Kinetics and mechanism of the reduction of VIb transition metal oxides and their oxysalts

. .)urnal of the Less-Common Metals, 54 (1977) 243 - 261 0 Elsevier Sequoia S.A., Lausanne -Printed in the Netherlands

KINETICS AND MEC~~IS~ TION

METAL

OXIDES

AND

243

OF THE REDUCTION OF VIb TRANSITHEIR

OXYSALTS*

J. HABER Research Lu~o~tories Kmk6w (Potand)

of Catalysis and Surface Chern~s~,

Polish Academy

of Sciences,

Summary Results of studies on the reduction of VIb t~nsition metal oxides and their oxysalts are reviewed. The mechanism of the reduction of Moos and WOs in hydrogen was studied by photoelectron spectroscopy. In both cases four different doublets corresponding to 3d or 4d electrons respectively were observed depending on the degree of reduction. On reduction, isolated M4’ ions are at first formed, then clusters of edge-sharing octahedra on shear planes are formed in which pairing of l@ ions occurs due to metal-metal bonding. UPS spectra show that the M4’ ions constitute two types of donor levels in the forbidden energy gap. At a very advanced stage of the reduction metallic molybdenum appears at the surface. Similar spectra were obtained when molybdates with layer structures such as koechlinite were reduced. It was shown that reduction proceeds by the formation of shear planes in the MO-O layers, the reduction of cationic layers being observed only at a very advanced stage of the reduction. The shear plane formation mechanism ensures a very rapid transport of oxide ions in the lattice, the adsorption of the reducing agent becoming the rate-determining step in these conditions. Zero order kinetics are thus followed in quasi-isobaric conditions. Reoxidation follows the kinetic equation derived under the assumption that the incorporation of oxygen into surface oxygen vacancies is the ratedetermining step. When molybdates form structures composed of isolated MO-O polyhedra, their reduction results in the formation of new phases of lower valent molybdenum compounds. In some cases, e.g. cobalt or nickel molybdate, the kinetics are first order against hydrogen and follow sigmoidal curves which are well described by the equation of the autocatalytic reaction in a shrinking sphere model. The autocatalytic effect is due to traces of metallic cobalt, formed in the course of the reaction, which activate hydrogen.

*Presented at the Conference on “The Chemistry and Uses of Molybdenum”, University of Oxford, England, 31 August - 3 September, 1976, sponsored by Climax Molybdenum Co. Ltd. and the Chemical Society (Dalton Division).

244

Introduction Reactions involving the reduction of Group VIb transition metal oxides and their oxysalts have been a subject of considerable interest [ 1 - 111 in recent years for several reasons, one of them obviously being the importance of their industrial application. As examples of the reaction type solid( 1) + gas(l) = solid(2) + gas(2), they offer several advantages for studying the correlation between kinetics of solid state reactions and their mechanism, and in particular the relation to the chemical properties of the reacting solids which is the basic problem of the theory of solid state reactions. Removal of oxygen from the lattice of the VIb transition metal oxides results in the formation of ordered arrays of oxygen vacancies, followed by a very facile rearrangement of the layers of initially comer-linked metal-oxygen octahedra into an arrangement of edge-linked octahedra, resulting in the formation of a shear plane [12]. It was suggested [13] that this process might be responsible for the very rapid transport of oxygen observed in lattices of such oxides as compared with the diffusion of oxygen in close-packed structures of transition metal oxides. However, the mechanism of generation of shear structures and their transformations in the course of the reduction is of great importance for the understanding of many basic properties of IVb, Vb and VIb transition metal compounds. Last but not least, we have pointed out [14] some time ago that the easy evolution of one oxygen ion on transformation from the corner-linked to the edge-linked arrangement of M-O octahedra in the process of formation of a shear plane may be related to the ability of these structures to insert oxygen into the organic molecule in selective oxidation processes of hydrocarbons. Molybdates, tungstates, vanadates, etc. are known as active and selective catalysts of these processes. Oxidation reactions proceed with the redox mechanism, the lattice oxygen ion from the catalyst being inserted into the hydrocarbon molecule and the catalyst then being reoxidised by oxygen from the gas phase [ 151. In steady state conditions, a certain degree of reduction of the catalyst is established, depending on the ratio of the rate of reduction of the solid to the rate of its reoxidation. The steady state degree of reduction may have in turn a pronounced influence on the selectivity of partial oxidation. Information on the kinetics and mechanism of the reduction and reoxidation of such systems is therefore of considerable interest.

Kinetics of topochemical

reactions

Reduction of an oxide system with a reducing gas such as hydrogen or a simple hydrocarbon is an example of a topochemical reaction of the type G, + S1 = G2 + Ss. In the case of more complex systems such as oxysalts, several solid products may be formed as a result of the reduction. The characteristic feature of a topochemical reaction is the fact that it is localised at the interface between solid substrate and solid product, which is called

245

the reaction interface, in contrast to a homogeneous reaction which proceeds in the whole volume. It is therefore convenient to divide the discussion of a topochemical reaction into two parts: (a) a discussion of the flux of the reaction per unit surface area of the reaction interface; (b) a discussion of how the reaction interface changes in the course of the reaction. The total rate R, of the reaction, i.e. the rate measured experimentally, may thus be expressed in the form &(T,cbt)

=

(1)

r(Ci,TF(t)

where r is the specific rate of the reaction, i.e. the rate per unit surface area of the reaction interface, and S is the reaction interface. As the latter changes usually with the time of the reaction, the total rate of the reaction is also a function of the reaction time, i.e. of the degree of conversion. Let us first consider the flux of the reaction occurring at unit surface area of the reaction interface. The overall process of reduction is complex and comprises the transport phenomena to and from the reaction interface as well as chemical transformations such as adsorption and reaction. The following steps may be specified: the external diffusion of hydrogen from the gas phase to the surface zone; the internal diffusion of hydrogen in the pores of the solid; the chemisorption of hydrogen at the surface of the solid substrate; the chemical transformation at the reaction interface with the participation of lattice oxygen ions, resulting in the reconstruction of the lattice; the desorption of water and the internal and external diffusion of water. A kinetic analysis of a series of such consecutive steps is usually based on the steady state approximation. The condition for this approximation to be applied is that in the series of consecutive steps considered the rate constant of one step must be much smaller than the rate constants of all the other steps so that it becomes rate determining. It should be remembered that under steady state conditions all consecutive steps proceed at the same rate which is equal to the rate of the overall reaction. Under such conditions, the reaction flux per unit surface area of the reaction interface may be expressed by the equation r=

hififcj)

(2)

where hi is the rate constant of the rate-determining step and f;:(q) reflects its stoichiometry and depends on the respective concentration parameters; these may be concentration gradient in the case of external diffusion as the ratedetermining step, pressure in the case of adsorption, or surface coverage in the case of chemical reaction. The process of reduction may be described by the following chemical equations : Hz (gas, bulk)-+ Hz (gas, reaction zone)

diffusion

Hz (gas, reaction zone) $

adsorption

2H(ads)

246

2H(ads) + O(surf) + Hz0 (ads)

reaction.

HzO(ads) *

desorption

Hz0 (gas, reaction zone)

Hz0 (gas, reaction zone) + Hz0 (gas, bulk)

diffusion

where O(surf) represents the lattice ion at the surface. In the steady state the following equations must be fulfilled : (3) kaH,

ti

-

eH,

-

ke?i, +kaH,o(l keg

1

=DHo1

eH,0)2h,

-eH,

=

kdH,e&

-eH,Oh,O

+ ko i 7. = IZdH,OeH,O

PH,o ‘-P&o

(4) (5)

(6)

6

where D is the diffusion constant, 6 the thickness of the diffusion layers, p” and p the pressures of the reactants in the bulk of the gas and in the reaction zone respectively, k, and kd the rate constants of adsorption and desorption respectively (ha/k, = b, the adsorption coefficient), and k the rate constant of the reaction. The reaction flux per unit surface area of reaction interface is given by r=kei

(7)

2

By solving the system of eqns. (3) - (6) for different rate-determining steps, as proposed by Rozovski [16], and substituting the value of 8 thus obtained into eqn. (7) the following formulae for the specific rate of the reaction are obtained. For the case when mass transfer is rate determining, the rate is given by

where k, = D/S is the rate constant for diffusion. For the case when the chemical process (adsorption or reaction) controls the rate, the rate is expressed by kan,Pn,

r= ((1 + bH,o&,o)(km,

+ W1’2 + (kH,PH,)1’2~2

When adsorption is the rate-determining step, km, plifies to

r=(1 +

bH,oPH,o+(l/k)1'2(k.H,PH,)1'2}2

(9)

Q k and eqn. (9) sim-

(10)

The expression for the rate of the overall reaction when the chemical reaction at the interface is rate determining may easily be obtained from eqn. (9) if it is assumed that lz,,* S=k which is equivalent to the assumption that adsorption

247

B

0

t

t

Fig. 1. Two types of observed changes of the reaction interface in the course of the reaction, and the related models (a degree of conversion, t time). equilibria

r=

are

established at the surface of the solid. Equation (9) transforms to

kbn*Pn, (1 + bn*OpH*o + (bn pH )1’2}2

(11)

An analogous set of eqiations may easily be obtained for the case when no dissociation of hydrogen takes place but lattice oxygen is extracted directly by the hydrogen molecule. It should be borne in mind that the rate-determining step may change in the course of the reaction (as a result of the evolution of the reaction interface) owing to the change of texture of the reacting system. Let us now pass to a discussion of the second question concerning the changes of the reaction interface during the course of the reaction. They have their origin in several different phenomena. On contacting the solid with the gas phase reactants, the reaction starts and, after a certain time, the first crystal nuclei of the solid product appear. The reaction interface, which we may visualise as the interface between the nuclei of the product and the solid substrate, begins now to increase more and more rapidly by two processes: the growth of the nuclei already formed and the appearance of new ones. At a certain stage of the reaction the crystal nuclei of the product have grown at the surface of the substrate grains to such an extent that they begin to make contact with each other. From this moment the decrease of the reaction interface begins because of the overlapping of the product nuclei and the steady consumption of the substrate grains. A typical sigmoidal curve of the dependence of the conversion on time obtained is shown in Fig. l(a), in which a model of the reaction is also shown schematically. In the derivation of the equation describing the changes of the reaction interface in the period when it increases, two parallel processes must thus be taken into account: nucleation and growth of nuclei. The type of equation obtained depends on the mechanism of the nucleation as well as on the texture of the product nuclei and substrate grains. An excellent discussion of

248

these problems may be found in the monograph of Delmon [17] so we shall quote here only the final result. The volume of the nuclei at a moment t is given by the expression

(12) where (dG/dt), is the rate of nucleation (G is the number of nuclei) at time 7, u,,(t, 7) is the volume at time t of the nucleus which was born at time 7. The term under the integral thus represents the sum of the volume of the nuclei born between the times 7 and dT, and the integral u(t) represents the sum of all these quantities for 7 varying from t1 to t, t1 being the induction period. Substituting the proper expressions for dG/dt and u, and integrating, we obtain u(t) = C(t -

&y+*+l

(13)

where p = 1, 2 or 3 depending on whether the nuclei are mono-, bi- or tridimensional and 4 = -1, 0 or qP depending on whether the nucleation_ proceeds instantaneously, proceeds at constant rate or follows a power law. The constant C depends on the shape of the nuclei. Equation (13) may easily be verified experimentally. Let us take as an example a process in which tri-dimensional nuclei are formed and the nucleation proceeds at constant rate; thenp = 3 and q = 0. Taking into account that the degree of conversion 01is proportional to the volume of the product formed, we obtain (Y1’4= k(t - tl)

(14)

In many cases, however, a different picture is obtained, the reaction interface decreasing continuously from the beginning of the reaction when it has the maximum value. This is usually interpreted in terms of a very rapid nucleation which results in the total coverage of the substrate grain with a thin layer of the product in the first instant of the reaction (Fig. l(b)). The reaction interface then decreases as the substrate grain is consumed in the course of the reaction. This model is often called the shrinking sphere model and the decrease of the reaction interface in the case of spherical substrate grains is given by a simple formula: s = S,(l

- @2’s

Kinetics of the reduction

(15)

of Group VIb transition

metal oxides and oxysalts

A review of the kinetic data [ 1 - 111 of the reduction of Group VIb transition metal oxides and oxysalts shows that there are processes in which the reaction interface develops during the course of the reduction and which proceed initially at a constant reaction interface but then decrease

of , , , , 0

10

20

30 Conversion

4_0

, / 50

1’1oI

60

70 Time

(mini

Fig. 2. The rate of reduction of a-CoMoOd in 30 Torr of hydrogen as a function of the conversion. Temperature of reduction: 1, 480 “C, 2, 500 ‘C; 3, 520 “C; 4, 540 “C (after ref. 8). Fig. 3. Degree of reduction of b-CoMo04 in 30 Torr of hydrogen at 500 “C as a function of time: sample 1 was heated in oxygen at 540 “C for 2 h, cooled to 500 “C and outgassed; sample 2 was outgassed at room temperature and heated in vacuum; sample 3 was heated in vacuum to 300 “C and then in hydrogen to 500 “C (after ref. 8).

because of the conversion of the substrate grains. As an example of the first type of behaviour, Fig. 2 shows results of the measurements of the reduction of a-CoMo04 [S] . The experiments were carried out in the circulation reactor to eliminate the diffusional limitation. The observed rate is plotted as a function of the conversion. Similar results were obtained with b-CoMoO,. The shape of the curves indicates that the reaction interface increases initially because of the nucleation and nuclei growth and then decreases as a result of the decrease in the surface of molybdate grains. The induction time depended strongly on the pretreatment conditions of the sample. This is illustrated in Fig. 3, in which kinetic curves of the reduction of b-CoMoO, after different pretreatments are shown. Curve 1 was obtained after heating the sample at 540 “C for 2 h in oxygen, cooling it to 500 “C, outgassing and then introducing hydrogen. Curve 2 represents the sample which was outgassed at room temperature for 1 h, heated in vacuum to 500 ‘C, outgassed at this temperature for 30 min and then exposed to hydrogen. For curve 3 the sample was outgassed at room temperature and heated in vacuum to 300 “C. Hydrogen was then introduced and the sample was heated to 500 “C. As may be expected, heating in oxygen very effectively removes the nuclei of the reduced phases from the surface of molybdate grains, whereas an initial evacuation results in their generation. This process is greatly enhanced by preheating in hydrogen. In order to find out whether the reaction is controlled by adsorption or by chemical reaction at the interface, the dependence of rate on hydrogen pressure was analysed.

250 0.16

i

0,14-

.z T

0,13-

c

7 0)

g 2% N % 2 a

g

O.lZ-

x O.ll-

g

1.5-

0 e

O.lO-

% 2 9 ‘, ; t

2.Or

5

o.os0.08 -

0.10

0.11

0.12

0.13

0.14

0.15

(R,,.)*((cm3H,)*

0.16 0.17 0.18 mid

1

s '=

1.0 -

g 2 5

0.5-

5 0) e B I

O_ 0

Fig. 4. Effect of hydrogen pressure on the reduction, plotted according to eqn. (17). Fig. 5. Kinetics of the reduction at 20 Torr of hydrogen at 440 “C: I, Bi,MoO,; Bi2Mo209; III, Bi2(MoOp)a (after ref. 5).

As given by eqns. (1) and (ll), in the case when reaction is the ratedetermining step, the total rate k,, is given by R

max ={I

kbHzPH,

+ bH,O&,O + (&,p~,

)1/2}2 smax

II,

at the interface

(1’3)

The maximum value R,,, of the experimental rate is usually used for kinetic analysis in order to eliminate the variation of the second variable of eqn. (l), namely the reaction interface S. It is assumed that the maximum value of the reaction interface is practically independent of such experimental conditions as temperature or hydrogen pressure. The validity of this equation may be checked by rearranging it into the following form: (Rmax/p#2

= a - bRmax112

(17)

where u=

(kbH,S)l” 1 + bH,oPH,o

and b =

(bH, )1’2 1 + bH,oPH,o

a and b are assumed to be constant, because in a circulation reactor the pressure of water vapour may be assumed constant. as shown in Fig. 4, a When (Rmax/.2)1’2 is plotted against R,,,, straight line is obtained, the slope of which is proportional to the adsorption

251

15.

1000-

5 990-

v 5 ‘I g

980-

9io-

01

I

0

1

I

2

4 ,t*lmin)+

6

ll3

1!5

117

179

electronegativity

Fig. 6. Kinetics of the reduction of BizMoOs in 250 Torr of hydrogen: I, 420 “C; II, 480 “C; III, 510 “C (after ref. 5). Fig. 7. Mo=O

stretching frequencies in various molybdates (after ref. 22).

coefficient. It may thus be concluded that it is the reaction at the interface which is the rate-determining step of the reduction of cobalt molybdate. The sigmoidal dependence of the conversion on reaction time has also been observed in the reduction of such molybdates as MnMoO, or MgMoO, [18] as well as in some studies of the reduction of WOs [Z] and MOO, [19]. Although it might also suggest that in these cases the reaction interface develops due to nucleation and nuclei growth in the course of the reduction we shall see later that another explanation may sometimes be offered. Entirely different behaviour is observed when reduction of such compounds as the three bismuth molybdates, BisMoOs, BisMosOs and Bis(Mo04)a, is followed. Figure 5 shows the kinetics of the reduction of these phases carried out in a circulation reactor at 440 “C in quasi-isobaric conditions of 20 Torr of hydrogen [ 51. A linear dependence with time of the amount of hydrogen consumed was observed in all cases when the measurements were carried out at 20 Torr or lower pressures of hydrogen, indicating that the reaction interface does not change in the course of the reduction.

252 TABLE 1 Reduction of bismuth and cobalt molybdates in hydrogen Preparation

T(“C)

Rate ((mol0)

a-CoMo04

500 440 440 440

1.65 1.28 1.68 1.42

Biz(MoO& BizMo209 Bi2MoOG

x x x x

m-2 min -1) 1O-5 1O-2 1o-2 1O-2

Activation energy (kcal mol-I) 25.0 32.5 29.3 18.7

Apparently, the whole surface of the molybdate grains becomes instantaneously covered with nuclei of the reduced phase on contacting it with hydrogen. The degree of reduction attained in the experiments presented in Fig. 5 did not exceed several per cent and the effect of shrinkage of the molybdate grains could not yet manifest itself. The pseudo-zero order of the reduction with respect to hydrogen results from the quasi-isobaric conditions of the experiments. Measurements of the rate of reduction at 440 “C at different hydrogen pressures showed that the reduction is first order with respect to hydrogen. It may be concluded that adsorption of hydrogen is the ratedetermining step. It follows from eqn. (10) that when k, is very much less than k and the water pressure is very low the second and third terms in the denominator become much smaller than unity and the equation simplifies to r = k,p. This conclusion is supported by a comparison of the rates of reduction in hydrogen and in different hydrocarbons which shows that the rate of reduction is proportional to the energy of the hydrogen bonds broken in the course of reduction [20]. When reduction is carried out at higher hydrogen pressures, a linear dependence of the amount of hydrogen consumed with the square root of time is observed. This is illustrated in Fig. 6 which summarizes the results obtained by reduction of Bi2Mo0, in 250 Ton of hydrogen. They may be interpreted by assuming that the product forms a thin compact layer enveloping the molybdate grains. On increasing the pressure of hydrogen, adsorption is accelerated to such an extent that transport of reactants across the product layer becomes rate determining. Two possibilities must be considered here: inward diffusion of gaseous hydrogen to solid product/substrate interface or outward diffusion of oxygen ions from the solid substrate to product/gas interface. The fact that diffusion limitation appears only at high pressures indicates that it is the diffusion of oxygen ions which is important, since both diffusion and adsorption of hydrogen depend similarly on pressure and are accelerated to the same extent by an increase in pressure. The activation energy for the diffusion of oxygen ions, calculated from these experiments,

253

Fig. 8. Formation of a shear plane in the WOa structure (after ref. 12).

amounts to 30 kcal molll, in good agreement with the value of 32 kcal mol-’ obtained from measurements of the rate of reoxidation [6] and the value of 26 kcaI mol-’ calculated from the C?Os exchange experiments with MOO, [21]. of Table 1 summarizes the rates and activation energies of reduction bismuth and cobalt molybdates. It may be seen that the rates of reduction of bismuth molybdates are three orders of magnitude greater than those of cobalt molybdates. In contrast, the wave numbers of the stretching vibrations of double-bonded terminal oxygens Mo=O, as determined by Adzamov et al. [22] have the same value for these two molybdates (Fig. 7) indicating that the oxygen bond strength is the same. It should thus be emphasized that the rates of reduction cannot be taken as a direct measure of the oxygen bond strength, a procedure often adopted in literature. Linear kinetics were also observed in the case of the reduction of MOO, [lo] and WOs [7].

Mechanism

of the reduction

An important question arises at this point as to whether there are any features of the mechanism which could explain why certain kinetics are followed in the case of reduction of compounds of a given group. We shall start with a discussion of the mechanism of reduction of MOO, and WOs, which has been a subject of considerable interest in recent years because of the phenomenon of crystallographic shear [23 - 261. When oxygen ions are removed from the lattice of these oxides by reduction, the oxygen vacancies thus generated form ordered arrays along given crystal planes. After the concentration of vacancies has reached a certain critical value, they become annihilated by rearrangement of the initially comerlinked metal oxygen octahedra into an arrangement of edge-linked octahedra, resulting in the formation of a shear plane. This is shown in Fig. 8. taken from the classical paper of Wadsley [12].

3’5

40

Binding energy{& f

45

225

230 235 binding energy(eVt

t

240

Fig. 9, Changes of XPS spectra of W(4f) level in WOa in the course of reduction in hydrogen. (a) Fresh sample, outgassed at 270 “Cfor 12 h; (b) exposed to 5 x lo-’ Torr of hydrogen for 1 min; (c) exposed to 5 X 10m5 Torr of hydrogen for next 5 min; (d) exposed to 1 X 10e4 Torr of hydrogen for next 15 min; (e) metallic W foil (after ref. 31). Fig. 10. Changes of XPS spectra (Mo(3d)doublet) of MOOQin the course of reduction in hydrogen. (a) Fresh sample; (b) sample after heating for 2 h at 210 “C and bombarding with hydrogen for 3 min at 6 x 10B6 Ton; (c) sample after bombarding for 30 min at 4 x Torr for 60 min at 550 “C; -----10m5 Torr; (d) sample after bombarding at 4 X lometallic molybdenum thin foil (after ref. 30).

As the number of shear planes increases, they form ordered structures of eq~d~~t parallel planes separated by blocks of the initially formed MOs. These are known as shear structures and give rise to an infinite number of phases which may be described by several formulae: MnOsn-r, M, Osn _ s etc, where n is the width of the unperturbed initial blocks between the shear planes. The generation of shear planes and their ordering have been extensively studied by X-ray diffraction and electron microscopy in monocrys~s of both WOs [27] and Moos [28,29]. Interesting information on the changes in bond structure accompanying the formation of shear planes at various stages of the reduction of WOs and MOO, and its bearing on the mechanism of the reduction of their more complex compounds as oxysalts was recently obtained in our laboratories

255

using photoelectron spectroscopy [30, 311. Figure 9 shows the W(4f) region of the spectrum of a WO, sample at different degrees of reduction by hydrogen carried out in situ in the spectrometer [31]. Spectrum 9(a) is the initial sample of WOa and shows the characteristic doublet of W(4f,,s) 6+ ions with binding energies of 36.1 eV and 38.7 and W(4f,,,) electrons in W eV. Spectrum 9(b), obtained after a short reduction time, can easily be deconvoluted into two doublets: the initial W6’ ions and a second one with binding energies of 35.4 eV and 38.0 eV respectively. Deconvolution of spectrum 9(c) obtained after a longer reduction showed that besides the doublet of W6’ ions the spectrum is composed of three other doublets and is different from that observed in spectrum 9(b). Thus, five different tungsten species appearing at various stages of reduction could be identified. Results of an analogous experiment with MOO, are represented in Fig. 10, in which the Mo(3d) region of the spectrum is shown [30]. Spectrum 10(a) of the initial Moos sample shows the doublet of Mo(3ds,s) and Mo(3ds,z) electrons in Mo6’ ions with binding energies of 231.5 eV and 234.7 eV. Deconvolution of spectra 10(b) - 10(d) obtained after various periods of reduction shows that reduction of MOO, gives four different molybdenum species. For a series of simple compounds which consist of similar coordination polyhedra of the given cation and differ only in the valence state of the cation, i.e. in the way in which the coordination polyhedra are linked together, it may be assumed that a linear dependence exists between the binding energy of the core electrons and the valence state. This dependence was thus used for the assignment of the five different W(4f) doublets observed in the course of the reduction of WOs and the four different doublets obtained on reduction of MOO,. Two points are fixed in each case: the positions of W6’ or Mo6* ions and the positions of metallic tungsten or molybdenum, respectively. The results shown in Fig. 11 indicate that tungsten ions of apparent oxidation numbers 5+, 4+ and 2+ and molybdenum ions of apparent oxidation numbers 4+ and 2+ appear as intermediates. In view of the well-known instability of divalent tungsten and molybdenum ions in oxide systems, the appearance of an apparent oxidation number of 2+ must have a different meaning. It may be explained in terms of the following mechanism for the phenomena observed in the course of the reduction of WOs and MOO,. After contacting WOs with hydrogen, W5’ ions are first formed at the surface, probably due to chemisorption of hydrogen which results in the formation of OH groups. Under stronger reducing conditions isolated oxygen vacancies are generated by removal of lattice oxygen ions, the two electrons for each vacancy formed being localised at the adjacent W6’ ions which appear as isolated W4’ ions. Apparently, MOO, contacting with hydrogen results immediately in the generation of oxygen vacancies, Mo5’ ions being unstable under these conditions. When the concentration of oxygen vacancies exceeds a certain critical Value, a transformation of the corner-linked WOs octahedra takes place along the shear planes into the arrangement of edge-sharing octahedra. The W-W distance in the shear plane is considerably smaller than in the normal WO,

256

, 0

.

,

,

4 2 Oxidation number

r

) 6

12262

527.4 --

532.2 529.8

534.6

Fig. 11. Binding energy values of W(4f) (Curve 1) and Mo( 3d) (Curve 2) level as a function of the apparent oxidation number. Fig. 12. Bi(4f), Mo(3d) and O(ls) signals of BigMoOe samples reduced in hydrogen. I, initial sample outgassed at 120 “C; II, after outgassing at 476-‘C for 15 h; III and IV, after exposing to hydrogen at 470 “C for 10 min and 1 h, respectively (after ref. 34).

matrix [27 ] indicating that a metal-metal bond has been formed between the two paired W*’ cations of the adjacent edge-sharing octahedra, the four d electrons being localised on the molecular orb&& of the two cations, which both show the apparent oxidation number of*Z+. In the case of Moos, built of an array of corner-linked zigzag strings of edge-sharing Moos octahedra, crystallographic shear consists of the formation of units of these strings linked by edgesharing of the last three octahedra in each unit along the shear plane 1281. Clusters of edge-sharing octahedra are thus formed, where pairing of MO**ions may take place; this is rn~fest~ by the appearance of the oxidation number 2+ in the electron spectrum. It should be remembered at this point that structural analyses of the transition metal dioxides MOs crystallizing in the rutile-type structure have shown that two different metal-metal distances are observed in these lattices: successive pairs of metal atoms in a string of edge-sharing octahedra are brought alternately closer together and further apart [ 32 f . The distances between adjacent metal atoms in pairs are so short that they indicate the formation of metal-metal bonds between central atoms of edge-sharing octahedra. For the group of dioxides with the Moos structure there is a steady decrease of the distance between the metal atom pairs with increasing number of valency electrons of the metal atoms which are not engaged in met&oxygen bonds [ 331, indicating multiple bond character. The appearance of the Mo(3d) doublet, at binding energy values identical with those of the doublet assigned to paired MO*+ions in the electron spectrum of reduced BisMoOs (Fig, 12), may be taken as an indication

257

that in the case of certain molybdates reduction may also result in the formation of shear planes [34] . In such structures as BisMoOs, built of layers of corner-linked MOO, and BiO, polyhedra, this process can be visual&d as consisting of a transformation from comer-linked to edge-linked octahedra within the MO- layers without any changes of the Bi-0 layers. It seems that the lattice can withstand a considerable concentration of shear planes; the low temperature experiments of reduction with butene have shown that the reduction of Bi2M00s stops after 0.5 mol of oxygen ions has been removed, the solid then having the composition Bi2M005.5 [35]. At higher temperatures this stage cannot be separated and the reduction leads to the reconstruction of the MO-O layers into MOO, nuclei and the collapse of Bi-0 layers with precipitation of metallic bismuth. In fact, as seen from Fig. 12, only reduction of molybdenum is observed in the first stages of the reduction, metallic bismuth appearing later. All these processes may have a pronounced influence on the kinetics of the reduction. It may be expected that the reconstruction of an MO-O layer, in which a large concentration of shear planes has already been formed (resulting in the formation of clusters of edge-sharing octahedra,)in the lattice of MoOa, will be comparatively easy. Nucleation of Moos may be thus very rapid, the reduced phase uniformly covering the surface of the BisMoOs grains and metallic bismuth readily forming small droplets, since it is liquid and very mobile at the temperatures used in the reduction experiment. Under these conditions the reaction interface remains practically constant at the beginning, which explains the observed linear dependence of the conversion with time. It also becomes understandable that activation of hydrogen is the most difficult step, adsorption thus being rate determining. The results of the X-ray analysis indicate that reduction of Bi2(Mo04)s and Bi2M0209 phases leads initially to the formation of Bi2Mo0, and MoOa as the main products [6]. Reduction of BiaMoOs leads to the formation of metallic bismuth and Moos, in agreement with the results of photoelectron spectroscopic experiments mentioned earlier. These processes may be described by the equations Bi2(Mo04)s

reduction

-

reduction

BisMosOa

A

Bi2MoOG

A

reduction

BisMoOs + MoOa BisMoO,

+ MoOa

Bi + Moo2

XPS studies of the three bismuth molybdate phases have shown that their surface is always enriched in MOO, [34] . This may explain why, ,on reduction of BisMoOs, as in the case of the other two molybdates, the nucleation of the reduced phase MOO, is very facile and rapid and therefore the reaction interface remains constant.

258

1

o# /

0

100 Time (mini

1

I

200

0.1

0.2

I

0.3

I

0.4

1

0.5

a

Fig. 13. Reduction of a-C!oMoOd at 500 “C in hydrogen at different pressures: I, 4 Torr; II, 13 Torr; III, 30 Torr (after ref. 9). Fig. 14. Kinetics of the reduction of a-CoMoOd in 360 Torr of hydrogen at different temperatures plotted in terms of eqn. (19) (after ref. 9).

Conversely, reduction of cobalt molybdate results in the formation of molybdite, CoZMo30s, and a phase of the spine1 type, CozMoOI [S]. It may be described by the following equation: 4CoMo04 + 4Hz = Co2M030s + Co2MoOd + 4Hz0 This process requires the comple reconstruction of both cationic and anionic sublattices, although nucleation of the new phases may be difficult. It may be expected that the reaction interface will develop slowly during the course of reduction, thus producing the characteristic sigmoidal kinetic curves. Summarising this part of the discussion, we may advance a general hypothesis that enrichment of the surface of many molybdates and tungstates in MOO, or W03, respectively, and the ability of the latter to form shear structures gives rise to a very facile and rapid nucleation of the product phases in the course of reduction. As a result, the maximum reaction interface is immediately formed and the kinetics may be interpreted in terms of the shrinking sphere model, adsorption being the rate-determining step. The same model explains the linear kinetics of the reduction of MoOa and W03 themselves. Yet in some cases, as already mentioned, sigmoidal kinetic curves were observed by certain authors [2, 191. We shall see, however, that sigmoidal curves do not necessarily mean the evolution of the reaction interface.

259

Catalysis in the reduction of oxides We shall first describe the results of a series of experiments in which the reduction of cobalt molybdate of high surface area was followed with a high temperature X-ray camera [9]. The finely ground sample was spread in the form of a thin layer 0.5 mm thick on the platinum sample holder and was heated to the temperature of the experiment in nitrogen; the rate of reduction was measured in the 1:l mixture of Ns + H2 passed at atmospheric pressure. Optimum conditions were thus created for a very rapid nucleation. The results presented in Fig. 13 could imply that we are dealing with the evolution of the reaction interface. It was found, however, that the reaction is first order with respect to hydrogen, indicating that dissociative adsorption of hydrogen is the rate-determining step of the reaction. The sigmoidal character of the kinetic curves in such cases could be due to the autocatalytic effect of traces of metallic cobalt formed in the course of reduction as highly dispersed clusters, cobalt being known as a good catalyst for activating hydrogen through its dissociative adsorption. The presence of trace amounts of metallic cobalt in the products of the reduction of CoMo04 was indeed observed in the course of microscope studies [36, 371. In order to confirm the autocatalytic model of the reduction of cobalt molybdate, an experiment was carried out in which CoMoO, was mixed mechanically with 5% of Co0 which, in a separate test rwI, was shown to reduce completely to metallic cobalt within a few minutes under the experimental conditions. The results of this experiment clearly showed that the metallic cobalt formed in the first minutes of reduction completely eliminated the induction period. This indicates that a certain limiting surface concentration of active centres of metallic cobalt that activate hydrogen exists, the reduction proceeding through a spill-over effect [9]. Under these conditions a shrinking sphere model may be assumed; the decrease of the reaction rate is due to the decreasing surface area of the reduced grains and is described by eqn. (15). The rate of reduction may thus be expressed by

(18) or, after rearrangement, y=

l

(1 -c+~‘~

1 da p--=h(a, pu,S dt

+acr>

(19)

where k is the rate of the adsorption, a0 the initial concentration of the catalyst required to start the reaction, a(~the concentration of the catalyst after a time t when a conversion (Yhas been attained and S the specific surface area of the solid. Figure 14 shows the experimental data plotted in the form of Y as a function of (Y,described by eqn. (19), for different temperatures. A very good fit of experimental points is obtained for a broad range of conversions from 0.05 to 0.5.

260

The catalytic effect of the addition of various metals such as platinum, palladium and nickel on the reduction of MOO, has been observed by several authors j38 - 431. Recently, Delmon [ 19 ] reported that the addition of cobalt oxide completely eliminates the induction period owing to the catalytic effect of metallic cobalt formed in the first minutes of the reduction. A question may thus be raised as to whether the sigmoidal character of the kinetic curves, observed on reduction of Moos and WOs, could not be explained by the auto~~lyti~ effect of the small amounts of low valent MO or W ions formed at the surface in the course of reduction. The ability to activate hydrogen is not unknown among the low valent cations of Group VIb, CrsOs being used as the hydrogenation catalyst.

References 1 P. Barret, in G. M. Schwab (ed.), Proc. 5th Int. Symp. on the Reactivity of Solids, Elsevier, Amsterdam, 1965, p. 492. 2 P. Barret and L. C. Dufour, C. R. Acad. Sci., 258 (1964) 2337. 3 Y. Lapostolle and L. C. Dufour, C. R. Acad. Sci., 270 (1970) 970. 4 K. Aykan, J. Catal., 12 (1968) 281. 5 J. Beres’, K. Briickman, J. Haber and J. Janas, Bull. Acad. Pol. Sci., Ser. Sci. Chim., 20 (1972) 813. 6 K. Briickman, J. Haber and J. Janas, Bull. Acad. Pol. Sci., Ser. Sci. Chim., 21 (1973) 762. 7 M. ~ittingh~ and P. G. Dickens, in M. W. Roberts (ed.), Proc. 7th Int. Symp. on the Reactivity of Solids, Chapman and Hall, London, 1972. 8 J. Haber and J. Janas, in P. Barret (ed.), Reaction Kinetics in Heterogeneous Chemical Systems, Elsevier, Amsterdam, 1975, p. 737. 9 J. Haber, A. Kozlowska and J. Sloczynaki, in Proc. 8th Int. Symp. on the Reactivity of Solids, Gothenburg, 1976, in the press. 10 J. von Destinon Fortmann, Can. Metaii. Q., 4 (1965) 1. 11 M. J. Kennedy and S. C. Bevan, J. Less-Common Met., 36 (1974) 23. 12 A. D. Wadsley, in L. Mandelcorn (ed.), Nons~ichiometric Oxides, Academic Press, New York, 1964. 13 M. O’Keeffe, in van Go01 (ed.), in Fast Ion Transport in Solids, North-Holland Publ. Co. Amsterdam, 1973, p. 233. 14 A. Bieladski and J. Haber, in C. Amberg (ed.), Surfaces of Non-metals, Reidel, in the press. 15 J. Haber, Z. Chem., 13 (1973) 241. 16 A. Ya. Rozovski, Kinet. Kataf., 2 (1961) 374; 3 (1962) 572. 17 B. Dehnon, Introduction a la Cinitique Hiterogene, Edition Technip, Paris, 1969. 18 J. Janas, Ph.D. Thesis, Krakow, 1976. 19 J. M. Zembala, P. Grange and B. Delmon, C. R., Acad. Sci., 279 (1974) 561. 20 C. R. Adams, in Proc. 3rd Int. Congr. on Catalysis, Vol. 1, North-Holland Publ. CO., Amsterdam, 1965, p. 240. 21 V. S. Muzy~n~v, K. C. Cheshkova and G. K. Boreskov, Kinet. K&al., 14 (1973) 432. 22 K. Ju. Adzamov, in Proc. 3rd Conf. on Heterogeneous Oxidation Catalysis, Acad. Sci. Aserbajdshon SSR, Baku, 1976, p. 52. 23 J. S. Anderson, in M. W. Roberts and J. M. Thomas (eds.), Surface and Defect Properties of Solids. Vol. 1, The Chemical Society, London, 1972, p. 1.

261 24 R. J. D. Tilley, in M.T.P. Int. Rev. Sci. Ser. 1, Inorg. Chem., Vol. 10, Butterworths, London, 1972, p. 279. 25 J. S. Anderson and R. J. D. Tilley, in M. W. Roberts and I. M. Thomas (eds.), Surface and Defect Properties of Solids, Vol. 3, The Chemical Society, London, 1974, p. 1. 26 R. J. D. Tilley, in M.T.P. Int. Rev. Sci., Ser. 2, Inorg. Chem., Vol. 10, Butterworths, London, 1975, p. 73. 27 J. G. AIlpress, R. J. D. Tilley and M. J. Sienko, J. Solid State Chem., 3 (1971) 440. 28 L. Kihlborg, Archiv. Kemi., 21 (1963) 443. 29 G. Liljestrand, in Proc. 8th Int. Symp. on the Reactivity of Solids, Gothenburg, 1976, in the press. 30 J. Haber, W. Marchewski, J. Stoch and L. Ungier, Ber. Bunsenges. Phys. Chem., 79 (1975) 970. 31 J. Haber, J. Stoch and L. Ungier, J. Solid State Chem., 19 (1976) 113. 32 A. F. Wells, Structural Inorganic Chemistry, Oxford Univ. Press/Clarendon Press, Oxford, 1962. 33 A. Magneli and C. Andersson, Acta Chem. Stand., 9 (1955) 1378. 34 B. Grzybowskia, J. Haber, W. Marczewski and L. Ungier, J. Catal., 42 (1976) 327. 35 G..C. A. Schvit, J. Less-Common Met., 36 (1974) 329. 36 J. Haber and J. Sloczynski, Proc. Intern. Coll. on Materials Science, Krakow, 1975, Papers of the Commission on Ceramics, Acad. Pol. Sci., 21 (1974) 19. 37 J. Sloczyhski, Bull. Acad. Polon. Sci., Ser. Sci. Chim., 24 (1976) 61. 38 J. E. Benson, H. W. Kohn and M. Boudart, J. Catal., 5 (1966) 307. 39 J. Masson, B. Delmon and J. Nechtschein, C. R., Acad. Sci., 266 (1968) 928. 40 K. M. Sancier, J. Catal., 23 (1971) 298. 41 G. C. Bond and J. P. B. Tripathi, J. Less-Common Met., 36 (1974) 31. 42 D. K. Lambier, T. T. Tomova and G. V. Samsonov, Powder Metall. Int., 4 (1972) 17. 43 G. V. Samsonov, D. K. Lambier, V. Kh. Panagarova and T. T. Tomova, in Proc. 3rd Conf. on Powder Metallurgy, Zaopane, 1971, PWN Warsaw, 1972, p. 325.