with water and centrifuged several times. The tungsten gel was prepared as follows. Hydrochloric acid solution (3.8 mol 1-l) was dropwise added to an ammonium paratungstate ~(NH,)l,H2W,,0,2~H20J solution (6.5 lop3 mol 1-') at room temperature. The pH of the solution was monitored up to 1-1.5. A clear, yellow solution was obtained.After stirring the solution, it progressively became turbid and turned to a gel after some hours. This gel was aged for 48 h and then separated from the mother solution by centrifugation at 2000 rpm for 10 min. The pH of the silica gel was decreased to 1-1.5 with the hydrochloric acid solution. Then the two gels were rnechan- ically mixed and left together for 12 h before freeze-drying them. Two samples containing 7 and 10.9 wt.% WO, were pre- pared and are referred to hereafter as A7 and A10, respectively, whereas the pure tungsten oxide and silica oxide is denoted as W and silica A. Mixture of Silica Gel and Ammonium Puratungstute Salt The method of preparation of the silica gel was the same as described above.Before the gel was mixed with the ammonium paratungstate solution (8.1 x lo-, mol 1-'), its pH was increased to 9.5-10 with an ammonia solution (6.6 rnol 1-'). This sample is referred to as silica B. The gel and the tungsten solution were then stirred and kept for 24 h. After centrifugation (1000 rpm for 10min) the gel was freeze-dried. This sample is referred to as X6. Impregnation of Si02with Tungsten Salt Commercial SiO, (Degussa FK 700) was impregnated with an aqueous solution of ammonium paratungstate (9.6 x lop3 mol 1-l). The impregnation was carried out using 4.2 ml of solution per gram of SO2. Water was evaporated at 40°C in a rotary evaporator and the sample was dried at 100°C for 2 h. This sample is referred to as 18.All the samples were calcined at 450 and 900°C for 4.5 h and 1.5 h, respectively, and the W03 content of the solids was evaluated by atomic absorption (AA) after dissolution of the samples in hydrofluoric acid at 200°C. The samples are denoted by the method of incorporation of the tungsten, i.e. the content of WO, and the temperature of calcination (for example 18-900 represents the impregnated sample with 8 wt.% WO,, calcined at 9OOOC). Physico-chemical Characterization Powder XRD measurements were carried out with a Siemens D 500 diffractometer using Cu-Ka radiation. The specific surface area was measured gravimetrically by nitrogen adsorption using the BET method. The measurements were made in a vacuum microbalance Setaram MTB x,the fresh and calcined samples first being outgassed to constant weight at 100 and 200 "C, respectively.XPS measurements were carried out with an SSX-100 spectrometer (Surface Science Instruments 206) equipped with an A1 X-ray source working at 200 W. The powdered samples in their original form were pressed into small inox-holders and the C Is, 0 Is, Si 2p and W 4f photoemission lines were recorded for each sample. Binding energies (Eb)were deter- mined with reference to the C 1s line at 284.8 eV. The cali- bration of the spectrometer energy scale was performed using Au 4f,,, (E,=84 eV). A flood gun, with an energy of 6 eV, was used to eliminate differential charging of the samples. XPS atomic ratios were determined by using the total inte- grated areas of the W 4f5/2,7/2 (spin doublet) and the Si 2p photoelectron lines.These ratios were estimated using the formula: NdNsi =(lw/Isi (Ssi/Sw ) (ESi/Ew where Ni,Eiand Siare, respectively, the number of atoms of the element i, the kinetic energy corresponding to a given line and the Scoffield photoelectric cross-sections. LRS measurements were performed using a Raman micro- probe (Mole from Jobin-Yvon). The exciting light source was an Ar' laser emitting the 488 nm line with the power at the sample kept as low as possible, at ca. 1mW, to avoid transformation of metastable phases. TPR experiments were carried out in a home-made dynamic apparatus. A mixture consisting of 5% hydrogen in argon was used as the reducing gas (25 ml min-').The temperature was raised to 840°C at a rate of 10°C min-'. The weight of the sample (20-40 mg) was adjusted in order to have the same content of tungsten in the reactor. The hydrogen con- sumption was measured with a thermal conductivity detector. The electron microscopy studies were performed on a JEOL Temscan lOOCX microscope equipped with a Kevex 5100C energy-dispersive spectrometer for X-ray microanalysis. The samples were dispersed in water using an ultrasonic device and then deposited on carbon films supported on copper grids. These were studied in the conventional transmission mode (CTEM) and by analytical electron probe micro-analysis (EPMA). Results Table 1 shows the surface areas of the catalysts before and after calcination at 450 and 900°C.Heating the catalysts at 450°C has practically no influence on the surface area, while a sharp decrease in surface area is found upon calcination of the catalyst at 900 "C. J. MATER. CHEM., 1994, VOL. 4 Table 1 WO, content and specific surface area of the samples BET surface area/m2 g-calcination temperature/"C WO,sample (wt.%) N,/N,,x lo2 fresh 450 900 ~~ A1 7 1.92 797 702 213 A10 10.9 3.08 705 690 77 X6 6.5 1.77 427 482 277 I8 8.3 2.33 349 352 4 silica A 0 - 803 - 311 silica I 0 - 584 - 11 XRD All the X-ray diffractograms of the samples are shown in Fig. 1 and 2. The XRD pattern of W-25 (Fig. 1 ) is character- 0: 1O( 220 021 c I. I I 9 25 41 57 2gdegrees Fig.1 X-Ray diffraction pattern of the pure tungsten oxides (indexing according to the orthorombic WO, cristallographic data): (a) W-25, (b)w-900 A1 0-900 A10-450 n I, ----4-..----18-450 =X6-450 18-900/ 41 81 1 41 81 2Hdegrees Fig. 2 X-Ray diffraction pattern of the WO,/SiO, samples calcined at 450 "C and 900 "C.The XRD peaks of orthorombic WO, and of silica are referred to in (a) and (d),respectively. 0,Christobalite. J. MATER. CHEM., 1994, VOL. 4 istic of an amorphous sample, and the XRD features of the calcined sample at 900 "C characterize the orthorombic WO, phase (Fig. 1)according to JCPDS data (no. 20-1324). The diffractograms of uncalcined or calcined silica-based samples show a broad band at ca.28=22', which is due to amorphous silica (Fig. 2). The X-ray diffractograms of the A7 and A10 catalysts calcined at 450 'C and 900 "C are shown in Fig. 2[(a) and (b)]. The main characteristic peaks of WO, could be detected in the pattern of A7 calcined at 450°C on the underlying peak of silica. These peaks are more evident with increasing tungsten loading. However, they are not observed in the pattern of A7-900. This is probably due to a modification in the sample of this oxide upon calcination, as will be shown by LRS. After calcination the X6 samples did not exhibit any lines of a crystalline phase. By comparison, the XRD pattern of sample 18-450 [Fig. 2(4] indicated the presence of WO,, whilst after calcination at 900°C these features are present with the diffraction peaks of cristobalite (21.92", 36.18', 31.46').XPS The binding energies of the 0 1s and Si 2p levels for all the catalysts agreed with the values obtained for silica (0Is= 532.9 eV and Si 2p= 103.6 eV). Fig. 3 shows, for all the cata- lysts, the W 4f XPS doublet characteristic of the presence of W6+ species in an oxygen environment. For the uncalcined A7 and A10 catalysts the peaks are narrow and the position is similar to that observed for the tungsten gel W-25 (36.1 and 38.1 eV). Well defined lines at 35.7 and 37.8 eV, character- istic of tungsten trioxide, are observed for the solid W-900, whereas a broadening of this doublet occurs for the calcined A7 and A10 samples. This broadening is even more important for X6, whilst the well resolved lines of bulk WO, are observed in the solid I calcined at 900 "C.The effect of the temperature of calcination on the superficial tungsten concentration is illustrated in Fig. 4 where the XPS atomic ratios (W 4f/Si 2p) are plotted against the temperature. 42 38 34 42 38 34 EdeV Fig. 3 XPS spectra of the W 4f levels for fresh and calcined samples 0 500 10000 500 1000 TIT Fig. 4 (lw/Isi)XPS atomic ratio for fresh and calcined samples: (a) A7, (b)A10, (c) X6, (d) I8 LRS The Raman spectra of the dried silica powder (Fig. 5) show only a broad line at around 1100 cm-'. Thermal treatment of these silica gels at 900°C causes the system to evolve towards vitreous silica, as indicated by the Raman features at 500 and 1100 cm-', in agreement with literature data."*'2 LRS is as a very useful technique for characterizing tungsten oxides or oxotungstate species.Under our experimental con- ditions, using a very low laser beam power to avoid phase transformations, the spectrum of the tungsten gel shows a line at 960 cm-I and a broad band at around 600-720 err-'. This could be assigned by reference to literature data" to the existence of the hydrate WO3.2H,O, of which the particle size and/or the non-ordered nature are likely to produce the spectral distortion and line broadening. In spite of the low pH of preparation, the existence of silicon heteropolyt unsgtate 720 8r 640 A-900 1100 IIIIIIIIII 200 400 600 700 1000 Aijlcm-Fig.5 Raman spectra of reference samples: (a): silica, (h): tungsten oxide. The plasma lines of the laser are indicated by P. J. MATER. CHEM., 1994, VOL. 4 should be pre~1uded.l~ After calcination of the catalysts, intense lines at 810 and 720 cm-', corresponding to bulk WO,, are 0b~erved.l~ The main features of the tungsten oxide gel, identified as WO3.2H,O are also observed on the uncal- cined samples, whereas the Raman features of WO, are mainly seen after calcination at 450°C (Fig. 6) with the broad band of vitreous silica (490 and 11OOcm-I). Although the SiO, support may exhibit a line at 980 cm-', significant intensity variations between the lines are noted on the Raman spectrum of the A7-450 sample and confirm that this line is the W-0, stretching mode of a polytungstate species, where the subscript t denotes a terminal W-0 bond.I5 The relatively poor quality of the Raman spectra is probably due to the poor crystallinity of the supported samples, as shown by XRD measurements.The other vibrational modes are always weaker and are hardly detectable above the background. After calcination of the catalyst at 900 "C the tungsten oxide is the only detectable species (Fig. 6); however a close comparison with the Raman spectrum of bulk oxide indicates that there are some major differences. These occur mainly in the shift and broadening of the main high-frequency lines at 800 and 687 cm-' and in the lattice lines region (100-400 cm-') which reflects the crystal- linity of the sample.These differences suggest that this WO, species shows little crystallinity and/or interacts with the support. Similar LRS features were previously obtained by Pham Thi and Velasco on W03 thin films sputtered on various supports.16 Raman results for samples X and I are reported in Fig. 7. With the Raman features of the vitreous silica, the X6-25 sample exhibits a Raman line at 960 cm-' with a broad band on its low-wavenumber side as well as very broad bands at 560cm-1 and in the 200-400cm-' spectral range. These features are consistent with the presence of a polymeric oxotungstate entity interacting with the silica gel, similar to those previously described for W/Al,O, system^.'^ After calci- nation of the catalyst the features of WO, appear at 810 and 710 cm-'; however, the main Raman line of the precursor at I 810 I P A7-25 200 400 600 800 1000 1200 AVcm-' Fig.6 Raman spectra of fresh and calcined A7 and A10 samples. The plasma lines of the laser are indicated by P 18-9004 720118081X6-900 200 400 600 800 1000 1200 A,/cm-' Fig. 7 Raman spectra of fresh and calcined X6 and I8 samples. The plasma lines of the laser are indicated by P 960 cm-', is still observed. In view of the very high diffusion cross-section of WO, compared to the supported oxoanions, the band of WO, present after calcination at 450°C should not be interpreted as being due to the presence of large amount of WO, resulting from the transformation of the oxotungsten entity.In contrast, the Raman spectrum of the 18-25 sample shows well defined lines at 972 and 891 cm-' as well as bands in the low spectral range which could be assigned to a well defined but weakly interacting polytungstate entity. However, it is not possible to ascertain the exact nature of this oxide species. After calcination of the catalyst at 450 "C the crystal- line oxide is the only species detectable by LRS, in agreement with the XRD results. As the Raman microprobe allows analysis of selected particles, heterogeneity is detected. Thus, after calcination of the sample at 900 "C, the Raman spectrum of some particles shows evidence of the presence of well crystallized WO, (not reported here) whereas the Raman spectrum of other particles shows a broad asymmetric line of vitreous silica beyond 500cm-', as well as the WO, lines.However, the Raman spectrum of cristoballite12 was not observed. As this silica form was evidenced by XRD, this implies that segregation has occurred during calcination at 900°C and, owing to the spatial resolution of the analysis, cristobalite was not detected by Raman spectroscopy. TPR The TPR profiles of the bulk WO, prepared by the gel method are shown in Fig. 8. When the gel was not calcined (W-25), the TPR profile showed two reduction peaks, at 622 and 710 "C. However, when the sample was calcined at 900 "C, the reduction occurred in one step at a higher temperature. The same results were found when mechanical mixtures of bulk oxide and silica were reduced.The TPR profiles of samples A7 and A10 (Fig. 9) show that reduction becomes J. MATER. CHEM., 1994, VOL. 4 440 840 TPC Fig. 8 TPR profiles of the tungsten oxides: (a) fresh gel (W-25) and (b)gel calcined at 900 "C (W-900) Imb2X6-25 11111 11III 440 840 440 84(TI'C Fig. 9 TPR profiles of fresh and calcined samples prepared by mixing tungsten gel and silica gel (A7 and A10); impregnation of silica gel with a solution of tungsten salt (X6)and by impregnation of commer-cial Si02 with tungsten salt (18) more difficult after calcination and that sample A7-900 is practically unreduced. Fig. 9 also shows the reducibility of the X6 samples. For these, it is evident that the hydrogen con- sumption is much lower than for the other samples.In order to correlate the amount of reduction with the different oxidation states of tungsten, some XRD analyses were performed after reduction of the samples at 840°C. The presence of tungsten metal (Wo) is always observed, but the corresponding peaks are weakly detectable on sample A7-900. In the case of bulk tungsten oxide gel (W-25), the XRD spectra taken after stopping the reduction at 650 "C indicates the prcsence of both W02 and W metal. The analysis of the samples A7-900 and X6-900 by CTEM and EPMA indicates the presence of some glassy particles consisting of silica and tungsten, which are smaller in the X6-900 sam2le. Discussion The influence of the method of preparation and the calcination temperature will be discussed separately.Influence of the Method of Preparation The tungsten gel is identified as an amorphous hydrate of tungsten oxide, and this hydrate is clearly identify in the A7 sample. The TPR pattern of the A samples, similar to the fresh unsupported gel, is slightly shifted (by 30 "C) to a higher reduction temperature. This should probably be attributed to the fact that this oxide is trapped in the silica particles. This suggests that, in this preparation, the oxide particles are very small and retain their chemical identity, as would be the case in a mechanical mixture of very fine particles. This explanation is supported by the XPS atomic ratio, which is almost the same as the W: Si atomic ratio. Sample A10 shows the same behaviour.The W 4f XPS doublet of sample X6 is weakly resolved; such broadening has been previously reported for silica- or alumina-supported oxides and assigned to a strong interaction between the oxide and the ~arrier.',~,'~ In agreement with this interpretation, the Raman spectrum indicates that a well dispersed polymeric oxotungstate species is interacting with the silica gel. In contrast, the lowering of the (W/Si)xps intensity ratio of sample I8 may be explained by an inhomogeneous dispersion of a polytungstate species on top of the silica, due to the low level of interaction between the oxotungstate ions and the hydroxy groups of the silica surface. Therefore, this isopoly- tungstate ion transforms easily when the sample is heated into bulk WO,, as will be seen below by LRS. The TPR reduction show two peaks, the higher of which corresponds to the reduction of larger particles of WO,.Effect of Calcination Temperature The presence of tungsten trioxide was ascertained by LRS and XRD in sample A7-450. The TPR behaviour is intermedi- ate between that of a gel and well crystallized WO,. This difference in reducibility may be explained by the incomplete transformation of the interacting gel into WO, and the existence of a new phase characterized in LRS bj a broad line at 980cm-'. Even after calcination of the cxtalyst at 900 "C the tungsten remains homogeneously dispersed in the silica, as shown by XPS. But some modifications of tlie Raman spectrum of the A7-900 sample, with respect to biilk WO,, have been reported in the above paragraph.No crystalline tungsten trioxide is observed by XRD, implying that it is amorphous, microcrystalline or present in undetectable amounts. So in relation with the method of prcparation starting from gel, it is possible to assume that the loss of reducibility should be due to the very small size of the oxide particles which are embedded in the support and intcract with it. It has been demonstrated that small oxide particles are reduced less easily than larger ones." However, the rate of reduction obviously decreased for larger particles. The structure of the solid A10 is similar to A7 described above. Upon calcination of the sample at 450"C, d shift to higher temperature of the TPR peak and a shift and broaden- ing of the XPS lines show that, as in the case of AT-450, the W is distributed in two different species, one interacting with the support, whereas the other, which tends to agglomerate, is identified as crystalline particles of WO,.After c;ilcination of the catalyst at 900"C, due to the higher tungsten content, the heterogeneity increases and some WO, particles should be bigger than those in sample A7-900, in agreement with the decrease in the XPS intensity ratio. Their distribution in the matrix becomes irregular and the surface area decreases drastically. In all these observations, it is striking that the reducibility of these samples calcined at 450°C is controlled by the effect of dispersion in the support and interaction with it.In contrast, the reducibility of the sample calcined at 900 "C is controlled by the crystallite size of the W03 particles. In relation to the X6 samples, calcination increases the surface concentration of tungsten as well as its irr teraction J. MATER. CHEM., 1994, VOL. 4 with the silica, as shown by the broadening of the W 4f lines. These results and the weak reducibility of the samples both before and after calcination could be explained if we assume that the method of preparation had allowed the incorporation of tungsten as isolated oxoanions, giving a higher dispersion and interaction with the silica gel. This preparation allowed the formation in sample X6-900 of small particles of silica and tungsten oxide which seem to have the characteristics of a glass, the nature of which is not defined.Calcination at 450 "C of the impregnated sample allowed the formation of crystalline W03, but the broad XPS W 4f line of sample 18-450 also suggests the existence of oxotungsten species interacting with the carrier. A similar observation was reported by Kerkhof et a1.* On the other hand, cristobalite formation at 9OO0C, with the corresponding decrease in specific surface area, is due to a segregation of the tungsten and silica. In relation to the heterogeneity detected by LRS, these results may explain the increase of the XPS intensity ratio and the narrowing of the W 4f XPS lines.Conclusions The three different methods of introduction of W03 on an SiOz support, and the two different calcination temperatures influence the structure, texture and behaviour of the solids. Mixing two separate gels (tungsten and silica gel) gives well dispersed small WO, particles which, upon calcination at high temperature, give an amorphous phase. Impregnation of ammonium paratungstate on silica gel improves the chemical interaction of isolated tungsten oxoanions with the silica. Upon calcination of the sample, small bulk W03 particles are formed but dispersed oxotungsten species remain on the surface of the support. The interaction of the oxotungstate species with the surface of the support is weak when the solid has been prepared by impregnation of silica with ammonium paratungstate.Upon calcination, both phases seem almost independent. References 1 L. Rodrigo, K. Marcinkowska, A. Adno, P. C. Roberge, S. Kaliaguine, J. M. Stencel, L. E. Makowsky and J. R. Diehl, J. Phys. Chem., 1986,90,2690. 2 Y. Okamoto, T. Imanaka and S. Teranishi, J. Phys. Chem., 1981, 85, 3798. 3 0.Takehiko, A. Masakazu and K. Yutaka, J. Ph.vs. Chem., 1986, 90,4780. 4 P. Biloen and G. T. Pott, J. Catal., 1973,30, 169. 5 L. L. Murrel, D. C. Grenoble, R. T. K. Baker, E. B. Prestidge, S. C. Fung, R. Chianelli and S. P. Cramer, J. CatLil., 1983,79,203. 6 Y. V. Plyuto, J. Stoch, I. V. Babytch and A. A. Chuyko, J. Non Cryst. Solids, 1990, 124,41. 7 F. P. Kerkhof, J. A. Moulijn and A. Heeres, J.Elcctron Spectrosc. Relat. Phenom., 1978, 14,453. 8 F. P. J. M. Kerkhof, J. A. Moulijn, R. Thomas and J. C. Oudejans, in Preparation of Catalysts II. Scient$c Basis for the Preparation of Heterogeneous Catalysts, ed. B. Delmon, P. Grange, P. Jacobs and G. Poncelet, Elsevier, Amsterdam, 1978,p. 77. 9 B. M. Reddy, K. S. Prasad Rao and V. M. Mastikhin, J. Cutul., 1988,113,556. 10 E. Etienne, E. Ponthieu, E. Payen and J. Grimblot, J. Non Cryst. Solids, 1992,147 & 148, 764. 11 A. Bertoluzza, C. Fagnano, M. M. Morelli. M. Gugliemi, G. Scarini and L. Maliavski, J. Raman Spectrosc.. 1988,19,297. 12 J. Etchepare, M. Merian and P. Kaplan, J. Cht)m. Phys., 1978, 68, 1531. 13 M. F. Daniel, B. Desbat, J. C. Lassegues, B. Gerand and M. Figlarz, J. Solid State Chem., 1987,67,235. 14 R. Thouvenot, M. Fournier, R. Franck and C. Rocchiccioli-Deltcheff, Inorg. Chem., 1984, 23, 598. 15 D. Ouafi, F. Mauge, J. C. Lavalley, E. Payen, S. Kasztelan, M. Houari, J. Grimblot and J. P. Bonnelle, Card. Today, 1988, 4, 23. 16 M. Pham Thi and G. Velasco, Solid State Ionics, 1984, 14, 217. 17 J. A. Horsley, I. E. Wachs, J. M. Brown, C;. H. Via and F. D. Hardcastle, J. Phys. Chem., 1987,91,4014. 18 Y. Barbaux, A. R. Elamrani, E. Payen, L. Gengembre, J. P. Bonnelle and B. Grzybowska, Appl. Catal., 1988,44, 117. 19 J. L. Lemaitre, in Characterization of Heterogeneous Catalysts, ed. F. Delannay, Marcel Dekker, New York, 1984, ch. 2. Paper 3/06366F; Received 25th October, 1993