Preparation of High Surface Area Reduced Molybdenum Oxide Catalysts BY ROBERT BURCH* Department of Chemistry, The University, Whiteknights, Reading RG6 2AD Received 20th March, 1978 High surface area molybdenum oxides have been prepared by the thermal decomposition and reduction of molybdenum (VI) oxalate. It has been found that the initial trioxide formed is highly oxygen deficient and that the composition depends on the method of preparation, varying from for an oxide prepared in vacuum. The rate of reduction was found to be dependent on: the method used to decompose the oxalate ; the temperature of the decomposition ; the partial pressure of hydrogen ; and the partial pressure of water vapour. Vacuum-prepared trioxides are reduced in a single stage to MOO^-^. Nitrogen-prepared trioxides reduce first to Mo401 or MozOs depending respectively on whether " wet " or '' dry " conditions are used.In most cases the reduction-time curves are essentially sigmoidal in character. Possible reduction mechanisms and rate-determining steps are discussed. for an oxide prepared by decomposing the oxalate in nitrogen, to ~ ~~ ~~ The platinum metals have been used for many years as catalysts for reforming and isomerization reactions of hydrocarbons. Recent attempts to improve the existing platinum catalysts have centred on multimetallic catalysts of which the platinum- rhenium reforming catalysts are perhaps the best known and most important. Both platinum and rhenium, however, are rare and expensive metals, so there are good reasons for trying to find alternative materials which could be used to catalyse reforming and isomerization reactions.Recently published work on clean metal films indicates that the group 6 transition metals have high activity for catalysing some hydrocarbon reactions. Supported oxides of these metals are difficult if not impossible to reduce to the metallic state,2 and unsupported oxides, although they can be reduced, usually have low surface areas. Consequently, little is known about the catalytic activity of these metals in a highly dispersed form. The reduction of low surface area molybdenum trioxide has been investigated previously by a number of workers 3-9 but no results have been published on the reduction of high surface area oxides. Low surface area oxides reduce very slowly, and molybdenum metal can only be prepared in reasonable times by using high temperatures (typically > 900 K) which, of course, produce a low surface area material. The thermal decomposition and reduction of molybdenum (VI) oxalate produces high surface area oxides.lo* These high surface area unsupported molybdenum metalfmolybdenum oxide materials have been shown previously l2 to have high activity and selectivity for the isomerization of straight chain hydrocarbons to the corresponding branched chain isomers.The origin of the catalytic activity in these materials and the mechanism of the isomerization reaction are currently being investigated. In this paper we describe the preparation of the high surface area reduced oxides and show how the kinetics of reduction are affected by the experimental conditions.2982R . BURCH 2983 Subsequent papers will describe the physical structure of the solids,13 and how this is created during the decomposition and reduction reactions, and the variation in cat- alytic activity with the physical and chemical properties of the materials. EXPERIMENTAL MATERIALS Molybdenum (IV) oxalate (Climax Molybdenum), nominal composition H2(M003(C204)(H20)) H20, was used as a starting material. This was crushed and sieved, and the powder having particle sizes between 150 and 250 pm was used in all the subsequent experiments. Hydrogen and nitrogen (B.O.C. 99.999 % purity) were passed through a liquid nitrogen trap before use. No further purification appeared to be necessary judging by the fact that the reduced samples were invariably extremely pyrophoric.EQUIPMENT Most of the thermal decomposition experiments were performed in situ in a Cahn vacuum Samples varying in weight from 20 to 100mg were loaded into a platinum crucible Occasional batches of trioxide were prepared by decomposing about 10 g of the oxalate micro balance. (diameter 8 mm, depth 5 mm) suspended by a fine nichrome wire from the balance arm. in a tube furnace either under vacuum or under static nitrogen. RESULTS Several different experimental parameters have an effect on the rate of reduction of molybdenum trioxide. These are : (i) the method used to decompose the molyb- denum oxalate; (ii) the temperature of reduction; (iii) the partial pressure of hydrogen ; (iv) the partial pressure of water.The effect of these various parameters is summarised in fig. 1. METHOD USED TO DECOMPOSE THE OXALATE Decomposition at 573K in nitrogen produces a small degree of oxygen de- ficiency MOO^.^^), whereas decomposition at 573 K under vacuum gives a compound of nominal composition MOO^.^^. At 673 K, decomposition in nitrogen produces a compound of composition MOO^.^^. These different " trioxides " also have important structural differences from each other. Decomposition at 573 K in nitrogen (or in hydrogen) gives a mainly meso- porous material with a B.E.T. surface area of about 100 m2 g-l. The oxide pre- pared under vacuum contains mainly micropores, and has a B.E.T. surface area of 40m2g-l. It should be noted that, because of the microporosity of both the nitrogen (and vacuum) prepared oxides, the B.E.T.surface areas must be regarded as upper estimates of the true surface area. This will be particularly true for the vacuum-prepared oxide, for which the true surface area may be perhaps 20 % smaller than the B.E.T. surface area. In contrast to the porous trioxide prepared at 573 K, decomposition at 673 K gives an essentially nonporous oxide with a small surface area (10 m2 g-l). The difference in surface area between the nitrogen (and vacuum) prepared tri- oxides might be expected to result in a similar difference in reactivity during reduction with hydrogen. In fact, as can be seen from fig. 2(a), exactly the reverse trend is observed. The vacuum-prepared oxide is significantly more reactive than the nitrogen- prepared material even allowing for the fact that the compositions of the starting2984 MOLYBDENUM OXIDE CATALYSTS 10 m2g-1 FIG.1.-Products formed from the decomposition and reduction of molybdenum oxalate under different experimental conditions. OF---- time/h FIG. 2 . 4 2 ) Reduction of vacuum-prepared (curve A) and nitrogen-prepared (curve B) molybdenum trioxide at 573 K and lo5 Pa hydrogen. (b) Reduction of nitrogen-prepared trioxides at 673 K (curve C) and 573 K (curve D) and lo5 Pa hydrogen. (c) Reduction of molybdenum trioxide at 573 K at various hydrogen pressures : E, 91.3 ; F, 70.7 ; G, 49.4 ; H, 23.2 kPa. (d) Reduction of molybdenum trioxide at 673 K at various hydrogen pressures : J, 100.2 ; K, 71.4 ; L, 48.4 ; M, 17.7 kPa. (e) Reduction of molybdenum trioxide in dry conditions at 573 K and lo5 Pa hydrogen.ct is the extent of reduction to molybdenum dioxide.R . BURCH 2985 materials are different. Thus, the time required for reduction from 20 to 100 % to occur is 3.2 h for the nitrogen-prepared oxide and 2.2 h for the vacuum-prepared oxide. Clearly the reactivity is not related only to the surface areas of the samples. The vacuum-prepared oxide must contain a greater concentration of active sites at which nucleation of the product phase can occur. This may well be associated with the microporosity of these materials and with the high degree of oxygen deficiency in the initial trioxide (MOO,. 2). The large oxygen deficiency in the vacuum-prepared trioxide does not appear to arise from any kinetic effects during the thermal de- compositions of the oxalate because, in separate experiments in which a nitrogen- decomposed oxide is subsequently evacuated at 573 K for several hours, the maximum oxygen deficiency corresponds only to MOO^.^^.A very high degree of oxygen deficiency appears only to be possible if it is incorporated into the oxide during the decomposition of the oxalate. EFFECT OF CHANGING THE REDUCTION TEMPERATURE Fundamental changes in the physical structures of the samples caused by heating at higher temperatures prevented the evaluation of the activation energy for the reduction reaction. For example, it was observed that on heating a nitrogen- prepared oxide sample from 573 to 673 K the surface area decreased from 100 to about 10 m2 g-l, and the porosity of the oxide was completely eliminated.The consequences of these structural transformations in terms of the reactivity during reduction are illustrated in fig. 2(b), which compares the rate of reduction for two samples taken from the same batch of nitrogen-decomposed oxalate. Although the oxide at 673 K has a much higher oxygen deficiency the rate of reduction at 673 K is only marginally faster than at 573 K. In fact, if we compare the reduction curves for the two oxides which have similar oxygen deficiencies, that is curve A in fig. 2(a) and curve C in fig. 2(b), we observe that the reduction is faster at 573 than at 673 K. Although these two samples have approximately the same initial chemical com- position they have very different physical structures. The 573 K oxide is micro- porous whereas the 673 K oxide is nonporous. We presume that this indicates that the oxygen deficiency is accommodated in different ways in the two oxides.In the 673 K oxide the deficiency may be contained almost exclusively by shear plane formation MOO^.^^ corresponds to the shear phase Mo,02,), whereas in the 573 K oxide there may be a high concentration of oxygen vacancies at or close to the surface of the micropores. This would suggest that the rate of reduction is controlled by a surface process and is therefore dependent on the number of nucleation sites. If oxygen diffusion was the rate controlling step we would expect to find a much higher rate of reduction at the higher temperature. Other evidence for a surface-controlled reaction will emerge later.EFFECT OF HYDROGEN PRESSURE A series of experiments was performed on fresh samples taken from a batch of nitrogen-decomposed oxalate, in order to determine the effect on the rate of reduction of varying the hydrogen partial pressure. The rate of reduction both at 573 K [fig. 2(c)] and at 673 K [fig. 2(4] increases linearly with the hydrogen pressure. We shall discuss later the implications of these results in terms of reaction mechanisms.2986 MOLYBDENUM OXIDE CATALYSTS EFFECT OF WATER VAPOUR An experiment was performed in which the sample pan in the microbalance was surrounded by about 20 g of previously outgassed high surface area alumina. The purpose of this experiment was to see whether the rate of reduction would be affected by removing, as efficiently as possible, all the water molecules in the vicinity of the sample.The reduction curve obtained in this experiment is shown in fig. 2(e). Instead of the normal sigmoid curve obtained with earlier samples we observe a double sigmoid curve. Furthermore, at point X, when the reduction had proceeded to MOO^.^, the microbalance was evacuated with the sample at 573 K. Hydrogen was reintroduced, and the rate of reduction was initially much faster than previously. The evacuation treatment may have created new active sites on the surface of the reduced oxide, or it may have removed adsorbed water which was inhibiting the reduction reaction. The physical structure of the reduced oxide prepared under dry conditions is also quite different from the structure of any of the other reduced materials in that it contains a large concentration of micropores.The presence of water during reduc- tion seems to have three effects. It affects the shape of the reduction curve, and presumably the mechanism of the reduction reaction, it affects the pore size distribu- tion of the reduced product, and it appears to inhibit the reduction reaction. DISCUSSION OXYGEN DEFICIENCY I N MO03-x The first unusual aspect about this work which requires some explanation is the high degree of oxygen deficiency which is observed in all the trioxide starting materials, but particularly in the case of the vacuum-decomposed oxalate samples. For low surface area molybdenum trioxide samples which contain isolated oxygen vacancies the limiting composition is about Higher levels of oxygen deficiency are accommodated by the elimination of vacancies with the creation of shear ~1anes.l~ Various shear compounds of molybdenum oxides are known (Mo401 M O ~ O ~ ~ , Mo9OZ6 etc.) but none of these would be expected to form merely by heating moly- bdenum trioxide under vacuum at 573 K.A few disordered shear planes might be created under these conditions but the overall composition would not fall below, say, M002.g9. On thermodynamic grounds, no lower oxide should be formed in the absence of a reducing agent. One possible reason for the high degree of oxygen deficiency of our samples could be the reducing action of hydrocarbon molecules originating from the grease used on the ground joints of the microbalance.We do not think that this is correct for a number of reasons. First, the maximum oxygen deficiency is observed on samples which have been heated for only a few minutes at 573 K. Other samples have been heated for longer periods and yet have not been reduced to the same extent. Second, the degree of oxygen deficiency appears to be determined mainly during the decomposition reaction and not as a result of outgassing at 573 K. Third, in all cases the weight of the oxygen-deficient samples becomes constant after only a few minutes at 573 K, and then remains constant for several hours at least. If reduction by hydrocarbon vapours was important we would not expect a reaction which was initially very rapid to suddenly stop completely. An alternative, and we think a more acceptable, explanation of the high degree of oxygen deficiency is that this arises because of the high surface area and porosity of our materials.We consider that the oxygen deficiency i s accommodated partlyK. BURCH 2987 by the formation of shear planes and partly by the creation of oxygen vacancies in or near the surface of the oxide particles. The evidence in support of this proposal is twofold. First, we observe the highest oxygen deficiency for the vacuum-prepared trioxide which in addition to being very microporous is also the most reactive during reduction. Second, even if we assume that for the compound (surface area 40m2 g-') all the oxygen vacancies occur in the surface this would still be reasonable because of the large surface area of the oxide.It would be equivalent in terms of the overall composition of having a composition of M002.995 for an oxide with a surface area of 1 m2 gl. Further support for the concept of having a large number of oxygen vacancies in the surface of the trioxide is found in the published l i t e r a t ~ r e . l ~ - ~ ~ Both Haber et aZ.15 and Cimino et a1.16 have used X-ray photoelectron spectroscopy to study the surface of molybdenum trioxide after very mild reduction treatments. The con- clusions of both sets of workers are essentially the same. They have shown that in the very early stage of the reduction lattice oxygen ions close to the surface are removed to create isolated oxygen vacancies, with the electrons being localised on an adjacent Mo6+ ion forming Mo4+. When the concentration of oxygen vacancies exceeds an unspecified critical value, strings of edge-sharing MOO octahedra are created from the original corner-sharing octahedra present in stoichiometric MOO,.(This is the precursor state for the development of the shear compounds which are characteristic of bulk MOO,.) The essential feature of this published work is that it shows that the surface of molybdenum oxide need not necessarily have the same chemical or physical structure as the bulk oxide. Even a very mild treatment can create in a few minutes a surface reduction which would require many hours to produce throughout the bulk oxide. This is further substantiated by the observation l5 that a thin layer of molybdenum metal is formed on low surface area Moo3 under conditions of temperature (823 K), hydrogen pressure (5 x Pa) and time (1 h) where the rate of reduction is so slow that no bulk metal would be formed.We conclude from this that although only a very small concentration of oxygen vacancies can be accommodated in the bulk of MOO,, a much higher concentration of oxygen vacancies can be present on the surface of an oxide particle. If, as in our case, the oxides have large surface areas and provided that the oxygen deficiency is built in during the genesis of the trioxide, then large oxygen deficiencies are possible. We may contrast the compositions of the nitrogen-prepared trioxide MOO^.^,) and the vacuum-prepared trioxide MOO^.,^) and speculate that MOO^.^^ cor- responds to the limiting composition due to shear plane formation in a high surface area oxide, and that therefore the lower oxygen content of reflects a higher concentration of oxygen vacancies at the surface.This, as we shall see presently, should result in the two types of oxide having different reactivities. SHAPE OF REDUCTION AGAINST TIME CURVES There are several features of the reduction against time curves shown in fig. 2 which require discussion. Curve A has a smooth sigmoidal shape which is charac- teristic of a reaction in which the rate is controlled either by the nucleation and growth of the product phase, or by an autocatalytic reaction. Curve B has a discontinuity at a composition in the vicinity of MOO^.,^, which would correspond to the shear phase Mo4011, and it may be that on reduction MOO^.^, goes first to Mo4OI1 and then to Moo2-,.We note also that the reduction does not stop at MOO, but con- tinues until a composition of about MOO,., is reached, at which point a much slower2988 MOLYBDENUM OXIDE CATALYSTS linear reduction is obtained. Reduction towards bulk molybdenum metal does continue, nevertheless, even at these low temperatures. Curve A does not show a discontinuity at MOO^.,^. This may simply be because it is difficult to observe a discontinuity so close to the start of the reduction of On the other hand it is also possible that the reaction does not pass through Mo,Oll because of the different structure of the starting material. An unusual reduction against time curve is obtained [fig. 2(e)] when the reaction is performed at 573 K in dry conditions.The reduction seems to proceed through a highly oxygen deficient Mo205 phase as an intermediate compound rather than Mo,O, , as before. We appear to have observed three types of reduction process; Moo2.,, going directly to MOO,-, in a single step; MOO,.,, going first to Mo40,, and then to Moo2-, under wet conditions; MOO^.^^ going to Mo205 and then to MOO,-, in dry conditions. The differences between the reduction of Mo02.B2 and M002.g8 axe perhaps not surprising bearing in mind the different structures, surface areas, and chemical compositions of the starting materials. However, the differences between the wet and dry reductions of MOO^.^^ are certainly unexpected. The product of the dry reduction is much more microporous than the product of the wet reduction. This indicates that water vapour is rapidly removed from the oxide surface. Possibly the efficient removal of water vapour from the surface stimulates the oxide to lose further oxygen and thus by-pass Mo4OI1.Reduction at 673 K (curves C and J) appears to give a parabolic curve, but we suspect that this is only because the initial acceleratory step is very short and not easily detected. Thus, when reduction at 673 K is performed at lower partial pres- sures of hydrogen (curves L and M) we again observe a smooth sigmoidal reduction curve. In all cases the reduction proceeds in a single step from MOO^.,^ to MOO,-,. A further interesting observation in the reduction experiments at 673 K is that at low hydrogen partial pressures where the rate of reaction is slow the final product (taken to be the composition at which the reduction curve almost levels off) is much more highly reduced than are any of the other products obtained at higher hydrogen pressures.This is a reproducible effect. We have here the unusual situation that an initially slower reduction reaction can eventually result in a more highly reduced material. We do not have any certain explanation of this effect, although it is possible that, because reduction is slow, water vapour produced as a product is more readily desorbed from the surface, and therefore causes less damage to the active surface of the reduced product so that it in turn becomes reducible. MECHANISM OF REDUCTION It has been pointed out earlier that there are significant differences in the physical structures of the three types of trioxide which we have been investigating.We must be cautious, therefore, in analysing the kinetic data in terms of a single reaction mechanism since the rate-determining step may change from one structure to another. Bearing in mind this reservation it is at least possible to produce a plausible interpreta- tion of the kinetic data based on a common rate-determining step. However, before discussing our results further it will be useful to summarise what is known * at the atomic level about the mechanism of the reduction of molybdenum trioxide. In the first stages of reduction a high concentration of oxygen ion vacancies are created in the surface of the oxide. These vacancies diffuse relatively slowly into the bulk at the temperatures used in the reduction experiments.X-Ray photoelectron spectroscopy shows that Mo4+ (but not Mo5+) ions are created at the same time asR. BURCH 2989 oxygen vacancies are produced. Clusters of edge-sharing octahedra containing Mo4+ ions are formed, and some metal-metal bonding between Mo4+ ions is indicated by the appearance of Mo2+ signals in the electronic spectrum, and by the nature of the MOO, structure which places alternate pairs of metal atoms close together. Further surface reduction to metallic molybdenum occurs relatively easily. The large concentration of shear planes in Moo3-,, is taken by Haber to indicate that the creation of the MOO, lattice, which also contains edge-sharing octahedra, should be facile. He therefore proposes that nucleation of MOO, will be rapid, and the rate-determining step is considered to be the dissociative chemisorption of hydrogen.The results of our work described earlier which are pertinent to a discussion of the reduction reaction are; the observation of a sigmoidal reduction against time curve ; the dependence of the rate of reduction on the hydrogen partial pressure both at 573 and at 673 K ; the high activity of a highly oxygen deficient vacuum-prepared oxide ; the inhibition by water vapour ; and the effect of ageing on the reactivity of the trioxides. All these factors point to a surface process being rate-determining, rather than say oxygen diffusion. Nucleation and growth of a product phase would be the most usual interpretation, but the dissociative adsorption of hydrogen coupled with an autocatalytic reaction would also account for the observed results.On the assumption that nucleation is rate-determining we can explain the vari- ations between the various samples as follows. We suggest that for the vacuum- prepared trioxide there are a large number of oxygen vacancies at the surface, which can migrate and combine to germinate a nucleus of the reduced product. Rather fewer oxygen vacancies exist in nitrogen-prepared samples, and even fewer survive after raising the temperature from 573 to 673 K. The overall result is that reduction of a vacuum-prepared trioxide at 573 K is fast because it contains a large number of nuclei, reduction of a nitrogen-prepared trioxide at 573 K is slower because it contains fewer nuclei ; and reduction of the same trioxide at 673 K is slow because it contains very few nuclei, and the higher inherent activity of each nucleus only just compensates for the much smaller number of nuclei.On the other hand, if we assume that the chemisorption of hydrogen is rate- controlling we would expect a linear reduction against time curve. However, the observed sigmoidal curve could be produced by the autocatalytic effect of small amounts of low valent, or even metallic molybdenum, formed on the surface during the earliest stages of reduction. It is certainly possible that low valent molybdenum will have the ability to activate hydrogen. We do not think that it is possible, from the information available, to distinguish between nucleation and hydrogen adsorption as possible rate-determining steps.We are grateful to the S.R.C. for an equipment grant. We are grateful to the Climax Molybdenum Company, who first suggested this research, for their continuing support. We thank Mr. N. B. Mason for his assistance with some of the early experiments. J. F. Taylor and J. K. A. Clarke, 2. phys. Chem., 1976,103,216. J . R. Anderson, Structure of MetaZlic Catalysts (Academic Press, London, New York, 1975). J. von Destion-Forstmann, Canad. Metallurg. Quart., 1965, 4, 1. Ph. A. Batist, C. J. Kapteijns, B. C. Lippens and G. C. A. Schuit, J. CcztczZysis, 1967, 7, 33. Kh. Vasilev and T. Pencheva, Khim. Ind. Sofia, 1970,18,202. M. J. Kennedy and S. C. Bevan, Proc. 1st Int. Conf. Molybdenum, Reading, 1973, ed. P. C. H. Mitchell, p. 11.2990 MOLYBDENUM OXIDE CATALYSTS 0. Bertrand and L. C. Dufour, Compt. rend. C, 1974,278,315. Chemical Systems, ed. P. Barnet (Elsevier, Amsterdam, 1975), p. 696. T. Fransen, P. C. van Berge and P. Mars, React. Kinetics Catalysis Letters, 1976, 5, 445. lo G. A. Tsigdinos, C. J. Hallada and R. W. McConnell, U.S. Patent 3,912,660 (to Amax:Inc.), October 14, 1975 ; German Patent 2,451,778 (to Amax Inc.), May 28, 1976. l1 G. A. Tsigdinos and W. W. Swanson, Ind. andEng. Chem., to be published. I2 R. Burch and P. C. H. Mitchell, J. Less-Common Metals, 1977, 54, 363. l 3 R. Burch, J.C.S. Faraday I, 1978,74,2991. l4 L. A. Bursill, Proc. Roy. SOC. A, 1969, 311, 267. l5 J. Haber and J. Janus, in Reaction Kinetics in Heterogeneous Chemical Systems, ed. P. Buret (Elsevier, Amsterdam, 1975), p. 737 ; J. Haber, W. Marczewski, J. Stoch and L. Ungier, Ber. Bunsenges. Phys. Chem., 1975, 79, 970 ; J. Haber, Proc. 2nd Int. Con$ Molybdenum, Oxford 1976, ed. P. C. H. Mitchell and A. Seaman, p. 119 ; J. Haber, A. Koslowska and J. Sloczynski, in Reactivity of Solids, ed. J . Wood, 0. Lindquist, C. Haelgesson and N.-G. Vannerberg (Plenum Press, London and New York, 1977), p. 331. * A. Castellan, J. C. J. Bart, A. Bossi and N. Giordano, Reaction Kinetics in Heterogeneous l6 A. Cimino and B. A. de Angelis, J. Catalysis, 1975, 36, 11. l7 J. Grimblot and J. P. Bonnelle, Compt. rend. C, 1976, 282, 399. (PAPER 8/520)