J O U R N A L O F C H E M I S T R Y Materials EVects of preparation parameters on oxygen stoichiometry in Bi4V2O11-d† Isaac Abrahams,*a Alexandra J. Bush,a Franciszek Krok,b GeoVrey E. Hawkes,a Keith D. Sales,a Peter Thorntona and Wladyslav Boguszb aDept. of Chemistry, Queen Mary and Westfield College, University of L ondon,Mile End Road, L ondon, UK E1 4NS bInstitute of Physics, Warsaw University of T echnology, ul.Koszykowa 75, 00–662, Warsaw, Poland The eVects of various preparation parameters on vanadium reduction in Bi4V2O11-d have been investigated using EPR, 51V MAS solid state NMR and UV diVuse-reflectance spectroscopies and also SQUID magnetometry and powder neutron diVraction. The results confirm that a greater amount of vanadium reduction is observed in rapidly quenched samples and that significant oxidation occurs when samples are slow cooled.Evidence for spin–spin dipolar coupling is seen in the EPR patterns while uncoupled V4+ spins contribute to weak paramagnetic behaviour. Band gaps of around 2 eV from the UV data suggest there may be a significant electronic component to low temperature conductivities. The 51V NMR data are not inconsistent with the presence of mainly distorted octahedral and tetrahedral coordinations for vanadium.The relationship of the three major polymorphs in Bi4V2O11 Introduction can be explained with reference to a mean orthorhombic cell Fast oxide ion conducting bismuth oxide based compounds am#5.53, bm#5.61 and cm#15.28 A° ;9 for the c-phase a= have recently become of interest as solid electrolytes for b#am/Ó2, c#cm; for the b-phase a#2am, b#bm, c#cm; for applications in a variety of solid state ionic devices.1,2 The the a-phase a#3am, b#bm, c#cm. The true crystal system of materials typically used in such devices are stabilised zirconias the a-phase however has been the subject of some discussion. which possess good chemical stability but have relatively high It has been found that low levels of cationic impurities present operating temperatures, around 1073–1273 K.Certain bismuth in commercial samples of V2O5 result in an orthorhombic oxide based compounds show high conductivities with low phase for a-Bi4V2O11. However, use of high purity V2O5 yields activation energies and in some cases have shown conductivit- a phase with a small monoclinic distortion which has been ies comparable to stabilised zirconias but at significantly lower crystallographically characterised with a cell a#bm, b#cm, temperatures. c#6am, b=89.756°.10 Bismuth vanadate, Bi4V2O11, is the parent compound of an Bi4V2O11 is very sensitive to preparation conditions.extensive range of substitutional solid solutions which have Preparations by slow cooling and air quenching yield products become known as the BIMEVOX family.3–6 Bi4V2O11 shows which are visually diVerent in colour.These colour changes complex polymorphism but essentially has three main poly- are likely to be due to diVerences in the electronic structure morphs a�b (720 K) and b�c (840 K).7 Substitution of V by caused by diVering amounts of V4+.The eVect of reduction of a host of iso- and alio-valent cations leads to stabilisation of V is to increase the total number of oxide vacancies and hence the highly conducting c-polymorph to room temperature. increase the tetrahedral coordination of V. The incorporation Conductivities in the order of 10-1 S cm-1 have been reported of additional vacancies by vanadium reduction may have at 873 K for the parent compound7 and a number of the important eVects on the ionic conductivity.The true formula substituted BIMEVOXes such as BICOVOX8 and is therefore better written as Bi4V2O11-d. BICUVOX.9 The value of d has previously been investigated.12 Heating The idealised structure of Bi4V2O11 (Fig. 1) consists of samples at 1073 K in air results in a compound with d= alternating layers of [Bi2O2]n2n+ and [VO3.5&0.5]n2n-, where 2.5×10-2, while this increases to 5×10-2 under argon.It has & represents oxide ion vacancies. The [Bi2O2]n2n+ layers have basal edge-shared BiO4 square pyramidal groups with the oxygen atoms forming the basal plane and bismuth in the apical position. The vanadate layer in the idealised structure is made up of VO6 octahedra which corner share in the equatorial plane.This layer is distorted in the real structure of a-Bi4V2O11.10 In our recent defect structure determination of the Co doped material BICOVOX,11 it was found that the vacancies in the vanadate layer are concentrated in the equatorial positions around V. In addition distortion of the apical positions yields a structure in which the majority of sites are in fact distorted tetrahedral.In the structure determination of a-Bi4V2O11 both tetrahedral and octahedral V coordinations were found.10 Fig. 1 Idealised layered structure of Bi4V2O11. Small shaded circles are V, large shaded circles are Bi and unshaded circles are O atoms. † Presented at the RSC Autumn Meeting, 2–5 September 1997, University of Aberdeen, Scotland.Equatorial vacancies are not shown. J. Mater. Chem., 1998, 8(5), 1213–1217 1213been found that samples prepared under a reducing atmosphere Results and Discussion of Ar and 10% H2 yield a compound with one-third of the V Visible colour changes are observed between slow cooled and reduced, i.e. Bi4V2O10.66 which has a crystal structure best rapidly quenched samples. Samples which are furnace cooled described as Bi6V3O16.13 Under suitable conditions complete in air or oxygen show a deep red colour compared to the conversion to VIV is possible as in the structure of Bi4V2O10.14 quenched samples which are brown.The UV diVuse-reflectance However in this structure all the V are found in square spectra for samples prepared using conditions i–iv is shown in pyramidal coordination.Fig. 2. All samples show strong absorption in the blue region In this paper various preparation parameters for Bi4V2O11, but weaker absorption in the red region. That for the air such as slow cooling or air quenching, have been investigated quenched sample (i) has a stronger absorption in the red with respect to their eVects on V reduction.We have used region, thus accounting for its darker brown colour. The results SQUID magnetometry, neutron diVraction, UV diVuse reflecshow that for slow cooled samples (ii–iv) there is a clear tance spectroscopy, EPR and 51V solid state NMR to investiabsorption edge around 600 nm whereas in the air quenched gate the structural consequences of V reduction in Bi4V2O11-d. samples this absorption edge is less well defined.Clearly the results suggest a diVerence in electronic structure between slow Experimental cooled and rapidly quenched samples. We believe that this diVerence is caused by small changes in the oxygen stoichi- Preparation ometry and hence the oxidation state of V. In the slow cooled Bi4V2O11-d was prepared from Bi2O3 (Avocado 99%) and samples clear band edges are visible.For these samples band V2O5 (Aldrich 99.6%) by conventional solid state techniques. gap energies were calculated as 1.99, 2.04, and 1.98 eV for Synthesis was carried out by heating a well ground mixture of samples ii, iii and iv respectively. The air quenched sample (i) appropriate molar quantities of the starting materials at 923 K did not show a clear band edge and we were therefore unable for 6 h in a gold boat and then overnight at 1123 K.In the to calculate a band gap energy for this sample. The band gaps synthesis of samples, four types of preparation and cooling of around 2 eV compare with semiconductors such as CdSe conditions were adopted as follows: (i) preparation in air and and CdS (1.74 and 2.42 eV respectively at 300 K17).This rapid quenching from 1123 K; (ii ) preparation in flowing suggests that these materials probably show significant elecoxygen and exponential slow cooling to room temperature; tronic semiconducting behaviour at lower temperatures and (iii ) preparation in air and exponential slow cooling to room that the low temperature conductivities may have a significant temperature; (iv) preparation in air and linear slow cooling at electronimponent.It should be noted however that a rate of 25 °C h-1. measurement of oxygen transport numbers between 720 and Phase purity was confirmed by X-ray powder diVraction. 1120 K yield a near unity value suggesting that oxide ion conduction predominates at high temperatures.7 Crystallography The nominal +5 oxidation state of V is rarely achieved universally in vanadium oxides, with significant amounts of High resolution neutron diVraction data on samples i and ii V4+ in most commercial samples of V2O5.Therefore one were collected on the HRPD diVractometer at the ISIS facility, expects that in Bi4V2O11 a proportion of the vanadium will be Rutherford-Appleton Laboratory. Data were collected at room in the lower oxidation state V4+ with an electronic configur- temperature in back-scattering mode in the TOF range ation 3d1.The EPR spectra (Fig. 3) confirm the presence of 20–120 ms. The samples were placed in a V-can in the 1 m unpaired electrons in the system. position. Structure refinement was carried out using the In the structure determination of BICOVOX we described Rietveld method.All calculations were performed using the likely coordination for V, viz. distorted tetrahedra and GSAS.15 A starting model for refinement was based on the distorted octahedra.11 The tetrahedral sites arise from equa- idealised orthorhombic mean cell in space group Aba2.16 This torial vacancies in the idealised vanadate layer. The axial approach ignores the weak superlattice reflections but allows oxygens are distorted away from their ideal positions, however for a satisfactory refinement of unit cell contents.the total number of oxygens in the axial position is not less than 2 per V. Therefore all oxide ion vacancies are concentrated Spectroscopy in the equatorial layer. If only tetrahedral and octahedral 51V magic angle spinning (MAS) solid state NMR data were coordinations occur, and in Bi4V2O11 the defect structure again collected at 157.8 MHz on a Bruker AMX-600 spectrometer only shows equatorial vacancies, then the total number of using a 4 mm outer diameter rotor, and a spin rate of 12 kHz.The pulse width was 0.7 ms and 4k points were acquired for each transient, with an acquisition time of 0.02 s and a relaxation delay of 0.5 s.Typically 5000 transients were accumulated for each spectrum. Chemical shifts are reported with the high frequency positive convention and are referenced to external VOCl3 (=0 ppm). UV diVuse-reflectance spectra were collected on a Perkin- Elmer 330 spectrophotometer equipped with a dual channel diVuse reflectance attachment. Relative reflectances of low loaded samples were measured against a white reference.EPR data were collected on a Bruker 200D X-band spectrometer employing 100 kHz modulation, magnetic field markers from an NMR Gaussmeter and an external microwave frequency counter. All measurements were carried out at room temperature. Magnetic measurements Fig. 2 UV diVuse-reflectance spectra for Bi4V2O11-d prepared by (a) SQUID measurements were performed on a Quantum Design quenching in air (sample i ), ( b) exponential slow cooling under a MPMS-7 with a magnetic field of 2000 G.Measurements were dynamic oxygen flow (sample ii), (c) linear slow cooling in air (sample iv) and (d) exponentially slow cooled in air (sample iii ) carried out on samples i, ii and iv. 1214 J. Mater. Chem., 1998, 8(5), 1213–1217than in the air quenched sample.In the EPR patterns for samples that were slow cooled linearly (sample iv) the higher magnetic field signal was not observed. A half field line was seen in all samples which is likely to result from spin–spin coupling of V4+. This signal at low field was found to have a g value of 4.459 in the air quenched sample and this g value did not vary significantly between samples i–iv.The relative intensities of the resonances vary between samples and reflect the spin concentration. Table 2 summarises the g values for samples i–iv and are in good agreement with a previous study where g values of 1.9543 and 4.3449 were observed.19 The general features of solid state 51V MAS NMR spectra of oxovanadium(V) compounds have been described by Crans et al.20 51V is a quadrupolar nucleus (spin I=7/2) and the MAS spectra are a superposition of the sharper central transition (mI=+1/2�mI=-1/2) and the six broad satellite transitions. The central transition appears as a central line flanked by spinning side bands, and the intensity pattern for the powdered samples is dominated by the 51V chemical shift anisotropy.21,22 Second order quadrupolar eVects cause some distortions to the central transition, giving a shift of the central line away from the isotropic chemical shift as well as distortions of the band shape. However, it is expected that the second order quadrupole eVects are minimised in spectra measured at Fig. 3 EPR spectra of Bi4V2O11-d (a) sample i, (b) sample ii, (c) sample the highest magnetic field strengths as obtained here (14.1 T).iii, (d) sample iv The 51V MAS NMR spectra for samples i–iv are shown in Fig. 4. There are two principal centre band resonances ( labelled tetrahedral sites can be calculated as 0.5 per V. This means A and B; position invariant with MAS rate) with one weaker that the tetrahedral5octahedral ratio in the idealised V5+ high frequency spinning side band ( labelled *) from resonance system is 151.Reduction of the V is likely to introduce further A (there is the possibility of a second high frequency side equatorial vacancies and therefore increases the tetra- band). A weak low frequency spinning side band from resonhedral5octahedral ratio. ance A is overlapping with resonance B. Other features in the The powder neutron refinements are summarised in Table 1.spectra are the weak broad signals of the spinning side band As expected sample ii shows a higher oxygen content than manifold due to the partially excited satellite transitions.21 The sample i, indicating a greater degree of oxidation in the oxygen lack of a widespread manifold of spinning side bands for either slow cooled sample.The diVerence in unit cell volume between A or B indicates that these resonances have modest values for the two samples is minimal. This can be explained by consider- the chemical shift anisotropy (200–300 ppm). Resonance B is ing that any increase in cell volume through incorporation of additional oxygen is balanced by a reduction in size of the Table 2 g values calculated from the EPR spectra of samples i–iv vanadium ionic radius in changing from V4+ with an ionic radius of 0.46 A° to V5+ with an ionic radius of 0.355 A° .18 sample half field line g# 2 line From the EPR patterns shown in Fig. 3 it can be seen that the air quenched sample, i, which shows the greatest amount air quenched (sample i) 4.459 1.957 of V reduction, has the strongest signal at a g value of 1.957.oxygen slow cooled 4.485 2.235 exponentially (sample ii) Slow cooling in oxygen (sample ii) appears to shift the position slow cooled in air 4.465 1.960 of this signal to a g value of 2.235 which is accompanied by a exponentially (sample iii) change in line shape. In the case of the two samples slow slow cooled in air linearly 4.487 — cooled exponentially in the furnace (samples ii and iii ) the (sample iv) signal at higher magnetic field, g#2, becomes much weaker Table 1 Refined atomic parameters from the room temperature neutron diVraction profiles of (a) Bi4V2O11-d quenched in air [sample (i)] and (b) Bi4V2O11-d exponentially slow cooled in oxygen [sample (ii )] (estimated standard deviations are given in parentheses) atom x/a y/b z/c occupancy Uiso/A ° 2 (a) sample (i)a Bi(1) 0.4968(8) 0.1688(2) 0.000(-) 1.00(-) 0.0346(9) V(1) 0.000(-) 0.000(-) 0.0508(-) 1.00(-) 0.025(-) O(1) 0.243(3) 0.2502(5) 0.263(2) 1.00(-) 0.0215(9) O(2) 0.335(3) 0.506(3) 0.308(3) 0.55(2) 0.085(3) O(3) -0.064(1) 0.1005(6) 0.046(3) 1.00(-) 0.085(2) (b) sample (ii )b Bi(1) 0.493(1) 0.1690(2) 0.000(-) 1.00(-) 0.0156(9) V(1) 0.000(-) 0.000(-) 0.0508(-) 1.00(-) 0.025(-) O(1) 0.243(3) 0.2458(8) 0.263(3) 1.00(-) 0.007(1) O(2) 0.316(2) 0.511(2) 0.303(3) 0.63(2) 0.056(3) O(3) -0.067(1) 0.0996) 0.051(3) 1.00(-) 0.056(3) aRWP=17.13%, RP=14.37%, REX=2.13%, for 6020 data points and 570 reflections.a=5.5827(2), b=15.2283(6), c=5.5073(2) A ° , V=468.20(5) A ° 3. bRWP=10.36%, RP=8.26%, REX=1.92%, for 6020 data points and 572 reflections.a=5.5840(3), b=15.2218(9), c=5.5080(3) A ° , V=468.16(7) A ° 3. For definition of R-factors see ref. 26. J. Mater. Chem., 1998, 8(5), 1213–1217 1215Fig. 4 Solid state 51V NMR spectra for Bi4V2O11-d, (a) sample i, (b) sample ii, (c) sample iii, (d) sample iv. Resonances for distorted tetrahedral (A), octahedral (B) and square pyramidal (C) vanadium sites are indicated.clearly broader than A and there is evidence of splitting of B vanadium. It is known that in the structure of Bi4V2O10, i.e. the fully reduced system, all the vanadium is five coordinate [Fig. 4( b)–(d)] in some of the spectra. The possibility exists that this apparent splitting is the result of second order in square pyramidal coordination,14 and it is highly likely that in the partially reduced system some five coordinate vanadium quadrupole eVects on the band shape for a single resonance or, more likely, it is the result of two quite similar environments.will be present. These assignments are in contrast to those of Hardcastle Previously collated 51V NMR data have shown that, typically, resonances from tetrahedral vanadium display lower et al.24 who in a study of the composition range 151–6051 Bi2O3–V2O5 assigned a peak at approx.-425 ppm to BiVO4 values for the chemical shift anisotropy than octahedrally coordinated vanadium,23 however these values depend criti- and a peak at -510 ppm to the tetrahedral site in Bi4V2O11. Our diVraction evidence suggests that there are not significant cally upon the symmetry around vanadium.Therefore, the isotropic 51V chemical shift alone is not an absolute indicator amounts of BiVO4 present in the sample and therefore both peaks are due to the main phase Bi4V2O11. Their original of vanadium coordination number. Crystallographic data10 indicate the presence of both four and six coordinate vanadium assignment was based on the assumption that the structure of Bi4V2O11 contained V in entirely tetrahedral coordination.It and the observed uncorrected shift for resonance A is -423 ppm, which is similar to the isotropic shift reported for has since been shown that in the low temperature monoclinic form of a-Bi4V2O11 both tetrahedral and octahedral V sites BiVO4 (tetrahedral vanadium).22 The observed uncorrected shift for resonance B occurs at -510 ppm and compares with are present.10 Susceptibility plots for samples i, ii and iv, derived from the literature values for distorted octahedral oxovanadium of -500 to -536 ppm.22 We therefore assign resonance A to SQUID data, are shown in Fig. 5. Samples i and ii show classic paramagnetic behaviour with low overall magnetic susceptibil- tetrahedral vanadium and B to octahedral vanadium.The chemical shift anisotropy for resonance A is within the range ities. This paramagnetism is attributable to a small number of uncoupled V4+ spins. A small kink in the curves is seen in the derived by Crans et al.,20 for distorted four or five coordinate vanadium sites, while distorted six coordinate sites were 50–70 K region due to residual oxygen in the sample holder. This eVect is normally swamped in concentrated spin systems reported by these authors to have somewhat higher values for the chemical shift anisotropy in the range 500–700 ppm.The but is observed here due to the relatively small magnetisation. The relatively low magnitude of susceptibility observed in relative magnitudes of the chemical shift anisotropies for sites A and B indicate that both these sites are somewhat distorted sample iv suggests that in this material there is a low uncoupled spin concentration.The half field lines in the EPR data which away from regular coordination geometry, which is consistent with the crystallographic evidence. are indicative of spin–spin coupling were observed in all samples including sample iv, where the main high field signal In slow cooled samples a third resonance, C, is observed centred at around -398 ppm. We believe that this third was absent.Therefore it can be concluded that although the SQUID data reveal information on the nature of the uncoupled resonance may be due to low levels of five coordinate 1216 J. Mater. Chem., 1998, 8(5), 1213–1217square pyramidal vanadium appearing in slow cooled samples. The relatively small band gaps of around 2 eV suggest that low temperature conductivities may have a significant electronic component.We gratefully acknowledge the EPSRC for a project studentship to A.J.B. and for use of the ISIS facility at the Rutherford-Appleton Laboratory. We would like to thank Professor P. Day and Dr. S. G. L. Carling at The Royal Institution of Great Britain for use of the SQUID magnetometer, Dr.A. Aliev using the ULIRS Solid State NMR 600 mHz service at QMW and Dr. D. Oduwole at the ULIRS EPR service at QMW. References Fig. 5 Susceptibility plots of Bi4V2O11-d, synthesised by various preparation parameters, derived from the SQUID data 1 N. Q. Minh, J. Am. Ceram. Soc., 1993, 76, 563. 2 J. B. Goodenough, A. Manthiram, M.Paranthaman and Y. S. Zhen, Mater. Sci. Eng. B, 1992, 12, 357. 3 F. Abraham, J. C. Boivin, G. Mairesse and G. Nowogrocki, Solid V4+ spins in this weak system they cannot be easily related to State Ionics, 1990, 40/41, 934. the overall V4+ concentration because of the extent of spin– 4 G. Mairesse, in Fast Ion T ransport in Solids, ed. B. Scrosati, spin coupling. A. Magistris, C.M. Mari and G. Mariotto, Kluwer, Dordrecht, 1993, p.271. An important consequence of significant V reduction is that 5 J. C. Boivin, R. N. Vannier, G. Mairesse, F. Abraham and it imposes a lower solid solution limit in the substituted G. Nowogrocki, ISSI L ett., 1992, 3, 14. BIMEVOXes when compared to the ideal situation of a fully 6 S. Lazure, Ch. Vernochet, R. N. Vannier, G.Nowogrocki and oxidised system. Considering BIMEVOX solid solutions of G. Mairesse, Solid State Ionics, 1996, 90, 117. general formula Bi2V1-xMxO5.5-3x/2 (where M is a divalent 7 F. Abraham, M. F. Debreuille-Gresse, G. Mairesse and metal); in the ideal case where V is fully oxidised to V5+ and G. Nowogrocki, Solid State Ionics, 1988, 28–30, 529. 8 F. Krok, W. Bogusz, W. Jakubowski, J.R. Dygas and assuming that the solid solution mechanism involves creation D. Bangobango, Solid State Ionics, 1994, 70/71, 211. of only equatorial vacancies, as observed in the BICOVOX 9 E. Pernot, M. Anne, M. Bacmann, P. Strobel, J. Fouletier, R. N. structure,11 the limit for solid solution formation will be when Vannier, G. Mairesse, F. Abraham and G. Nowogrocki, Solid State all the possible vacancies are introduced, i.e.when all the V Ionics, 1994, 70/71, 259. sites are tetrahedral. This can be calculated to occur at x= 10 O. Joubert, A. Jouanneaux and M. Ganne, Mater. Res. Bull., 1994, 0.33. Generally for divalent substitution a lower solid solution 29, 175. 11 I. Abrahams, F. Krok and J. A. G. Nelstrop, Solid State Ionics, limit of around x=0.25 is usually observed.3 However, Lee 1996, 90, 57.et al.25 have shown that for M=Co this can be extended to 12 M. Huve, R. N. Vannier, G. Nowogrocki, G. Mairesse and G. Van the maximum of x=0.33. Observed lower solid solution limits Tendeloo, J.Mater. Chem., 1996, 6, 1339. in other systems suggest that further vacancy creation does 13 O. Joubert, A. Jouanneaux and M. Ganne, Nucl.Instrum. Methods not occur and that a possible explanation is that many of Phys. Res. B, 1995, 97, 119. these additional vacancies have already been created through 14 J. Galy, R. Enjalbert, P. Millan and A. Castro, C. R. Acad. Sci. Paris, Ser. II, 1993, 317, 43. V reduction. However, it is unlikely that these lower solid 15 A. C. Larson and R. B. Von Dreele, Los Alamos National solution limits can be explained entirely by this mechanism as Laboratory Report No.LAUR-86–748, 1987. this would imply unreasonably high V4+ concentrations. It 16 International T ables for Crystallography, Volume A, ed. T. Hahn, may well be the case that thermodynamic considerations are IUCR, Kluwer, Dordrecht, 1992. the predominant factor in determining the solid solution limit. 17 C. Kittel, in Introduction to Solid State Physics, John Wiley and Sons, Chichester, 1976, 5th edn., p.210. 18 R. D. Shannon and C. T. Prewitt, Acta Crystallogr., Sect. B, 1969, Conclusions 25, 925. 19 A. Aboukais, F. Delmaire, M. Rigole, R. Hubaut and G. Mairesse, We have shown that oxygen stoichiometry in Bi4V2O11-d is Chem. Mater., 1993, 5, 1819. greatly aVected by preparation parameters. Air quenching of 20 D. C. Crans, R. A. Felty, H. Chen, H. Eckert and N. Das, Inorg. samples preserves a high V4+ concentration, with slow cooling Chem., 1994, 33, 2427. methods in air or oxygen increasing the amount of oxidation. 21 R. H. H. Smits, K. Seshan, J. R. H. Ross and A. P. M. Kentgens, J. Phys. Chem., 1995, 99, 9169. While we have not been able to directly measure the value of 22 H. Eckert and I. E.Wachs, J. Phys. Chem., 1989, 93, 6796. d in this study, our results suggest that there is a significant 23 O. R. Lapina, V. M. Mastikhin, A. A. Shubin, V. N. Krasilinikov amount of V reduction and is reflected in the observed and K. I. Zamaraev, Prog. NMR. Spectrosc., 1992, 24, 457. diVerences in colour, magnetisation, EPR signal strength and 24 F. D. Hardcastle, I. E. Wachs, H. Eckert and D. A. JeVerson, unit cell contents. 51V NMR spectroscopy is not inconsistent J. Solid State Chem., 1991, 90, 194. with the presence of mainly distorted tetrahedral and distorted 25 C. K. Lee, G. S. Lim and A. R.West, J.Mater. Chem., 1994, 4, 1441. 26 H. M. Rietveld, J. Appl. Crystallogr., 1969, 2, 65. octahedral coordination geometry for vanadium in all samples irrespective of preparation conditions with small amounts of Paper 8/01614C; Received 25th February, 1998 J. Mater. Chem., 1998, 8(5), 1213–1217 1217