J. MATER. CHEM., 1991, 1(1), 23-28 Investigation of Defect Ordering in the Perovskite System SrCro,, Feo,903 -,,by Mossbauer Spectroscopy Terence C. Gibb School of Chemistry, The University, Leeds LS2 9JT, UK The oxygen-deficient perovskite system SrCro.lFeo,903- has been studied by Mossbauer spectroscopy and X-ray powder diffraction techniques. The material obtained is strongly dependent upon the conditions of preparation. Partially oxidized materials annealed in air and quenched are simple cubic perovskites with considerable cation and oxygen vacancy disorder. Annealing at 1200°C in vacuo and then cooling step-wise produces an ordered brownmillerite superlattice with the chromium ordered onto octahedral sites. However, rapid cooling gives a microdomain-textured material containing a random distribution of cations; the introduction of chromium stabilizes the high-temperature state of Sr2Fe205 which has never been quenched to room temperature.Annealing in an argon stream at 1200°C gives a partially oxidized crystalline brownmillerite in which the chromium (and extra oxygen) are incorporated in the tetrahedral layers. The magnetic ordering temperatures are correlated with the observed site distribution of the chromium. Keywords: Mossbauer spectroscopy; Defect ordering; Brownmillerite; Microdomain; Perovskite The perovskite SrFe03 is of particular interest because it is one of the few oxide phases to contain iron in the +4 oxidation state.' It readily loses oxygen, ultimately producing the Fe3 + oxide SrFe02,5 (usually referred to as Sr,Fe205), which has been shown to have a grossly oxygen-deficient perovskite- related structure of the brownmillerite with alternate layers of iron in six and four co-ordination to oxygen. Work in this laboratory4 and subsequently has revealed the existence of the mixed-valence phases SrFe02.875 and SrFeO,,,,, both having an ordered arrangement of oxygen vacancies, although there remains considerable confusion over the precise atomic arrangement.Material quenched from high temperatures to room temperature seems to contain an intimate mixture of at least two of these four phases, probably with considerable intergrowth. There is considerable e~idence~-~ to support the existence of a cubic phase at high temperatures over the whole composition range.This phase, which may well have a microdomain texture, is accompanied by the formation of an averaged valence state for the iron at a temperature which rises with increasing Fe3+ content. The cubic form of Sr2Fe205 is believed to be a microdomain-textured brownmill- erite,6-8 although it has proved impossible to preserve this phase by quenching. Work in this laboratory on the brownmillerite Sr,CoFe05 has shown that the cobalt can be incorported either at random or preferentially on the tetrahedral sites, depending on the conditions of preparation.'.' Partially oxidized material, pre- pared by quenching, contained microdomains of the brownmil- lerite with excess oxygen accommodated in the domain walls. SrFe03-, and SrCro~,Fe0.,O3-, were ground together in a ball mill, pressed into a pellet, and initially fired in a platinum crucible at 1200°C for 11 days with two intermediate grindings.Aliquots of this material were then annealed for several days under a variety of conditions, as detailed in the text. Initial characterization in each case was by X-ray powder diffraction recorded with a Philips diffractometer using nickel- filtered Cu-Kcr radiation. Chemical analyses for nominal Cr4+/Fe4+ content were carried out as described previously.' ' Mossbauer data were collected in the temperature range 78 < VK <600 using a 57Co/Rh source matrix held at room temperature; isomer shifts were determined relative to the spectrum of metallic iron. Results and Discussion The initial choice of a 10% substitution of iron by chromium was prompted by the failure of the earlier investigation to observe a substituted brownmillerite, analogous to those observed in the equivalent calcium system at up to 28% substitution.' Furthermore, ordered compounds such as Ca2CrFe05 and Sr,CrFe05 do not appear to exist.Samples were prepared from the initial material by annealing for several days in air (or argon or in uacuo) at a controlled temperature before quenching into liquid nitrogen (from air) or cooling quickly. The chemical analyses and phase analysis by X-ray powder diffraction and Mossbauer spectroscopy are summarized in Table 1. More recently, a detailed study of the system SrCr,Fe, -x02.5+yPreparations in Air (x=O.25, 0.33, 0.50) has been carried out" which revealed a number of perovskite-related phases including a novel 15R- rhombohedra1 phase, the nature of which remains elusive.However, there was no evidence at that time to support the existence of a Cr-substituted brownmillerite phase. This paper reports a study of the composition SrCro.lFeo.903-, which has revealed the existence of a num- ber of unusual brownmillerite-related phases, and the effects of Cr substitution on the SrFeO3_,, system are discussed. Experimental Accurately weighed amounts of spectroscopic grade Fe203, Cr203, and SrC03, with stoichiometric ratios appropriate for The Mossbauer spectra at 290 K of equivalent pairs of materials for SrFe03-y and SrCro.,Fe0~,O3 -y quenched from 1200, 1O00, 800, 600"C, and cooled very slowly in air are illustrated in Fig.1-3. All the chromium-substituted materials gave a very sharp X-ray pattern that could be easily indexed as a simple cubic perovskite pattern, the cell parameter being given in Table 1. The oxygen content increases (y decreases) at lower temperatures, and the cell parameter decreases. The Mossbauer spectra (Fig. 1) for the 1200, lo00 and 800°C quenches comprise very broad magnetic hyperfine patterns that are completely different from those for the parent system shown in Fig. 2. The latter show a superposition of a typical magnetic brownmillerite pattern (from Sr2Fe205) and a cen- 24 J. MATER.CHEM., 1991, VOL. 1 Table 1 Chemical analyses and X-ray phase analysis for samples prepared under a range of experimental conditions SrFeO, -SrCrO. 1Fe0.~03-y conditions y phases" y phases" 1200°C air/quench 0.452 BM-(P) 0.353 P (a =3.932 A)1ooo"C air/quench 0.412 BM-(P) 0.314 P (a=3.934 A)800°C air/quench 0.337 BM-(P) 0.273 P (a=3.923 A)600°C air/quench 0.230 split-P 0.209 P (a=3.913 A)slow-cool in air 0.120 split-P 0.126 P (a= 3.905 A) C1200°C argon --0.422 BM .-1200°C in uucuo --0.500 BM u) (step-cooled) .-1200°C in uacuo --0.497 microdomains (fas t-cooled) E CI "Phases identified as: BM, brownmillerite; P, perovskite; split-P or microdomains (see text). . lllllllllll -12 -0 -4 0 4 8 12 velocity/mm s-' Fig. 2 The Mossbauer spectra at 290 K of samples of SrFeO,-, annealed in air at (a) 1200, (b) 1OOO and (c) 800°C and quenched into liquid nitrogen W Fig.1 The Mossbauer spectra at 290K of samples of SrCro,lFeo,903-yannealed in air at (a) 1200, (b) 1OOO and (c) 800°C \jand quenched into liquid nitrogen tral paramagnetic component from SrFe02., 5. The computed area ratios were used to estimate an oxygen content which agrees within experimental error with the chemical analyses, .-and these materials can best be described as multiphase with -: no signs to suggest the presence of microdomain intergrowths. 3 The chromium-substituted samples appear to be single- -----7, !\i?L'cphase, the broad magnetic patterns being indicative of con-siderable disorder in the oxygen vacancies.There are some indications of fine structure, which is consistent with a vari- ation in the iron co-ordination. There is no conclusive evidence to suggest the presence of any iron in oxidation states higher I I I I I I I than +3, which are required by the stoichiometry. The -4 -2 0 2 4 spectrum at 800°C is partially collapsed because of a lowering of the magnetic ordering temperature, and the 600°C quenches velocity/mm s-' and slow-cooled samples shown in Fig. 3 are paramagnetic Fig.3 The Mossbauer spectra at 290K of samples ofat 290 K. The two iron materials have chemical compositions SrCro,lFeo,903-y and SrFeO,-, annealed in air at 600°C andand STF~O~,~,~, very close to the nominal ones for STF~O~.,~ quenched into liquid nitrogen [(a)10% Cr, and (c)], or slowly cooled respectively. The X-ray patterns (which show superlattice in air [(b) 10% Cr, and (d)] J.MATER. CHEM., 1991, VOL. 1 lines) and the Mossbauer spectra are similar to those described confirming the identification. However, the equi- valent Cr-substituted samples are very different. The Mossbauer spectra show no evidence for the ordered-vacancy phases. There is still some doubt as to the correct analysis of the spectra for STF~O~,,~ and SrFe02.875 because of the difficulty in obtaining high-purity samples; the spectrum of the latter (at the bottom of Fig. 3) shows clear evidence for contamination by the former. However, it is widely accepted that the ideal phases contain 1:l mixtures of the nominal cation states Fe3 +/Fe4+ and Fe3.' +/Fe4', respectively.An analysis of the nominal iron states in the Cr-substituted material is not possible, but it is evident (and confirmed by numerical evaluation) that despite a similar level of oxidation, the mean isomer shift of the iron has shifted significantly to a higher value. This implies a significant reduction of the iron at the expense of chromium which may be presumed to be oxidized to at least the 4 + state. This observation is consistent with the results at higher Cr content.'' It will be clear from the above that the substitution of chromium has suppressed the tendency to form ordered defect phases, and that the products are highly disordered and therefore difficult to study in detail.However, the materials annealed under argon or in U~CUOproved to be much more amenable, and produced some surprising results which are now described. Preparations in vacuo A sample of the Cr-substituted material was annealed in U~CUO at 1200°C for 6 days before annealing at 100°C intervals for 24 h down to 600°C and then cooling to room temperature. The product was a brownmillerite without additional oxygen, and the good quality X-ray pattern was indexed with the orthorhombic cell a =5.665(2)A, b =15.509(5)A, c = 5.523(2)A and a cell volume, r! of 485.2(5) A3, compared with a= 5.672(2) A, b =15.561(5) A, C= 5.525(2)A, V=487.6(5) A3 for the parent Sr2Fe20S. The parameters for Sr2Fe2O5 are in good agreement with published data.2y3 The substituted material shows a reduction in all parameters consistent with the introduction of the smaller Cr3+ ion.The Mossbauer spectrum as a function of temperature is shown in Fig. 4.The pattern at 78 K is clearly that of a brownmillerite, but with a substantial reduction in the inten- sity of the lines from the octahedral sites. The area ratio of the octahedral/tetrahedral sites was computed to be 0.84(9), compared with an ideal figure of 0.80 if all the Cr is on octahedral sites and 1.25 if on tetrahedral sites. Therefore, within experimental error the chromium is effectively ordered onto the octahedral sites. The spectrum at 290 K is particularly interesting because it shows considerable fine structure, which can be interpreted directly in terms of the individual nearest-neighbour environ- ments. Such effects have not been reported previously in a brownmillerite system, although they have been observed in the ~rthoferrites.'~,'~ A magnified portion of a typical data analysis is shown in Fig.5, and shows clear evidence for three octahedral and two tetrahedral sites with different hyperfine parameters. To simplify the analysis it was assumed that the linewidths, r,isomer shifts, 6, and quadrupole perturbations, E, are the same for a given co-ordination to oxygen. The computed parameters and area ratios are shown in table 2, together with the ideal probabilities, calculated assuming that the Cr is exclusively on octahedral sites. It can be seen that there are in fact only five significant combinations for this model, and that the observed areas are in reasonable agree- ment with prediction.Most of the discrepancy is almost certainly due to a difference in the recoilless fractions. Our I l l l l l l l l l l -12 -8 -4 0 4 8 12 velocity/mm s-' Fig,4 The Mossbauer spectra at various temperatures of SrCro,lFeo,902,5annealed in V~CUOat 1200°C and then step-cooled to 600°C. T/K: (a) 600, (b)570, (c) 550, (d)500, (e)400,(f)290, I I I I I I I -10 -9 -a -7 -6 -5 -4 -3 -2 velocity/mm s-' Fig. 5 Part of the Mossbauer spectrum at 290 K of SrCr,,lFeo,902~5 as shown in Fig. 4. The curve-fit shows the existence of three distinct octahedral and two tetrahedral iron sites produced by the ordering of the chromium onto the octahedral sites J.MATER. CHEM., 1991, VOL. 1 Table 2 The Mossbauer data analysis at 290 K for SrCr,., Fe,,, 02,5prepared by step-wise cooling in uucuo octahedral sites' tetrahedral sitesb no. Cr neighbours ideal probability observed area BIT no. Cr neighbours ideal probability observed area BIT 0 0.182 0.175 48.8 0 0.356 0.394 40.8 1 0.182 0.162 46.5 1 0.178 0.199 37.9 2 0.068 0.070 43.9 2 0.022 --3 0.01 1 4 0.001 "6=0.38, E= -0.35 mm s-l, r=0.43 mm s-'; bd=0.16, E= +0.30mm s-', r=0.43 mm s-'. own data on various brownmillerites suggest that the recoilless fraction is usually marginally greater at the tetrahedral site, and it is reasonable to assume that introducing a smaller atom in the octahedral layers may accentuate this difference.The temperature dependence of the spectrum is otherwise as expected for a solid solution system. The collapse seen immediately below the Neel temperature of 575 K is typical, and there is no suggestion of a second phase present. Above 575 K, the spectrum comprises a quadrupole doublet; the superimposed doublets from the two site symmetries are not resolved because of nearly identical quadrupole splittings. In conclusion, this sample represents a single-phase brown- millerite with ordering of the Cr onto the octahedral sites and a Neel temperature of 575 _+ 5 K. A second aliquot of material was annealed in uucuo at 1200°C in the same way, but was cooled quickly in the anticipation that this might preserve a cation-disordered form.The X-ray pattern was completely different, with intense broad lines in the correct place for a simple cubic perovskite. However, there were also many weaker and broader features. Close comparison with the pattern for the ordered brownmill- erite suggested that these features were indeed derived from the superlattice lines of the brownmillerite cell. A possible explanation is that the material contains domains of the brownmillerite lattice, which are small enough to cause aver- aging of the X-ray scattering, i.e. with a mean size of the order of 200A. This conclusion was reinforced by the Mossbauer characterization. The Mossbauer spectrum as a function of temperature is shown in Fig.6. The patterns strongly resemble the expected brownmillerite type, but with additional broadening and unsymmetrical lineshapes. It is difficult to determine the octahedral/tetrahedral area ratio accurately, but various com- putations suggest that they are approximately equal as would be expected for a cation-disordered structure. The temperature dependence is similar to that in Fig. 4, except that the Neel temperature has increased to 600+5 K. Once again, the paramagnetic phase shows only a single doublet without resolution of the two sites. There are no signs of extensive relaxation in the magnetic hyperfine spectra, as seen pre- viously9,' for the microdomains found in Sr2CoFe05 +y and Ca,LaFe,O, +y.One can deduce that the magnetic cation lattice is coherent over much greater distances than the X-ray lattice, which shows an averaging of the brownmillerite struc- ture over distances of the order of 200A. The possibility of magnetic coherence in microdomain materials containing oxygen vacancies and microdomains has been discussed pre- viously.' It would appear that the introduction of Cr into Sr2Fe205 has at last enabled the successful preservation of the high- temperature 'cubic' form at room temperature. The brownmill- erite cell is determined by the order of the oxygen vacancies along a [1lo] axis of the perovskite cell to give an orthorhom- bic supercell in which the cation layers normal to the b axis have their spins aligned along the c axis,2 although the Mossbauer spectrum is insensitive to any rotation of the spins I l l l l l l l l l l -12 -8 -L 0 4 8 12 velocity/mm s-' Fig.6 The Mossbauer spectra at various temperatures of SrCro,lFeo,902,5annealed in uucuo at 1200°C and then rapidly cooled. T/K: (a) 610, (b)600, (c) 580, (d) 500, (e)400,(f) 290, (g) 78 in the a-c plane. Microdomains can be produced by a periodic change in the choice of the [llO] axis for vacancy ordering, This will imply that there must also be a rotation of the spin axis across at least some of the domain boundaries associated with the effective rotation of the crystal axes, and will contrib-ute to the broadening seen in the spectrum. A study by electron diffraction should throw more light on the nature of the microdomain behaviour, and further work in this direction will be pursued.Preparations in Argon It is possible to prepare Sr2Fe205 under an argon atmosphere, and an attempt was made to obtain the equivalent Cr- substituted material in the same way by heating for 5 days at 1200°C under an argon stream before cooling by switching off the furnace. The product was, however, partially oxidized in accord with earlier experience at higher Cr concentrations. Nevertheless, the X-ray data showed a crystalline brownmiller- ite with the lattice parameters a =5.635(2)A, b =15.652(5)A, c =5.532(2)A, V=487.9(5) 81,. The b parameter is significantly larger than that for the in uucuo preparation, or indeed for the parent Sr2Fe20S, and this is convincingly shown by the substantial movement of the (080) reflection to lower angle.J. MATER. CHEM., 1991, VOL. 1 1 1 I 1 - 1 I I 1 I I I I -12 -8 -4 0 4 8 vetocity/rnrn s-' Fig. 7 The Mossbauer spectra at various temperatures of SrCro,lFeo,902,5,8annealed in flowing argon at 1200°C and then rapidly cooled. Note the reversal in the relative intensities of the outer lines compared to Fig.4 caused by the transfer of chromium from the octahedral to the tetrahedral sites layers. T/K: (a) 650, (h)630 (c) 62.5, (d)620, (e) 600, (-0500, (8)400, (h) 290, (i) 78 The Mossbauer data in Fig. 7 are again consistent with a brownmillerite structure, but with a further increase in ordering temperature to 630+ 5 K, and an apparent decrease in the fraction of tetrahedrally co-ordinated iron.There is no evidence to suggest that the extra oxygen is accommodated by producing a second phase as has been found in the parent system. Ignoring for the moment the question of the additional oxygen, various computed fits to the data at 78 and 290K give the fraction of tetrahedral sites to be only 0.43 compared to an ideal probability for complete order of Cr onto the tetrahedral sites of 0.444. Furthermore, there is evidence to suggest that there are at least two distinct types of octahedral site. At 78 K a fraction of ca. 0.38 of the iron occupies sites with a field of 52.4 T and parameters generally consistent with octahedral co-ordination in a brownmillerite, compared to a prediction of 0.356 of the iron with six Fe nearest neighbours.A further 0.20 of the iron has an average field of 53.6T but with a much smaller quadrupole perturbation, and presum- ably represents 0.178 of sites with one Cr neighbour on tetrahedral sites (plus 0.022 with two Cr neighbours). There appears to be less distinction between the 'tetrahedral' sites. There is no evidence for higher oxidation states of iron, and the excess oxygen corresponds approximately to that required for oxidation of chromium to the +4 state (a repeated preparation from new starting material gave y =0.446 and very similar spectra). The excess oxygen can be presumed to occupy the tetra- hedral layers, and be responsible for the increase in the length of the b axis.This is believed to be the first time that evidence has been found for the introduction of a significant excess of oxygen into the brownmillerite structure without forming a textured intergrowth or microdomain phase. The octahedral sites adjoining the Cr4+ sites presumably experience a signifi- cant change in electric field gradient, which is seen in the Mossbauer spectra. It is possible that an added oxygen ion tends to be situated between a pair of the smaller Cr4+ ions in 5-co-ordination, and this would leave the iron co-ordination in these tetrahedral layers unchanged. A similar preparation at 1000°C under argon showed a higher oxygen content, but was clearly a mixed phase with the introduction of a cubic perovskite phase.Effects of Order-Disorder on the Nkel Temperature Further support for these explanations can be gained by a closer examination of the critical temperatures for magnetic ordering. The Neel temperature of Sr,Fe,O, is 700K. The partial replacement of some of the iron by another cation will affect the exchange interactions and modify this temperature, usually depressing it significantly. The exchange interactions are essentially short range between nearest cation neighbours via the oxygen anions, and an increasing dilution of the iron with a diamagnetic cation will effectively begin to isolate some iron atoms completely from the long-range exchange. These effects have been expressed quantitatively by Gilleo,' and demonstrated in the related perovskite system EuFe03, for Co and Cr sub~titution.'~~'~ The Fe-0-Cr interaction is apparently rather weak compared with the Fe-0-Fe inter-action, and to a first approximation can be ignored as if it were a simple diamagnetic replacement.This may reflect the competition between ferromagnetic and antiferromagnetic coupling which is strongly dependent on the bond angle.18 The Sr2Fe205 lattice is more complex because of the two co- ordinations, but a similar treatment can be given to investigate the effects of ordering of the cations on these sites. For an arbitrary site occupation by Cr, the formula can be written as Sr,(Cr,Fe, -.Jo(Cr,Fel with an average of five inter- actions per magnetic cation.Following Gilleo, the probability that an ion of co-ordi- nation n is linked with rn magnetic ions when the occupation probability is x is given by n! -x)" (1)PAM)=rn! (n-rn)!x-(l The probability of linkage to only 0 or 1 Fe ions (which precludes magnetic coupling to the bulk material, assuming that the couplings to Cr are very weak) is E =nxn-l( 1-x) +xn and the proportion of Fe ions contributing to magnetism is only (1 -x) (1-E). The proportion of octahedral sites mag- netically active is thus No=(1-~)(1-6x5 +5x6) and the proportion of tetrahedral sites magnetically active is NT =(1-y)(1-4y3+3Y4) The total of 10 exchange pathways comprise four oct-O-oct, four oct-O--tet, and two tet-O-tet interactions.Thus, the number of active interactions is 4N02 +4NoNT+2NT2. The Neel temperature as a function of composition is thus given by 4N; +4N0NT+2N-f700_-TN = NO +NT 5 (2) If all the Cr cations are on octahedral sites (x=O.2, y=O) then TN=603 K, if they are completely random (x =0.1, y = 0.1) then TN=629 K, and if they are all on tetrahedral sites then TN=655 K. Thus simple theory predicts a variation of Neel temperature of some 50°C with change in site occupation. It is significant to note that the predicted magnitude and variation in the Nee1 temperature is in good agreement with the observations made on the three phases. This is not unreasonable since the magnetic exchange interactions are short range and dominated by the Fe-0-Fe linkages, making other factors such as coherent domain boundaries and Cr4+ defect clusters comparatively ineffective. Conclusions The primary result of this work is the observation that chromium can be incorporated into the Sr,Fe,O, lattice in three very different ways according to the experimental con- ditions.Annealing in mcuo at ca. 600°C to prevent oxidation and facilitate cation ordering produces a brownmillerite with chromium concentrated onto the octahedral sites. A similar preparation with rapid cooling from 1200°C results in a random site occupation and the stabilization of a high-temperature 'cubic' phase containing microdomains of brown-millerite with a coherent cation interface. Annealing in argon at 1200°C to allow a limited degree of oxidation introduces chromium and presumably excess oxygen into the tetrahedral layers and causes elongation of the b axis.Preparations allowing a higher degree of oxidation produce a cubic perovsk- ite with considerable vacancy and cation disorder, and the introduction of chromium seems to suppress the formation of the ordered vacancy phases which are known in the parent SrFeO, -y system. I am grateful to Mr. A. Hedley for chemical analyses, and to SERC for financial support. J. MATER. CHEM., 1991, VOL. 1 References 1 P. K. Gallagher, J. B. MacChesney and D. N. E. Buchanan, J. Chem. Phys., 1964, 41, 2429. 2 C. Greaves, A. J. Jacobson, B. C. Tofield and B. E. F. Fender, Acta Crystallogr., Sect. B, 1975, 31, 641.3 M. Harder and Hk. Muller-Buschbaum, Z. Anorg. Allg. Chem., 1980,464, 169. 4 T. C. Gibb, J. Chem. SOC., Dalton Trans., 1985, 1455. 5 Y. Takeda, K. Kanno, T. Takada, 0. Yamamoto, M. Takano, N. Nakayama and Y. Bando, J. Solid State Chem., 1986,63,237. 6 M. Takano, T. Okita, N. Nakayama, Y. Bando, Y. Takeda, 0. Yamamoto and J. B. Goodenough, J. Solid State Chem., 1988, 73, 140. 7 L. Fournes, Y. Potin, J. C. Grenier, G. Demazeau and M. Pouchard, Solid State Commun., 1987, 62, 239. 8 J. C. Grenier, N. Ea, M. Pouchard and P. Hagenmuller, J. Solid State Chem., 1985, 58, 243. 9 P. D. Battle, T. C. Gibb and S. Nixon, J. Solid State Chem., 1988, 73, 330. 10 P. D. Battle, T. C. Gibb and P. Lightfoot, J. Solid State Chem., 1988, 76, 334. 11 T. C. Gibb and M. Matsuo, J. Solid State Chem., 1990, 86, 164. 12 T. C. Gibb and M. Matsuo, J. Solid State Chem., 1990, in the press. 13 T. C. Gibb, J. Chem. SOC., Dalton Trans., 1983, 873. 14 T. C. Gibb, J. Chem. SOC.,Dalton Trans., 1983, 2031. 15 T. C. Gibb, J. Solid State Chem., 1988, 74, 176. 16 P. D. Battle, T. C. Gibb and S. Nixon, J. Solid State Chem., 1989, 86, 86. 17 M. A. Gilleo, J. Phys. Chem. Solids, 1960, 13, 33. 18 T. C. Gibb, J. Chem. SOC., Dalton Trans., 1984, 667. Paper 0/02644A; Received 13th June, 1990