~

ELSEVIER Applied Catalysis A: General 157 (1997) 173-193

APPLIED CATALYSIS
A: GENERAL

Vanadium phosphorus oxides for n-butane oxidation to maleic anhydride

M i c h e l Abon*, J e a n - C l a u d e Volta
lnstitut de Recherches sur la Catalyse-CNRS, 2 Avenue Albert Einstein, 69626 Villeurbanne C~dex, France

Abstract This publication presents recent developments of the knowledge of the vanadium phosphorus oxides used as catalysts for n-butane oxidation to maleic anhydride. We emphasize particularly the nature of the active phase, the active centres and the role of the redox and acido-basic properties contrasting informations between catalysts presenting different degree of disorder. The nature of the active oxygen species and the mechanistic aspects of the reaction are discussed from studies conducted in our laboratory and by comparison with the present knowledge.
Keywords: V-P-O catalysts; n-Butane selective oxidation; Maleic anhydride

I. Introduction

Maleic anhydride is an important intermediate for chemical industry, particularly for the production of unsaturated polyester resins, agricultural chemicals, lubricating oil additives and pharmaceuticals [1-4]. Productivity from n-butane has now largely replaced productivity from benzene due to a lower cost and for environmental reasons. Different industrial technologies are presently existing which involve fixed-bed, fluid-bed as well as circulating fluid-bed riser reactors [3]. They are all based on the unique vanadyl pyrophosphate (VO)2P207 catalyst. Research on this catalytic system has been largely developed in the last decade and many publications and reviews have been devoted to this formulation [1,3,5-7]. The catalytic performances strongly depend on the characteristics of the vanadium phosphorus oxides (VPO) catalyst which derive from the preparation of the VOHPO4.0.5H20 precursor. The morphology of the crystallites of the
* Corresponding author. 0926-860X/97/$17.00 1997 Elsevier Science B.V. All rights reserved. P I I S 0 9 2 6 - 8 6 0 X ( 9 7 ) 0 0 0 1 6- 1

174

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

precursor can be controlled by following different preparative routes. In the aqueous medium route, V205 is reduced by a mineral agent (HC1, N2H4...) in water and phosphoric acid is added [8]. In the organic medium route V205 can be reduced by an organic reagent like methanol, tetrahydrofuran or isobutanol prior to the addition of phosphoric acid [9,10]. The main precursor compound prepared is, in both cases, the vanadyl phosphohemihydrate VOHPOa.0.5H20 (V 4+) [10]. An alternative to the organic route is to reduce a vanadium phosphate (V 5+) like VOPO4.2H20 by alcohols [1 l]. By changing the nature of the alcohol, it is possible to control the morphology of the final precursor or even to change its nature [11]. The precursors prepared according to the organic route give catalysts which present, after calcination, a higher specific area [15-50 m 2 g-l) as compared to the aqueous medium route (10 m 2 g-a). The preparation of the VPO catalysts has been the subject of extensive studies. Many parameters have been examined particularly the nature of the reducing agent and solvent, the P/V ratio at the level of the preparation and the activation conditions [3,6,7]. All these parameters strongly influence the final phase composition of the catalysts and particularly the distribution of phosphorus and vanadium and its oxidation state. Presently, there is an agreement on the specific role of (VO)2P207 but the real nature of the active vanadyl pyrophosphate as function of the activation conditions and the characteristics of its surface composition is still a subject of discussion. It clearly appears that the surface chemistry and the bulk properties of the active catalyst change with the time of activation. This allows to stress the differences observed between catalysts for short period of activation (100 h) (non-equilibrated VPO catalysts) and catalysts which have been working for longer periods in industrial plants (equilibrated VPO catalysts). Kinetics and mechanism of the reaction will strongly depend on these conditions. Before presenting recent results obtained in our laboratory concerning the active phase, the active sites, the redox and the acido-basic properties of the VPO system, we summarize some general informations on this system.

1.1. Structure of the VPO phases

There are different VPO phases with vanadium in the +5, +4 and +3 oxidation state and for which structures have been resolved [12,21]. The V(5+) phases correspond to hydrates like VOPO4.H20 [13], VOPO4.2H20 [14,15], phosphates VOPO 4 (Oq, Oqi , /~, "(, (5) [12,13,16]. The V(4+) phases correspond to hydrogenophosphates like VOHPO4"0.5H20 [ 10,17], VOHPO4.4 H20 [ 10,18] or VO(H2PO4)2 [ 19], to pyrophosphate (VO)zP207 [9], to metaphosphates VO(PO3)2 [12,20]. The V(3+) phases correspond to gPO4 [12,21], V(PO3)3 [20]. The vanadyl hydrates and hydrogenophosphates are generally considered as catalysts precursors. The structure of VOHPOa.0.5H20 is constituted of pairs of

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193 VO 6

175

octahedra sharing a common face. Couples of octahedra are connected together through PO4 tetrahedra, forming the (0 0 1) planes. In one octahedra pair, the V=O bonds are in cis position. Between the (0 0 1) planes, HzO molecules are connected through hydrogen bonds. (VO)2P207 [12,22,23] presents a structure in which two octahedra are joined by edges. Octahedra pairs are connected by PO 4 tetrahedra which gives a layer structure in the (1 0 0) plane. In opposition to VOHPO4.0.5H20, the V=O bonds in the octahedra pairs are in trans position, and the layers are connected together by pyrophosphate groups. The transformation VOHPO4.0.5H20 to (VO)2P207 occurs according to a reaction which has been proposed to be topotactic [13] with the elimination of two water molecules. The V(5+) phases are built with isolated tetrahedra connected by PO4 tetrahedra. The structures of 6- and "7-VOPO4 have not yet been resolved while it exists in some proposals [12,24]. 1.2. Reaction scheme The limiting step for n-butane oxidation has been proposed to be the cleavage of a C-H bond from a methylene group [25]. The transformation of n-butane to maleic anhydride should be done according to an olefinic route in which the intermediates are at equilibrium with the gaseous atmosphere [26,27]. For other authors, an alkoxide route was proposed in which the intermediates are connected to the VPO catalyst by V - O bonds all along the reaction scheme [28,29]. A concept of V 5+ ensembles on the (VO)2P207 matrix was then proposed. Their size, their number and their location was considered to be very important to control the local VS+/V 4+ density and the catalytic performance. The rate of desorption of the olefinic intermediates and furan was observed to be low as compared to the rate of their transformation to MA [30,31] which implies that MA should be rapidly desorbed to avoid its combustion. The experimental conditions (temperature, composition of the flow, contact time, time of activation) will determine the nature of the active sites. This was illustrated by comparing a (VO)2P207 catalyst (basically V 4 phase with almost no V 5+) and a non-equilibrated VPO catalyst (disordered (VO)2P2OT, V 4+ phase with some residual V 5+ entities) [28]. The presence of residual V 5+ entities on the V 4+ matrix was observed to favour the n-butane transformation to MA. But the optimal VS+/V4+ ratio associated to the best catalytic performance is not known. 1.3. Role of acidity This aspect has been considered to be important all along the reaction scheme and particularly for the activation of the n-butane molecule [26]. A concerted mechanism has been proposed implying the abstraction of two hydrogen atoms on V4+-O 6- sites which play the role of Lewis acid and base, respectively. It has been

176

M. Abon, J.-C. Volta/Applied Catalysis A: General 157 (1997) 173-193

also proposed that the C-H cleavage on n-butane needs the cooperation between a Lewis acid site (V 4+) and an acidic Br6nsted site (P-OH) [32] Poisoning by NH3 or K strongly decreases the n-butane conversion to MA [27,33]. An infrared study of the acid sites using NH3, pyridine and acetonitrile as probe molecules showed the existence of Lewis and Br6nsted sites [34,35]. A correlation was observed between the selectivity to MA and the number of strong Lewis acid sites. This was confirmed by dehydration of isopropanol on these catalysts [32]. It has been proposed that Lewis sites are associated with surface coordinatively unsaturated (CUS) V 4+ ions while Br6nsted sites are associated with P-OH [26,35,36]. The contribution of these two types of acid sites as function of the redox properties has to be clarified.

1.4. Redox properties

Active and selective VPO catalysts present a mean oxidation state of vanadium slightly higher than 4.0 usually associated with a P/V ratio slightly higher than 1.0. They are characteristic of the (VO)2P207 phase. [37-40]. The redox role of V 5+ and V 4+ sites during the catalytic process has been proposed [41]. Further works have considered that the active phase is constituted by the main (VO)2P207 phase (V 4+) with some VOPO4 (V5+) [42-45]. However, the exact nature of the V 5+ species, their location and their connection with the (VO)2P207 matrix and its microstructure is still debated.

1.5. Nature of active oxygen species

The participation of lattice oxygen (Mars-Van Krevelen mechanism) is well known for the catalytic oxidation of hydrocarbons on metal oxides [46]. It can be partly considered to explain the catalytic oxidation of n-butane to MA on VPO catalysts. According to some authors [47], three types of oxygen entities have to be taken into account for the formation of MA: bulk oxygen species for the O atom of the cycle, chemisorbed labile oxygen species for lateral oxygen of the cycle. The same oxygen species should be involved for the formation of CO and CO2. A "pseudo-ozonide" structure has been proposed for the initial activation of n-butane with the chemisorption of 02 on the (VO)aP207 structure [48]. The consequence is that MA should not be formed in the absence of dioxygen which appears to be unlikely since MA is formed in these conditions [49-51]. Using the TAP (Temporary Analysis of Products) technique, two types of oxygen species have been proposed: (i) Bulk oxygen species which should be involved in the n-butane/I-3 butadiene oxidative dehydrogenation and for O insertion in the MA cycle; (ii) adsorbed activated O species O* should be responsible for oxidation to CO and CO2 and for the oxidation of furan to MA [30]. It thus appears that there are many discrepancies in the literature about the role and the nature of the active oxygen species. An explanation for such a problem has

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

177

been given which makes a distinction between fresh (non-equilibrated) and longtime working (equilibrated) VPO catalysts [3]. "In the fresh catalysts, lattice oxygen is very easily lost due to defects in the (VO)2P207 structure", while "with an equilibrated catalyst, lattice oxygen is not as labile" and can perform only specific steps in the reaction scheme which goes from n-butane to MA [3]. In a recent review devoted to the vanadyl pyrophosphate catalysts [7] G. Centi, B.K. Hodnett and G.J. Hutchings stressed the importance of a better definition of the active phase and of the active sites on a molecular scale. The role of the redox and the acido-basic properties were considered as key questions for this system. We were of the opinion that the high complexity of the VPO system made necessary the cooperation between different techniques in order to approach the bulk and the surface properties and reach the local structure of the VPO catalysts, when possible, in situ conditions. With this strategy in mind we developed a research at the Institut de Recherches sur la Catalyse to better reach these goals. We report here the results recently obtained in this laboratory on this matter.

2. N a t u r e o f t h e active p h a s e - a c t i v e c e n t r e s - r e d o x a n d a c i d o - b a s i c properties

The use of complementary physicochemical tools has made possible a better approach of this question. An important problem is relative to the atomic surface composition regarding phosphorous and vanadium and particularly the oxidation state of vanadium and the VS+/V4+ distribution. The role of the structural disorder of (VO)2P207 has been considered also to be very important. In order to point these questions, two main approaches were considered in our laboratory: (i) the comparison of the catalytic and physicochemical properties of different vanadyl pyrophosphate catalysts (obtained by calcination under nitrogen at different temperatures) with the catalytic and physicochemical properties of a poorly crystallized VPO catalyst (obtained by activation under the catalytic reaction flow at lower temperature); (ii) the study of the evolution of the physicochemistry of the VOHPO4, 0.5 H20 precursor in the course of the nbutane oxidation, simultaneously with the control of the catalytic performances. We used the complementarity of several techniques to analyse the evolution of the corresponding vanadium phosphorous oxide materials, like X-Ray Diffraction (XRD), MAS-NMR of 31p,31 p NMR by spin echo mapping, Raman Spectroscopy (in situ conditions). The nature of active oxygen species and the redox properties were studied by oxygen isotopic labelling, electrical conductivity and TPD of oxygen. The surface composition (P/V ratio) was studied using XPS and Low Energy Ion Spectroscopy (LEIS). Indeed, XPS also provided informations on the oxidation state of surface vanadium. Acidic properties were investigated using the adsorption of probe molecules (ammonia and pyridine) by volumetry, microca-

178

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

lorimetry, TPD and FTIR. In addition the different catalysts were compared using the test reaction of isopropanol conversion

2.1. Comparative study of (VO)2P207 and VPO catalysts prepared by activation under the reaction mixture
The precursor VOHPO4.0.5H20 prepared in an organic medium was calcined under nitrogen at 450C, 500C, 600C, 750C and 880C for 40 h to obtain five (VO)2P207catalysts denoted PYRO-450, PYRO-500, PYRO-600, PYRO-750 and PYRO-880, respectively. Fig. 1 presents the XRD spectra of the corresponding catalysts. They are all characteristic of (VO)2P207 [19] with a crystallinity which increases with the temperature of activation. Four catalysts were prepared from the same precursor by calcination at 400C under the reaction mixture (nCaHlo/O2/ He=1.6/18/80.4) (gas flow: 2.4 1 h -1 - VSHV=1500 h -1) for different times on stream (0.1, 8, 84 and 132 h). They are denoted VPO-0.1, VPO-8, VPO-84 and VPO-132, respectively. Fig. 2 shows that their XRD spectra are characteristic of a poorly crystallized (VO)2P207 phase [52]. 31p NMR by spin echo mapping is a powerful technique to analyse the oxidation state of vanadium connected to phosphorous atoms in VPO catalysts [53,54]. The

~
5 10 1$ 2 20 25 30 THETA/degree 35

PYRO 600

PYRO ,150

40

F i g . I . X R D s p e c t r a o f p y r o p h o s p h a t e catalysts.

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

150
s 100 - .

179

.f ~ 5o
0

5 10 15 20 25 30 35 40 45 50

Loo _
"~ 50 0 5 10 15 20 25 30 35 40 45 50
150

150

1oo.
'!. 50.
0

5 10 15 20 25 30 35 40 45 50 5 0 ~

'~.~1 s 100 ~... ~" 50

0

5 I0 15 7.0 25 30 35 40 45 50 2 THETA/degree

Fig. 2. X R D s p e c t r a o f V P O c a t a l y s t s .

pyro-catalysts show a signal in the range 2400-2600 ppm [55] which is progressively displaced from 2400 up to 2600 ppm for the highest temperatures of calcination. This signal is characteristic of P connected to V 4+ atoms in well crystallized (VO)2P207. On the contrary, the VPO catalysts (Fig. 3) display three features: 1. A signal at 2400 ppm, typical of (VO)EP20 7 poorly crystallized [44,53,54]. Note that this signal increases with the time on stream, from VPO-0.1 to VPO132, showing a progressive reduction of the solid towards the V 4+ pyrophosphate phase; 2. A signal at 0 ppm typical of P surrounded by V 5+ in the VOPO4 phases; 3. An intermediate signal in the range 200-1200 ppm. This signal has been tentatively attributed to P in a complex environment including both V 4+ and V 5+ ions in a poorly crystallized structure [52]. An examination of the corresponding catalysts by 31p MAS-NMR allowed to discriminate between the VOPO 4 phases (6- and oqI-VOPO4) and indicated a progressive decrease of the 6-VOPO4/~n-VOPO4 ratio with the time of activation,

180

M. Abon, J.-C Volta~Applied Catalysis A: General 157 (1997) 173-193

18
16

2400

14

12 ~. =,

,g
/ o-"

. ......

o/

...

, 2950

, 2350

, 1750

, ItS0

, 550

, -50

"" -650

3550

8 (ppm) / H3PO4 Fig. 3. 3]p NMR spectra by spin echo mapping of VPO catalysts (spectra normalized with respect to the peak at 0 ppm).

vli

Precursor
1: Oxydehydratafion 2 :Topotactic transformation

Catalyst
3,5: Reduction V(V) to V(W) 4: Isovalence transformation

Fig. 4. Scheme of the evolution of the VPO catalyst with activation time under the reaction mixture.

simultaneously with the reduction to (VO)2P207 [52]. The whole set of result allowed to present a comprehensive scheme of the evolution of a VPO catalyst with activation time, as shown in Fig. 4. It may be recalled that previous studies by in situ Raman spectroscopy [56,57] have ascertained the presence of 6- and a i r VOPO4 in a VPO catalyst activated under a butane (1.6%)/air mixture. Fig. 5 shows a typical Raman spectrum of a VPO catalyst after 360 h on stream, with bands characteristic of (VO)zP207, t~- and oqI-VOPO4. The same VPO catalysts have been examined by Kiely and coworkers using electron microscopy and electron microdiffraction [58]. The hemihydrate precursor exhibit a lozenge shape exposing (0 0 1) and (2 1 0) faces. The transformation of the precursor to (VO)2P207 occurs according two routes: (i) A direct epitaxial transformation along the (2 1 0) facet planes leading to the development of a pyrophosphate fringe at the edge of the precursor crystallites; (ii) An indirect transformation occurring in the interior of the hemihydrate platelets leading first to t~-VOPO4 which then leads to rectangular shaped (VO)2P207 crystallites.

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

181

800

i000

v(cm-1)

1200

Fig. 5. Raman spectrum of a VPO catalyst after 360 h under the reaction mixture.

These observations at a microscopic scale are then in quite good agreement with the scheme of evolution of VPO catalysts (shown in Fig. 4), pointing to a progressive reduction with time of the V 5+ phases with therefore an increasing formation of a poorly crystallized (VO)2P207 phase. It was indeed interesting to see if the evolution of a VPO catalyst, as revealed mainly by 31p-NMR, could be evidenced by a surface sensitive technique such as XPS. However, it is well known that the quantitative estimation of the vanadium oxidation states is not a trivial task because the separation in the V2p3/2 binding energy of the V 4+ and V 5+ state (about 1 eV) cannot be achieved. To tackle this problem, we have used a pyrophosphate prepared at 880C and 7-VOPO4 as standards for the calibration of the V 4+ and V 5+ binding energies measured, respectively, at 516.9 and 518eV, in agreements with others [59-61]. This calibration was then used to measure the V4+/V 5+ ratio on the VPO series from the decomposition of the V2p3/2 level. The V4+/V 5+ ratio steadily increases with activation time, from 1.1 (VPO-0.1) to 1.7 (VPO-132), thus confirming a progressive reduction of the catalyst. Indeed, the butane conversion and the selectivity for maleic anhydride formation greatly increase with time of activation and these changes can be correlated with the reduction of the catalyst surface (XPS V4+/V 5+ ratio) as shown in Fig. 6. These results show that the surface concentration of V 5+ species must not be too high to ensure good catalytic performance. On a real industrial VPO catalyst activated for 2000 h on stream, it is already well known, that only (VO)2P207 is usually detected with a mean oxidation state of vanadium close to 4. V 5+ are oxidizing agents with probably a conflicting dual role: active in the oxygen

182

M. Abon, J.-C Volta~Applied Catalysis A: General 157 (1997) 173-193

70

60

till 30
~o
o

40

S(CO)

i!!

s~co2,

s,M ,

ii::
::!i . .

1,2

1,3

1,5 V4+/V5+

1,7

Fig. 6. Evolution of the catalytic performance of a VPO catalyst as function of the surface V4+/V 5+ ratio determined by XPS.

insertion steps (Vs+/V4+ redox couple) but also in the combustion of maleic anhydride and surface intermediates. Fig. 6 suggests that the relative percentage of V 5+ achieved after 132 h of activation is still too high (about 37%). On the series of pure (VO)2P207 catalyst, we found that the V4+/V 5+ ratio, measured in the same way by XPS, increased with the temperature of calcination. The catalytic performance, as shown, in Fig. 7, goes through a maximum for a v a + / v 5+ surface ratio lying between 5 and 7.

I 0 5"

n-C4

[] AM

"=-", .~." E ~ ~
~2=

4.

3-

2.

1"

Pyro450 Pyro-600

Pyro-750

Pyro-880

I
o

I
; ; ;

I
)
V4+/V5 +

I
; ; ~o

Fig. 7. Intrinsic activities of pyrophosphate catalysts (n-butane conversion and selectivity for MA) as a function of the surface V4+/V 5+ ratio determined by XPS.

M. Abon, J.-C. Volta/Applied Catalysis A: General 157 (1997) 173-193

183

Though additional measurements are required to confirm and precise these data, these experiments strongly suggest that the steady surface concentration of V 5+ must be kept at a fairly low level for best catalytic performances.

2.2. Surface composition of VPO catalysts

Numerous XPS studies agreed on an excess of phosphorus on the surface of these catalysts with a (P/V) atomic ratio around 2. However, it has been claimed by Misono et al., on the basis of an XPS analysis of "VPO glasses" [62], that the actual (P/V) ratio would be close to 1 due to erroneous XPS sensitivity factors. We prepared a VPO glass [55] by heating at 1040C for 2 h a mixture of V205 and H3PO4 (85%) with a bulk (P/V)=I. 1. Quenching the liquid mixture ensures an amorphous solid as checked by XRD. This solid was broken and XPS analysis on inside fragments, using the usual Scofield cross sections and the corrections for the mean free path of photoelectrons led to P/V= 1.9 [55]. We propose that the excess of phosphorus could be a general phenomenon on the surface of phosphates as it has been also recorded on iron phosphate by Bonnet [63] and on VO(H2PO4)2 and VO(PO3)2 by Sananes et al. [21,64]. According to Morishige et al. [65], the excess of phosphorus would be due to the presence of VO(H2PO4)2 on the surface of the precursor leading after thermal treatment to VO(PO3)2 over (VO)2P207 crystallites. As VO(H2PO4)2 is soluble in hot water, we put the precursor VO(HPO4), 1/2H20 in hot boiling water for 15 min before filtration of the boiling mixture using a btichner. The solid was then rinsed with boiling water, dried at l l0C overnight. XPS analysis on this washed precursor gave a (P/V) ratio of 1.72 (instead of 1.85 on an untreated precursor). After calcination up to 750C under nitrogen, the obtained (VO)2P207 solids (previously washed or not) were quite similar with respect to their (P/V) surface ratio: 1.73 and 1.77, respectively. It turns out then that the excess of phosphorus would not be related to the presence of a few layers of VO(PO3) 2 on the surface of the pyrophosphate phase. LEIS is known to be the most surface sensitive technique. Fig. 8 compares the LEIS spectrum given by the VPO-400 solid at the very beginning of the analysis (5 min) and after 130 min under the 4He+ beam (1 keV, 20 nA). Both spectra are similar with the notable exception of the peak resulting in the scattering of 4He+ ions by vanadium. The V peak has grown very significantly after 130 min under the ion beam which is expected to provoke some sputtering. This evolution can be seen more clearly in Fig. 9 which shows the changes with time of the A(P)/A(V) ratio, A(P) and A(V) being the area under the phosphorus and vanadium peak. This ratio decreases in a fairly similar way on both catalysts: PYRO-750 and VPO-400 previously activated 132 h on stream. It is well known that LEIS is very sensitive to shadowing effects and the results strongly suggest that vanadium ions are first partly masked by phosphorous ions. Then this masking effect is progressively removed likely as a result of the sputtering of excess surface phosphorus species.

184

M. Abort, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

300

400

500

600

700

800

Kinetic Energy (eV)

Fig. 8. LEIS spectra for 5 and 130 min under the 4He+ ion beam (1 keV, 20nA, 69=142).
~8
i =50hA .41 I-',

~7 0,6!

~ 0~
~0,4 0~ ~2

'~"

- . . . . . . . ~ro-750 I

10

15 20 25 Time of sputtering (rain)

30

35

Fig. 9. Concentration depth profile with time of sputtering under the 4He+ ion beam (1 keV, 50 nA, 69 =142). A(P) and A(V) are the area under the P and the V peaks, respectively.

The present LEIS analysis is therefore in good agreement with XPS results pointing to an excess of phosphorous on the surface. As already proposed by Ebner and Thomson [66], a possible explanation would be that the surface terminates with pendant groups of pyrophosphates, as also recently depicted on the idealized model given by Cavani and Trifiro [67]. Very recently, the surface composition of VPO catalyst has been re-examined by Coulston et al. by XPD [68]. These authors determined an experimental scaling factor A allowing to determine the (P/V) ratio from the experimental intensity ratio (/p/Iv). This factor A was obtained from a calibration derived from five organometallic complexes with a nominal (P/V) stoichiometry in the range 0.5-2. Using this calibration, a (P/V) ratio close to the bulk value was measured on/3-, 7- and 6-

M. Abon, J.-C. Volta/Applied Catalysis A: General 157 (1997) 173-193

185

VOPO4, on (VO)2P2Ov, VO(PO3)2 and VO(H2PO4) 2. These perplexing results

show that this question is still open to discussion.

2.3. Acido-basic properties

We studied acidic properties by ammonia adsorption at 80C, using differential microcalorimetry coupled with volumetric measurements [69], on three solids: PYRO-880, PYRO-750 and VPO-400 (activated 132 h on stream). VPO-400 and PYRO-750 are characterized by medium-strong acidic sites with a heat of adsorption decreasing from 150 to less than 20 kJ mo1-1 with increasing NH3 coverage. The strength and the number of acidic sites are lower on PYRO-880. As shown in Table 1, the intrinsic activity for the formation of maleic anhydride from n-butane increases with the number of acidic sites, which suggests the participation of these sites in the reaction mechanism as also proposed by Busca et al. [26,27] and Cornaglia et al. [35]. The surface acidity has been confirmed by TPD of ammonia (adsorbed at 80C) [55], yielding a number of acidic sites in fairly good agreement with volumetry measurements, with a ratio (N/V) ranging from 0.3 to 0.9 depending on the temperature of preparation of (VO)2P207 (N----number of adsorbed NH3 molecules, V-number of surface vanadium atoms). The activity of the catalysts was also compared using the test reaction of the conversion of isopropanol performed at 140C in the absence of oxygen [55]. As shown in Fig. 10, the activity for propene formation, and then the number of acidic sites, are maximal on VPO-400 and decrease, for the pyrophosphates series, with the temperature of preparation. Besides propene, diisopropylether was detected as a minor product but acetone formation in small amount was only evidenced for VPO-400, the catalyst characterized by the higher percentage of V 5+ species (presence of 6- and oqI-VOPO 4 as previously indicated). It may be concluded that the basic sites responsible for acetone formation must be either scarce or very weak on efficient catalysts. Furthermore, the results suggest that basic sites could be associated with the presence of V 5+ phases. This assumption was ascertained by the fact that, on "y-VOPO4, the rate of acetone formation is about three times higher than on VPO-400. However, propene remains by far the main product. These observations are in general agreement with those reported by Harrouch Batis and Batis [32].
Table 1 Catalytic activity and number of acidic sites 1 Catalyst PYRO-880 PYRO-750 VPO-400 Ni (x 10 is mol m -2) 0.8 2 3.8 Ni/V 0.1 0.3 0.6 Ai ( 1018 mol 0.6 0.9 2.4
m

2)

Ni: Number of NH3 molecules irreversibly adsorbed at 80C, (Ni/V) ratio: (Ni/number of vanadium atoms on the surface, Ai: intrinsic activity for maleic anhydride formation in the n-butane oxidation at 400C.

186

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

3,0

2,0!
"~
~ Q '~'"

o .....

......

I:::::::: vP ,00
Pyro-600
~

, , ~d2
0,0 / 0,0

'--'. ..... . ..... . ............ ~i~ a~.~a~.~.,.n~.ni~.~.n~.a

]72:1::: r,ro-8 00 [ Pyro-8

, . 3,0

, 6,0

, . 9,0

, . 12,0

, 15,0

Time

(h)

Fig. 10. Activity of pyrophosphate and VPO catalysts in the isopropanol conversion to propene at 140C.

In order to further ascertain the absence of significant basicity, we tried to adsorb acidic probe molecules, CO2 and SO2, using differential microcalorimetry and FTIR. Both techniques were unable to detect a significant adsorption of these molecules on VPO-400 as well as on PYRO-450. FFIR was also used to characterize the nature of acidic sites by ammonia and pyridine adsorption on very thin self-supporting disks of VPO-400 or PYRO-450 [55]. On both catalysts, Lewis and Brtnsted sites have been evidenced in agreement with previous studies [27,35], as shown for example in Fig. 11 relative to pyridine adsorption on VPO-400. After adsorption at room temperature, pyridine was progressively desorbed under vacuum at increasing temperature. A temperature of at least 300C is required to completely desorb pyridine, in agreement with the presence of medium-strong acidic sites. A quantitative analysis showed that Lewis sites (band at 1450 cm -1) are roughly four times more numerous than Brtnsted sites (band at 1545 cm-1), at least for room temperature adsorption. As already proposed [26,27,35], Lewis sites are likely associated with coordinatively unsaturated (CUS) V 4+ ions. As a result of the structure of (VO)2P207, two vanadium pairs, with vanadyl in trans-position, are expected on the surface of the basal (1 0 0) face. Therefore, a theoretical number of one half of surface vanadium could be CUS V 4+ ions. Actually some of these ions could be oxidized to V 5+ or hydroxylated. Brtnsted sites are likely related to acidic hydroxyls, V-OH and P-OH.

2.4. Nature of active oxygen species

As already recalled in the present paper, the participation of lattice oxygen to the forrnation of maleic anhydride has been questioned [30,47,48]. We have reexamined this question using 1802 (98 at%) in a closed-circulation reactor with on-line analysis of products by mass spectroscopy [70]. The reaction was performed at 350C with a large excess of oxygen to comply with the usual industrial conditions. The 180 labelling of MA (maleic anhydride) starts from zero, as shown in Fig. 12, and then increases up to about 80% on a pyrophosphate

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

L

187

1450
L

1610

~i : ==

,<>,oD~\~ , /I:'~ ':~s

11 /l

i I,:,
i!
J,,

Iti
i
t

L t

'Jt J.'

300oc f
I

-"-"~%"~'/~--I I

1700

1600

1500

1400
Wavenumber

(cm "1)

Fig. 11. Evolution of the pyridine F r I R spectra with temperature under vacuum. Pyridine was first adsorbed at 25C on VPO-400.

catalyst (PYRO-750). A similar evolution was recorded on a VPO-400 catalyst. As the MA labelling clearly passes through the origin (Fig. 12), it can be concluded that only lattice oxygen is active for MA formation (we also checked that it is also valid for COx). The relatively rapid increase of the 180 labelling of products can be explained by the participation of lattice oxygen belonging to only a few surface layers, about 4 as estimated from the 180-oxygen balance (180 in the product and total amount of 180 consumed). The participation of lattice oxygen was confirmed by comparing the experimental evolution of the MA isotopic distribution with the calculated molecular mass spectra, assuming that the relative intensity of the four isotopes (mass numbers 104, 102, 100 and 98) is dictated by the surface labelling of the catalyst. As shown in Fig. 13, a good fit is obtained between the experimental and the calculated curves. When the steady-state labelling of MA is achieved, the surface of the catalyst is indeed expected to be strongly enriched in 180 in so far as 160 oxygen vacancies

188

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

80 60
40
O

,;,r,,, 20
* i t i

50

100

150

200

250

300

n-Butane consumed (10-6 moi)

Fig. 12. Evolution of the MA labelling with the course of the reaction (n-C4Hl0+lSO2) at 350C on PYRO-750. F(MA)%: percentage of laO in the MA molecule.

80It ----o--I {~) txp.

~60-

-~ 20"

20

40
-----o-.... e-,-

60
I {lO0) (~p, 1 {lO0) c a k .

80

g~40" .~
~ 20"

~
80" .... I--,

~
[ (1{~) ~ p . I(102) cak.

8o

..=
40-

2O 80

40

60 I {104} exp. 1 {104} o a k .

~60
40

.... I---

20"
t}

m
0 ,v

20

40
Fs(%)

60

80

Fig. 13. Evolution of the experimental and calculated isotopic distribution of MA with the course of the reaction - Fs%: percentage of 180 active oxygen species on the surface of the catalyst.

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

189

t~

2.( '60 F
,6o L
120

is 0
1

0l 300

400

~ 500

i 600

, 700

i 800

~ "

Kinetic Energy (eV)

Fig. 14. LEIS post reaction analysis - 4 H e + (1 keV, 20 nA, 69 =142 )

created by the reaction must be mainly replenished by reoxidation with gaseous 1802. The presence of 180 on the catalyst surface was ascertained by LEIS (4He+, 1 keV, 20 nA, 0=142C), as shown in Fig. 14. However, it is interesting to remark that there are still about 50% of 160 ions on the surface, giving then a direct evidence that only some peculiar oxygen species are active, which are involved in the redox reaction of oxygen insertion. Among these oxygen species, vanadyl oxygens (V=0) are frequently favoured but bridging oxygens, (V-O-V) or ( V - O P), are also proposed as discussed for example by Michalakos et al. [71]. TPD of oxygen on PYRO-750, performed in collaboration with Joly and Mehier, yielded additional information on the nature of active oxygen [72]. Following some surface reduction at 650C under vacuum, the amount of adsorbed oxygen increases with the temperature of 02 exposure from 3500C to 500C. The kinetics of dissociative oxygen adsorption is then activated. TPD spectra indicated a single form of labile surface oxygen which could be the lattice oxygen species active in the n-butane oxidation. It may be stressed on the fact that weakly adsorbed oxygen desorbing at low temperature was not evidenced. As discussed by Iwamoto et al. [73], selective oxidation catalysts including V205, MOO3, Bi2MoO6, do not contain weakly adsorbed oxygen. To end with, it is worth mentioning recent measurements performed by electrical conductivity [74]. As shown in Fig. 15, the electrical conductivity, measured at 400C, abruptly decreases with exposure to n-butane and steeply increases as soon as the hydrocarbon is replaced by oxygen. Fig. 15 shows that these cycles are quite reproducible, giving a direct evidence of a redox mechanism. As expected, an intermediate conductivity level is achieved in the presence of the reaction mixture. Such a behaviour was recorded on PYRO-450 and VPO-400 catalysts. The increase of conductivity induced by oxygen is typical of a p-type semiconductor where holes, or positons p+, would be the main charge carriers. The following electrochemical equilibrium can be postulated:
W 5+ ~ W 4 + -q- p+

(1)

Such an electron vacancy (or hole) can be filled by an electron hopping from a

190

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

-5

C o%

q",o2 q.,o+%

-613 0

8 ~

2"?o2'2
I
I I I I

-9

50

100

150

200

250

300

sequence

duration

(min)

"4'5/C4tH1~ 0 02

C4H10

02 C4H1 0+ Oz

,~-6.s

VPO 400

-7.S

T = 400C
-8.5
0
I I I 1 I

50

100

150

200

250

300

350

sequence duration

(rain)

Fig. 15. Electrical conductivity changes measured at 400C on PYRO-450 and VPO-400 under sequential exposures to n-butane(11 Ton'),oxygen(125Tort) and the reaction mixture.

neighbour anion
0 2- + p+ ~ O -

(2)

Such a phenomenon might be relevant to the mechanism of butane activation involving a C-H bond cleavage:
C4H10 +

p+ ~

C4H 0 + H +

(3)
(4)

If one considers the equilibrium given in Eq. (2), Eq. (3) is equivalent to
C4H10 + O - ~ C4I-I~9+ O 2- + H +

It is interesting to remark that the addition of Eqs. (1) and (2) yields to
V 5+ -~- O z- +--4,V 4+ -~- O (5)

M. Abon, J.-C. Volta~Applied Catalysis A: General 157 (1997) 173-193

191

This equilibrium is known as a charge transfer reaction able to generate a couple V4++O - which could be crucial for the n-butane activation and may be also further steps of the mechanism. Indeed O - is believed to be a reactive oxygen species and CUS V 4+ displays a peculiar reactivity owing to its Lewis acidity as previously discussed. The steady-state surface concentration of reactive (V4-O -) pairs is likely low, but it is usually accepted that the sites responsible of butane activation are scarce but quite active as alkane activation is not easy.

3. Conclusions

The combination of several physicochemical tools which approach both the surface and the bulk properties of the vanadium phosphorus oxides was necessary to gain information on such a complex catalytic system. The surface of the catalyst still poorly understood can be better characterized by the combined use of X-Ray Photoelectron Spectroscopy (XPS) and Low Energy Ion Scattering (LEIS). The bulk properties can be better investigated by techniques such as Laser Raman Spectroscopy (LRS), solid state Nuclear Magnetic Resonance (NMR) (especially 31p NMR by spin echo mapping). Other techniques can be also very valuable like Extended X-Ray Absorption Fine Structure (EXAFS) [75]. It is now demonstrated that the active phase is more complex than the simple (VO)2P207 model and that its surface structure is dependent on the activation conditions under the n-butane/air atmosphere which controls the local VS+/V 4+ distribution. The V 5+ species (with a low density) on the (VO)2PeO7 matrix participate to the reaction scheme. The influence of the oxidation state of a VPO catalyst on the catalytic performance appears to be important as evidenced by Schuurman et al. [76,77]. Besides the redox properties, the acido-basic features appear to be also relevant to the mechanism. It is worth to notice that the VPO catalysts exhibit a fairly strong acidity without any significant basicity. ~SO labelling studies, in agreement with recent Temporary Analysis of Products (TAP) experiments by Kubias et al. [78] show that no adsorbed oxygen species seem to be involved in the formation of maleic anhydride. These results are at variance with previous observations by Gleaves et al. [30]. Further developments are expected from a better knowledge of the microstructure of the VPO catalysts. For example, Gai and Kourtakis have shown structural modifications induced by a reducing environment with O vacancies located in the V - O - P bridges [79], while Wachs et al. have shown that the amorphous layer terminating the (1 0 0) face of (VO)2P207 progressively disappears with time of activation. [80]. Electrical conductivity measurements [74] allowed to propose the V4+-O - pair as possible active sites responsible for n-butane activation.

192

M. Abon, Z-C Volta~Applied Catalysis A: General 157 (1997) 173-193

Since the paper of Schiott et al. [81], theoretical chemistry is active in the field of n-butane oxidation on VPO catalysts. For example, a recent study in our laboratory by Robert et al. [82] deals with the role of mixed valence state of vanadium and the participation of lattice oxygen to the formation of maleic anhydride.

Acknowledgements The authors are indebted to Dr. Kossi B~r~ for its efficient collaboration to this paper.

References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] M. Malow, Hydrocarbon Process., 11 (1980) 149. H. Haase, Chem. Ing. Tech., 1-2 (1972) 80. G. Cenil, F. Trifiro, J.R. Ebner and V.M. Franchetti, Chem. Rev., 88 (1988) 55. F. Cavani and F. Trifiro, Chemtech, (1994) 18. R.A. Varma and D.N. Saraf, Ind. Eng. Chem. Prod. Res. Dev., 18 (1979) 7. B.K. Hodnett, Catal. Rev. Sci. Eng., 27 (1985) 373. G. Centi (Ed.), Vanadyl Pyrophosphate Catalysts, vol. 16, no. 1, Elsevier, Amsterdam, 1993. J.P. Harrison, US Patent 3 985 775 (1976), assigned to Chevron Research Co. R.A. Schneider, US Patent 3 864 280 (1975), assigned to Chevron Research Co. J.W. Johnson, D.C. Johnson, A.J. Jacobson and J.E Brody, J. Am. Chem. Soc., 106 (1984) 8123. I.J. Ellison, G.J. Hutchings, M.T. Sanan6s and J.C. Volta, J. Chem. Soc., Chem. Coma-nun., (1994) 1093. E. Bordes, Catal. Today, 1 (1987) 499. E. Bordes, P. Courtine and J.W. Johnson, J. Solid State Chem., 55 (1984) 270. G.Z. Ladwig, Z. Anorg. Chem., 338 (1965) 266. H.R. Tietze, Austr. J. Chem., 34 (1965) 266. R. Gopal and C. Calvo, J. Solid State Chem., 5 (1972) 432. C.C. Torardi and J.C. Calabrese, Inorg. Chem., 23 (1984) 1308. M.E. Leonowicz, J.W. Johnson, J.E Brody, H.E Shannon and J.M. Newsam, J. Solid State Chem., 56 (1985) 370. Y.E. Gorbunova and S.A. Linde, Dokl. Akad. Nauk. SSSR, 245 (1979) 5864. B.C. Tofield, G.R. Crane, G.A. Pasteur, R.C.J. Sherwood, Chem Soc., Dalt. Trans. 1806 (1975). M.T. Sanan6s, G.J. Hutchings and J.C. Volta, J. Catal., 154 (1995) 253. T. Okuhara and M. Misono, Catal. Today, 16 (1993) 61. P.T. Nguyen, R.D. Hoffman and A.W. Sleight, Mater. Res. Bull., 30 (1995) 1055. E Ben Abdelouahab, R. Olier, N. Guilhaume, E Lefebvre and J.C. Volta, J. Catal., 134 (1992) 151. M.A. Pepera, J.L. Callahan, M.J. Desmon, E.C. Milberger, P.R. Blum and N.J. Bremer, J. Am. Chem. Soc., 107 (1985) 4883. G. Busca, G. Centi and E Trifiro, Appl. Catal., 25 (1986) 265. G. Busca, G. Centi, E Trifiro and V. Lorenzelli, J. Phys. Chem., 90 (1986) 1337. Y. Zhang-Lin, M. Forissier, R.P. Sneeden, J.C. V6drine and J.C. Volta, J. Catal., 145 (1994) 256. Y. Zhang-Lin, M. Forissier, J.C. V6drine and J.C. Volta, J. Catal., 145 (1994) 267. J.T. Gleaves, J.R. Ebner and T.C. Kueehler, Catal. Rev. Sci. Eng., 30 (1988) 49. G. Centi, G. Fornassari and E Trifiro, J. Catal., 89 (1984) 44. N. Harrouch Bails and H. Batis, J. Claim. Phys., 90 (1993) 491. S.J. Puttock and C.H. Rochester, J. Chem. Soc., 107 (1985) 7757. G. Busca, G. Centi and E Trifiro, J. Am. Chem. Soc., 107 (1985) 7757. L.M. Cornaglia, E.A. Lomhardo, J.A. Anderson and J.L. FielTO, Appl Catal. A, 100 (1993) 37.

M. Abon, J.-C Volta~Applied Catalysis A: General 157 (1997) 173-193

[36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82]

193

E Cavani, G. Centi and E Trifiro, J. Chem. Soc., Chem. Commun., (1985) 492. G. Poli, I. Resta, O. Rugged and E Trifiro, Appl. Catal., 1 (1981) 935. B.K. Hodnett and B. Delmon, J. Catal., 88 (1986) 43. B.K. Hodnett, Ph. Permanne and B. Delmon, J. Catal., 6 (1983) 231. H.S. Horowitz, C.M. Blackstone, A.W. Sleight and G. Tenfer, Appl. Catal., 38 (1988) 193. E. Bordes and E Courtine, J. Catal., 57 (1977) 236. B.K. Hodnett, Catal. Today, 1 (1987) 477. J.C. V6drine, J.M. Millet and J.C. Volta, Faraday Discuss. Chem. Soc., 87 (1989) 207. J. Li, G.L. Lashier, G.L. Schrader and Gernstein, Appl. Catal., 73 (1991) 83. N. Guilhaume, M. Roullet, G. Pajonk and J.C. Volta, Stud. Surf. Sci. Catal., 72 (1992) 255. E Mars and D.W. Van Krevelen, Chem. Eng. Sci., 3 (1954) 41. G. Centi, E Trifiro, G. Busca, J.R. Ebner, J.T. Gleaves, in: M.J. Phillips, M. Ternan (Eds.), Proceedings of Ninth International Congress on Catalysis, Calgary, Ottawa, 1988, p. 1538. EA. Agaskar, L. De Daul and R. Grasselli, Catal. Lett., 23 (1994) 339. S. Szakacs, H. Wolf, G. Mink, N. Bertoti, B. Wiistueck, H. LOcke and Seeboth, Catal. Today, 1 (1987) 27. R. Contractor, J. Ebner and M. Mummy, Stud. Surf. Sci. Catal., 55 (1990) 553. G. Emig, K. Uihlein and C.J. H~icker, Stud. Surf. Sci. Catal., 82 (1994) 243. M. Abon, K.E. Bere, A. Tuel and E Delichere, J. Catal., 156 (1995) 28. M.T. Sanan6s, A. Tuel and J.C. Volta, J. Catal., 145 (1994) 251. M.T. Sanan6s, A. Tuel, G.J. Hutchings and J.C. Volta, J. Catal., 148 (1994) 395. K.E. Bere, Thesis no 038-96, University of Lyon, I (1996). E Benabdelouahab, R. Olier, N. Guillaume, E Lefebvre and J.C. Volta, J. Catal., 134 (1992) 151. J.C. Volta, K.E. Bere, Y.J. Zhang, R. Olier, in: S.T. Oyama, J.W. Hightower (Eds.), Catalytic Selective Oxidation, ACS Symposium Series 523 (1993) 217. A. Burrows, C. Kiely, GJ. Hutchings, K.E. B6r6, J.C. Volta, M. Abon, J. Catal, submitted. L.M. Cornaglia, C. Caspani and E.A. Lombardo, Appl. Catal., 74 (1991) 15. T.E Moser and G.L. Schrader, J. Catal., 104 (1987) 99. Ph. Bastians, M. Genet, L. Daza, D. Acosta, E Ruiz and B. Delmon, Stud. Surf. Sci. Catal., 72 (1992) 267. T. Okuhara, T. Nakama and M. Misono, Chem. Lett., (1990) 1941. E Bonnet, Thesis no 263-94, University of Lyon I, 1994. M.T. Sananes, GJ. Hutchings and J.C. Volta, J. Chem. Soc., Chem. Commun., (1995) 243. H. Morishige, J. Tamaki, N. Miara and N. Yamazoe, Chem. Lett., (1990) 1513. J.R. Ebner and M.R. Thompson, Catal. Today., 16 (1993) 51. E Cavani and E Triftro, Chemtech, (1994) 18. G.W. Coulston, E.A. Thompson, N. Herron, J. Catal., in press. K.E. Bere, M. Gravelle and M. Abon, J. Chim. Phys., 92 (1995) 1521. M. Abon, K.E. Bere, E Delichere, Catal. Today, in press. EM. Michalakos, M.C. Kung, I. Jahan and H.H. Kung, J. Catal., 140 (1993) 226. J.E Joly, C. Mehier, K.E. Bere, M. Abort, in preparation. M. Iwamoto, Y. Yoda, N. Tamazoe and T. Seigawa, J. Phys. Chem., 82 (1978) 2564. J.M. Herrmann, E Vemoux, K.E. Bere, M. Abon, J. Catal., submitted. M. Lopez Granados, J.C. Conesa and Fernandez-Garcia, J. Catal., 141 (1993) 671. Y. Schuurman and J.T. Gleaves, Ing. Eng. Chem. Res., 33 (1994) 2935. Y. Schuurman, J.T. Gleaves, J.R. Ebner and M.J. Mummy, Stud. Surf. Sci. Catal., 82 (1994) 203. B. Kubias, U. Rodemerck, H.W. Zanthoff, M. Meisel, in: Fifth European Workshop on Selective Oxidation by Heterogeneous Catalysis, Berlin, 1995, Dechema. EL. Gai and K. Kourtakis, Science, 267 (1995) 661. V.V. Guliants, J.B. Benziger, S. Sundaresan, N. Yao and I.E. Wachs, Catal. Lett., 32 (1995) 379. B. Schiott, K.A. Jorgensen and R. Hoffmann, J. Phys. Chem., 91 (1991) 2297. V. Robert, S.A. Borshch, B. Bigot, J. Mol. Catal., in press.