Structural Characterisation 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 reduced molybdenum oxides have been prepared by the thermal decomposition and reduction of molybdenum (w) oxalate. Nitrogen physisorption isotherms, analysed by the a,-method, have been used to determine the physical structure of the oxides. The effect of tempera- ture, water vapour pressure and the degree of reduction, on the physical structure of the oxides has been studied. It is found that reduction creates initially slit-shaped micropores, which on further reduction, or in the presence of water vapour, broaden into mesopores. The changes in structure which accompany the passivation or the annealing of reduced oxides are described. High catalytic activity for the hydrogenation, dehydrocyclisation, isomerization, etc., of organic compounds is restricted mainly to the platinum group metals. Platinum itself is a very effective catalyst for a wide range of hydrocarbon reactions, but it is expensive so there is considerable incentive to search for alternative catalysts.There are two approaches to this problem. One is to prepare more efficient, more stable platinum alloy catalysts. Another is to consider whether other transition metals might have useful catalytic activity. Binary compounds of chromium, molybdenum and tungsten (oxides and sulphides) have moderate activity for catalysing hydrocarbon reactions, but very little is known about the catalytic properties of the corresponding highly dispersed metals.This is mainly because of the difficulty in preparing dispersed high surface area materials. In the case of supported catalysts there is a strong interaction between the metal ions and the support (silica or alumina) which makes it extremely difficult to reduce the supported metal ions to the metallic state.l It may be possible to overcome these problems by using different supports, for example graphite, or by deactivating the oxide supports before impregnation with solutions of the metal ions. Before embarking on such research it would seem appropriate to determine whether or not these transition metals have any useful catalytic properties which could be exploited by using new or modified supports.Published work on clean metal films indicates that a number of the early transition metals have high activity for catalysing hydro- carbon reactions. In order to determine whether this high activity could be generated under less stringent experimental conditions it was decided to investigate the prepara- tion of high surface area, unsupported molybdenum metal/molybdenum oxide catalysts, prepared by the thermal decomposition and reduction of molybdenum(v1) ~ x a l a t e . ~ ~ These materials have been shown previously to have high activity and selectivity for the isomerization of n-hexane to branched chain isomers. The origin of the catalytic activity and the mechanism of the isomerization reaction are currently being investigated. The kinetics and mechanism of the thermal decomposition and reduction of molybdenum(vI) oxalate have been described elsewhere.6 We describe in this paper the way in which the pore structure and the surface area of the reduced molbydenum oxides evolves during the progressive reduction of 29912992 MOLYBDENUM OXIDE CATALYSTS molybdenum trioxide at low temperatures.We also describe how the physical structure of the materials is affected by the experimental conditions used during the preparation. Changes in structure have been followed by measuring nitrogen adsorption isotherms. EXPERIMENTAL MATERIALS The molybdenum metal/molybdenum oxide samples were prepared by the thermal decomposition and reduction of molybdenum(v1) oxalate, Ht [MoO3(C2O4)H2O] . HzO (Climax Molybdenum).Details of the kinetics of the reduction reaction have been described elsewhere.6 Hydrogen for reduction, and nitrogen for the adsorption isotherms, were B.O.C. 99.999 % purity, and were used without further purification, except to bleed the gases slowly into the apparatus through a liquid nitrogen trap. The reduced oxides were invariably extPemely pyrophoric, which indicated that the purity of the gases was completely satisfactory. EQUIPMENT All the decomposition, reduction and nitrogen adsorption experiments were performed in situ in a Cahn R.G. electrobalance. Samples of molybdenum oxalate were ground and sieved (< 150 pm mesh size) and portions of about 100 mg placed in the bottom of a platinum crucible (8 111111 diameter by 5 mm deep) suspended by a nichrome wire from the arm of the microbalance.The microbalance was fitted with a vacuum jacket and could be evacuated to Pa using an Edwards E02 oil diffusion pump. For the nitrogen adsorption isotherms pressures were measured with a Bell and Howell pressure transducer. The temperatwe of the liquid nitrogen bath was measured periodically with a nitrogen vapour pressure thermometer. To ensure that the temperature of the sample corresponded as cIosely as possible with the temperature of the cryostat, the sample was suspended just above the flat bottom of a narrow glass hangdown tube (outside diameter 14 mm). The hangdown tu& in turn was immersed to a constant depth of 25 cm in the liquid nitrogen. RESULTS AND DISCUSSION DECOMPOSITION PRODUCTS OF MOLYBDENUM OXALATE The decomposition of molybdenum oxalate was performed by warming the samples slowly to 573 K in nitrogen (sample MX-l3N), hydrogen (MX-l2H), or vacuum (MX-29V).In the latter case the rate of heating was controlled so as to maintain the pressure below 5 x Pa. The trioxide formed in each case is slightly oxygen deficient, and the compositions determined from the weight increase observed after heating the samples to constant weight in air at 800 K were MOO^.^^ (nitrogen), Moo2+ (hydrogen), MOO^.^^ (vacuum). When decomposing the oxalate in hydrogen the reaction was stopped at M002.g6 by quenching the sample. The nitrogen adsorption isotherms obtained for these oxides are shown in fig. l(a). The isotherms have little in common except for the presence of a hysteresis between the adsorption and desorption branches of the isotherm, indicating some mesoporosity in all the samples. However, both the size and particularly the shape of the hysteresis loops are quite different.The hysteresis loop for MX-13N approxi- mates to type A in the de Boer classification which would indicate that the pores were cylindrical, whereas the hysteresis loops for MX-12H and MX-29V are more like type B corresponding to slit shaped pores. The difference between MX-13N and MX-12H may be due to the higher oxygen deficiency in MX-12H. Although the bulk compositions only differ by 0.02 in the oxygen to molybdenum ratio the surface may have undergone a proportionately much larger change.R. BURCH 2993 The overall shape of the isotherms again show differences.The isotherm for MX-13N is almost type I1 in the Brunauer Deming, Deming and Teller classification,* MX-12H is type IVY and MX-29V is almost type I. These differences in shape reflect differences in the pore size distribution. The isotherm for MX-13N rises steeply throughout the range of relative pressures 0.3 c P/P, c 0.95, indicating the presence of pores right across the size range from small micropores (diameter (2.0 nm) to large mesopores (diameters up to 50 nm). The isotherm for MX-12H has a similar upward curvature at low relative pressures, but levels off ai high vdues of PIP,. The pore distribution in this case is shifted towards smaller pores. For MX-29V the whole balance of the pore distribution has tipped towards microporosity and there are now very few mesopores.The subtle differences in the pore size distributions of these trioides can be described far more precisely by using the a=-method of isotherm analysis. I I I I 0 4 08 0.k 0.8 PiPo FIG. 1 .-Nitrogen physisorption isotherms for molybdmum oxides : (a) molybdenum trioxides prepared by decomposing molybdenum oxalate in hydrogen (A), in nitrogen (B), and in w c u m (C) ; 0 adsorption, x desorption ; (b) samples MX-3(a) and MX-3(b) ; (c) sample MX-7 ; (4 sample MX-10 ; (e) standard isotherms for iron molybdate (1) and non-porous molybdenum trioxide (2) ; (f) samples AM-1N and AM-2N ; (g) samples MX-1 l(a) (curve E), and MX-1 l(c) (curve F) ; (h) passf- vated and active reduced oxides : G, active sample ; €3, passivated sample ; J, annealed sample.2994 MOLY BDE N UM OXIDB CA'IALY S TS (&-METHOD The as-method 9-13 provides a simple graphical method of comparing the various features of the isotherm of an uncharacterised material with the isotherm obtained under the same experimental conditions for a non-porous standard reference material.A suitable reference material is one whose surface chemistry corresponds as closely as possible to the surface of the unknown sample. This ensures that the interaction between adsorbate molecules and the external surface of the unknown sample will be similar to the interaction between adsorbate molecules and the surface of the non-porous reference. Differences between the two isotherms can therefore be attributed to differences between the physical structure of the unknown sample and the non-porous reference material.From the standard isotherm the amount adsorbed (WJ at each value of relative pressure (PIP,) is converted into a " reduced " adsorption by dividing W, by Wo.4, where Wo.4 represents the amount adsorbed at a reference pressure P/Po = 0.4. The values of Ws/Wo., thus obtained are called as-values. The isotherm for the unknown sample (2) is analysed by plotting Wz against Q. Deviations of this a,-plot from linearity are interpreted in terms of micropore filling or capillary condensation depending on the nature of the deviations. The choice of a good standard isotherm is crucial to the as-method of isotherm analysis. As a check we have used two isotherms, one for an iron molybdate sample and one for a molybdenum trioxide sample which was prepared by decomposing ammonium dimolybdate. Both isotherms appear to have the correct characteristics [fig.l(e)]. The shapes of the isotherms and the absence of hysteresis are typical of non-porous materials. The samples have low surface areas (14.5 and 11.0 m2 g-l, respectively) which is also consistent with a non-porous structure. Furthermore, the surface of both of the reference materials should be fairly similar chemically to the surfaces of our molybdenum samples. Qs-CURVES FOR MOLYBDENUM TRIOXIDES The a,-curves for samples MX-l3N, MX-12H and MX-29V are shown in fig. 2(a) together with the as-curve for sample AM-1N. This latter sample, which has been referred to earlier, was prepared by the thermal decomposition of ammonium dimolybdate in nitrogen at 623 K.The linear as-plot for AM-1N confirms that the isotherm for iron molybdate is a satisfactory standard. Curve C shows a small upward deviation from linearity at PIP, = 0.15 which indicates micropore filing in pores whose diameters range from very small through into the mesopore region. This is followed by some capillary condensation in mesopores but the upper limit where the mesopores are all filled only extends to a diameter of 7 nm. The final almost flat portion of curve C corresponds to multilayer adsorption on a small external surface. Curve A is similar to curve C in exhibiting on upward deviation at PIPo = 0.15, but the upper limit on the mesopore diameter is slightly higher at about 10 nm. Multilayer condensation on a small external surface is again observed.Curve B exhibits an upward deviation at much higher values of relative pressure (PIP, = 0.35), so the proportion of micropores is very much smaller in this sample. The upper limit on the pore diameter in the mesopore range is now >20 nm. There- fore, in sample MX-13N there is a distribution of pore sizes ranging from micropores right through to very large mesopores. Some of the properties of these trioxides are summarised in table 1. A comparisonR. BURCH 2995 between SBET and Sp, and between Vo.9J and Vp, confirms that the degree of micro- porosity increases in the order MX-13N < MX-12H < MX-29V. The external surface area decreases in the reverse order, with the decrease from MX-13N to MX-12H being particularly noticeable.The external surface area of MX-13N (table 1) is 26 m2 g-l. From X-ray line broadening we have estimated an average particle size for this material of 40-45 nm which corresponds to a surface area of 30m2g-l for cubic particles. The agreement between these two surface areas is very good. From the nitrogen adsorption isotherms on the trioxide samples we can obtain valuable information on the way in which the preparative conditions affects the structure of the trioxide. In wet nitrogen an essentially mesoporous trioxide is US FIG. 2.--as-plots for molybdenum oxides : (a) A, sample MX-12H ; B, sample MX-13N ; C, sample MX-29V; D, sample AM-1N; (b) reduced oxides x, MX-10; 0, MX-7; 0, MX-3(a); 0, MX-3(b) ; (c) samples MX-ll(a), (curve E), and MX-1 l(c), (curve F) ; (d) active (curve G), passivated (curve H), and annealed (curve J) samples; (e) samples AM-1N (lower curve) and AM-2N (upper curve).TABLE PROPERTIES OF THE TRIOXIDES MX-l3N, MX-l2H, MX-29V, AM-1N surface areafmz g-1 pore volumefcm3 8-1 sample sBBTa SETb Sme VP* YO.95. MX-13N 99 26 100 0.12 0.12 MX-12H 119 9.5 90 0.09 0.11 MX-29V 41 3.6 20 0.02 0.03 AM-1N 11 11.6 - - a SBET is the surface area, calculated from the nitrogen isotherm using the B.E.T. method. S ~ T is the external surface area, calculated from the as-plots. C Sp is the surface area of all the pores except micropores, calculated from the nitrogen desorption isotherm. d Vp is the volume of the mesopores, calculated from the nitrogen desorption isotherm. e Vo.9s is the volume of all the pores, calculated from the nitrogen isotherm.2996 MOLYBDENUM OXIDE CATALYSTS obtained.In wet hydrogen a slightly more oxygen deficient trioxide is obtained which contains some mesopores but also a significant number of micropores. Under vacuum conditions a very microporous material with a small external surface area is obtained. We conclude from this that hydrogen reduction generates slit-shaped micropores in mesoporous molybdenum trioxide, and that water vapour causes a significant proportion of the micropores to be widened out to mesopore dimensions. DEVELOPMENT OF STRUCTURE DURING HYDROGEN REDUCTION REDUCTION I N THE PRESENCE OF WATER VAPOUR 1. OXALATE-DERIVED POROUS TRIOXIDES. Fig. l(b), (c) and (d) show the nitrogen adsorption isotherms, and fig. 2(b) shows the corresponding a,-plots, for reduced samples prepared from " nitrogen-decomposed " oxalate according to the conditions of temperature and time summarised in table 2.These samples were prepared by reduction in static hydrogen so that water vapour produced in the reaction would remain in the vicinity of the samples. TABLE 2.-sAMPLES OF REDUCED MATERIALS PREPARED IN THE PRESENCE OF WATER VAPOUR heat treatment time at temperature time at higher mol % sample 573 K/h raised to/K temperaturelh Mo metal MX-7 23 - 18.4 MX-3(a) 20 773 1 52.0 MX-10 8 673 1.5 52.6 - a MX-3(b) 798 5 90.0 aMX-3(b) was prepared by further reduction of MX-3(a). bA~~urning only Mo and Mooz are present in reduced oxides. Both the temperature and time of reduction affect the structure of the reduced oxides, although the temperature is more important.Thus MX-7 and MX-3(a) have almost identical isotherms and a,-plots, whereas MX-10 is less microporous and more mesoporous. We note also that the degree of reduction does not appear to be a critical factor in determining the physical structure of the reduced oxides, at least up to a nominal composition of 50 % Mo metal. The main differences between MX-7, MX-3(a) and MX-10 appear to originate in the different thermal treatments. The development of the pore structure during reduction appears to occur in the following way. The initial reduction creates slit-shaped micropores in the trioxide. Providing the temperature is kept low (573 K) these micropores are widened into a range of pore sizes by a surface reconstruction which is catalysed by water vapour.If the temperature is raised to 673 K the average pore size is increased and the effective external surface area is also increased. If the temperature is raised to 773 K some sintering occurs with a concomitant loss of external surface area and a decrease in the number of mesopores. When reduction is continued at temperatures above 773 K there is a considerable loss of mesoporosity but fresh reduction retains the micropore content of the sample. The overall effect, shown by sample MX-3(b), [fig. l(b) and 2(b)], is to reduce the surface area of the sample and to increase the relative proportion of micropores. The aS-plot for MX-3(b) shows very little additional pore filling above PIPo = 0.5, which indicates an upper limit on the pore diameter of about 5 nm.The reduced product has a structure, therefore, which depends on temperature and, at higher reduction levels, on the degree of reduction. Low temperature reduction produces a porous material containing mainly micropores but someR . BURCH 2997 mesopores. Reduction at intermediate temperatures causes the average pore diameter to increase. Prolonged reduction at high temperatures results in sintering with a loss of external area and some blocking-off of the mesopores, but, because new micropores are continuously being created by further reduction, the micropore content remains high. Table 3 summarises some of the relevant properties of these reduced oxides. Comparisons between SBET and Sp and between V,,.95 and Vp confirm the general pattern of the structural changes described above. TABLE 3.-PROPERTIES OF REDUCED OXIDES sample SBET a SEXT a S p a VP" v0.95 " MX-7 158 11.9 110 0.095 0.144 MX-3(a) 144 11.9 110 0.096 0.140 MX-3 (b) 78 7.1 31 0.026 0.058 MX-10 162 26 106 0.107 0.159 MX-ll(a) 174 9.5 68 0.054 0.115 MX-ll(C) 111 8.3 62 0.040 0.070 a See caption to table 1.2. NOWPOROUS TRIOXIDES. The trioxide samples described above had porous structures and although it has been implied that hydrogen reduction initially creates microporosity this does not show up very clearly because of the complexity of the structure of the starting material. The effect of hydrogen can be demonstrated more clearly by considering the reduction of a non-porous trioxide (AM-1N). On reduction at 733 K a sample (AM-2N) containing nominally 6 % molybdenum metal was obtained. The isotherms in fig.l(f) and the a,-plots in fig. 2(e) show very clearly the creation of micropores but the complete absence of mesopores. It does not appear from sample AM-2N that micropores broaden into mesopores during the normal reduction process. Mesopores may be created by a secondary structural transformation involving the coalescence of adjacent micropores. The presence of mesopores in the high surface area reduced oxides, but not in the low surface area oxides, presumably arises because in small particles the average spacing between the micropores is small and so there is a good chance of finding micropores close enough together to coalesce. A simple 3-dimensional model made to represent as closely as possible the physical properties of sample MX-10 shows that between 30 and 40 % of the micropores would be close enough to another micropore to coalesce.REDUCTION IN ABSENCE OF WATER VAPOUR A set of experiments was performed in which the sample pan was surrounded by previously outgassed alumina pellets to maintain a dry environment around the sample. The nitrogen isotherms obtained for the samples reduced under dry conditions, at the temperatures shown in table 4, are plotted in fig. l(g), with the corresponding a,-plots shown in fig. 2(c). Some of the physical characteristics of these samples are summarised in table 3. There are important differences between, for example, MX-ll(a) and MX-10 after approximately the same heat treatment.MX-ll(a) is very much more micro- porous, it has fewer mesopores, and a much smaller external surface area. Further reduction of MX-ll(a) to MX-ll(c) causes a decrease in the B.E.T. surface area, a decrease in microporosity, but very little change in the mesoporosity.2998 MOLYBDENUM OXIDE CATALYSTS The differences between samples MX-ll(a) and MX-10 suggest that in the presence of water vapour some micropores are broadened into mesopores, possibly by coalescence as described earlier. A similar effect of water vapour was noted earlier when we described the differences between the trioxides decomposed under vacuum and in hydrogen. The dry oxide which was prepared under vacuum was much more microporous. TABLE 4.-sAMPLES OF REDUCED MATERIALS PREPARED UNDER DRY CONDITIONS heat treatment sample 573 K/h raised to /I( temperaturelh Mo metal time at temperature time at higher mol % MX-ll(a) 4 623 1.1 24.4 MX-ll(c) a 673 22.0 80.0 a MX-1 l(c) was prepared by further reduction of MX-1 l(a).PASSIVATION OF REDUCED OXIDES The reduced oxides described above are extremely pyrophoric. They can be passivated by treatment with very dilute oxygen+argon mixtures. It is of interest to examine the effect of passivation on the physical structure of the samples. Fig. l(h) shows the nitrogen isotherms (curve G is a desorption isotherm) before (curve G) and after (curve H) a passivation treatment for a sample having a nominal composition of 59 % Mol41 % Moo2.* The corresponding a,-plots are shown in fig. 2(d). The similarity between curve G [fig.l(h)] and the earlier isotherms for samples MX-11 and AM-2N shows that the pyrophoric material has a structure which is dominated by microporosity. Curve G [fig. l(h)] is almost exactly 50 % larger than curve H across the whole range of relative pressures. This shows that the passivation treatment has affected all of the surface features of the sample equally. Approximately one third of (i) the external surface area, (ii) the number of mesopores, and (iii) the number of micropores, has each been eliminated by the passivation reaction. The B.E.T. surface area is also reduced by a similar amount, from 124 to 84 m2 8-l. EFFECT OF ANNEALING After heating the passivated sample in nitrogen at 823 K for 19 h the adsorption isotherm shown as curve J in fig. l(h) was obtained.The corresponding a,-plot, in fig. 2(d) shows that there have been significant changes in the structure as a result of this annealing treatment. The a,-plot has a slightly shorter initial linear portion and at a, = 0.8 the curve deviates upwards from linearity. This indicates that the balance between microporosity and mesoporosity has shifted in favour of meso- porosity. The annealing treatment has caused the elimination of many of the micropores, and possibly a small proportion of the micropores have been widened to mesopore dimensions. However, the external surface area has not been much affected by the annealing treatment, as evidenced by the close similarity in the a,-plots for samples H and J over the range 1.6 < a, < 2.5. SUMMARY The evolution of the structure of molybdenum oxides obtained from the thermal decomposition and reduction of molybdenum oxalate can be summarised as follows.* I am grateful to the Climax Molybdenum Co. Ltd for supplying a sample of the passivated material, and for the desorption isotherm on the active material.R. BURCH 2999 In the absence of water vapour decomposition of molybdenum oxalate at 573 K produces an essentially microporous material, but in the presence of water vapour both micropores and mesopores are obtained. Reduction of the trioxide at 573 K creates micropores initially, but further reduction, or again possibly the presence of water vapour, causes some pores to be widened into mesopores. In contrast, the reduction of a low surface area trioxide produces only micropores.Passivation of the active surface of a reduced product results in almost identical decreases in the number of micropores, the number of mesopores, and the B.E.T. surface area. Finally, annealing reduced samples at 823 K results mainly in a loss of microporosity and some increase in mesoporosity. I thank the S.R.C. for an equipment grant. I am grateful to the Climax Molybdenum Co. Ltd who first suggested this work, for their continuing support. We are grateful to Prof. K. S. W. Sing for many stimulating discussions concerning this work. J. R. Anderson, Structure of Metallic CataZysts (Academic Press, London and New York, 1975). J. F. Taylor and J. K. A. Clarke, 2. phys. Chem., 1976, 103,216. ’ 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 Arnax Inc.), May 28, 1976. G. A. Tsigdinos and W. W. Swanson, Ind. and Eng. Chem., to be published. ’ R. Burch and P. C. H. Mitchell, J. Less-Common Metals, 1977, 54, 363. R. Burch, J.C.S. Faraday I, 1978,742982. ’ J. H. de Boer, The Structure and Properties of Porous Materials (Butterworth, London, 1958), p. 68. S. Brunauer, L. S. Deming, W. S. Deming and E. Teller, J. Amer. Chem. SOC., 1940, 62,1723. K. S. W. Sing, in Proc. Int. Symp. Surface Area Determination (Butterworth, London, 1970), p. 25. lo F. S. Baker, J. D. Carruihers, R. E. Day, K. S. W. Sing and L. J. Stryker, Disc. Faruday SOC., 1971, 52, 173. M. A. Alaria Franco and K. S. W. Sing, AnaZes de Quim., 1975,71,296. K. S . W. Sing, Ber. Bunsenges.phys. Chem., 1975, 79, 724. (Academic Press, London and New York, 1976). l 3 K. S. W. Sing, in Characterisation ofPowder Surfaces, ed. G. D. Parfitt and K. S. W. Sing (PAPER 8/521) 1-95